how to design fast bench supply with CC and CV?

[Boscoe]

I recently made a PSU that is almost exactly like yours. 0.25mA accuracy is very tricky, I’ve got about +/-1mA.

https://github.com/geowal19/PSU-Hardware

The finished item can be seen here:

gswdh.co.uk/gallery.html

[exe]

I implemented current sinking.  So, now it can source and sink. Here are some changes I've made.

1) I chose OPA192 as CC/sinking opamp because it has good precision, low input bias current and not so slow. It also has good phase margin (I think so, because it can drive capacitive load, and because I did simulation).

2) I chose LT1468 for CV because it is fast and it has LTSpice model. It is supposed to be low-noise too (it its class, according to Design Note 355).

3) Output is push-pull now, no LT1010 anymore. I hope this will let me go all the way to +-1A output. Practically speaking, even +-200mA would be enough for me.

Phew, what a ride. It took me a lot of time to make it working. Still, I need to verify that it is stable. I tested it under different conditions (e.g., varying load currents, voltages, etc) in the simulator, but I don't have confidence in my skills. How much do you think it would cost if I hire a consultant to design/verify compensation in this circuit?

[exe]

I realized using instrumentation amplifier to set voltage is not a good idea because of high common mode voltage. Websimulator [1] shows that only AD8825 and very few others inamps can do what I need. Probably, AD8476 has the same problem, but it has low input impedance (10k common-mode input). I need to think about it.

If I want switchable shunts, I also need to make a bi-directional switch. Previously, for sourcing-only, I used a few single mosfets. Now I problably need a back-to-back pair of mosfets or something. This thread should help: https://www.eevblog.com/forum/testgear/designing-a-simpler-current-ranger/

[1] https://tools.analog.com/en/diamond/?doc=AN-1401pdf#difL=0&difR=2.5&difSl=0&gain=5&l=0&pr=AD8225&r=12&sl=0&tab=3&ty=2&vn=-3&vp=16&vr=0

[exe]

One thing I forgot is to include diode in feedback loop. It helped improving phase margin.

I found that with a big shunt resistor and large output capacitance (10n+) it starts oscillating. Solution is to increase feedback resistor. I think I'll make it switcheable because I have a bunch of bidirectional mosfet optocouplers laying around. Those were not cheap (3euro/pc), so I better use them .

Got some accuracy calculations. I was secretly hoping to achieve 6.5 digit accuracy, but that's turned out to be much more challenging than I expected. Well, that's actually a good news for me, I can relax use some cheaper and more common parts. I hope to get some 4.5digit precision (whatever that means). For low currents the limit is bias current of opamps, and leakage of shunt switches. If I do it right, there will be some tens of nA parasitic currents. For voltage sensing the limit is accuracy of in-amp (AD8422). And, of course, reference itself (I have ltc6655 and max6226, not sure yet which to use) plus INL of ADC. I think I'm over-engineering this part and I should read less voltnutery.

I'm currently busy with current switching. Turns out to be not as easy as I hoped if sub-uA accuracy is required. There are two problems: 1) gate-source leakage of mosfets 2) gate leakage as the current flows through the shunt (can be a problem at elevated temperatures for fets with integrated gate zener). I could drive fet with photovoltaic drivers, but those are expensive and I'm not sure about life span (but probably that's not a problem because I found that I don't need to drive optoisolators at full 10 or 20mA current, they often work with much lower current which should increase reliability).

[exe]

Good evening, my friends!

Will this circuit work? I want to be able to switch the shunt off. I don't understand the DC path for gates. Mosfets are in series, so when M1 is not conducting then M2 can't be discharged. In simulating it sorta works, but not sure if it's going to work in reality. What do you think?

I think it may only work reliably when there is output load connected, but I don't think it will reliably work without load.

[Zero999]

Good evening, my friends!

Will this circuit work? I want to be able to switch the shunt off. I don't understand the DC path for gates. Mosfets are in series, so when M1 is not conducting then M2 can't be discharged. In simulating it sorta works, but not sure if it's going to work in reality. What do you think?

I think it may only work reliably when there is output load connected, but I don't think it will reliably work without load.

It's not clear what you're trying to do. Why is there a sine wave going to the MOSFET gate? A square wave is required to simulate the switching action.

Presumably you want to switch the current sense resistor? Use a relay. Switching a MOSFET is much more complex and the relatively high resistance is an issue. A DPST relay can be used to switch between two different resistor values.

Have you built any prototypes yet?

[exe]

It's not clear what you're trying to do. Why is there a sine wave going to the MOSFET gate? A square wave is required to simulate the switching action.

It's just for testing.

Presumably you want to switch the current sense resistor? Use a relay. Switching a MOSFET is much more complex and the relatively high resistance is an issue. A DPST relay can be used to switch between two different resistor values.

Relays are quite bulky, I wanted to save some board. Also wanted three current ranges. Although, relays don't leak as much current, and don't have drain-source capacitance.

Have you built any prototypes yet?

Nope, working on first pcb. Finalizing analog part. Currently stuck at range switching. I also don't expect first board to work 100% correctly.

[magic]

Will this circuit work? I want to be able to switch the shunt off. I don't understand the DC path for gates. Mosfets are in series, so when M1 is not conducting then M2 can't be discharged. In simulating it sorta works, but not sure if it's going to work in reality. What do you think?

I think so.

When the gate is high, body diode of M1 clamps the joint sources and ensures turn-on.
When the gate is low, current / leakage through both FETs charges the joint sources up and ensures turn-off.

I think it may only work reliably when there is output load connected, but I don't think it will reliably work without load.

Nevermind, that's a good point. Without load, you will not be able to apply low voltage to the gate. It will kinda work as long as M1 conducts, but once it's off V3 will simply charge M1 drain to a higher voltage using the gate as reference because Cds < Cgs. If M2 has higher threshold voltage, it will still be on. Leakage will slowly resolve this problem.

Connect V3 to the supply rail if possible.

[Zero999]

It's not clear what you're trying to do. Why is there a sine wave going to the MOSFET gate? A square wave is required to simulate the switching action.

It's just for testing.

Presumably you want to switch the current sense resistor? Use a relay. Switching a MOSFET is much more complex and the relatively high resistance is an issue. A DPST relay can be used to switch between two different resistor values.

Relays are quite bulky, I wanted to save some board. Also wanted three current ranges. Although, relays don't leak as much current, and don't have drain-source capacitance.

You could use two switches, which bypass the highest value resistor, with lower values.

Regarding your question about the gates: MOSFETs work in both directions, but there's a diode in reverse parallel with the drain and source, so both MOSFETs in your circuit will turn on. Why are you using two, rather than just one? Two MOSFETs are required when it has to block current in both directions, but your circuit already has a pass transistor which can turn off, to block current.

Another problem with MOSFETs is the gate voltage needs to be higher than the output, by enough to turn it on and it must never exceed the output, by the more than the maximum gate-source rating.

Have you built any prototypes yet?

Nope, working on first pcb. Finalizing analog part. Currently stuck at range switching. I also don't expect first board to work 100% correctly.

I wouldn't recommend getting a PCB made, without at least building a rough prototype on strip-board first.

EDIT:
So I take it the latest design and both sink and source, hence the idea of using two transistors, but I take it the current limit isn't designed to work on the sink? If so, there's no point in having the reverse blocking transistors.
https://www.eevblog.com/forum/projects/how-to-design-fast-bench-supply-with-cc-and-cv/msg3438060/#msg3438060

[MegaVolt]

Will this circuit work? I want to be able to switch the shunt off. I don't understand the DC path for gates. Mosfets are in series, so when M1 is not conducting then M2 can't be discharged. In simulating it sorta works, but not sure if it's going to work in reality. What do you think?

A parasitic diode is always present in field effect transistors. As soon as you add it, you will understand how the circuit works and why there are exactly 2 transistors, although it would seem that one is enough.

[exe]

Thanks everyone for your comments and suggestions!

So I take it the latest design and both sink and source, hence the idea of using two transistors, but I take it the current limit isn't designed to work on the sink? If so, there's no point in having the reverse blocking transistors.
https://www.eevblog.com/forum/projects/how-to-design-fast-bench-supply-with-cc-and-cv/msg3438060/#msg3438060

I'll answer other questions in the evening. Concerning current sinking, the bottom opamp from the message you linked provides constant current sinking. That's why I need a bidirectional switch. It wasn't part of original specification to sink current, but it seems it's "easy" to add it.

Why I wanted a bi-directional switch: it's be able to turn-off output completely. Otherwise it will sink current.

A bit more thoughts on relay vs back-to-back fets. There is one or two aspects that I don't like about fets, those are leakage current, and drain-source capacitance (100-1000pF). Those may compromise performance when sourcing small currents. I'll do more simulation to see if this is a problem or not.

[David Hess]

A parasitic diode is always present in field effect transistors. As soon as you add it, you will understand how the circuit works and why there are exactly 2 transistors, although it would seem that one is enough.

The parasitic body diode is always present however it does not have to be connected to the source.  There are 4-wire power MOSFETs available like the Vishay Siliconix Si3831DV and Micrel (now Microchip) MIC94030/94031.

[exe]

I added a 1MEG resistor for return path. This should solve the issue. The downside is, it steals some shunt current when switch is on.

Optocouplers are an interesting solution, but expensive and I have a reliability concern. So, may be I'll one optocoupler just to drive the lowest range where this 1M resistor will introduce a noticeable error. Although, for me it's hard to say if this makes much sense because there are three or four opamps connected to the output and their input bias introduces some error too.

Another issue is leakage currents. I'm not sure how measure it, tried to use uCurrent Gold, but numbers were jumping around quite a lot in mV range. I tried to connect my an8008 dmm directly as shunt in mV voltage (hoping it has ~10M input impedance), but got confusing reading. Also, turned out uA range on that DMM doesn't work. So, I used ut61e in uA range.

I connected  TSM210N02CX fet, applied 20V across drain and source with zero gate voltage, ut61e show 0.00uA current. Then I heated it up to some 70-80C (crudely heating with a 100C heatgun) , I got leakage of some 0.10uA or so. So, I guess, it's good-enough for me, I just need to keep them below, say, 50C. I didn't measure the gate leakage, but I have so many fets around, that if it becomes a problem I'll install another one, preferably without protective zener on the gate.


PS forgot to ask, does reverse-biasing of the fet's gate helps with drain-source leakage?

[exe]

One of very big challenges for me was to choose ADC. I hoped that I can just pick a cheap sigma-delta ADC and be done with it. It turned out to be a much bigger challenge than I thought. Long story short, I ended up with ads1220 because it seems to have very low drifts, very good PGA and reasonable INL (which I hopefully can compensate with multi-point calibration). It has only one sampling channel (with multiplexing), but that, I hope, it will be fine for my application.

Other solutions I considered are ad7175 (overkill, no PGA), LTC2492 (low sampling rate, otherwise seems to be very good), ads131a04 (drift specifications are not great). Apart from that, I read tons of other datasheet, but only the parts I mentioned above made into short-list.

Why I wanted fast sampling: I want to see if I can do some sort of power profiling. The downside is, without switching shunts it will have a limited dynamic range. But may be I'll be able to join two channels, one set to low current range, another one is for high current range. But that's just a vague idea that probably won't work. Still will be interesting to test this.

Now I'm finalizing schematic, then I'll make a board. I want to make the whole board for tests, not just analog part. The reason is, the first revision will probably have a few mistakes, so I want verify that digital part is also done right (probably not).

[prasimix]

For power profiling simultaneous sampling is welcome. Therefore something like mentioned ADS131A04 or ADS131E04 should be a nice candidates.

[exe]

For power profiling simultaneous sampling is welcome. Therefore something like mentioned ADS131A04 or ADS131E04 should be a nice candidates.

I have one idea about it. I assume that we do power profiling in CV mode. Therefore, there is no need to sample voltage. At least not as often as current. So, I may try to sample only current in power profiler mode, and sample voltage only if power supply is in current-limiting mode. Although, still not sure what is required sampling frequency for that. It seems 10Ksmps is fine, but ads1220 gives on 2K sample rate.

Anyway, I've seen a few circuits, they all imply fast switching shunts with comparators, which I don't have, and not sure how fast I can do switching with mcu. Also, on nordic PPK2 shunts are also bypassed with 10n capacitors (presumably to allow for surge currents) which I'm not sure how would affect performance of my power supply. If I have board space I might make it switcheable.

[prasimix]

Commercial solutions starts with 1 KSPS (Qoitech) and go up to 500 KSPS (R&S). I'm gonna try 32 KSPS simultaneously on 4 channels (for efficiency profiling) or 64 KSPS on 2 chanels (for power profiling).

[exe]

I just wanted to check-in that the project is still alive. I'm struggling with routing the pcb. But it seems I'm getting there.

[Zero999]

I just wanted to check-in that the project is still alive. I'm struggling with routing the pcb. But it seems I'm getting there.

Have you built a prototype yet? If not, I strongly recommend doing so, before making a PCB.

[exe]

I just wanted to check-in that the project is still alive. I'm struggling with routing the pcb. But it seems I'm getting there.

Have you built a prototype yet? If not, I strongly recommend doing so, before making a PCB.

Thanks, Zero999, I appreciate your help. My reasons to start with pcb are:

1. There are hundreds of parts on the boards, a lot of digital circuitry. I need to prototype that too. Like, I never worked with many components that I'm going to use (such as cmos switchs, crystal oscillators, usb, it's my first multi-layer board, my first project with separate grounds, my first precision circuit), so I need to prototype that too.
2. Analog part looks fine to me. I left solder jumper where I expect problems or I have doubts about configuration.
3. The main opamp has 90MHz bandwidth (LT1468), I'm not sure if I can put it on breadboard
4. Routing is critical for this project. So, it's not just about getting schematic right, but also route it right.
5. Manual soldering is prone to errors for me. So, even for prototypes I still use EDA to figure out how to route wiring to reduce errors.

Considering the cost of a four-layer board ($18, which is nothing comparing how much time I spend on the project), I wanted to start with pcb. I'll try to make it working, then, if there are critical mistakes, I do another spin. How does it sound? Do you still suggest building analog part separately?

[Zero999]

I just wanted to check-in that the project is still alive. I'm struggling with routing the pcb. But it seems I'm getting there.

Have you built a prototype yet? If not, I strongly recommend doing so, before making a PCB.

Thanks, Zero999, I appreciate your help. My reasons to start with pcb are:

1. There are hundreds of parts on the boards, a lot of digital circuitry. I need to prototype that too. Like, I never worked with many components that I'm going to use (such as cmos switchs, crystal oscillators, usb, it's my first multi-layer board, my first project with separate grounds, my first precision circuit), so I need to prototype that too.
2. Analog part looks fine to me. I left solder jumper where I expect problems or I have doubts about configuration.
3. The main opamp has 90MHz bandwidth (LT1468), I'm not sure if I can put it on breadboard
4. Routing is critical for this project. So, it's not just about getting schematic right, but also route it right.
5. Manual soldering is prone to errors for me. So, even for prototypes I still use EDA to figure out how to route wiring to reduce errors.

Considering the cost of a four-layer board ($18, which is nothing comparing how much time I spend on the project), I wanted to start with pcb. I'll try to make it working, then, if there are critical mistakes, I do another spin. How does it sound? Do you still suggest building analog part separately?

Why are you planning to use such a fast op-amp? It's more likely to oscillate, if you don't slow it down, because the pass transistor(s) form an additional pole.

I recommend just prototyping the analogue part. Use strip board, with the tracks cut as short as possible, for the op-amp. There are several options for prototyping SMT ICs, on strip board, which depend on the pitch of the pins. The part can be glued to the board upside down, with the connections made using enamel wire. SMT to DIL breakout boards are very easy to use. If you're stuck and the pitch is 0.05", the copper strips on ordinary strip board can be cut in half, so they match the pins.

[Neomys Sapiens]

I just wanted to check-in that the project is still alive. I'm struggling with routing the pcb. But it seems I'm getting there.

Have you built a prototype yet? If not, I strongly recommend doing so, before making a PCB.

Thanks, Zero999, I appreciate your help. My reasons to start with pcb are:

1. There are hundreds of parts on the boards, a lot of digital circuitry. I need to prototype that too. Like, I never worked with many components that I'm going to use (such as cmos switchs, crystal oscillators, usb, it's my first multi-layer board, my first project with separate grounds, my first precision circuit), so I need to prototype that too.
2. Analog part looks fine to me. I left solder jumper where I expect problems or I have doubts about configuration.
3. The main opamp has 90MHz bandwidth (LT1468), I'm not sure if I can put it on breadboard
4. Routing is critical for this project. So, it's not just about getting schematic right, but also route it right.
5. Manual soldering is prone to errors for me. So, even for prototypes I still use EDA to figure out how to route wiring to reduce errors.

Considering the cost of a four-layer board ($18, which is nothing comparing how much time I spend on the project), I wanted to start with pcb. I'll try to make it working, then, if there are critical mistakes, I do another spin. How does it sound? Do you still suggest building analog part separately?

In my experience, the LT1468 is as care-free as it is performant. I never experienced instabilities with this OPA.
Otherwise, I'm with Zero999 regarding construction. also, this is a good partitioning for separation of modules. It would give you testable interfaces on both sides.

[Zero999]

In my experience, the LT1468 is as care-free as it is performant. I never experienced instabilities with this OPA.

Was that just on its own, with nothing but the usual resistors in the feedback loop? My concern isn't the op-amp in isolation, but according to the other schematics posted in this thread, it will have a pass transistor on the output, inside the feedback loop. The extra delay due to a transistor, especially a big slow Darlington pair, can cause oscillation, with a fast op-amp. It's possible to avoid this by slowing the op-amp down, with a capacitor between the inverting input and output, but why not just use a slower op-amp in the first place?

[Neomys Sapiens]

In my experience, the LT1468 is as care-free as it is performant. I never experienced instabilities with this OPA.

Was that just on its own, with nothing but the usual resistors in the feedback loop? My concern isn't the op-amp in isolation, but according to the other schematics posted in this thread, it will have a pass transistor on the output, inside the feedback loop. The extra delay due to a transistor, especially a big slow Darlington pair, can cause oscillation, with a fast op-amp. It's possible to avoid this by slowing the op-amp down, with a capacitor between the inverting input and output, but why not just use a slower op-amp in the first place?

I used this component in multiple applications which included power control and dynamic ADC/DAC circuits.
I concede that complex feedback loops can lead to instabilities, BUT:
- these might show also with an OPA which is slower by a power of ten
- I found that I could tame instable loops by replacing OPAs, which are in the same speed group by this one.
And many instabilities are triggered by OPAs with a not-so-solid settling behaviour.
That said, it was the selection of the builder.

[exe]

In my experience, the LT1468 is as care-free as it is performant. I never experienced instabilities with this OPA.

Was that just on its own, with nothing but the usual resistors in the feedback loop? My concern isn't the op-amp in isolation, but according to the other schematics posted in this thread, it will have a pass transistor on the output, inside the feedback loop. The extra delay due to a transistor, especially a big slow Darlington pair, can cause oscillation, with a fast op-amp. It's possible to avoid this by slowing the op-amp down, with a capacitor between the inverting input and output, but why not just use a slower op-amp in the first place?

As of why the CV opamp needs to be fast, it needs not to introduce much phase lag as it is driven by CC opamps. So, CV opamps needs to be much fast than CC opamps.

Stability is checked in ltspice. In fact, it was the biggest challenge for me. Output stage is a plain push-pull stage with just two transistors, no darlington. This is why I also have to limit maximum output current. I hope I can get 1A, but we'll see. I understand with higher frequencies bjt gain drops. Assuming hfe is 100, getting 1A output requires 10mA drive current from the opamp. Is this too much for lt1468? Typical output current claimed to be 22mA, but, as I understand, higher output current slows opamp down.

I'll make the circuit on a perfboard and report results here.