This post will describe the building of a complete instrument based on the earlier investigation of the Tektronix SG505 Signal Generator described here:
I'm going to use the same enclosure as I used for the SG502, which is a TEKLA KL-22. This is a black plastic enclosure with a metal front panel and back panel. I would have liked to use a full metal enclosure but the ones I found where either too small, too low, or too large and mostly all of them had the wrong dimensions to use a front panel.
To get a feel for the proper handling of the knobs and the overall layout, I typically create a mock-up using a paper front panel with a card board backing functioning as the PCB.
I made a number of changes to the initial layout to make sure I can turn and switch everything without cumbersome finger acrobatics. I also want to group controls together as much as possible, and make the positions logical. Also the position of the BNC output and the main frequency control contraption needs to be positioned correctly.
Here is that mock-up in the final version, after spending a few days relocating and printing and mounting the controls again. (this is not the final version (see the PCB below), some text got moved a bit here and there)
Here is the view from the back:
And here the detail for the mounting of the main potmeter contraption.
This is an earlier version, the mounting holes for the bracket line up in the final one. (;-))
After finishing this, I could finelize the design of the Front panel PCB for The KL-22 , so it has a nice looking feel for the instrument. Here is the 3D viewer result of the final version:
The rectangle in the top is the cut-out for the OLED display that will show the frequency. There are no mounting holes on the OLED display board itself, you have to find a way to somehow glue it to the back of the front panel.All the holes are isolated from the front and back ground poor to add some more strength and EMI blocking. I will order the black solder mask with white silkscreen lettering as I usually do for all of my front panels.
New PCB Layout
I have also started to create a new layout for the main board, and added the relay's for the range switching, as well as the output amplifier.
Here are the circuits for the frequency range selection, using a simple rotary switch located on the front panel.
This is the frequency setting for the phase shift amps.
Here are the dampening circuits for the AGC and the Peak Detector.
And here is the circuit for the frequency multiplier rotary switch:
A new Power Supply?
I'm very happy with the Jung SuperReg, but it is more than a bit of an overkill for this instrument. I'm now looking into testing a dual shunt supply that I also very recently used for my Twin-T notch filter and the Victor Mickevics oscillator. I'm using Victor's design to start with. I've used LTSpice to learn a bit more about the principle and I think I now know enough to be dangerous and feel brave enough to give it a try. I also need to add the 12V rail for the relay's and feed the Arduino and LCD display.
This will decide whether I'm going to add the components to the main PCB, or create a separate one, and then add the Arduino counter circuit to it.
Well, after some fooling around with prototypes, I finally got it going on two protoboards.
Here is the schematic of the supply:
The prototype below does not have the transformer snubber and the bridge capacitors implemented.
The power LED is optional. If you're going to use the OLED display, it will show that there is power. During testing, I used an LED to tell me that power is on, otherwise I can't see it visually, and before I added it, to my dismay, I left the unit powered on for a whole night. Not good.
The Shunt Supply principle is quite simple, almost literally taken from my Twin-T notch filter design and I only added trimmers for the key voltages and tweaked the current distribution. I'm now also using a much smaller transformer that I can mount on the PCB.
Here is the first result:
For some reason that I do not understand yet, there are more harmonics of the principle visible and you can also see a bit more of the mains frequency. I'm hoping that a real PCB and complete mains and transformer filtering will reduce that.
The top protoboard on the left is the raw supply circuit fed with the ac from the 2x 28VAC transformer. It has the rectification and filtering plus the LM317HVT that feeds the 44V to the dual shunt supply, located on the lower board. The two boards are connected together using the two 91 Ohm resistors that provide the shunt supply current headroom.
I'll do some more long term testing and look at the temperatures, but it looks like this supply is adequate for this application.
I finished the layout of the new power supply. It will be mounted up-side down from the top of the enclosure. There is a normal transformer located on the bottom for the +/- 17V rails, and the top rectangular in black is the 12V DC power module. That circuit is completely isolated from the +/-17V rails.
And finally, here is a 3D picture of the main board:
The 12V supply for the relays, the Arduino and the OLED display is completely separate from the generator circuits itself.
A reconsideration for the front panel
While working on the final layout for the main board and the front panel, I became more and more agitated with the two screws that mount the potmeter reduction unit that are visible on the font panel.
After some hemming and hawing, I bit the bullet and redesigned the hardware construction such that there are no more screws visible on the front panel. It meant that I had to move some parts on the main board out of the way, and add another bracket and holes to the main board. I also used this moment to slightly change some positions and changed some of the silkscreen text.
This is how the front panel looks now:
And here is the new construct for the potmeter and the reduction unit:
The reduction unit is now no longer mounted on the front panel, but has it's own bracket mounted on the main PCB. The two brackets are made of an L shape aluminum 1mm thick and is 20x30mm. Commonly available in DIY stores.
Here is the main PCB:
On 22-aug-2025 I uploaded the three PCB fabrication files to PCBWay and asked them to produce them under the sponsoring agreement. They already passed their review successfully. I hope to get their approval for the sponsoring soon in which case they can be here in 7-10 days.
Construction Details
The three boards have arrived from my sponsor PCBWay and as usual, look great. Especially the front panel came out very well with the black color of the solder mask and the crisp white silkscreen. Compared to other fab houses, PCBWay uses a special matt black color that is especially interesting for front panels because does not shine or mirror and does not leave finger prints. It is more difficult to clean though, so don't spill any flux on it. I use it on all my front panels and I quickly counted 5 of them on different instruments.
This time I forgot to ask them to put their manufacturing number on the back side of the board. By default, they put it on the front, as you can see here in the bottom left. It's not hindering though.
The Power Supply
I already populated the power supply board (top left above) the same afternoon. The next day I tested the vitals and all the voltages where there.
One important notice that I now also added to the schematic is that with this supply, you always need to connect a nominal load to the output of the shunt supplies, otherwise the TL431's will get too hot. They will try to reduce the voltage by sinking all the current. I used two 680R 1/4 W resistors as a load, connected to the board as you can see in the picture below.
I could easily adjust the voltages, the trimmers have a good range. The LED that I mounted on the board, to have a visual indication of the applied power, was too bright. I changed the series resistor from 10K to 47K to reduce the brightness. Now that I know that the supply is working, I can give it a good soak & scrub.
The Main Board
After knowing that the power supply works, I can start on the main board. I normally keep the prototype in a functioning state, but there are so many special parts on it that I can't do that. I'll simply transfer the parts from the prototype to the new board.
The challenge will be that I used the Tek part identifiers on the prototype, but started anew for the new board. I will have to use the iBOM plugin for KiCad on both PCB's to figure out which is which and where.
Here's a very valuable tip. If you don't have this iBOM plugin installed yet, you're missing out on a fantastic addition. Highly recommended! My buddy Bud looked at it when I told him, and initially was not impressed. However it took only a little explanation of the features to make him a true believer as well. Now he swears by it.
Here is how that looks like:
The top screen shows the iBOM windows for the prototype with the parts list on the left, and the layout on the right. Everything is linked, so when you select a part in the layout, iBOM parts list or iBOM layout, the cursor will go there and highlight the part(s).
The bottom screen shows the Version 2 board, using a second browser. So I select a certain part value on the prototype, and also select the same part value on the V2 and then know where it is located on the prototype and where it needs to go on the V2 board. I first apply solder paste on the V2 board, and then use my heat-gun to remove the part from the prototype and place it on the V2 board. When everything is done, I can use the heat-gun or reflow plate to solder all the parts on the V2. Easy-peasy in theory, not so easy in practice, but after discovering and correcting a few errors, I could also add the other parts.
In the parts list, there are checkmarks you can fill in so you can keep track of where you are.
The main board is now ready for a reflow soldering process on my new hot plate. This is the largest board with a lot of parts, so yet another good test of the hot plate design.
Wave soldering went well, although I had to nudge a few tiny (0402 diodes - I ran out of the 0805 types) components in place. I'm now soaking the boards for several hours so I can clean them properly in order to inspect them better with my microscope. With the amount of flux and very tiny solder ball bearings in the way, that's not really possible.
Aftermath of reflow soldering
This is a little deviation, but I wanted to share it with you nonetheless. I will probably move this information to a new post, but at least you know why I did not present any progress for about a week.
After a reflow process, you end-up with a lot of flux that was part of the solder paste. Unfortunately, you need to remove that completely because it will have influences on hf circuits, and is hygroscopic so will attract water that will result in corrosion. Another side-effect of using solder paste is that you end-up with a number of so called ball-bearings. They are very tiny solder balls that are the result of the part sucking down to the PCB and that will push solder from beneath the part to the sides of the component.
Below is a good example of this issue. Mind you, this is after cleaning and scrubbing the board twice already!
The whitish goo is the solder flux and you can clearly see several ball bearings.
The remedy is to soak the board in a solvent for a while, and then vigorously scrub the PCB with a brush to remove the unwanted flux and the ball bearings.
Sounds easy, but is not so. The selection of the most optimum solvent is a science by itself, and is highly depending on the solder paste you are using. Some have water base paste, most have not. When you soak the board with a solvent, and use a brush, you typically end up with a very sticky board. To remove that, you can use IPA 99% to rub the PCB some more, and then generously poor fresh IPA on the board to let is drop of the board, and hopefully take the sticky goo with it. After that process, you can use clean (distilled to avoid calcium deposits) water to wash the remaining IPA off, and then dry the board. You can use a hairdryer, or use a solder heat gun, as I do, because that will allow you to blow the remaining liquid out of small pockets.
That's the process, and I have been using that for a long time.
This time however, it did not go well at all. First of all, the PCB was the largest I ever used on my reflow hotplate and that meant that the solder in the middle was overly done, and on the edges just about. There was an abundance of flux and I have never seen so many ball bearings. Cleaning as I describe above did not give a good result. Even after soaking for several hours, the goo stayed and the ball bearings with it.
Rather frustrated, I finally bit the bullet on purchasing an ultrasonic bath. I had been eyeing them for a while, but thought I could do without yet another box and also save the money. So I bought one for 2 liters that is just large enough for this PCB (160mmx115mm). It fits, but not in the wire-bucket.
After trying it with my solder paste solvent in the pure state, I first set the heating of the solvent to 50C and tried it for 5 minutes. That was much, much better, but still not good enough so I tried it for another 10 min. after adding some liquid dish washing soap to further breakdown the flux. That resulted in about a 90% successful cleaning. The edges of the board were not as clean as the middle, but that's because the single transducer is in the middle, and I'm pushing the size limit.
In the process, I noticed that one of the tiny 0402 diodes that I soldered by hand after the reflow soldering got loose (note the circle below). It must have been a cold solder joint, but I'm actually glad that happened because finding that error would have been difficult and I'm warned for the next time.
You can also clearly see the cloud of the solder paste solving into the liquid.
In any case, I'm happy with the results of the cleaning session now.
In the meantime, I may have discovered the reason for my reflow issues. It turns out that my solder paste is probably expired. I used SN4258 (Sn42/Bi58/Ag1), 138C melting point with a mesh (size of balls) 20-38um, or size #4. Unfortunately, there is no expiration date.
The effects that I see are exactly what is described when using expired solder paste. I always keep mine in the fridge, but it explains a lot. Unfortunately, I can't find the purchase date anymore but must be less than a year. I ordered it when I was working on my reflow hotplate. Lesson learned...
First Power Up
I connected the 12V rail to my Lab Power Supply to keep an eye on the current and used a jumper to select the frequency ranges, and listened for the relays clicking, which they did.
I then added the +/- 17V rails again using my Lab Supply and set the current to the expected value of 35mA. That didn't work, the supply went into current limiting, so something was wrong. Upon inspection, it turned out that I goofed with the location of two of the NE5534 Opamps, they were rotated 180 degrees. The marking is very difficult to see on them.. Unfortunately, trying the power again did not result in a lower current. I probably fried them, because after replacing them, I got the expected current result.
By then I could use the current shunt power supply for the SG505 and use that from now on. While checking the voltages, I noticed that the +17 could no longer be adjusted higher than 16.4V, so I lowered the -17V to also 16.4V to have symmetry. I need to investigate that in more detail later on, but I now had the 12V rail and the +/- 16.4V rails.
I then added the three potmeter connections and used my DSO to look at the generator output. Nothing, just some weird very low frequency flopping of the trace. That smelled of a bad connection in the overall loop so I hunted for poorly soldered parts. It took a while adding some flux (here we go again) and using either the heat gun or my solder iron to go over all the possible parts. Eventually I found the culprit and was presented by a beautiful sine wave. The frequency setting worked, the vernier worked and the frequency ranges worked as well.
With that result, I can now investigate and test the output amplifier and that's next on the list.
Wired-up all the required parts to make the amplifier functional, and eventually got it all working.
Oops!
I found one serious issue that I didn't see when using the prototype. I never tested the counter and then also looked at the sinewave distortion. In addition, I also used a slightly different pick-up point to tap the output frequency for the counter. That's a mistake, the 1K series resistor loads the feed-back loop of the output amplifier way too much due to the clamping of the diodes on the other side. I even tried 10K and that's still not good enough. Higher values result in no signal at the diodes.
The fix is rather easy and only involved a single trace cut from RV2 to the 1K of the squarer circuit and connect it now via a jumper wire to the Sine Out test point. I will fix this in the schematic and layout of course.
Tuning the highest range
While it is still lying on my desk and easily accessible, I decided to tune the X10K range because the frequency was a little too high compared to the lower ranges. I replaced the 100pF capacitors for C16 and C26 to 140pF. I didn't have that value, so I used a 150pF probably an X-version. That resulted in a frequency of 105KHz instead of 100KHz. The 140pF COG should get it closer.
Looking into the rail situation
I also investigated the fact that I could not get the positive rail up to +17V, while the -17V worked. The raw supply is also a little lower than I hoped for with the transformer I'm using, so I adjusted that output to 40.0V. With that, I adjusted the rails now to +/- 16.4V so they are equal. I don't think it will matter much. The original SG505 used +/- 15V rails, I decided to use the SG505 Option 2 rails that are +/- 17V which will give a bit more head-room for the Opamps.
I have not made any distortion measurements yet, I'll wait until everything is inside the enclosure and I have shortened all the leads that go to the front panel to the appropriate length.
Using the front panel
The next step is to mount the controls on the front panel and the PCB in the enclosure. I also need to make a few mechanical adjustments to the front panel because the holes do not match they way I designed it, and I made a mistake with the size for the main potmeter. These turned out to be a KiCad footprint issues and the result is that the holes are too small. Easy to fix but I also need to move the hole for the BNC connector a bit. Obviously, all the changes will be updated in schematics and the the Gerber files that I will publish.
Putting it in the enclosure
I finished the assembly of the front panel and put the main PCB inside the enclosure. It all works as advertised, but I had to make several changes to the front panel design.
Here is a view from the back that shows a few more details.
The power supply is still outside the enclosure as you can see, and I connected it with flying leads that still look untidy.
Now that I have the main potmeter reduction unit installed, the adjustment of the frequency is very precise and feels great. If you're in doubt of getting one, do it, you won't regret it.
Adding the frequency counter
I also fixed the counter loading problem that I mentioned earlier so I can now start to add the Arduino Nano, the opto-coupler and the OLED display.
The installation went pretty easy, also because I used tape to secure the display to the front panel.
However, making it all work was more involved than I anticipated, even though I built a prototype that (seemed) to work very well. After a few hours of struggling, I ended-up making several hardware and software changes to make it work reliably throughout the frequency range.
I eliminated the squarer circuit and replaced the BJT with a MOSFET to reduce the load on the frequency generator tapping point. I could now also use a much larger series resistor, and placed that very close to the generator output again, instead of using the output amplifier because that signal can be attenuated.
The trace that goes from the generator output to the MOSFET for the counter is very long, so I wanted to reduce the current and hence the possibility for cross-talk.
I also had to power the opto-coupler open collector output with the Arduino 3V3 instead of the 5V rail to reduce the 5V p-p hammering on the port. That made a significant improvement to the precision of the counter.
Lastly, I tried to get a bit more fancy with the frequency count display, and wanted to add decimals for frequencies below 1KHz. After trying for some time, I realized that the library I'm using is optimal for the higher frequencies, but not so much below 1KHz. For those frequencies, to get a decimal digit, I needed to select a gate time of 10s which makes the counter unresponsive for this application. Selecting a library that fixes that will not work well with the higher frequencies, and I have not figured out how to use both libraries and select them dynamically. It is what it is. The solace is that the counter is a lot better (more precise) than the frequency dial of the original SG505. If you need a higher precision, use a real counter!
The DMM shows the frequency from the SG505 output and can be compared with the counter display. Not a perfect match, but good enough!
Next step is to mount the power supply in the top half of the enclosure and create the correct length for the power interconnects. By then, it's time for some distortion measurements, some tuning/calibration and see if I need to add shielding to the inside. I think I will...
Oh boy, oh boy...
I finished with putting everything inside the enclosure, calibrated the 0dBm output level and found out that I need to make changes to the -10dBm switch circuit.. I have a new set of values that I tried and worked for the Pi attenuator but will need to take the board out to replace the resistors. That can wait.
I wanted to make a distortion measurement to see where we are:
The 50Hz mains is everywhere, I kind of expected some of it because there are two transformers in the box but was hoping some serious shielding would help, but this is out of the question. The good news is that the principle has pretty good harmonic distortion results. THD h2-12 is 0.00063%, but's it's overwhelmed by the hum.
It's abundantly clear that the power supply has to move out. There is no place for both of them in this house. Bummer.
That means going back to an earlier idea and that was to use a separate enclosure to house the power supply and keep it away from the generator. That's pretty easy to try with some longer connection leads for the power rails.
Unfortunately, even with the generator in a metal box, the hum and the AC-DC supply generated noise is still there.
Here is the result with the +/-17V connected, but without the 12V supply connected. There is no sine wave because the relays don't work. This is kind off the noise floor.
It seems that the shunt supplies are working OK, it's the 12V supply and circuits.
This is with my lab supply feeding the 12V rail, activating the relays.
And below is the same measurement but with the Arduino Nano out of it's socket. I also did a quick trimming of the H2 reduction, and that worked very well. Went from -110dB to -126dB.
The silver lining
Without the 12V AC-DC convertor, and without the Nano/LCD display, the generator works fine.
THD-N is 0.0033%, THD H2-12 is 0.00020%, with H2 at 0.000050% and H3 at 0.00014% That's good to know and very good news! Now I need to find some solutions.
Finding the root cause
We already established that it is mainly the AC-DC 12V supply, but is there more to it? I tried it also with the Nano powered, but pulling the LCD connector. This dramatically reduces the noise so it's not the Nano itself, adding little peaks at 15KHz and 19KHz, but more the i2c signals and the power devices on the OLED board itself:
Using a 1uF blocking capacitor and using the E-MU to look at the AC-DC 12V rails, it sees this:
And without the OLED display this:
So the dominant culprit is the OLED display, and there is also seems to be some back-coupling from the generator into the 12V rail. On a hunch, I disconnected the generator output signal to the counter input circuit and presto, we have proof that the harmonics are from the Nano and counter input circuits.
When I pulled the opto-coupler, the harmonics went away. Putting it back in and removing the 3V3 for the open collector pull-up did not make a difference. Disabling the Arduino loop reading the counter did also nothing. It sure looks I need to cut the power to the Arduino and the display to make that circuit quiet.
The two part solution:
Powering the 12V section of the Arduino counter with a SPST switch, I can now eliminate the generated noise from that circuit. It works great, when I power the 12V with my lab supply. Have a look:
With the Arduino Nano circuit turned on, it looks like this:
Mind you, the noise is still below 120dB, but the root of the problem is the AC-DC 12V supply. That will need some serious filtering, or another solution.
The second part of the solution
I need to filter the AC-DC supply much better, or create a different solution. Seriously reducing that noise is going to be a challenge to force it below 120dB. I first need to profile that supply better.
The 12V supply draws 46-48mA when everything is powered.
We're looking mostly at 50Hz mains related hum, plus many harmonics in the higher frequencies.
I can try to filter the hum and noise, but the whole reason for the AC-DC convertor feeding the separate 12V supply and to avoid coupling to the generator fell flat on its face. That's not the solution.
So, when the coupling is there anyway, I might as well do away with the 12V circuit and tap the 12V rail from the main supply. I quickly added a few components to the 5V section of a bare Jung supply PCB and modified the output for 12V. It's actually 11.5V but that has no consequence. When powering the 12V rails with the new supply, we're making real progress. This is when the Nano supply is switched off:
I still want see if I can get rid of the mains related hum on the left, but the part on the right is now clean.
To give myself some peace of mind, I also tried the Jung SuperReg again for the +/-17V rails. Surprisingly, with the current configuration, it produced more harmonics. Go figure.
So, all in all, I decided to redesign the power supply by removing the AC-DC converter, and add a third rail fed by the 40V Vreg supply, to provide the 12V supply. The 12V rail does not connect to the common ground, so it stays floating from the rest.
I tried many things to reduce the mains hum and was partially successful by using more capacitance on the main supply.
The next step is to work on the grounding/shielding of the generator. Right now, the front panel is isolated from GND, and I need to change that now, and also ground the switches and potmeters on the front panel. The whole generator will continue to be isolated from earth ground.
Major changes and a departure from this version
After considerations off the issues described above, I have now decided to make a significant change to the generator, now that the power supply is no longer going to be in the same enclosure as the generator. I'm now going to use a full metal enclosure for the generator and put the noisy parts of the power supply in another enclosure.
This means that with the other changes I already had to make/want to make, I will go through a redesign of the main PCB, the Power Supply and the Front Panel.
The sponsoring agreement with PCBWay requires a new Blog for every set of PCB's that they sponsor, so I will start a new Blog for the new Version 3 of the DIY SG505 rebuild. This one was already too long, so it's a good reason to start a new one anyway.
A Github repository is available here it will be updated with information during the project when I have verified the correct operation. I'm still working on it, so there is very little information there at the moment so please be patient.
This post will describe the DIY build of a new 1MHz, a 10MHz and a 100MHz differential probe with several power alternatives.
In 2017, my buddy in crime Bud and I (well, mostly Bud) build a DIY X10 100MHz differential probe that has been popular because so many other makers have build it, or even changed it to their liking. We know of a 1x and a 100x design modification. Our earlier design was described here:
Unfortunately, that 2017 design has some hard to get and rather expensive parts, especially the Opamps, the voltage regulator and also some hard to get capacitors.
In early 2025, Bud has started a new project to revisit the design and look into making the probe easier to build and make some desired and some needed improvements. He investigated the whole concept again and came-up with a main probe PCB that can be used to build a 10x 100MHz, a 10Mhz and a 1MHz version. The 10 and 1 MHz probes produce less noise. The other constraint on the earlier design was the external power supply. Now that we have USB-C PD capabilities available to us, Bud added a number of power options that can be added to power the probe to your liking.
When testing my diff probe, we found that only GaN technology, low power (<35W) good quality USB-C PD adapters are quiet enough to be used with the diff probe. Be aware! Below we explain why.
Another main issue that he wanted to address was a 3D printable enclosure and also making the probe slimmer so it's easier to hold while probing around with it.
There are some more improvements and refinements and if you're interested, you can follow along his very detailed design and test efforts here:
Bud has built several probes and verified and tested everything. Now that he is done, I will also build two probes and can verify the BOM's and describe the building procedures here and also add some more information that will hopefully help others building this very useful instrument.
When I'm done (and only then), the details like schematics, pictures, the BOM's and Gerber information will be available in a new Github project. I will also enter the information in the Shared Project list of my sponsor PCBWway so PCB's can be easily obtained there too.
If you're looking to build a diff probe, all I can say at the moment is to stay tuned for a little while longer. I will most likely be finished in October/November of 2025, so if you can hold-off a little longer, I recommend you do so.
What I'm going to build
I already have the earlier 100MHz probe, but wanted to build another one so I can test it with my gear that is a little bit different from what Bud has or uses. More specifically, my DSO is a 300MHz version, and while he has a modern and much more improved and new DSO (12-bit, touch screen, etc, etc. I'm jealous!) it is a 100MHz version. He has a function generator, but that only goes to 60MHz. I have a fast edge (<1ns) pulse generator, and also a nanoVNA.
So I completed two purchase orders that is mixed between LCSC and DigiKey to get all the parts for the 100MHz version, most of the different parts for the 10MHz version, and then the USB-C based power supply items.
I also ordered the circuit boards from my sponsor PCBWay, who graciously supply 5 each of the boards to me without cost.
Here is a complete set:
From top left to right: the discrete VREG board, the Buck convertor VREG board and the LDO SOIC-8 VREG board. Below it both sides of the diff probe that can be used for the 100, 10 and 1MHz versions.
On the back side of the probe is the voltage splitter circuit that turns the input voltage to a pos and neg rail.
I have been using PCBWay for years now, and the quality of the boards are always very good, with crisp silkscreen text that is easy to read. The finer details, when looking with a microscope are clean.
This is especially important when you start to use 0402 parts, where the quality and crispness of the footprint and the solder mask is becoming increasingly important. You'll find that out when during reflow soldering the parts stay in place or move to their intended place without thumb stoning.
Building the 100MHz probe
The schematic is largely based on the previous design, and basically has the same specifications, but with a number of refinements and changes. The input attenuator is changed so we have equal resistors, and that allows us to have equal capacitors that are now parallel to every resistor.
Specifications:
Input impedance: 20MegOhm//1.25pF - differential, 10MegOhm//2.5pF for each terminal to GND.
Differential Gain = 1/10V/V. Any lower than this and most hobby DSO's cannot resolve a 1V input signal with any clarity.
Maximum AC Common Mode Voltage (with 50V differential input) = 350VAC
CMRR > 90dB @ DC, ~60dB @ 1MHz.
100MHz: Differential Voltage Range > +/-25V for 240VAC common-mode, +/-25V for 0V common-mode.
3dB bandwidth >= 100MHZ (depends on signal amplitude)
DC offset < 1mV (trimmed)
Noise: 30mVpp or lower at output.
Power supply: 5.25V +/- 0.25V. This can be a USB-C PD supply plus a regulator, see the power options.
Cost: ~$50 not including shipping and handling costs. Not including 3D printed enclosure.
As with the old probe, during the verification and calibration, you will have to add capacitors to the not placed (NP) C11 and C12 so the trimmers C15 and C16 can properly adjust the AC compensation. There are extra values listed in the BOM that need to be ordered so you have the possible values at hand.
The next change is the first gain stage. The positive output now has an optional offset adjustment that can be installed when you need it. When you don't need it, don't install it because it degrades the output a little bit. Both outputs from the gain stage go to a summing amplifier. Unfortunately, Bud was unable to find suitable alternatives for the two Opamps, which may be a little difficult to get.
The rail splitter Opamp is replaced by a device that is easier to get, and there are several alternative devices that can be used. Some refinements were necessary because we no longer use the difficult to get small Tantalum capacitors. The ones we use now need a tiny resistance (act as a snubber circuit) to make everything stable.
Next additional and optional circuit deals with a power on LED. Depending on the way you power the probe, or what power supply you use, you can populate these two parts to get a visual power indication.
Last major change is the power input circuit. Note that there is no "real" connector on the PCB, but only two pins. The idea behind it is that you can add a selection of tiny power modules (called daughter boards) to the bottom of the probe, based on your particular taste or need. More about that later in the power section.
Lastly, as you can see from the picture at the beginning of the post, there is now a 3D printable enclosure available.
Here is the schematic for the 100MHz probe:
The PCB looks like this in the 3D viewer:
You will note that it's a much slimmer design, that will be easier to hold and maneuver when you're probing around in a circuit.
Note also that the middle pin of the front-end will have to be removed to create the creeping space. The 3D model for the header is not modified, so it shows all three pins.
The bottom of the PCB houses the rail splitter circuit, and is also the place where the power supply PCB will be added. That's where those four square holes are for. Bud calls these little boards Daughter Boards.
Below is my version of the probe:
The calibration went without a glitch and I added two 47pF trimmer caps to get the AC gain in the range.
The new Power Supply Options
There are two components to the power supply. One is the input, and one is the regulator. The three probe models require different input voltages, 5.3V for the 100MHz and 15V for the 10 and 1MHz probes.
The USB-C input board
The input voltage for the probe regulator can come from a variety of supplies. The most optimum is a USB-C PD Trigger Board board that fits inside the probe enclosure. These Trigger Boards are very inexpensive, are widely available, but you have to select a certain type and size.
These Trigger Boards communicate with the USB-C Wall-Wart supply, and request the desired voltage. There is a CPU on the Trigger Board that takes care of the rather complicated communication protocol.
These tiny boards are available in different voltage configurations, like 9V, 12V, 15V and 20V. Although the voltage you order is fixed for the board, you can still change it by closing or opening a bridge (a 0402 0 Ohm resistor or a solder blob). So 9V can be configured to 12V by adding a bridge, and 20V can configured to 15V by removing a bridge.
The board in the picture is a 20V version, and if you remove the resistor in the top right, it will be configured for 15V.
The 100MHz probe requires the 9V Trigger Board version, and the required voltage regulator will bring that down to 5.3V. The 10Mhz and 1MHz probes require the 20V Trigger Board version and the required voltage regulator will drop that to 15V. I ordered a number of each of them, they typically come in lots of 5.
These Trigger Boards need to be securely fixed (glued) on to the probe regulator board. The voltage regulator board is fixed to the probe by soldering 4 power/support pins. The Trigger Board with the USB-C connector needs to have a firm connection to the Trigger Board, because of the force that will be applied when you push the USB-C cable into the connector.
The output of the Trigger Board needs to be connected to the probe regulator board by two wires that are soldered on the pads.
Below is a picture of Bud's version.
This construction will fit snugly into the 3D printed enclosure.
Obviously, these USB-C Trigger Boards need to be fed by a USB-C cable going to a Power Delivery (PD) supply.
The diff probe does not draw a lot of power (about 50mA) such that a 20..30W supply for mobile phone chargers will do.
WARNING
I was planning on using a 100W supply myself, because I wanted to use it for many other applications. The one I ordered turned out to be a mistake! Don't get a high power one, it may not work well with the probes. Read on to find out why.
The probe voltage regulator options
Bud create a number of voltage regulator boards to satisfy your particular need or preference. There is a switched power regulator, an LDO linear regulator and a discrete voltage regulator.
As mentioned earlier, there are two output voltages required for the probes from the regulator, 5.3V for the 100MHz probe, and 15V for the 10MHz and the 1MHz probes. The 100MHz probe requires an input voltage of 5.3V (+/- 5%), to create accurate +/-2.62V positive and negative rails for the Opamps.
The 10MHz and 1MHz probes require an input voltage of 15V to create the +/-7.5V (+/- 3%) positive and negative rails.
I decided to build a linear regulator (the VREG SOIC-8) and also try a VREG discrete regulator for the 100MHz version with a 9V output, and another VREG discrete regulator for the 10MHz version with 15V output.
The VREG SOIC-8 daughter board
The VREG SOIC-8 Daughter Board (DB) is intended to be used with the 100MHz probe, because it outputs 5.3V and requires an input voltage of 9V.
It can be configured with a number of LDO voltage regulators. I have decided to use the LP2951 LDO voltage regulator. It comes in an SOIC8 package, and Bud designed a Daughter Board specifically for these devices.
Here is the schematic:
As you can see, there are a number of voltage regulators that can be used on this PCB.
Here is the 3D picture of the PCB for it:
The parts are located on the bottom of the PCB, here is the top:
The USB-C Trigger Board and the Voltage Regulator combo
I populated the SOIC-8 low drop regulator board, and connected it to a 9V Trigger Board. That all worked well, but I ran into a gotcha that took quit some time to investigate.
These two boards need to be firmly glued together, but I first wanted to see what the power-up sequence was for the whole chain (USB-C PD supply, 9V Trigger Board and the 5.3V SOIC-8 regulator. It turns out that it performs really well with a nice and gradual power-up. The output voltage was a perfect 5.3V
Next step is to glue the boards together after aligning them with the enclosure.
Detecting a nasty surprise...
After I installed the board combo to the probe and verified the calibration and while further testing the probe, I noticed nasty spikes on the probe output signal. Up until now, I had used my Lab Supply to power the probe and running the calibration and tests. The output was noisy, as expected, but otherwise clean.
Eventually, comparing notes and measurements with Bud, we found out that the root cause of my problem (he does not have these switching glitches) is the 100W USB-C wall-wart I purchased. When I looked for one, I wanted to make sure that it produced all the voltages and I could not find many that listed them. I ended-up buying one that was advertised as a laptop charger and that may have been a mistake.
I purchased mine from Amazon that is called a Basicvolt 100W laptop charger. This is what it shows on the label:
Note the second caution. Is that the give-away?
It turns out that this supply is very noisy. So much so that it is unusable with our probe. The hf switching transition stuff comes through everything. Below is the wall wart output at 20V. The ripple is a non issue, the regulator will take care of that without even blushing, but the hf stuff at several 100's of mV is another story. This screenshot is from the supply, the probe is not connected.
The output of the connected 9V Trigger Board looks like this. Note the 2V p-p hf noise:
The Trigger Board itself does not contribute of create the hf noise. It only tells the USB-C supply what to do.
Here is what the probe output looks like on my DSO:
It's a little hard to see, but the hf is very present on the GND output signal of the probe. The circuits on the regulator and the probe itself filter the hf noise a lot, but it is not sufficient.
The net-net is that I need to change my USB-C supply and see if I can get a better one. The trouble is that almost all of these power adapters are specified for laptops or phones, and they don't care.
Bud is using 20W and 30W wall-warts and they are clean, so I ordered a 20W one as well.
Mine is in the mail and will arrive this evening, so I can report more about this issue.
In the meantime, I tested a genuine Apple PD charger that supports 5 and 9V outputs.
This is the result.
The switching noise is not even that high at 150mVp-p, but the minute 60mA load is already enough to cause a ripple. Not good enough for the diff probe.
Just for kicks, I then tested my Dell laptop supply, also a USB-C PD version.
Unusable!
By now I was getting a bit worried...
The one I ordered is this small power supply, hoping that would be the answer:
This is a simple USB-C PD supply. However, below is the zoomed-in switching noise riding on the output:
Note the V/Div scale of 1V! This is a terrible supply and certainly unfit for the diff probe. Even when you use it to just charge a phone, the noise it generates will be everywhere. Very badly designed. How did that get through any certification? I returned it right away.
After more consulting with Bud, it transpired that he is using two GaN technology wall-warts. Could this be the answer? With the GaN technology, the switching frequency can be up to 10MHz. The switching noise is then much easier to tame.
So, I found a seemingly suitable one and ordered it:
When it arrived, I found that even the packaging was very well done. It felt solid with a surprisingly "heavy" weight of 138 grams. My hopes were going up. It produced this:
Notice the much higher switching frequency, and the V/Div scale at 200mV/Div.
However, I was still disappointed, so with some more consulting from Bud, I tried to tame this tiger by adding a snubber circuit of a 4.7 Ohm resistor and a 4.7uF capacitor, mounted right on the 9V Trigger Board. I then also used two 1K at 100MHz ferrites to connect to a 510 Ohm 1W resistor that acted as the 60mA load. All four added parts are 0805 size.
With this addition, the switching noise is completey gone. All that is left is the residual noise of the adapter, my environment and that of the DSO.
Time to take a step back, and realize what we have here. Bud already put a 10uF/1 Ohm snubber on the probe in addition to two ferrites. Is that already sufficient?
So the next step was to use a clean 9V Trigger Board, connect it to the diff probe and measure the noise again.
Below is the setup I used.
The new GaN USB-C PD supply is connected through a good quality USB-C cable of 1 mtr to the USB-C monitor connected to the 9V Trigger Board. I did not want to modify the battery supply connection to the probe, so from there it connects to a 9V battery adapter that is plugged into the 9V battery holder connector. The battery holder is connected to the discrete VREG Daughter Board mounted on the probe through at least 35cm of plain but flexible 26AWG wire. The diff probe is outside the shielded enclosure. The input of the probe has no signal, but is terminated by a 50 Ohm feed-through. This must be a realistic but worse case condition setup.
And this is the output of the diff probe itself on my DSO.
Apart from the inherent noise generated by the diff probe amplification, there is no switching noise from the adapter visible. So the on-board measures are already sufficient to remove the switching noise!
The net-net is that we now highly recommend that you use a good quality GaN technology low power PD (<35W) adapter powering the diff probe. The designed-in filtering on the DB regulators and diff probe should be sufficient to eliminate the adapter noise.
However, if you end-up with one that is a bit noisier, you can still add the snubber circuit right on the Trigger board as I tested earlier.
Problem identified and solved.
A battery + VREG power option
Because of the noise mishap, and while waiting for the new wall-warts to arrive, I switched to using another power option for the probe. Obviously, the most optimum low noise power for the probe is by using a battery. I decided to keep it simple and selected a 9V rechargeable battery to power the VREG board.
Because I will also build an 18V supply by putting two 9V cells in series for the 10MHz probe, I wanted to add the high voltage protection that Bud designed for the discrete VREG option. When I tried it, it nicely cuts out at about 13V and drops the output to about 3V, so no harm is done when I accidentally use the 18V battery supply with my precious 100MHz probe.
Below is the construction. I use a 9V battery holder that has a power switch. They are widely available. I then soldered two flexible 26AWG wires to it and they were soldered to the VREG board +VIN and -VIN solder pads. The VREG board lowers the 9V to the required 5.3V for the probe with more than enough head-room for a decaying cell voltage.
Because there is no strain relief for the wires, I decided to solder the wires coming from the other end of the VREG board, so the folding of the wires is over the board and I hope that will create a form of strain relief.
The VREG Buck daughter board
This board is Bud's favorite due to the very low heat it produces, and also, and I really suspect it is the main reason, he designed switching regulator chips for a living. This board is designed as a companion for the 100MHz probe that needs a power input of 5.25V +/-0.5V. The input to the board can be a USB-C PD supply plus a Trigger Board set for 9V.
Unfortunately, when I ordered all the parts, I forgot to order the inductor that is needed on this PCB. Some of them are available locally, but with a minimum ordering qty of 5,000 or even 10,000, so it will have to wait for another PO.
There is a long list of alternative switching regulators available for this board, I selected the NEX40400 myself. There are also a number of Schottky diodes available (not required with the NEX40400), and a number of the inductors. Take your pick.
And here is the PCB for it.
When I have built it, I will report on it here.
Shielding the probe
When the power supply worked well, I continued putting the probe together with the required shielding. The probe functions without it, but your fingers and near equipment will have a large effect on the signal fidelity due to the close proximity to the very sensitive attenuation parts at the input of the probe. Any additional disturbence gets amplified and will show itself on the DSO. Unfortunately, Bud found out that shielding at the inside of the enclosure ruined the calibration and functioning, so he elected to add the shield the outside.
If you're experienced with this, you could try a conducting paint spray to cover the insides of the enclosure and figure out a way to connect that layer to the earth ground. I have no experience with that, so I skipped it.
When Bud put together his probe, he used a copper foil trace to go from the SMA connector to the shielding in the front, and wrapped the shielding around the closed probe. He then used shrink-wrap tubing to make it looks nice and prevent accidental grounding of the shield (earth ground!) to the DUT.
I wanted to try another approach, that would allow me to open up the probe without damaging the shielding.
Here is the result of what I did:
I added the copper foil shielding to both halves of the enclosure in two sections. A smaller one for the front part just past the bend to the thicker part, and one larger section just overlapping the front part by about 3-4 mm. As you can see, I folded the shielding around the edges to the inside of the enclosure. I then used a sharp knife to cut it nicely.
The top half copper foil has an extra tab and I soldered the ground wire to it before pressing it to the plastic, to avoid warping the plastic enclosure. The other end of the ground wire was soldered to one of the SMA connector ground pins.
When you close the probe, the foil gets pressed together at the seams and makes surprisingly good contact.
The covered trimmer holes need to be opened on the one side, and the screw holes on the other. Note that you need to cut away the shielding around the two holes for the capacitor trimmers, because you can create a short to ground if you use a metal screw driver. I use the plastic trimmer tool in the picture that has a metal insert the size of an 0603 part as the tip to reduce the metal influences.
The resistance from the SMA connector to the very beginning of the probe tip was about 0.5 Ohm to both halves, so a perfect connection. My "finger" effects on the output signal are now completely gone. It works better than I expected and hoped for.
When I calibrated the probe again while in this shielded enclosure, I was able to get a true flat line for the CMRR and a 0mV offset adjustment. Unfortunately, the capacitance trimmers are now at the edge of their range so I need to change the 47pF caps to 39pF, which I don't have in the 0603 size yet. They are on their way.
I will not yet add the shrink-wrap tubing yet because I will need to open the probe a few more times. At least one time to change the trimmer capacitors. I also want to experiment more with the USB-C PD power option.
The calibration procedure and some measurements with pictures will follow soon, check the calibration procedure further down.
Building the 10MHz probe
The basic schematic is the same as the 100MHz probe, so it can be built on the same PCB, but there are several part and part value differences.
One of the major reasons to build a 10MHz and a 1MHz probe is to reduce the output noise of the probe, which is quite substantial for the 100MHz version. For these versions of the probe, Bud was able to find two alternative Opamps, the OPA2810 and the OPA810, that are 1/3 the cost of the LTC6268/9 pair of the 100MHz probe.
Specifications:
Input impedance: 10MegOhm//2.5pF - differential, 5MegOhm//5pF each input to GND.
Differential Gain = 1/10V/V.
Maximum AC Common Mode Voltage (with 140Vpp differential input) = 350VAC
Maximum input voltage = +/-600VDC or 424Vrms.
CMRR > 85dB @ DC, ~50dB @ 1MHz.
Differential Voltage Range > +/-70V for 240VAC common-mode, +/-70V for 0V common-mode.
Power input 15V +/-0.5V. Can be USB-C PD, see power possibilities.
Cost: ~$30 not including shipping and handling costs. Not including 3D printed enclosure.
After I got the parts, I started adding the tiny solder paste droplets, and then continued placing the parts. Did I mention earlier that 0402 is really tiny? They are! After placing all the parts on the front side, I used my reflow oven to solder everything in place. There were a few parts that needed a little nudge to put them back into their intended place, but it went great. I'm so glad I build a reflow oven. Imagine using a heat gun on those 0402 parts. I did that earlier on the first version 100MHz probe, and that was not easy. You can easily blow a part from the board and never ever find it again. (I always buy a few extra 0402 parts, just in case they disappear in nowhere land)
There is so much flux goo on the board that I could not really use my microscope so I first cleaned it in my new ultrasound cleaner. I am now using a mixture that I concocted earlier of 50% IPA, 50% water (to avoid combustion!) and a good squirt of liquid dish cleaner to break the surface tension. Setting the cleaner to 40 degrees C and 20 minutes resulted in a squeaky clean board. Using the microscope showed a tiny solder bridge between the legs of the summing amplifier that was quickly remedied.
A first power-on using my lab supply at 15V and 20mA to avoid damage resulted in a current limit, but raising the limit showed the about 16mA current, the LED came on and everything looked good. The DSO that I connected to the output showed a steady line at about ground level.
Connecting my pulse generator at 1KHz showed the expected waveform. At this moment, you can try if the AC adjustment with the c-trimmer C15 is sufficient to create a flat edge towards the lower range of the trimmer. (if you use the enclosure with the copper foil shielding, it will create some extra capacitance) If not, you need to change the compensation capacitors C11 and C12. I already installed them with a 39pF value, but that was not enough to adjust the edge. At this point you can also see if the offset adjustment works, if it does, you should first install the correct values for the trimming capacitors.
Here is C15 at the maximum of the adjustment range:
And here with the minimum setting - obviously not enough, even when it will be in the enclosure later on:
The C-trimmer I'm using (LCSC p/n C22468123) has a range of 8pF to 28.5pF, so the adjustment range is 20pF. It seems I need about the same value in addition to the 39P that I already installed, so I'm going to try 56pF.
This resulted in a flat edge with a little adjustment room to spare. A value of 62pF would have been better, but I don't have them. I have 68pF, but I'm going with the 56pF for the time being.
Bud has a more scientific method to get the optimum values for the trimmer capacitors on his Hackaday site. If you're interested, have a look there.
Now it's time to fully calibrate the probe. Go to that section below for details.
During the calibration procedure, I found that the value of the trimmer capacitors were still too small, so I switched to 68pF values, the lower leg needed to increase to 100pF.
Building the 10MHz Probe Supply
The 10MHz and also the 1MHz probes require a supply of 15VDC. There are a few options Bud designed that you can select. I decided to use the USB-C wall-wart supply, a 20V Trigger board, and the de-populated version of the discrete VREG regulator.
We've found (see further up) that the wall-wart supply has to have some important specifications for it to function with the diff probes. It has to be quiet, at least quiet enough so the native filtering on the diff probe can clean it up and there is no switching residue visible at the probe output signal. Bud and I tried a couple different "chargers" and found that only the ones that have a low wattage (<50W) and have the GaN (Gallium nitride semiconductor) technology (higher switching speed) are acceptable.
Below is the UGREEN 3-port 45W charger that uses GaN technology that I'm using. It is sold by Amazon. On the Hackaday site, Bud shows what he's using. Unfortunately, his versions are not sold by Amazon in Europe.
Below is the Trigger board I'm using. They come pre-configured for 20V but can be re-configured for 15V as well. They typically come in sets of 5, and Amazon sells them.
The 20VDC output of the Trigger board goes to the discrete de-populated (no over voltage protection and a simpler regulator) regulator.
The simplified de-populated version schematic looks like this:
Note that we changed the value of R1 from 12K to 10K from earlier versions.
It can be configured on the same PCB as the full boat version, you just need to omit some parts and change some values. Make sure you use the recommended TL432 from the list. I used an SPX2431 and a BC847C. The circuit needs a B or better a C version for a current gain of >200.
My VREG board produced an output voltage of 14.98V. Using my DC load, I measured no voltage drop between no load and up to 100mA.
The discrete VREG board connects to the diff probe by 4 supporting pins, of which two also supply the power to the probe.
When you're happy with the performance of the board, you can now solder the four supporting pins to the VREG board. I used a plastic protoboard to keep the pins in place while soldering. Use as little solder as you can, because the protruding end of the pins need to be cut off, and filed or sanded almost flush with the PCB so the Trigger board can sandwich on top of it and glued in position.
The Trigger board and the regulator board are glued together back to back and are inter-connected with two wires. Before you attempt to glue, solder the two power wires in place. The positioning of the two boards relative to each other is a little tricky but once you know how it is supposed to fit together, it's rather easy. Practice the fit a few times before you permanently glue or solder things together.
This time I used 1 second glue to affix the boards together, so no drying time (but also no adjustment possible, you need to get it right the first time!) Did I mention that you need to practice the position a few times before you use the glue?
Below is the power supply sandwich in detail. Note that the Trigger board and the VREG board are aligned at the front, above the back of the SMA adapter. This allows the USB-C connector to protrude through the case. The sides of the board need to be aligned flush too so together they will fit in the slot for the boards in the space of the enclosure you see top right. You kind of hook the probe into the enclosure.
Try the positioning of the sandwich and the probe before you solder the 4 supporting pins.
When you did that, you will need to cut a bit from the length of the pins on the component side of the probe, so they don't stick out too much. You can see that in the picture above. Again, try the positioning a few times because after soldering the four pins, it will be very difficult to make changes.
Below is the bottom view of the probe with the power sandwich mounted.
The "rough" surface of the enclosure edges are there on purpose. There are tiny dimples on the edges that actually help to press the two halves together and create a very good connection between the foil on either halve.
And here is the completely mounted probe inside the enclosure with the copper foil shielding.
The cutouts in the copper are there so you can't short a metal screwdriver from the grounded shield to above all the C-trimmers.
Once you got this far, you can now continue to calibrate and profile the probe. The Calibration Procedure section is just below.
Building the 1MHz probe
As with the 10MHz probe, this version can also be built on the same PCB, and again there are part value differences to make it a 1MHz probe with the least amount of output noise. It uses the same Opamps because Bud could not find less expensive ones that had the right specifications.
At the moment, I have no plans to build this version myself.
Specifications:
Input impedance: 10MegOhm//2.5pF - differential, 5MegOhm//5pF each terminal to GND.
Differential Gain = 1/10V/V.
Maximum AC Common Mode Voltage (with 140V differential input) = 270VAC
Maximum input voltage = +/-600VDC or 424Vrms.
CMRR > 85dB @ DC, ~50dB @ 100KHz.
Differential Voltage Range > +/-70V for 240VAC common-mode, +/-70V for 0V common-mode.
10MHz: Differential Voltage Range > +/-80V for 240VAC common-mode, +/-80V for 0V common-mode.
Bandwidth >= 1MHZ (-3dB)
DC offset < 1mV (trimmed)
Noise: 0.5mVrms (3mVpp), input referred.
Power input 15V +/-0.5V. Can be USB-C PD, see power possibilities.
Cost: ~$30 not including shipping and handling costs. Not including 3D printed enclosure.
The Calibration Procedure
After you have selected the optimum trimmer capacitor values for C1and C12, to get the C-trimmer C15 in about the middle of the range to get a flat edge without under- or over-shoot, we can now run the complete calibration procedure. Bud also created a calibration procedure on his Hackaday site, which in principle has most of the calibration steps that I'm showing here. I added a few profiling steps.
First we recommend that you build a little circuit that will help in the calibration of the probe. If you have a pulse generator that can provide nice flat edges at 1KHz with an amplitude of 25-30V and a rise time between 150nS and 250nS, you can use that instead and move on to the calibration procedure.
The DIY calibration pulse generator
In order to create a pulse that can be used to calibrate the probe, Bud designed a little circuit that can be put together easily on a protoboard. It uses parts that should be in your stock already, or are commonly available. More information can be found on his Hackaday postings.
Calibrating a differential probe is a little different from a normal X10 scope probe. We would like to create a large enough voltage differential to put some common mode stress on the probe.
Note that an earlier version of this circuit used a MOSFET for the output but that circuit was depending too much on the MOSFET type. The version below is more forgiving.
Here is the simple circuit:
It uses commonly available parts to convert the Scope Cal signal of about 1KHz at 3Vpp and a rise time of 2.5uS to a pulse with a desired voltage level of between 25-30V and a suitable rise time of about 250nS. Not too fast to introduce artifacts and not too slow to properly select the trimming capacitors and perform the AC calibration.
Building it on protoboard like I did is fine.
I'm powering the contraption with my Lab Supply at 30V to create a nice clean 30Vp-p pulse.
The output of the generator circuit can be set to 25-30Vpp with the power supply and will have a rise time of about 370nS with 47pF on the output. Without a capacitor, the rise time is 270nS as in the above picture.
Note that with my 10x scope probe connected to the output, the slope will already change due to the capacitive load of the scope probe, in my case 10pF. Keep that in mind.
You should only use the rising edge from the pulse generator for the trimmer capacitor selection and the calibration.
Connecting the diff probe to the pulse generator
To connect the diff probe to the signal generator, I used my grabber lead contraption from my earlier 100MHz diff probe to connect the calibration signal to the diff probe. There are 510 Ohm resistors in series of both leads to reduce reflections due to the inductance of the leads.
If you don't have the above flying lead contraption yet (I highly recommend you build it!), you can also configure the pulse generator circuit on the protoboard such that you can plug the diff probe directly into the protoboard to connect to the pulse output, and avoid any of the additional artifacts. You will need to hold the probe into position by hand at all times though, or use a vice or circuit board clamp to hold it in position.
With the pulse generator at the ready and connected to the probe, you can now proceed to calibrate the probe.
Calibration and Verification Steps
Step 1 Calibration setup
Connect the diff probe inputs to the generator (I used the grabber clips shown above), and connect the diff probe output with a shielded cable to CH2 of your DSO.
Connect CH1 of your DSO with a 10x scope probe also to the output of the generator, and trigger from that channel so you have a stable trigger signal and can compare the two signals.
This is about what you should see on your DSO. CH1 in yellow with a 30vpp input signal, and CH2 in blue connected to the output of the diff probe with about a 3Vpp signal (/10) with the same polarity.
Step 2 Calibrate the DC CMMR
In this step we adjust RV1 so both of the attenuation arms (positive and negative) get the same resistive value to compensate for tolerances.
Connect the negative probe input to the same connection the positive signal is already connected to. (both probe inputs to the output of the square wave generator)
You should see something like this:
You may need to lower the V/Div setting, until you get something like the above picture. The calibration is to get the positive portion and the negative portion at the same DC level. Adjust RV1 for a flat line, meaning the two arms of the attenuator are the same in value and at the same level. Ignore the peaks for this step and also ignore the fact that they could align but not to GND.
With the adjustment done, you should see something like this:
Step 3 Calibrate the DC offset
At this step we're going to calibrate the DC offset at the output of the probe relative to GND.
There are two methods, lets start with the DSO.
Go to an even lower V/Div setting for CH2. You should probably see the base line move up (a positive offset) or down (a negative offset). In the picture below, the offset is +5mV. Adjust RV2, the offset trimmer, for a baseline which is the same as the GND indicator on the very left on my DSO. This is the little blue arrow with the 2 inside. You can move CH1 out of the way a bit to see both GND markers.
Adjust the baseline of CH2 with RV2 to match the CH2 GND marker.
The second method is to connect the diff probe output directly to your DMM using a BNC to banana convertor and center the reading close to 0mV to eliminate the offset.
That concludes the DC calibrations.
Step 4 Calibrate the AC positive signal path
We're now going to calibrate the AC response of the probe. Set the CH2 to 500mV/Div.
Disconnect the negative output from the probe to the output of the pulse generator and connect it to the GND of the generator.
Move the trace from CH1 away to see the whole CH2 trace.
Change the channel offset to have a picture like this:
Use C15 to create a flat rising edge of the blue CH2 signal, without over- or under-shoot.
In my case, I already adjusted the trimmer while finding the optimum trimmer capacitor values. However, with the probe fully warmed-up, and having completed the DC adjustments, I'm now finding that I can no longer get a flat edge. The trimmer capacitor value is obviously still a little too small, to I changed them to a 68pF value.
Note: if you are going to use the diff probe with the 3D printed enclosure and wrap it with copper foil, it is OK if you cannot get a flat line at the end of the range. The additional capacitance of the probe inside the enclosure will bring you more into the trimming range.
With the 68pF values, I now have a nice trimming range either way for the positive input and that is even with the shielded probe.
Step 5 Calibrate the negative AC signal path
Now we're going to calibrate the AC response for the negative portion of the input attenuator.
Swap the positive and negative inputs of the probe inputs to the signal generator output. (Neg to plus, Pos to GND)
You should see something like this after you adjusted the vertical offset of the DSO CH2 to get the complete pulse on the display.
Adjust the positive going edge for a flat top (blue trace) by adjusting C16.
I found that I could not do that. I tried increasingly larger capacitor values for C12, and had to go all the way up to 100pF before I could adjust it. I must have gotten some parts out of toleration.
Here it seems perfectly adjusted.
However, there is another, and more precise calibration method, and that is to connect the positive input also to the output of the generator, so both inputs of the probe are connected to the generator output.
This connection allows you to go to a much lower V/Div setting to zoom in on the edges. You can now adjust C16 for a more precise flat line.
So even though I already calibrated the flatness in the previous step, with this setup you get a lot more resolution.
This is the best calibration I could do:
Note that this is at 2mV/Div where my DSO contributes to the noise.
This was taken with the 10MHz probe inside the shielded enclosure. There is no significant difference for the 100MHz probe.
We are now done with the calibration, but we can record some additional parameters
Step 6 Record the noise floor
The noise floor, or the probe induced noise is important to know when you measure other signals. Especially the 100MHz probe has a pretty high noise floor as you will see.
To measure this, short both probe inputs together. I used a 50 Ohm in-line terminator because I have an SMA connector connection for it.
The noise floor of my 10MHz probe is recorded as about 4mVpp by my DSO. When I move the 50 Ohm in-line connector to the end of the coax cable going to my DSO, it reports 1.2mV so the probe adds the difference. However, this is with the unshielded probe!
With the shielded 10MHz probe, and with the unfiltered full 300MHz bandwidth of the DSO, the noise floor is now only 2mVpp. That's pretty excellent.
The noise floor of the shielded 100MHz probe is significantly higher at around 30mVpp. This was to be expected based on the higher gain.
You can use the averaging feature of the DSO to clean-up the signal and cut through the fog.
Step 7 Record the rise time and calculate the bandwidth
You normally would use a sine wave generator to feed the probe, and record the frequency where the drop in amplitude is 3dB lower at the 10% and 90% points of the edge The frequency at that point would determine the bandwidth of the scope. I don't have a 100+Mhz sine wave oscillator, but there is another way to determine the bandwidth at the -3dB point.
To record the rise time, and infer the bandwidth of the probe, you need to have a pulse generator that has a fast edge. At least 1nS for the 100MHz probe, could be a bit less for the 10MHz version.
Risetime derating
Here is a diagram from page 2.7 of the PG506 manual that shows the risetime derating. This is the difference between the DUT risetime and the risetime of the measuring device(s) and the accuracy consequence.
In our case, the 100MHz diff probe version should have a risetime of 3.5ns and my 300MHz DSO should have a rise time of 1.15ns. That is a factor of 3.0. Go to the graph above and select the 3 on the ration horizontal line. From there, go up to the diagonal line, and from the crossing, to the left that shows the accuracy. That shows that the setup I'm using should have a 5.2% accuracy for the measurements. A 200MHz DSO would result in a 1.75nS rise time, resulting around a whopping 20% inaccuracy. The rule of thumb recommendation is to select the measurement device to be 5x better than the DUT. That would mean a 500MHz DSO.
How to derive the bandwidth
I must admit that I forgot some of the things I learned in the 70's when in my early years at Tektronix, I was repairing and calibrating oscilloscopes. At the time, I had a constant amplitude sinewave oscillator that would go a few 100MHz on my bench (don't remember which device it was, but it was very old). When the PG506 1MHz scope calibrator came out, it was replaced. The PG506 not only had a readout, but also had a fast risetime circuit that could produce positive and negative pulses with a rise time of at least 1nS. Several years after I left Tek, I was able to obtain one of those fast rise boards and stored it several decades for later use. A number of years ago I put that into my DIY 20Hz-20MHz pulse generator. (described on my Blog here).
The updated procedure to derive the bandwidth (by me), is based on the procedure on page 1-6 in the PG506 manual (here), to find the bandwidth of an oscilloscope, is as follows:
You connect the fast rise pulse to the DSO, using a good quality BNC cable and a 50 Ohm in-line terminator, or 50 Ohm input termination. Set the DSO to 5ns/Div and 100mV/Div.
Obtain a stable signal. Due to the inherently noisy V/Div settings, you need to set the fast rise output voltage to about 500mV-550mVpp. At this level, the risetime of the fast rise is at its best.
Record the rise time of the signal. This is the risetime for the fast rise pulse (1ns), the cabling and the DSO. Tek calls this the Trc. Record that value.
Disconnect the BNC cable from the DSO and connect the fast rise output pulse through an inline 50 Ohm termination to the diff probe input. Connect the output of the diff probe to the same DSO channel as before. The output will be 10X lower, so set the DSO to 10mV/Div.
Obtain a stable signal, and again record the risetime. This risetime is called Trs by Tek.
Calculate the risetime of the diff probe as follows:
Tr (diff probe) = [(Trs)2 - (Trc)2]SQR
This will subtract the PG506 fast rise pulse (1ns)+DSO+Cabling rise time from that of the diff probe.
With that, you can calculated the bandwidth of the diff probe by using Tr in the following formula:
For some high-end DSO's, the 0.35 factor can be changed to 0.40 or even 0.45, but that is based on the design of the front-end amplifiers and if they used a steeper frequency roll-off than the standard -3dB. I'm using 0.35 for my Rigol DS2302A based measurements because I don't know what the roll-off is.
The first step is to obtain and record the risetime of the fast rise pulse, the DSO and the cabling.
I'm using a different acquisition mode for the DSO, to show the sampling challenge for my DSO at these high frequencies. Note how few dots there are on the edge. The recorded rise time is 2.0nS.
Let's now measure the 100MHz probe and see what we get.
I had to average the signal due to the noise, caused by the probe and the lower V/Div setting. The DSO records a 5.65ns rise time for the diff probe (Trs).
Using the formula:
Tr (diff probe) = [(Trs)2 - (Trc)2]
Tr (diff probe) = [(5.65)2 - (2)2] SQRT or (31.9 - 4)SQRT or (27.9)SQRT = 5.28ns
Bandwidth (GHz) = 0.35 / 5.28ns = 66MHz
/ Rise Time (ns).
This is not what we designed it for, and is an outright disappointment.
Hunting for the problem
The funny thing is, Bud's version works fine, so we started to hunt for the issue. Unfortunately, it's very difficult to probe the various parts of the probe while powered, due to the introduced noise and the very low signal levels. So I gave up on that and started to verify the possible culprits that Bud indicated one by one.
Obviously there are a few capacitors in the signal path that are suspect of lowering the risetime. C21 in the feed-back circuit of U2, C18 in the offset circuit, and C17 and C18 which are in the feed-back circuits of the two U1 amps. When removing C21 or C18, there was no change. These are normally the capacitors that limit the bandwidth.
However, when I removed C17 and C18, the 0.2pF capacitors, the risetime jumped to 3.15ns. Presto.
Using the formula again without the two capacitors:
Tr (diff probe) = [(Trs)2 - (Trc)2]
Tr (diff probe) = [(3.15)2 - (2)2] SQRT or (9.61 - 4)SQRT or (5.61)SQRT = 2.43ns
Bandwidth (GHz) = 0.35 / 2.43ns = 144MHz
So we now know the untamed probe is capable of 144MHz.
If it's not the probe, what what then?
So what's the issue? Is the 0.2pF really a 0.2pF?
I first suspected that I could have swapped the parts inside my containers or swapped the covers for the 0.2pF and the 0.4pF capacitor containers. I normally only have one open at a time, but I could also have swapped the capacitors coming out of the bag, although I'm always extra careful especially with unmarked parts. So a swap is highly unlikely but can easily be proved. Did I swap them and did I install a 0.4pF capacitor? When I tried an 0.4pF for C17, to verify a swap, the risetime jumped to 6ns, so that's not it and my 0.4pF is at least larger than whatever value my so called 0.2pF are.
I then checked where I got these two capacitors from and both were from LCSC. They are supposedly from a reputable supplier (muRATA for the 0.2pF and YAGEO for the 0.4pF), unfortunately, I have no way of verifying the capacitance or tolerance of an 0402 size sub pF value with my otherwise good LC meter.
Because one of the 0402 0.2pF capacitors jumped out of my tweezers into a never-been-seen-again place, I tried a new capacitor, unfortunately, with the same poor result. Did I get a poor batch, or maybe not the real thing?
When I was investigating my 0402 capacitor container issue again, I noticed an unmarked container that has some of the parts in it from the earlier 100MHz probe, and long and behold, there also was a single 0.2pF with a tolerance of 0.005pF in there.
Handling that precious one with the utmost care and soldering it in the C17 position, produced this:
So it's definitely the capacitor batch that I received from LCSC.
We're now at a bandwidth of 119MHz.
For the time being, I'm running the probe without C18 in the feedback loop of U1b installed. C17 and C18 correct a phase shift caused by the feedback resistors (R16 and R17) against the input capacitance (0.4pF) of the amplifier. They create a pole around 200MHz but don't do anything else and should not degrade the frequency response, but are. And the ones that should, C19 and C21, don't.
I suspect both capacitor values and need to replace them by ordering from DigiKey this time. Unfortunately, that will have to wait due to the large S&H costs.
Verification of the issue:
Bud ran some simulations and found out that the LTC6269-10 indeed oscillates when C17-18 are less than 0.08pF. The recommendation is to use at least the minimum value available, which is 0.1pF. Keep in mind that with a higher bandwidth comes more noise, and this 100MHz version is already on the noisy side. Hence you need good quality COG/NPO capacitors for C17, C18, C19 and C21 with tight tolerances. I'll put a note in the BOM to that effect.
Signal fidelity
Besides the risetime and bandwidth, what is most of the time much more important when you use the probe, is the pulse signal fidelity, which is excellent:
Let's now do the 10MHz probe:
I lowered the DSO timebase to 20ns/Div to see the complete edge:
The reported rise time of the diff probe (Trs) is 32.2ns.
Using the formula:
Tr (diff probe) = [(Trs)2 - (Trc)2]
Tr (diff probe) = [(32.2)2 - (2)2] SQRT or (1036 - 4)SQRT or (1032)SQRT = 32.14ns
Bandwidth (GHz) = 0.35 / 32.14ns = 10.89MHz
Just as Bud designed it for.
Step 8 Record the propagation delay
The propagation delay or also called signal latency is the "delay" that the probe circuitry introduces to the signal while travelling from the input to the output.
This is an important number to keep in mind when you are comparing the signal from a different source with that of the differential probe. Of course, the best method is to use two differential probes. (;-))
To measure that propagation delay I fed the probe with a <1nS pulse that also went to CH1 of my scope. To preserve the signal fidelity, I also terminated the signal at the DSO input to 50 Ohm. CH2 was connected to the diff probe output.
The difference between the input in yellow and the probe output in blue is about 17nS.
Note that the input pulse rise time and fidelity is somewhat compromised by the double 50 Ohm terminations and the BNC cables in a T-connection. Also, at the low level of 10mV/Div for CH2, and the still unshielded nature of the probe, I used averaging to get a clearer picture.
Here is the 100MHz probe:
It has a propagation delay of about 3nS.
Skew factor
This propagation delay is sometimes called skew, and in the newer DSO's you can even input the skew factor of an input channel, in effect eliminating the diff probe skew compared to the other input channel to make time measurements easy. That's why it's handy to record that value for your probe.
Step 9 Record the maximum differential input voltage
This is the voltage differential between the probe Neg and Pos inputs, before the amplifiers in the probe start to clip. I'm using the Collector voltage from my Curve Tracer because the triangle waveform shows the point of clipping very well.
Here is the 10MHz probe measurement:
The measured input voltage is +/-80.8Vpp, which means a total differential of approx. 160V and at that point the probe starts to clip the output signal which is why it is only 7.92Vpp, and not 8V.
Here is the 100MHz probe measurement:
The measured input voltage is +/-27.8Vpp at clipping for a total differential of approx. 55V.
As an aside, the maximum input voltage differential is mostly caused by the rail voltage for the Opamps, which is 5.3V (+/- 2.65V) for the 100MHz probe and 15V (+/- 7.5V) for the 10MHz probe.
Step 10 Record the frequency response flatness
Using a NanoVNA, you can let it sweep a set frequency range and see what the response by the diff probe over that range is.
The 10MHz probe:
Here I let the NanoVNA sweep from 50KHz all the way to 20MHz.
The frequency response starts to roll-off very gradually and reaches the -3dB point at
about 11MHz. Excellent!
The 100MHz probe:
In this case, I set the NanoVNA to start at 50KHz and let it sweep up to 200MHz.
There is a little dip at 70MHz in both the S1 and S2 inputs, most likely a reflection from the probe input circuitry. This is almost the same as with the previous 100MHz probe. Could also be caused by my SMA to 2-pin header to connect to the probe.
The overall response is very flat with a gradual roll-off starting at 120MHz. The -3dB is at about 150MHz. Also excellent!
That concludes the calibration and verification procedure.
Note that you may have to do the AC calibration again, once the probe is in the enclosure, and you applied the copper tape EMI shielding. That shielding, even applied on the outside of the enclosure will change the parasitic capacitance values a bit of the very sensitive front-end. That's why Bud designed the enclosure such that you can calibrate the probe while in the enclosure.
With the 10MHz version, there was no significant change so I could leave the trimmer caps as they were. The 100MHz is more sensitive and hence more prone to parasitic changes due to the shielding.
Just in case...
I am using the probe with the copper foil as it is for a while, so I can easily open the probe up again. You may want to cover yours with shrink-wrap tubing or tape, to avoid a short while using the probe in circuits, because the copper foil is earth grounded. You could also get a shock while holding the probe, and it touches part of a live circuit. But you already knew that, right?
Wrapping-up
With the design verified and tested, all relevant files are now added to a Github project so you can start to build the probe. The files have also been uploaded to the PCBWay Shared Project files.
Of course we like all our babies and as a parent, we should not single one out, but...
Both Bud and I really like the 10MHz probe best. If you look at the specifications, it is very versatile, and also very important, it has a very low noise output. This is especially important with our hobby DSO's. There you have it. The rest is up to you.
It was a very long journey that started with an operation on Bud's ankle and he desperately needed something to do to keep his sanity. This was early this year, and it took us all the way up to November to get it right. We are proud with the way this project turned out, and we hope you like it too!
Enjoy building and using these probes, we're sure you'll like them.
