Show by Label

Friday, March 19, 2021

Building a Curve Tracer - Version 2

Building a Curve Tracer - Version 2

Although Version 2 is a fully functional Curve Tracer, it has some limitations.

We don't recommend you build this version, but it shows the process we went through to get to the final version, described here: Version 3



After I built my Curve Tracer in 2017, as what I would now call a prototype, this project has gotten more interest.

A follower, Mark Allie, had contacted me if I would be available to help him build circuit boards with mostly SMD parts. His goal was to provide Curve Tracers for the students of his University. Of course I was keen to support his project. 

I fully endorse and support his activities. As far as I know, there are no decent DIY kits available on the open market anymore, which is why I started this project myself. It's even difficult to find a used commercial version, and if you do and it is broken, the parts are almost impossible to get. Many universities seem to suffer from this issue and are running out of working Curve Tracers. During this project, we got contacted by yet another university with exactly the same problem. A pity, because a Curve Tracer is such a great tool for learning about analog devices and it is almost a shame that in this digital age they are almost forgotten.

I want to revive the use of a Curve Tracer as much as possible.

Over the course of more than a year, Mark produced a set of boards and after solving some issues, made the instrument work.

Mark's latest schematics, BOM's, PCB details and Gerber files can be found on his Github site here : Mark's Github 

I would however caution you to start ordering PCB's and parts for this version 2, because although it is fully functional, it has some issues. We're now in the process trying to address them. If possible fix them in this version, and if not, in a new version that a small team of enthusiastic people are working on at the moment.

The development of version 3 will be described in yet another Blog posting: Version 3

Here are pictures of the very first prototype of Version 2 that Mark built. He since updated that version and it is this version (he calls it Version 1b) that is available and described on his Github site. 


Mark contacted me again to ask for help and sent me a set of his boards already populated with the SMD parts so I could complete the instrument. 

Here are two measurements that Mark took with his CT. 

First an NPN at 10uA/Step

And here a PNP at 10uA/Step


In the meantime, another interested hobbyist, Richard, also built this version. 


This was taken with a full triangle waveform per step, it clearly shows the two tracks. 
Changing the setup to use one flank per step cleared that display problem.

Look at the first CT blog in the section called Trade Offs for a detailed description of what we're talking about here. 

The three of us have been working hard on profiling the instrument and also trying to find the cause and solutions for some of the issues this version has. Besides fixing the issues, we're also working on several improvements and additions that we feel the original design really needed. Some of the issues will be described here, but the new design will be described in the Blog for version 3.

What were some of the issues that we have found.

I'll describe them here, also with the solutions we've found, although most of them will be implemented in the Version 3.

Component differences between my original design and Mark's version.

As I mentioned earlier, Mark Allie was so kind to sponsor me by sending a set of boards with the SMD parts already mounted.  When I started to complete the main board I ran a few tests on the partially finished portions. 

When I tested the triangle supply, I ran into an oscillation problem. It took a while to figure this out. Originally, I used a Darlington transistor as the series regulator, unfortunately, that TIP150 transistor is no longer available. Mark tried a few replacements like the BU323ZG but settled on a normal transistor the MJL328A. When I tried my last good old TIP150 first, the triangle supply worked as it should. I also changed some of the installed part values to the ones I had in my prototype, in order to try to make an identical replica. I noticed a small item. The trimmer for the zero Volt alignment for the triangle generator was at the end of the range, so I think it's better to change the 20K to a 50K version. I also had to tune the capacitor that sets the base frequency to a lower value because the triangle output was not linear in the highest voltage range. This is what I also found and described in my early version.

Initially, I only used the low volt (24VAC) winding of my special Collector supply transformer that I took out of my own CT.

I then moved on to the Front Board and got that going as well. Initially, I used short leads to connect the rotary switches to the board, so I could lay it flat on my bench to have better access to all the parts. That actually caused some noise, so I changed that by mounting the switches as they are intended. The connections from the main board to the front board were made by Dupont female-female leads so I could still lay it flat on my bench to have access to the parts.

A Design change

Mark and I went back and forth for quite some time about his way of switching the polarity. I was convinced he had it wrong, and I was missing a relay in his design, but it turned out he found a smarter way that actually saved a relays.

The "trick" Mark has found is to connect the Step Gen isolated Ground (ISO-GND) directly to the Emitter of the DUT. The Emitter already switches polarity from the CS- to the Common Ground (or CS+), so he could eliminate the relays on the Step Gen board, and accomplish everything with the same triple switch we already had.

The other change he made was to reference the BNC grounds to the Common GND. He also improved the X-Y amplifier section somewhat.

Another design change caused some problems for me

After I got most of the parts working, I added the high voltage supply and tested the adjustments for the maximum Collector voltage, and the current limiting settings. I then moved on to test a DUT, but by using scope probes, not the X-Y amp yet, and here I ran into issues. First of all the Step Gen would not work properly. That was due to a poor solder joint. I then started to destroy several of the 2N3904 transistors that I was using as the DUT. 

After a lot of head scratching, soul searching and going back and forth for several days with Mark and also Richard, it turned out that Mark inadvertently created a problem for me (well, actually for my transformer) with the way he created the Collector Supply. Mark optimized the circuit and made it work, of course, but although it was different from the way I designed it, Mark's version worked for him and Richard, so I thought of other issues first. It turned out that his design only works with real center-tapped transformers, or with separate windings with equal voltages but not with asymmetrical voltages, which is what my transformer has.

I use a special transformer with one winding of 24VAC, and another winding with 78VAC. Using that transformer with the current PCB caused the voltages to increase to very high levels. The result was that the automatic switching from the low to high supply created a large and destructive pulse and the relays started oscillating badly on the way down from high to low as well. This destroyed not only several poor 2N3904 transistors I used as the DUT, but also my last TIP150, the driving Opamp for the triangle supply and two of the BU323ZG series transistors, in addition to the voltage feed-back correction Opamp in the Step Gen. A real massacre! 

The untimely death of the Opamp in the Step Gen was a wake-up call, see below. This was most likely because one of the 2N3904's used as a DUT probably got shorted and the full collector supply was "presented" at the non-inverting input of the tiny SOT23 Opamp. Poor thing.

The remedy for the collector supply with special transformers is to use a full bridge for either winding like I used to have in my prototype. I first tried that Manhattan style and that worked as I had it before. The relay oscillation went away, however, since I was now concentrating on the relay switching, I noticed a problem that I did not discover in my prototype. There is a nasty spike during the transition that could be up to 150Vp-p with no load. The spike is reduced with a load attached but only went away completely with loads of less than 500 Ohm.


The lower blue trace is the RelayEnable signal that I used to trigger from. You can see that when the unit switches from the low to the high supply at around 30Vp-p, there is a large glitch with a duration less than 1mS but that can be up to 100V. It may very well have been in my original prototype, I just never saw it, and never had a DUT die on me that I could attribute to this issue. 

Moving from high to low is pretty smooth, just some relays bounce.

Obviously we need to find a way to get rid of this nasty glitch and if possible the relay bounce.

As a stop gap, we have found a fix that helps to reduce this problem, by using a diode and an R/C combination on the relays contacts that does the switching, to reduce the sparking. 

Collector supply issues

Richard, Mark and I all used different series transistor pairs and we all suffered from oscillation issues that I didn't have in my original version. When testing the circuit with my old TIP150 transistor, it worked pretty well, even though Mark used very different (faster) Opamps. Because I blew my Darlington transistors, the BU323ZG, I used the only other transistor type alternative to me, the MJL328A. It turned out that distributors like Digikey and Mouser were completely out of stock for the high voltage Darlington versions. This was probably caused by the current shortage of parts, especially in the car industries where these transistors are used in ignition systems. I put a back-order in and worked with the BU323ZG for the time being. The original design was for a Darlington and although it also worked with this normal transistor, it was not stable. I raised the Bias levels to give the transistor more current to work with, and also experimented with a driver transistor both in the Darlington and in the Sziklai configuration, but I was not happy. At this moment, Mark and Richard were only using 56VAC transformers that would give us 80V, but I wanted to increase the Collector supply to at least 180V. 

If you think about the demands we wanted to put on the supply, it has a daunting task. It needs to go from zero to 180V with currents ranging from zero to 2 Amp, with virtually no triangle distortion.

We now have developed a new version of the Collector supply that will be tested on a prototype soon and the results will be going into Version 3. The details will be described in the Version 3 Blog.

Initially, I had both boards and the transformer all wired-up flat on my bench so I had full access to the parts. Now that I fixed the Collector supply issue, I wanted to start to put the main board in the intended enclosure. I also got the parts that I needed to add the type C8 mains receptacle with a filter, the fuse holder and the mains switch. I added all that to the back panel. I also had to mount the transformer laying on it's back and mounted on the back panel, because there was not enough space to put it anywhere else. 

I could now finally move on and test and verify the operation some more. My unit is now fully operational and seems to be working fine, apart from the Collector Supply issue, that is more serious on my CT because of the transformer I use.

To be on par with the instruments Mark and Richard have, I ordered a similar transformer they used, a truly center tapped 56V AC 1.8A version and have been using that since.

Because the relay switching to the higher supply gave me some issues, I disabled the relays and continued to test and profile using the lower Collector voltage of less than 25Vp-p.

Avoiding another catastrophic event

When I had the glitching issues that were caused by my special transformer, I blew-up a number of parts. One of the casualties was the voltage follower that is used in the Step Generator voltage mode feed-back. My take on this was that the Collector-Base diode blew putting the full Collector voltage on the Base connection to the output of the Step Generator amplifier, and of course the poor unprotected Opamp. Opamps have a hard time surviving voltages on the inputs that are greater than the supply voltage. Because the DUT Collector voltage is only a silicon sliver with the thickness of a few microns away from the Base, this is something that can happen again. Besides, it's relatively easy to short the Collector jack with the Base jack, with the same effect.

As a first remedy, we have now added a 100K resistor in series with the input of the Opamp, and use two diodes to clamp the input to the rails. This is a quick-and-dirty solution for this version, but we need more.

I wanted to try to avoid that disaster from happening again and see if we could add good protection to the circuit. I called in the help of a good friend, who was a chip designer in a previous life, and he came up with an good concept that will be a very effective solution. The details will be described in the Version 3 Blog.

Another part related issue

We found that the final step (step 7) of the Step Gen was uneven compared to the rest of the steps. It turned out that this was caused by the selection of the buffer Opamp that was not a rail-2-rail version. We will fix that in the next version.

Adding a Z-axis blanking output

Richard uses an analog scope and he has found a way to pick-up a blanking signal from the triangle generator that he can feed into his scope. This will be permanently added to the next version.

Noisy traces on a DSO

Because I'm using a relatively inexpensive DSO, a Rigol DS2072A, most of the X-axis displays in the lower V/Div. settings are very noisy on my DSO. Both Mark and Richard use professional equipment and they don't have this issue.  There are two solutions. We can add another Opamp with a higher gain, but that will also amplify the noise from the source. The other solution is to use a higher value shunt that will increase the voltage. Both solutions will allow you to avoid the lower level Volts/Division settings of the DSO. We will adopt the dual shunt resistor solution using 1 Ohm that gives 1V/A and a 10 Ohm that will give 1V/0.1A and that could be added in this version, but will also be added in Version 3. 

This was taken with a 10mV/Div setting which is noisy but still OK. Lowering this setting however to 5mV/Div will make the picture very noisy and at 2mV/Div. unusable. Using the new 1V/0.1A will actually represent a 10x increase, allowing you to significantly increase the V/Div. settings to ranges that are a lot less noisy.

Here is a measurement using the 1 Ohm shunt viewing at 2mV/Div.

And this is the same measurement but with the 10 Ohm shunt now able to view at 20mV/Div.

Small offset on the Y-axis

Richard has found that on his CT, there is a slight offset of the X-Y picture. This is probably caused by part tolerances. He fixed it by shifting the output of the Opamp a little with a high value resistor. We're going to see if we can address that by using an Opamp that has an offset built-in.


Oscillation issues

We all had some oscillation issues. Some were coming from the Collector supply, and some from the Step Gen. The re-design of the Collector supply will resolve that. I have found a solution for the Step Gen oscillation by inserting a 10uH inductor in series with the Base connection to the DUT. That fixed the oscillation problem you see below.


So how does our Curve Tracer compare to the Tek 577?

Well, it will never be a replacement of course, but we would like to see if we can duplicate some measurements that are described in the 577 documentation. This is a reality check to see if we can/need to make some changes, or just call it a "no-can't do that".

Richard made a comparison with the way a particular measurement is described using a Tek 577, while using his Version 2 CT.
The details are available on my Github site here : CT Github site

We believe that the difficulties in setting the maximum current levels will be solved with the new voltage and current ranges that will be implemented for the Version 3.



I hope that we have identified and even resolved most of the issues with this version, although we're continuing to test and profile. 

Stay tuned!

If you like what you see, please support me by buying me a coffee:





paulv said...

Just some house-keeping information.
I will remove all comments that try to plug a service, so please don't pollute this Blog.
I will also remove all posts that contain personal information like an email address, for safety reasons.

JpKlitzke said...

Very Cool like to build one What was the estimated cost Paul

paulv said...

I would not build the Version 2 because of the issues, but only the Version 3. We did not make an estimate nor is it easy to make one, sorry.

Pierre said...

Which are the secondary voltages of the Transformers :T1, T2 and T3 in the diagrams

paulv said...

We do not support this version of the Curve Tracer because it has flaws, look at Version 3.