Simple 4-quadrant PSU with remote sense - please recommend topology/schematic

[Yansi]

Hi!

can you please recommend a topology or schematic or app. note how to build the simpliest possible 4-quadrant PSU?

Requsted properties are output voltage of up to about +-1V and +-5A with remote voltage sense on load. Being capable of controlling output voltage (and possibly current limit) and measuring real output U and I. (ADCs and DACs with common ground)

The setup should serve for PV cell characterisation on a local university. Non-comercial project, I have no profit from it. 've been just asked to help them.

I have been thinking the whole weekend how to do it and haven't thought out any usable circuit. :-/ I am stuck with no freaking ideas. Have come accross some interesting schematics, like that one Operational Power Supply from LM12 datasheet... but that does not use remote voltage sensing.

Can you help me please help the uni? :-)
Thank you,
Yan

PS: What the hell are doing those emitters of Q1 and Q2 in series with the caps?!

[samofab]

Google service manual for Keithley 228A, you'll find manual with schematics on xdevs. It's old, but quite simple.

[Marco]

Did you google yet? EDN has a pretty basic one which is a slight variation on a datasheet circuit to use LT1970 with external MOSFETs. Expensive IC, but who cares for a one off.

[timb]

Check the LT1970A datasheet out. There's a good example circuit on the last page for a +-5A version using two external FETs. Either use a cheap voltage reference and potentiometer to set the voltage/current limits or use a quad DAC + MCU.

I'm currently working on a supply based on this chip and it's coming along very nicely. (I'm also in the simulation stage of a +-100V/100mA discrete four quadrant PSU/amp.)

[amspire]

The LT1970A would be fine if you are using a resistive or inductive load. It will not be stable with a capacitive load larger then a few nanoFarads. If your load is a resistor, then it is a great chip.

[timb]

The LT1970A would be fine if you are using a resistive or inductive load. It will not be stable with a capacitive load larger then a few nanoFarads. If your load is a resistor, then it is a great chip.

Not true, at least in my case. It's been stable with large capacitive loads. They even talk about this in the datasheet.

[retrolefty]

Could someone give me/us a simple definition or example of a "four quadrant" PS unit? How does it differ from say a standard linear DC regulator.

[samofab]

"four quadrant" PS can source or sink current of either polarity. You can think of it as a battery that can be either charging or discharging (with a relay to reverse polarity). Quadrant 1 is for example positive voltage, positive current.

[timb]

Could someone give me/us a simple definition or example of a "four quadrant" PS unit? How does it differ from say a standard linear DC regulator.

As others have mentioned, it can source *and* sink current. When sinking current, it acts like an electronic load; when sourcing current, it acts like a power supply. It's also bipolar in either mode (meaning it can handle negative and positive voltage).

Here's a nice graph:



Most power supplies only work in Quadrant I (and statically in Quadrant III by swapping the + and - leads).

In short, they're basically big ass op-amps. Some times they're even sold as such, like the Kepco BOP series bipolar operational supply.

[Yansi]

Thank you for explanation, but I know quite well what those quadrants mean. 

Without remotesense, the supply circuit can be quite simple, as per the schematic from the LM12 datasheet I have posted in the #1 post.

With remote sensing, the "simply" is just gone. (When trying to avoid precision very very expensive ICs).

By the word "simple" I have meant more like "jellybean part like" solution, not a fancy unobtanium superexpensive IC with zero availability, like most LT stuff I have been working before.

So yes, I have tried to google something up, but found mostly nothing but bunch of that LT stuff which is not really needed here or many 4Q supply designs but without remote sensing.

Thank you, will look for that Keithley 228A manual, should be interesting.

EDIT: Found the Keithley schematic. Will be helpful, indeed! But am a little worried about precision of the diff amps (like the voltage sensing one when using jellybean parts like OP07...

EDIT2: Can someone please explain the Q1 and Q2 in the #1 post schematic with LM12 power opamp?

[timb]

@Yansi The explanation wasn't for you. It was for @remotelefty who asked what it meant.

Now, remote sensing is actually fairly easy to do for a four-quadrant supply.

Here's the simple way to do it with no extra parts:



Here's a slightly more complex way to do it using either a difference amplifier (such as the LT1990) or a standard op-amp and a quad resistor array:



In either method, you would want to place high value resistors between the force and sense terminals, to keep the supply from maxing out if the sense terminals weren't connected. You could also use a relay or analog switch (as shown in the second image) and have a MCU control it.

I don't know why you think the LT parts are unobtainium or expensive; they're not. (If they were Maxim, maybe!) They're available on e14, Digi-Key, Mouser, Farnell, etc.

The LT1970A combines an op-amp, power booster stage *and* an adjustable sink/source current limit all in a single package. For under $10. That's a really good price for what you get, as it reduces your BOM cost significantly and speeds design time. If you were building this into a product that you would be producing hundreds of thousands of, I'd say go discrete and save the money. But for a one-off? Use a monolithic solution.

You should go check out the prices of other power amplifiers on Digi-Key, $10 is about the starting price, moving up to $100-$200 (per chip) for the high voltage, high current stuff from Apex.

I'm not trying to get you to use the LT1970A; if another power amp fits the bill better for you, by all means use it. I'm just saying they're all fairly expensive. That said, once you've tried to boost a regular op-amp by bootstrapping FETs to it's power rails, or tried to build and compensate a discrete power stage, you'll wish you'd just spent the extra money on a power amp to start with. Been there, done that. It's not fun.

[Yansi]

Those two sensing examples are fine, but both depend heavily on the resistor precision and/or resistor trimming plus opamp offset. I did want to not use such. I don't like diff. amplifiers. Always the weakest point.

However, few experiments made clear, that OP07 with four common type resistors (one trimmed with a trimmer in series) will work good enough fort his job.

The actual schematic looks something like this and I'm in the process of breadboarding it. Due to several reasons, the -Force terminal will go straight to system GND and the sense shunt will be floating on the +Force terminal, sensed by a second diff. amp. A little change in my "no diffamp" rules, but seems it will work. At least good enough. They do not require extra precision, 1% should be enough for them.

Currently, I am more thinking about  the ouput current limiting. Can someone please explain the bizzare circuit in #1 post? Q1 and Q2 respectively, emitters to the caps. 

[quantumvolt]

...

The LT1970A combines an op-amp, power booster stage *and* an adjustable sink/source current limit all in a single package. For under $10. That's a really good price for what you get, as it reduces your BOM cost significantly and speeds design time. If you were building this into a product that you would be producing hundreds of thousands of, I'd say go discrete and save the money. But for a one-off? Use a monolithic solution.

You should go check out the prices of other power amplifiers on Digi-Key, $10 is about the starting price, moving up to $100-$200 (per chip) for the high voltage, high current stuff from Apex.

I'm not trying to get you to use the LT1970A; if another power amp fits the bill better for you, by all means use it. I'm just saying they're all fairly expensive. That said, once you've tried to boost a regular op-amp by bootstrapping FETs to it's power rails, or tried to build and compensate a discrete power stage, you'll wish you'd just spent the extra money on a power amp to start with. Been there, done that. It's not fun.

...

Agree. I have been working on a DC Power Analyzer since 2013. I have been prototyping old discrete HP / Tektronix / Fluke etc. stuff, general power amps and beefed-up op amp circuits. In the end I chose the LT1970.

OFF TOPIC: I have the 5 amp LT1970+FETs demo board http://www.linear.com/solutions/4288 . Due to age, health situation and obligations I have to strictly prioritize my projects. If anyone wants the board, I'll sell it for half of LT's price + shipping.

[Yansi]

Understand. This not a first PSU I build. I know how much trouble can be adding discrete voltage amplifying stages into OPamp FB loop. But this should be no trouble, as the required output voltage range is well in between the opamp supply rails, so a simple follower stage is needed.

I will use two power darlingtons and maybe two more trannies to reduce the opamp output current, and will try some linearization tricks to get rid of the hysteresis (base-to-base voltage of 2,4V or more) in the complementary pair to make it more stable.

[Alex Nikitin]

Can someone please explain the bizzare circuit in #1 post? Q1 and Q2 respectively, emitters to the caps. 

Nothing bizarre abouit it, Q1 and Q2 work as analogue switches, connecting a compensation RC network when either of the two current limit opamps is engaged.

Cheers

Alex

[Yansi]

Ah!! That is an interesting idea! 

I already know the idea of the "speed up" diodes D1 and D2, that does not allow to go the opamp to openloop and saturate, but the FB switching seems neat to also speed up the current limiter engaging. Veery clever.

[timb]

Those two sensing examples are fine, but both depend heavily on the resistor precision and/or resistor trimming plus opamp offset. I did want to not use such. I don't like diff. amplifiers. Always the weakest point.

Actually, it doesn't depend on resistor precision at all. It depends on resistor ratio. A quad resistor network with good matching isn't really that expensive, but if you're really cheap and it's for a one-off project you can simply hand select the four resistors and mount them in thermal contact.

Besides, the second example doesn't depend on external resistors at all, because you'd use a difference amp with built-in precision resistors. AD, TI and LT all make them. They're not expensive.

I'm not sure how difference amps are a weak point? To me, it would be a strong point, as they can often withstand very high voltages (the LT1990 has. +-250V common mode range, there's an AD part with similar specs) which is something you want in a 4-quadrant power supply. You never know what sort of voltage might accidentally get fed to it. Plus you don't have to mess around with trims for resistor ratio, etc. Offset can be compensated for easily by tying the output of a pot to the ref pin, though it's not really needed as most of these parts have under 1mV of offset.

[Yansi]

I think precision is the right word, as you need 2 resistor dividers with as was said same ratios. Thats achievable only by using precision resistors with tight tolerances, or by trimming the damn thing for match. And even then the TCO of the resistor will play a role, when one needs some good CMRR of the whole diff amp.

Yeah, sure, I can use precision opamp, IC, this, something... and buy an off the shelf 4Q PSU unit for $10 000. However this is not the case and the most "precision component" I will use here is OP07. As I've said, I will get no $ from this job and their required overall accuracy of the PSU is not too tight. 1% accuracy for PV cell measurement is good enough they've said. That has to be done using jellybean parts, hasn't it?

Why I call diff amps the weakest point? Because very rarely I use precision diff amp ICs, always build them with standard OPamps and using common 1% resistor types. The need for the trimming, the temperature drifts of offset voltages or resistor TCOs always make it hard to make the circuit up to reasonable spec, using such jellybean parts.

Lol, +-250V is not really what I need here. The requsted output voltage range was +-1 or +-2V at max. Sure, I understand the design should be as robust as possible, but I think withstand voltage of one tenth of those 250V will be more than enough for this experimental application.

In some short time I will report some results, as I've already started breadboarding some circuitry, and it seems promissing, so far...


With regards to the LM12 circuit and those two Q1 Q2 transistors: How will the circuitt behavior change when interchanging C and E in the circuit? Or why is the collector chosen to be connected to the opamp output?

[Marco]

Inamp/diff-amp doesn't feel like the natural topology for remote sensing a voltage, the exact matching of input impedance isn't all that relevant.

Is there perhaps a variation on high side current sensing topologies possible?

[Yansi]

Meh, i don't like it either. But I need it anyway, as I need to bring the exact voltage accross the load to an ADC.

[Siwastaja]

Inamp/diff-amp doesn't feel like the natural topology for remote sensing a voltage

So, what would you suggest?

[Marco]

So, what would you suggest?

As I said, something similar to current sensing, so differential voltage -> current -> voltage  ... something like INA170 for instance. Considering it further though all you would really accomplish is make common mode rejection independent from the resistors drifting, offset and amplification would still be drifting along. The only way to get rid of that is with chopping techniques I think which seems a bit too complex here.

PS. I'm not sure I actually understood the original problem BTW. Does the power supply actually have to regulate using the remote voltage? If it's just about sensing it use a differential ADC ...

[mazurov]

Take a look at the LT3091 datasheet, there is an example of voltage-controlled 4Q PSU made out of it plus LT3081. Methods of increasing output current/improving dropout at low voltages are spread across LT308x-LT309x datasheets.

[timb]

So, what would you suggest?

As I said, something similar to current sensing, so differential voltage -> current -> voltage  ... something like INA170 for instance. Considering it further though all you would really accomplish is make common mode rejection independent from the resistors drifting, offset and amplification would still be drifting along. The only way to get rid of that is with chopping techniques I think which seems a bit too complex here.

PS. I'm not sure I actually understood the original problem BTW. Does the power supply actually have to regulate using the remote voltage? If it's just about sensing it use a differential ADC ...

Yes. Of course it has to regulate using the remote voltage! That's the entire point of remote sensing (aka Kelvin Sensing, Four Wire Sensing) is to regulate the voltage at point of load, instead of at the power supply output. This is to make up for I-R drops in the output leads at high currents, or for precision applications at low currents.

A *difference* amplifier absolutely is the correct topology for this, as you obviously need to measure the difference between Sense + and -.  I'm not sure what you'd gain by converting the voltage to current and back to voltage in this case?

Anyway, this is how transistor power supplies have done it for the last 50 years. You *could* do it digitally, by sensing the remote voltage with a differential ADC, but that would only work if you're already controlling the output amplifier with a DAC. (So you'd have a local feedback loop with the output amp and a separate software remote sensing feedback loop.)


There are other alternatives; LT and TI both make parts that do voltage drop compensation. Some of them are "dumb" and simply increase output voltage as the current increase, based on a fixed parameter for the wire length. Others are smart and actually measure the wire to calculate the drop through some sort of voodoo magic.

[Siwastaja]

As I said, something similar to current sensing, so differential voltage -> current -> voltage  ... something like INA170 for instance.

Don't these current shunt monitor IC's rely on laser-trimmed matched resistors, too, or am I completely wrong here? It's only neatly integrated. So we are back to integrated difference amplifiers with integrated precision resistors, or using external precision resistors.

I can't really see the problem OP is having. Precision resistors and precision opamps are the most basic building blocks, cheap and easily available, rather easy to design in. Maybe you can be very clever and MacGyverize a precision instrument from really jellybean parts, but please share the details if you are able to do it!