Overengineered Lab PSU progress

[Berni]

Daves recent PSU projects remided me of my unfinished one and inspired me to finish it.Some might have seen the first attempt at this that worked but was unstable at times and i had a few complaints about it, also i got my hands on a bunch of little switchmode modules that can make a great preregulator and get rid of the heavy transformers. I was aiming for much lower specs but as design went on it ended up confirming to higher specs than first anticipated. What i wanted to do is basically 20V at 2A..



Final specs:
-Output voltage 0 to 40V (Down to 20mV without or with load)
-Output current 0 to 4A (Tested at 5A fine)
-Adjustable and non oscillating current limit
-Works as a electronic load too(Adjustable voltage and current)
-Low noise (Under the noise floor of my digital scope, multimeter says 1mVac)
-Starts up with output at 0V
-Fully digitally controlled (12bit DAC, 16bit ADC)
-Optoisolated serial interface

Final design:
Plan is to put 4 of these power supply in to a single case while having them communicate with the front panel over a optoisolated UART bus. Plan on the front panel is to have a large LCD along with rotary encoders and a USB to UART interface for the PC just because its easy to do. The interrupt line on the optoisolated bus is there to optionally shut down all the supply rails if one of them reaches the current limit. But for testing i just wrote a VB program to debug my PSU for now.



Circuit explanation:
The principle of operation is like any other lab PSU, but because everything is controlled by voltages the circuit becomes a bit more complicated. First opamp is just a basic differential amp that reads the voltage off the current shunt resistor, output of this is a nice voltage than can be fed to a ADC. Then it compares that to a max current signal from the DAC and that then pulls down the desired output voltage that also comes from the DAC. Bottom part of the analog circuitry is exactly the same except it works for sinking current.
The weird part is where the negative supply comes from, i tried to design it without one but it was too hard. Instead i came up with a cheep way of getting a negative supply. Between the negative end of the power source and ground i put a silicon diode that creates a 0.6V drop across it. This means that the negative side of the power source is now at -0.6V and we can use that as our negative opamp supply(But creates extra dropout voltage).

Power Source:
First plan was to use a transformer, but i got my hands on some nice tiny switchmode modules that convert 300V DC in to isolated 24V 6A or 48V 3A while being smaller than a pack of cigarettes. They also have a very nice feature of the output voltage being trimmable from 110% to 10% by the means of adding a external resistor. At first i wanted to use a digital pot to do that but the negative terminal is at -0.6V because of my negative rail cheat for the opamps, because of this i made my own variable resistor from cheep SMD mosfets and resistors that together create a 5bit variable resistor. This way the MCU can adjust the voltage on the power source so the linear regulator has the least amount of work
Datasheet for the modules:http://cdn.vicorpower.com/documents/datasheets/ds_375vin-micro-family.pdf

Problems with the design:
As always i had to bodge a few things to get it working properly. First issue of them all was that i put in some weird dual opamps from microchip that had a CS pin and that made half the thing not work while i was scratching my head.Then as with every analog design is changing out various resistor and cap values to get it tuned just right. Turned out i messed up the high side current shunt because i was doing 10x gain in there the opamp inputs got too close to the power rail and made it crap it self. As for the optoisolators those are a complete failure, first of all they ware SMD instead of trough hole like my board, but i got them to fit on to it. Then they started acting strange and that was due to the wrong footprint pinout again leaving the phototransistor side of the optoisolator backwards(Still sort of worked at times tho.) And at the end of it all i remembered UART has a idle state of 1 so the optoisolators should have been on the low side not high.(What sort of crazy shit was i on when i designed that part  )

So do you guys think this sort of PSU is a good idea or a complete waste of free time and components?

Will also post some test results soon along with scope screencaps.

[free_electron]

layout needs major work ...
This is unstability waiting to happen.
Didnlt you have problems with drill holes being too small ? like for the TIP142s ?

design problems :
- diodes in parallel do not evenly spread current ....

- The drive stage of the tip142 is bad. If you disable the output by turning on Q4 then output of opamp U1A goes to ground .. the base is never pulled to ground since the diodes block that.

[Berni]

The drill holes for it are a perfect fit, they look small because the pad around it is so huge. As for instability problems i tested it with a step response at various loads and settings, it didn't seam to oscillate with any of them. The most unstable was the current limit for the sink half, but i got it sorted with cap values and a pull up (Need to do more testing to make sure i solved that one).

For the diodes thing i agree its not the best solution, its mostly to make sure the negative rail doesn't develop too much voltage if something goes wrong. With beefy schotkeys it should hold the full current on 1 diode.

The TIP142 and its diode on the base is intensional in order to solve a design problem. When the supply is sinking current the output of this opamp goes to 0V while the emitter of the transistor might still be at say 30V, this creates a large negative voltage on the base that the transistor might not enjoy very much. Also as in the symbol the Darlington transistors has internal resistors between base and emitter to make sure it turns off.

[oPossum]

There is a better way to do high side current sense as explained by Bob Pease. That can be done with discrete components as explained in the article, or there are many little three pin chips that will do it such as the ZXCT1009.

[Berni]

Oh and free_electron, what do you not like about the layout?

I think Bobs current sense is quite a neat idea, it requires no trimming of the resistors, alto it requires a rail to rail input opamp(Hard to get cheep at these sort of voltages) and the opamp i have needs to be 1.8V away from Vdd. So i came up with a little two transistor cirucit that keeps the opamp up there with a 5V supply as the input varies from 6 to 50V or even more. This means the opamp can be a cheep 5V one. The regulation is not very precise as the voltage drifts from 4.7V to 5.05V as you go from 6V to 50V, but for giving the opamp some power its more then good enugh. The whole circuit draws about 1mA across the whole range so it shouldn't make anything noticeably warm.

So is this a good way to do it?

[nctnico]

The best stability test for a PSU is using a current sink for a load. IOW: an NPN transistor with the emitter to ground.

[codeboy2k]

I think Bobs current sense is quite a neat idea, it requires no trimming of the resistors, also it requires a rail to rail input opamp(Hard to get cheep at these sort of voltages) and the opamp i have needs to be 1.8V away from Vdd. So i came up with a little two transistor cirucit that keeps the opamp up there with a 5V supply as the input varies from 6 to 50V or even more. This means the opamp can be a cheep 5V one. The regulation is not very precise as the voltage drifts from 4.7V to 5.05V as you go from 6V to 50V, but for giving the opamp some power its more then good enugh. The whole circuit draws about 1mA across the whole range so it shouldn't make anything noticeably warm.

The beauty of an op-amp is that it doesn't care (too much!) about it's power supply voltage.  The output will be a stable gain up of the input. That power supply may affect the output somewhat, but you can mitigate that by choosing an op-amp with the best PSRR you can find.

Also see if you can find a small small 5V regulator to power the opamp instead of a zener and the transistors. One that can take 60v at the input.

And, isn't there something wrong on your drawing?  You've got the zener (vdd) on the high side of the 50V supply    your 50V can't go higher than the zener.

[Berni]

I was wondering how in the hell they are able to get a 150W+ SMPS in such a tiny form factor. Then I started reading. The module doesn’t include any line filter, rectifier and bulk capacitance or inrush limiters,TVS you have to buy that separately and its larger then the SMPS module and still doesn’t include a bulk input capacitor. So that’s how they do it they omit all the large heavy stuff. The supply isn’t constant power as well not uncommon but usually they a least provide constant power to 0.5 x Voutmax than current limit below that.

Still though not to bad saves you the most complex part that being the SMPS variable pre-regulator.

I would be interested to see the topology and the magnetic’s as well as the switching frequency. Judging by the size and the mediocore 72% efficency they must be up around 250kHz.

Yes these modules are missing the bit that connects it to the mains but i have a few 300V output 600W PFC switchmodes that should be in working order. And no they are not constant power at all, the 150W is at the rated output so a 24V one gives 6.25A max (Well they have a fair bit of headroom but you wont get like 10V 10A)

The best stability test for a PSU is using a current sink for a load. IOW: an NPN transistor with the emitter to ground.

What do you mean by that? What i used for testing is a little mosfet switch circuit i built. It switches GND between the two terminals when the switch is pushed. Problem was that using the switch directly caused too much contact bounce and created a lot of switching noise too. In all the transient tests i used this circuit to switch the load on and off cleanly.

The beauty of an op-amp is that it doesn't care (too much!) about it's power supply voltage.  The output will be a stable gain up of the input. That power supply may affect the output somewhat, but you can mitigate that by choosing an op-amp with the best PSRR you can find.

Also see if you can find a small small 5V regulator to power the opamp instead of a zener and the transistors. One that can take 60v at the input.

And, isn't there something wrong on your drawing?  You've got the zener (vdd) on the high side of the 50V supply    your 50V can't go higher than the zener.

Yes the diode is supposed to point that way. I need to keep the op amps 5V rails hanging in to the high input voltage. So with a 20V input the opamps ground is at 15V to create 5V across it. Not long after i realized i could have simply used a negative voltage regulator if i imagine the current shunt is 0V and the ground down where the output is is at -20V (From that perspective that is, in reality ground is 0V ofrcuse)

Scope screencaps of the tests are coming up real soon btw

[nctnico]

In theory a current source has an infinite impedance. The collector of a transistor behaves like a current source (or more accurate: a sink). The current will remain constant while the voltage is free to change. So if the power supply is prone to oscillating loading it with a current sink will most likely reveal that.

[Berni]

Ah yeah might try testing with a NPN transistor as a load then.

So here are the scope captures of the test i done on this PSU. The resistor and cap values ware tweaked slightly from the schematic to get speed while keeping stability(I hope so). As Dave has explained in the blog, these sort of regulators need tweaking to get them to be stable. Voltage wise they seam to work very nicely. Its just the current limits that start acting up when its shown a very low impedance load. That is certainly not surprising and its not a huge issue anyway.



Yellow trace: Output Voltage
Green trace: Voltage across the load switch(Trigger mainly)
Violet trace: Violet? What violet?...Oh doh thats not supposed to be turned on.
Transient load response
Here the PSU output goes from no load to full load in one quick step and so causing the largest output current transient it could see. The switchon transient looks pretty good(Note that thats a quite heavy 5A load) But the turn off has a bit of a problem with the inductance of the dummy load kicking back and causing weird stuff and that does seam to upset it a bit with the 5A load but with lighter loads its fine.


Current limit
This is the exact same test setup as before with the load suddenly being introduced, but the PSU is set to enter current limiting mode. With a small overcurrent condition(Upper Left) it does appear to be very slow so i might tweak the feedback values a bit there. But if you introduce it to a large overcurrent event it acts fast enough for my taste.I also tested it with a very low load of 0.1Ohm(Essentially a short circuit), it does seam to want to oscillate with that but it quickly dies down every time so not an issue there.


Switch on shape
When the board is given power it starts up with 0V output(Caps are empty and DAC is default 0V) so these transients are produced by giving the PSU the command to turn on over UART. There appears to be a tiny bit of overshoot but its nothing too big and its settles down quickly enough.The bottom ones is a test with a huge capacitor on its output. First is with no current limit at all so its trying to get to 5V as quickly as it can. But then the current limit is turned on it keeps oscillating with the cap in place, but at least it osculates exactly around the set current.This is likely very very difficult to solve anyway.


Current sink
Here the current goes in the opposite direction, making the PSU act as an electronic load. This had been  tested by connecting one side of the load to 12V and the other side to the output set to 5V. When the load switches on there is a bit of delay before the sink kicks in because the op amp has a too slow slew rate. But once it does it catches it nicely and keeps it at 5V. Bottom ones also show the current limit turned on, its not the fastest current limit out there but appears to be stable even with very low resistance loads, it does seam to want to oscilate with the 0.1 Ohm one but it also dies out quickly.

[aep9690]

This is unrelated to the general discussion but I actually worked for about 6 months at Vicor as a co-op student.  Its funny seeing the product I worked on making its way into the hobby scene because the are generally targeted at large companies like IBM.

[AndyC_772]

Bound to happen if you find the right people My bench supply is 2x Hewlett-Packard 6632B, each of which would be about £2000 new. I doubt they were ever intended to end up on a hobbyist's bench.

Then again, I doubt they were ever intended to end up on Ebay quite as inexpensive as these were either

[bingo600]

My bench supply is 2x Hewlett-Packard 6632B

Same here  (eB..), i got mine from Finland 
But they ought to come with ear-plugs in the box also.
My lab is besides the bedroom , and the wife refused to let me have the 6632B's on 24/7 

So for doing long time OCXO tests. I ended up getting some fanless also:
A defect HP 3610 that just had bad "Caps" (fixed), and a TTi-PL310.

But i absolutely love my 6632B's , and used one for calibrating the HP3610 (as load).
I could easily have used the 6632B as Voltmeter also , it has the accuracy for it.
But i chose the HP34401A


Btw: Andy ...
Thanx for the 6632B CV/CC excersise here https://www.eevblog.com/forum/beginners/first-bench-supply-ripplenoise-problems/msg141522/#msg141522

Interesting knowledge.

/Bingo

[T4P]

I see why HP would have put a louder fan
1) It was designed to go in 1U rack slot, how much height does a 1U rack slot? Not any other choice but to put a loud powerful fan.
And one reason the heatsink wasn't larger ( Possibly no extra space ) but also heatsinks work better vertically naturally cooled
That explains why!

[Berni]

Yes these vicor modules are designed for use in hugely expensive rack equipment so these switchmode modules likely cost a fortune. But i got the PSU modules for essentially free and decided to rip these tiny modules out of them.(They come 3 modules per PCB module and just have a few tiny noise suppression caps next to them)

A good example of ridiculously expensive equipment getting in to the hands of mere mortals is Mikes blog. He teared down a heap of special stuff from ebay and never really payed a whole lot for it.

As for this PSU project i kinda forgot about it a bit while working on job related projects, but i am going to fix the few issues with it and order the boards in final quantity and make a front panel for it.

[aep9690]

A word from the wise on those modules.  I don't know what model those are but Vicor makes those capable of supplying 320W output.  If the internal caps or FETs have been damaged and you try to turn them on they WILL explode violently.  Check to make sure the input and outputs are not shorted to anything before you do anything.

[Berni]

Well i did want to take a look inside one of these so meaby it exploding is not such a bad thing, does the hard work for me on its own

These ones are all rated for 150W but i do have a rather large one that is rated for 600W at 48V but i don't have a need for that sort of power right now. But since these can trimm down to 10% output voltage with a load i am thinking of putting a large dummy load on it and use it as a high power switching PSU.