Thursday, December 17, 2015
Rotary Encoders & Raspberry Pi
Well, no! The Pi is so much faster that the solution did not port or translate, see this post for Details on how I developed one for the Pi. https://www.raspberrypi.org/forums/viewtopic.php?f=37&t=126753&p=848012#p848012
Enjoy!
Demistifying Rotary Encoders
The new project needed a large selection method for frequencies and voltages, and traditional rotary switches became expensive and complex. So I decided to use a rotary encoder together with an embedded controller. It also solved the problem of a complicated frontpanel, because I now could use a display driven by the controller.
While researching rotary encoders, I learned a lot about decoding them, and eventually decided on a method that is adequate for my application.
I wrote two posts on the PicAXE forum to explain this in more details, and here is the link: http://www.picaxeforum.co.uk/showthread.php?28222-Demystifying-Rotary-Encoders-(one-more-time)-Part-1-2
Enjoy!
Monday, December 14, 2015
Using a single push-button to start/stop/powerdown the Raspberry Pi
Linear came up with a couple of chips that helps to solve this problem, and with the help of the Raspberry Pi foundation, an overlay was created to get a GPIO port that can signal the end of the Halt status.
Based on that work, I created a design that is well documented and rather easy to build. While I was at it, I came up with a couple more designs that uses this chip, the LTC2951-1, although there are several in this family. Unfortunately, these chips are hard to get, not in-expensive at about $5 each, and come in a tiny, very tiny SMD package. On top of that, MOSFET's are used to switch the power, and the right ones (with a low RDS-on) are also only available in SMD packages.
Eventually, I was able to come-up with yet another design that is even more simple, and only uses 4 resistors and 1 capacitor, in addition to a push-button.
Here is the link to the posts on the Raspberry Pi forum:
https://www.raspberrypi.org/forums/viewtopic.php?f=37&t=128019
Enjoy!
Monday, October 26, 2015
Adding an Analog Output to the Pi (DAC)
Here is that link : https://www.raspberrypi.org/forums/viewtopic.php?f=37&t=124184
Enjoy!
Adding an Analog Output to the Pi (PWM)
Here is that post : https://www.raspberrypi.org/forums/viewtopic.php?f=37&t=124130
Enjoy!
Adding an Analog Input to the Pi (ADC)
Here is the post : https://www.raspberrypi.org/forums/viewtopic.php?f=37&t=123962
Enjoy!
Tuesday, August 11, 2015
Controlling a temperature driven Fan with PWM
I don't like noise or hum, so I wanted to use a fan that would be off when not needed.
Typically, by using Pulse Width Modulation (PWM), rather than a voltage, you can precisely control the speed of the fan, and ramp it up when needed.
I was able to find some clever methods to implement this, and pieced together this design:
The building blocks are relatively simple. I use a DC-DC convertor, not a linear one, to bring the 24V DC to 5V, because a linear LM7805 would get too warm burning off the excess voltage.
I use a comparator with hysteresis to determine the starting point of the fan, based on the temperature reading (in ohms) of a thermistor that is mounted on the heat sink. I don't have the datasheet for the thermistor, but found that it gets from 10K at room temperature to about 5K when the heat sink is getting really hot.
I need 5V for the fan, because I have one that is only 10mm thick, and that is what I have room for.
The clever trick of this circuit lies in the fact that the Control Voltage (CV) input from a 555 timer is used to control the PWM.
The 555 is producing pulses, and the pulse width, and also a bit of the frequency is varied by applying a voltage to the CV input. The output of the 555 goes to a FET that drives the fan.
The whole thing works very well, although the low pulse frequency can be heard from the fan, so I needed to use C4 and C5 to remove that chirping sound.
There are two disadvantages of this design.
One is that you cannot regulate the full 100% of the pulse width. The minimum is OK at about 30%, it let's the fan spin very slowly, but the maximum is only about 70%.
The other disadvantage is that you cannot increase the frequency of the pulses above the hearing frequency of 20KHz, because then the effect of the thermistor on the PWM range is greatly reduced.
In my application, that is not good enough, I need to be able to get to the maximum fan speed in order to keep things cool.
There are special fan/motor controllers that allow you to do that, so I have a couple of TC648VPA chips on order. Stay tuned.
The chips arrived and I made two different circuits. One for a 12V DC Fan I also ordered, and one for the 5V DC fan. I used an Excel spreadsheet that is available on the MicroChip website to calculate the resistors (R1 and R2) to get the best starting and maximum fan speeds.
The TC648 works really well, and is a nice addition to my toolbox. Next step is to put the circuits on vero board and install them in the power supplies.
After playing and experimenting, I decided to make a few changes to the circuit. First, I implemented the VAS pin. The explanation says to set the Auto-shutdown threshold with this pin, but it also sets the turn-on threshold. The nice thing is that the fan will be driven to full speed for a short period, making sure it turns on right away. This was a problem I noticed with the earlier circuit.
Second, I deleted the resistor that was in parallel to the NTC. I felt that with a 10K NTC, it didn't do much. I also experimented with C7, which sets the frequency to see if I could remove the audible noises at lower speeds. That only worked with a 10nF cap, but then the PWM spread is limited and therefore also the speed control. So I resorted to using the 1uF value.
BTW, after a lengthy search, I found the information about the NTC I was using, because I wanted to know the temperature curves. Unfortunately, the shop I got them from didn't mention the type or manufacturer. I have the TDK B57045K0103K000.
I also found a much more elaborate data sheet for the device here: http://ww1.microchip.com/downloads/en/DeviceDoc/21755c.pdf
And here is the latest version of the schematic:
A start-stop button with shutdown and power-off features
In the meantime, I ordered some of the chips (LTC2951-1 : Start-Stop Push Button Controller) he used, and I was able to built a working circuit that does not even need Device Tree features. Here is the link to the post: https://www.raspberrypi.org/forums/viewtopic.php?f=41&t=117863
In this post you'll also find the links to the other ones.
Enjoy!
Tuesday, August 4, 2015
Dynamic DC Power Load (updated)
Here is my version:
Update : Aug 29, 2015
Over the last couple of weeks, several things have changed after the initial build. I received my new Volt/Amp meter, and modified that per the instructions by Peter Oakes. It now has a floating differential current sense circuit so I have put it in between the MOSFET's and the current sense shunt. I'm also very impressed with the accuracy of both meters. No more adjusting, and the decimal point and number of digits change like a "real" DMM, what it really is. On top of that, it has exactly the same accuracy as my DMM. Highly recommended!
I also installed a temperature controlled fan, it uses a nifty fan controller chip to do all that. Look at a separate post for details.
To get a better handle on the temperature at higher loads, I decided to use two MOSFET's in parallel. I was a little careless testing all this, thinking that this plus the fan would solve my heat problems. I was eager to start my tests of the two PSU's, which is why I built this tool in the first place. Well, I lost a MOSFET due to runaway thermal issues. Bummer!
Despite my search for information on how to put two power MOSFET's in parallel, all I could find (at the time), was the use of seperate gate resistors, to balance the two MOSFET's. Well, that didn't work very well. Due to the differences in the VGS for both devices, one took all the heat.
After swapping out the broken one, I was a little bit more careful, and noticed right away that there was a large difference in the thermal balance between the two. One was getting slightly warm, but the other too hot to touch. Apparently, the only real solution to the RDS(on) and therefore thermal balance issue is to build two circuits with separate op amps and sense resistors. I did not want to do that right away, and tried to see if I could get close enough. First I increased the value of the gate resistors, and also matched them to within 1 Ohm. That did not do anything to get a better balance. I then started to increase the value of the one one that got hot, but that did not produce the result I was after. In the end I used a 10K trimpot to set the VGS to a reduced level, and used my temperature probe to balance the two. I got to within a few degrees C after running the unit with a 1A load, such that the temperature was about 50 degrees C. It's not ideal, because I don't think the VGS delta has a linear relationship, but for now I'm happy.
Here is revision 3:
This design will most likely be further tuned while I'm going through the paces of testing and specifying my two power supplies, so stay tuned.
Update : August 30, 2015
Well, it didn't take too long for another refinement.
I kept mulling about the thermal balance challenge, and was not too happy about the comment I made earlier in the post above. While surfing for ideas for the parallel use of the power MOSFET's, I came across a comment that I took for face value, assuming the guy knew what he was talking about. I simply repeated his comment in the previous post that the "real" solution is to make two separate circuits. So to duplicate the op amp, FET and current sense resistor.
I kept thinking about that statement, and decided that it is misleading or even incorrect. Yes, you will get two circuits that together will share the load, but it does absolutely nothing to get a better load balance, or a better thermal balance. So, back to the drawing board.
Initially, I wanted to use two thermistors (or NTC's), thermally connected to each FET, and then use a comparator to decide if one was getting hotter than the other. The challenge was that I could not find an easy way to mount them on top of the MOSTFET's. Sure, I could have mounted them on the heatsink, but I would have lost some temperature sensitivity, with the heatsink acting as a dampener.
I then switched to using the junction temperature of transistors, but again could not find a good thermal connection to the FET's. I then went through my power transistor stack for ideas and found two excellent specimens. I happen to have quite some (very old, as in late 1970's) D40C5 darlington power transistors, that have a very flimsy heatsink. This turned out to be a perfect way to mount them with the same screw that also mounts the MOSFET, provided you also isolate the transistor tab, because it is the collector. I happen to have all the bits and bobs required, so I could continue.
BTW, the fact that I use darlington transistors, is only because I had them. You can use any TO22X kind of package, as long is you can thermally connect them to the MOSFET's.
In any case, whipping-up a small circuit to measure the delta temperature difference and drive a FET to drop the VGS on the gate of the FET that get's too hot was quite simple. By selecting a feedback resistor to create a small threshold, I managed to keep the temperatures within about 1 degree C. So whenever the temperature of Q2 id one degree C higher than that of Q1, the gate of Q2 is clamped and as soon as the temperatures are equal again, the gate is released again. Perfect!
Here is Revision 4:
Resistor R21 in the drain of the BS170 MOSFET is there to reduce the gate voltage, not to yank it to ground. With the resistor combination of R9, R14 and R21, the gate voltage drops from about 300mV, just reducing the load and therefore temperature of Q2, while not affecting Q1, and therefore the output.
Just in case you are interested to use this idea in other applications, you need to be aware of two constraints. With this particular application, I'm not concerned about the self-heating of T1 and T2. I'm interested in the delta temperature difference, not the absolute difference. Also, the fact that the load will move from both MOSFET's to mostly one for shorter periods is of little concern for me. Both MOSFET's are individually capable of handling the full load.
Enjoy!
After having used the DC load for about a year, I was a little annoyed about the fact that I could not obtain a zero load at the output, despite the trimmer on IC1 and using a small negative voltage for it. Although the potmeter (R6) was getting down to zero Volt, the output of IC1 remained at 3V3, resulting in a 0.0059 Volt over the shunt (R12).
I finally figured out a way to adjust the potmeter range at the lowest position to have an output of zero load. The trick was to apply a very small negative voltage at the lower part of R6, instead of connecting that to ground. I used a tap on the two diodes that created a small negative voltage for IC1, in an earlier attempt to obtain a zero Volt output, and fed that -0.7V to an added 10K trimmer. With R6 in the lowest position, the new trimmer is now adjusted such that the output is showing no load.
Here is the new schematic with that change:
June 13 2017
While going through the design of a new power supply (a new post will be coming), I was not satisfied with the dynamic switching of the load. The pulse that is coming in from the BNC input actually switches the load hard on and off, so the maximum load is used, and there is no way to limit that.
I also didn't really need a fully variable input to the load, a simple pulse would suffice. During the power supply design, I actually put a 555 based timer on a breadboard and fed that to the DC load external input. I also added a few parts to make the output variable. I liked the results, so I put everything on a little prototype board and added it to the internals of the DC load.
Here is the schematic of the Pulse Generator addition:
Anyway, I made Diptrace resemble the Eagle schematics a bit, because those still look visually pleasing.
OK, back to the Pulse Generator. I used a 555 based circuit that produces a pulse train with approx. 180mSec on, and approx. 70 mSec off periods. By just changing C3, you can change the frequency. At first, I used the 12V supply to power this circuit, but later changed my mind about that. I didn't want pulses with a 12Vp-p output roaring through my box, when I do sensitive measurements.
I changed the single throw switch that selected between static and dynamic inputs, to a double throw switch. The other side now switches the pulses off, by shorting the charging cap C3 to ground, such that the 555 pulses are off when the static mode is selected. It keeps things more quiet in the box. At first I switched the power to the 555, but turning it on produced an unwanted output glitch causing a high load on the output until the 555 settles.
The DC load will be switched on during the on-pulse, and fully off during the off-pulse. Because of the 5V supply, the output of the 555 timer is about 5V p-p. That signal is used for a trigger out signal. That signal now goes to the same BNC that was previously used as a pulse input. This signal typically goes to a scope to trigger it. This trigger output is an important feature because if you test a power supply, as an example, you want to see the effects of the load regulation while you switch dynamic loads on the power supply output. If the load regulation is any good, it will be very hard to see any effects on the output of the supply, so knowing where to look, based on the trigger signal, really helps a lot.
A second divider (R5 and the potmeter R6) attenuate the signal, so at the wiper of the potmeter, I will have 500mVp-p at maximum, and 0Vp-p at the minimum. A voltage level of 500mVp-p will turn the DC load fully on, which is the same as the static load adjustment potmeter does. I happened to have a special dual 10K potmeter, with separate controls (outer and inner) for each section. If you use other resistor values, you have to recalculate the dividers. Make sure that the maximum output on the wiper of the output setting (R6) potmeter is slightly above 500mV to be able to drive the load to the maximum.
The wiper of the output setting potmeter goes to the switch that selects the static or dynamic behavior of the DC load.
To create a minimum offset for the DC load in the pulsed mode, used to test Transient Response and Recovery Time Measurements for power supplies, the other half of the dual 10K potmeter is used to add a 0..100mV offset. This should be enough to create a "bias" current of up to 1Amp for the supply, and then drive it higher with the pulse. This allows you to see the "sagging" and negative pulse excursions of the supply output voltage better than letting the supply go to ground, or 0V when you don't add the offset.
If you now turn on the dynamic input, the pulsed output level (with or without out offset) can be set, and the Amp level will be displayed on the Amp panel meter display.
Because I had this special dual potmeter, all I had to drill was a single new hole in the front panel, and that just sits just above the static output level potmeter. The circuit is build on a small prototype board that is mounted on an extended post already mounting the main PCB.
Everything works really well, this simple DC Load is getting better and better, I'm happy.
Tuesday, May 19, 2015
Simple but precise 1KHz Distortion Analysis Tool
The most simple solution is to use a low distortion sine wave generator to feed the DUT, and at the output, filter the sine wave out again. What is left would be the distortion created by the DUT.
As it turns out, designing a very low distortion since wave generator is not that easy. Luckily I found a unit that is excellent for a reasonable price. It is called the Mickevich (the designer) Ultra Low Distortion (<0.00001%) 1kHz Sine Wave Generator.
What was left was to design a precise filter, the required power supply and the box to put it all in to.
In addition, I needed an external sound card interface that is used as a front-end for the PC-based FFT analyzer.
I described this project on a diyaudio forum where much more information can be found about the oscillator, and the filter I used : http://www.diyaudio.com/forums/equipment-tools/252751-simpel-1-2-diy-1khz-distortion-analysis-tool.html
In the meantime, I also purchased a high precision DIY capacitance meter, and also learned that the opamps I used are not the best. With time permitting, I will update the filter with better tuned filter capacitors, and change the OpAmps. Stay tuned.
Enjoy!
paulv
Triple function button to reboot/halt/reset/restart Pi
Here is the post on the raspberrypi.org forum : https://www.raspberrypi.org/forums/viewtopic.php?f=37&t=48455&p=412858#p412858
Enjoy!
Paulv
Monday, May 18, 2015
Failsafe(r) use of GPIO pin for critical applications
Here is a link to the post that went through the discovery process and with the help of some smart contributors, eventually led to a failsafe design in hardware and software. This resulting hardware and code has been running for about 18 months to my satisfaction.
The post was published on the raspberrypi.org forum : https://www.raspberrypi.org/forums/viewtopic.php?f=37&t=46889&p=432578#p432578
Enjoy!
Paulv
Simple UPS for Pi
here is the form post from raspberrypi.org : https://www.raspberrypi.org/forums/viewtopic.php?f=37&t=52860&p=500147#p500147
Enjoy!
Paulv
Detailed description of the popular DHT22 temp & humidity sensor
Here is the post on the raspberrypi.org forum : https://www.raspberrypi.org/forums/viewtopic.php?f=37&t=91326
Enjoy!
Paulv
Tuning a 0..30V DC 0..3A PSU DIY kit
Before you read on, you should know that after building two versions of the DIY kit, I was not very impressed about the stability, noise, but above all, the complete lack of any form of protection. Think about that before you connect something valuable to the supply.
Subsequently, I designed a more modern power supply that you can follow here: http://www.paulvdiyblogs.net/2017/07/my-new-power-supply.html
After building two of them as well, and learning a lot and have a lot of fun experimenting and building, I still decided to purchase a lab quality professional power supply. That should hopefully tell you a few things.
There are many things you can learn about the kit in this post, and also select possible improvements, but please understand that I can't really help you anymore. It's been too long ago.
Due to a move, I recently sold my large lab PSU, and needed a substitute. I wanted to be a little more flexible, and did not want a huge and heavy supply on my bench anymore.
Searching the web, I came across an inexpensive DIY kit that implemented a very popular design for a power supply. I seldom if ever need more than 1A, so I used the kit to tune it to my liking, and also added the latest modifications for the original design.
I added an LCD display and one addition to the original design is a current setting mechanism, using the display, so you can set the current limiting or constant current mode before connecting the DUT.
I have built two supplies and can connect them in parallel to get more current, or in series to get a true dual +0..30V *zero* -0..30V supply or a 0..60V supply. One is designed for 3A and one for 1A max.
After some fiddling, I also designed a simple dual tracking system when the two supplies are used in series, so one supply controls the other.
Enjoy!
paulv
Because I already have a voltmeter, the easiest method was to use that to show the current setting. However, showing the value on the current meter display with the unit I use is tricky.
To show the current setting on the voltmeter, all we really need is a convertor that translates the current limit setting to a voltage.
To show the relation of 1A = 1V, with R7 at 0.47R, we need a multiplication factor of 1/0.47 = 2.127.
By using an additional op amp (U5), we will make this circuit independent of the maximum current of the PSU.
If you look at the schematic, the circuit around U4 implements that function.
RV2 can be adjusted by setting P2 to the maximum value of the current, say 1A. You can measure the voltage at the wiper of P2 with a DMM and set P2 to read 1.00V on the DMM. If you implemented R18 in combination with a trimmer, adjust that trimmer first to show 1.00V with P2 at maximum. Push the CC set button and adjust RV2 to have the voltmeter of the PSU show 1.00V as well.
R24 = 1K
R25 = 240R
R26 = 10R
RV2 = 2K
RV3 = 200K or 250K (optional, see text)
U5 = TLE 2141
U6 = LM337
C 11 = 47uF/25V
C12 = 3300uF/50V
C13 = 22uF/10V
D13 = 10V 1W
D14 = Green LED
D15 = Red LED
Volt/Ampere panel meter
S1 double throw switch
S2 single throw push button
Modifying the PCB to the latest version of the supply
In the above text, I have given an overview of the changes to the components supplied with the kit, to make it work a little better.
First of all, we need to implement the supply changes to the opamps (through D13), and so a few traces need to be cut on the PCB. This will allow us to also switch to the TLE2142 opamps.
The photo below will show you what traces to cut (in blue) on the component side of the PCB:
1. The connection of the unregulated supply to the emitter of Q3
2. The connection of the unregulated supply to R19
3. The unregulated supply connection to U3 pin 7
To install the new 10V zener diode D13, you need to remove some of the lacquer on the positive supply trace, as indicated on the photo.
The cathode of D13 is then soldered on this spot, and the anode goes to the emitter of Q3 and also to the disconnected end of R19.
See this photo for a closer look:
The original zener D7 is not installed but C14 will be mounted in this location.
The LM337 will be mounted in place of R3, and I just figured out a way to make the connections to the ADJ pin and R25 and R26 to connections that are near. Make sure the (metal) body of the 337 does not connect to anything, it carries a voltage. Use heat shrink tube if needed. With only about 10mA current, it will not get warm at all.
Turn to the reverse side of the PCB, and look at this photo:
The new C10 is mounted on the reverse side of the PCB.
R10 is mounted on the back to make it easier to connect to the negative supply.
The pin 7 of U3 is connected with a wire to the anode of D13.
The following values of components from the kit are now changed:
R10 (from 270K to 1K),
R17 (from 33R to 68R),
R22 (from 3K9 to 1K5),
RV1 (from 100K to 5K) and
U1, 2 and 3 (from the TL081 to the TLE2141)
Despite what others have posted, I had to connect the minus supply of U2 to the negative supply, not to ground. The reason was that I could not get the output to go to 0 Volt with P1. It did go to 0V with the current limiter. With a negative supply of only -1.2V, it still does not go to 0V, but +25mV is close enough. (RV1 at 5K and R10 of 1K allowed the output to be adjusted from +43mV to + 25mV)
It has been stated that R15 and D10 have no purpose, but if you connect U2 to the negative supply, R15 and D10 remove any negative voltage from the output of U2 to the base of Q2.
Finally, if you only use the supply to about 1A, you can use a 220K value for R18 and you do not need to add RV3. If you use a 24V AC transformer, you probably don’t need to limit the maximum output to a precise 30V, and if so, you don’t need to install RV3 and R11 stays at 27K.
So with these changes and a few more parts, the kit can be modified and the total price will still be very attractive.
Latest update. August 4 2015
I was still not very happy with the CC mode of operation. Even with the above mentioned modifications, there is still too much noise and a mains ripple on the output during the CC/CL mode.
As it turned out, a lot of this noise comes from the Volt/Amp display I'm using. The switching regulator that is used on this display injects a lot of noise back into the supply. I also was still not satisfied with the ripple on the reduced supply (by D10) for U3, U5 and Q3, and connecting the display to this supply made it all worse.
So, to tackle these problems, I went back to using the LM7824 that was part of the kit, and used that instead of D10, the 10V zener that was used to create the supply to U3, U5 and Q3.
To counter the noise injection from the display, I now used D10 to reduce the raw supply and used that to power the display unit.
While on my quest to reduce noise, I also moved the display current shunt from the output terminal, to outside of the current feedback loop. This reduced some more noise, but also made the current setting more precise. (because the shunt was inside the feedback loop, the voltage over the shunt at higher currents created an error. Small because the shunt seems to be only 25 mOhm, but still)
In order to put the shunt there, you need to cut a PCB trace from the raw ground supply to R7 and connect the current meter shunt output at the supply side of R7. Make sure R21 and R17 are not measuring the current shunt of the meter, but only R7. The current meter shunt input goes directly to the connections of the anodes of D3 and D4 and the negative connections of C1 and C2.
To eliminate a possible ground loop, the ground supply lead for the display is no longer used. The ground for the display unit is coming from the shunt connection to the raw supply ground.
To eliminate large currents on the PCB as much as possible, I connected the collectors of Q4 and Q3 directly to the point where the cathodes of D1 and D2, and the filter capacitors C1 and C2 come together.
I also installed the "optional" trimmers to set the maximum output voltage (RV2) and maximum output current (RV3). It is important to set the maximum current limit, because the granularity of P2, a normal pot meter, is greatly increased allowing you to set the current level more precise.
C16 is used to eliminate some more noise.
Because the LED's D14 and D15 are now connected to the 24V rails, their current limit resistors (R27 and R23) need to double in value.
Lastly, the output capacitor C7 was enlarged from 10uF to 470uF. That seems a lot, but professional supplies actually use a lot more.
Here is the final schematic with the latest revisions:
The rise time of the supply is now about 5mSec and the fall time is just over 2 mSec at maximum voltage and current, measured with a dynamic electronic load, capable of 50uSec transients.
With all these modifications, the output noise is now 18 mV p-p across the voltage and current spectrum, and, more importantly, stays at that level in the CC/CL mode. To qualify that, the noise floor of my scope with the probe tip grounded is 12 mV p-p, and with the supply switched off, the noise floor is just below 16mV p-p. With a positive mind, you could deduct that the output now only adds 2 mV p-p noise. Mission accomplished!
One future mod I'll do is to add a parallel output transistor to Q4. My typical applications are low voltage, and this is the largest burden for the pass transistor, because it has to bleed-off the excess voltage. I'll rearrange the LM7824 on the heat sink to make room for the second 2SD1047. I'll use .22R emitter resistors (because I have them already) to pair them up.
And yet another update: Aug 14
Not only did I indeed install a parallel series transistor (2SD1047), I also modified one of my two supplies such that it could handle more current.
I'll continue to use one which is fed by a 24V 1.5A transformer, but that maximum output is limited with a current in excess of about 25V, when the regulation starts to falter because the raw voltage starts to collapse.
So, I needed a transformer with a higher voltage rating and a higher current rating to pull this off. Unfortunatly, the most common transformers are 15-0-15 or 30V at 3A or more, and that will produce a raw voltage that is too high for the choosen op amps. The TLE2141 can handle up to 44V, but 30V AC already translates into 30 * 1,414 = 42V. Without a load, even with the bridge diode voltage drops, that is still too much. More so, since two op amps are also fed with a negative 1.3 V supply. A 14-0-14 supply would be ideal, but I could not find one.
With the higher currents, you also need a fan to cool things, so that was added as well. See a separate post on a solution that I built. At a later date I'll include that circuit into the main schematic.
The transformer I ended up buying is a 15-0-15V AC at 3,3A. With 3,3Aac, I should be able to get a solid 2Adc, plenty for my purposes. I also changed the 4 diodes that were used in the full bridge configuration and selected a bridge with 600V 10A that can be mounted on a cooling fin. A bit overkill, but it was for the same price as an 8A version. You need some overkill because of the in-rush currents to the main filter caps. The two 3300uF filter caps are inadequate for these currents, so I installed two 10,000uF at 63V ones. I used a separate enclosure to put this all in, and use 4mm banana posts and jacks to connect the raw supply to the PSU. If you do that, remember to also feed an AC signal to the PSU because that is used to create the negative 1.3V rail. The enclosure is completed with a main switch, a main fuse and a power indicator. I also feed the AC 15-0-15 taps to banana jacks on the front panel, so I can use that for other purposes.
While running some more tests, I decided to put the ampere meter shunt back at the output. There was too much of an error in the measurement, because it included the currents of the actual supply itself.
The changed schematic for the new supply is as follows:
You'll notice that I departed from using the original way of showing all connections with wires. I now grouped the functionality so it's hopefully easier to understand.
Because the op amps are limited by their 44V rail-2-rail supply, I went back to using an LM317 to create a nice and steady 33V. This is just enough headroom to regulate the output to 30V. I used this supply to feed all op amps now, and that also required resistor value changes for the LED bias resistors. It also means that the supply modification with D10 needed to be undone on the PCB.
You'll notice that the bridge rectifier diodes are gone, and so are the filter caps and the bleed resistor. They all moved to the raw supply enclosure. I actually doubled the value of the bleeding resistor by putting two 2K2 2W resistors in series, because I found it was getting too hot with the additional voltage. I also changed D13, the Zener diode feeding the V/A display, to a more beefy 1W version, that I only had in a 22V version. I paid special attention to getting the main raw connections ( they are now a bit thicker in the schematic) to the required parts, and avoided going through the PCB as much as possible. C7, the 10uF on the output terminals is an anomaly, I just left it on the PCB, but is has little use compared to C10 which is mounted directly on the output terminals.
Other than that, there were no major changes, and the supply works really, really well. I now only need to install the fan controller but I wanted to play with the fan starting point a little more so it's quiet with small loads but kick in when needed.
Update Aug-28-2015:
I finally was able to find a simple but effective method to "tie" my two supplies together and create a tracking +30 0 -30V supply, or a +60V supply.
The principle is easy, if you connect the 0V output of one supply to the +0..30V output of the second supply, you actually can create a +30 0 -30V supply, or a 0..60V supply. You need to adjust both voltage potmeters to set the values, but if you want to measure a circuit with a variable voltage, you need a tracking mechanism. This can also be called a master/slave combination.
The trick is to make the voltage setting of one supply depending on the setting of the other supply. I experimented with various ways, but finally settled on the following circuit.
Let me explain.
The slave supply must be modified as follows. The connection of the wiper of the voltage setting potmeter (P1) must be disconnected, and fed to a switch. The switch connects back to the old wiper connection as you can see in the schematic. The other side of the switch goes to a voltage divider that sits between the positive output of the master supply and a resistor combination connected to the 0V of the slave supply.
To connect the two supplies together, the 0V of the master gets connected with a lead to the + output of the slave, and this becomes the new 0V. The above schematic should make that clear. If you want a 0..60V supply, the + is the + of the master, and the 0V is the 0V output of the slave.
The modification for the master is even easier. You need to add one resistor (R40) to the + output, and feed the other side to a connector such that it can be fed to the slave. As you can see on one of the pictures of my supplies in the beginning of this post, I originally used a 3-pole DIN connector to feed the 24V AC to the PSU. I have now switched to banana jacks, and have used the DIN connectors to tie the two together.
The trimpot R41 needs to be set such that the voltage setting on the master is the same as the voltage output on the slave. The signal going to the switch will be close to the reference voltage of 11V2.
I found that the best tracking accuracy can be obtained if both supplies are set to 30V in the +/- mode as in the schematic. You can then flip the switch to the Tracking mode, and you adjust R41 until the slave also reads 30V. You will notice that the tracking is pretty accurate (about 1%) until you go below 4-5V, it then gets increasingly out of sync to a few 100 mV at 1V. This must be due to the linearity difference in the gain of both the U2 op amps. All the other methods I tried were to eliminate this, but I did not succeed. On the other hand, this accuracy is good enough for me.
I have also added R43 as a security measure, to make sure the slave supply will not have an (undefined) output if the link between the sense resistor in the master is not connected to the slave or when the switch is moved from one position to the next.
You should also know that you need to set both current limits independently for both supplies, but if the master goes into current limit or constant current mode, the slave will follow suit, regardless of it's setting.
I will be doing some measurements of both supplies in a few days, to show some of the specifications and results. Stay tuned.
Enjoy!