Help with PSU design

[pitagoras]

Hello,
I'm working in a design of a LAB PSU. This is strongly based on @blackdog design, with some modifications in the pre-regulator, but also I want 2 additional requirements:

1) V and I controlled with a voltage (which will eventually be produced from an arduino based controller using PWM - as per Richard's idea)
2) remote sensing

This is work in progress but I attach the schematic anyway so to expose my problem.
As you might see, current sensing is wrong. In the original design, the ground of control circuit is just DC+ output. But for the remote sensing to work, it needs to be S+. But then, current sensing will not work properly (as there is the voltage drop on R27).
Of course I could add a differential opamp accross shunt R17 and then inject that into the I servo, as I've seen somewhere. But then I fear the CC control loop to become slow.
All "good" circuits I see having remote sensing use analog controls (PS-501, 503, PD-2020) or the ones having digital control use the differential opamp to measure the shunt drop voltage.
All comments are welcome, including other comments about the circuit.
Thanks !

[tombi]

What is too slow? Also will depend on what op amp you use.

The Agilent E3633A/34A used a differential amp to get the voltage from the sense terminals and solves the  current sense issue.

Might be a good schematic to study actually. There are service manuals with schematics available.

Tom

[pitagoras]

Thank you Tom, I was looking for such a "modern" circuit. I'll study it. Thanks

[Radiohead]

The design you are refering too is an early version of Blackdog's design. Here is his thread over at circuitsonline.nl; https://www.circuitsonline.net/forum/view/110029/1/co+0. The whole thread is in Dutch, which might be a problem for you.

In your starting post there's nothing about the output specs your trying to achieve in terms of Volts, Amps, cooling method, user interface, etc. If it's your first power supply design, it might be a better idea not go for high currents such as 5A immediatly, as you'll need to dissipate more heat with higher output currents. This power supply isn't as simple as it looks, which is why blackdog still hasn't finished it yet.

[pitagoras]

Thank you Radiohead.
I was using this design as a base http://www.bramcam.nl/NA/NA-01-PSU/NA-PSU-SCH-21.png?
It's not the last one?
I aim at 20V @ 2A maximum. The transformers I got are 24@3A so maybe I can go a bit further (if the pre-regulator really works as a power converter)
I want a somewhat compact design. I attach the front panel idea I've been working on, somewhat PS50x style.
As this has a pre-regulator, calculations for heatsink result in a rather small one. I'll put something well above calculations.
The enclosure will be built from perforated metal sheet (just as the sides decoration of this page). I am still in doubt about forced ventilation.
I'll build two identical so I can put them in series.
The idea for the UI is to design a control unit around an arduino mini that can provide two PWM precision outputs and implement a dual slope AD converter (I kind of have a good circuit to start, from an old TI app note) with switchable input channels to read actual V and I outputs.

[Kleinstein]

The circuit in  http://www.bramcam.nl/NA/NA-01-PSU/NA-PSU-SCH-21.png?
has a pre regulation based on phase angle modulation, thus not a full smps converter. So it would not improve the power factor and is not providing higher currents at low voltage. So already taking 2 A DC will push the 3 A transformer to it's limit, especially with large C2/C3. However one might get a little more than 20 V out. So 25 V, maybe even 30 V could work at lower current (e.g. 1 A).

I am not sure D17/D18 will work correct in this circuit, it might take two diodes in series to make is really work accurate.

Voltage adjustment with the pot is no solved very good, but this would naturally get better, if a µC and DAC is used instead.

I don't like the Sizlaki output stages very much. They can tend to oscillate and there is no real need for this here. Just a NPN Darlington would be good enough and less prone to oscillation.
For most of the regulator part it is the standard floating regulator. The current reading will not be super accurate though, as the control current is also flowing through the shunts.

[pitagoras]

Ok I started over "on paper" to design another circuit, this time based on HP E3633A.
So I'm studying this circuit.
I've drawn the basic voltage regulator on LT Spice as an essay of compensation (I try to learn from AN-1889 How to Measure the Loop Transfer Function of
Power Supplies  http://www.ti.com/lit/an/snva364a/s nva364a.pdf).

So far the circuit has no compensation but to start the app note says:

"Loop measurements can only be taken successfully on a design which does not oscillate (...) To achieve a stable design for measurements to be taken,
one can use the following trick on current mode control designs. Place a large capacitance from the
compensation pin to ground in designs where the error amplifier is a transconductance amplifier and from
the compensation pin to the FB pin in designs were the error amplifier is a standard voltage to voltage
error amplifier. A capacitor value of 1 µF will typically work well. "

Where should I place the capacitor in the circuit attached?

Thanks

[ZeTeX]

I can recommend you this channel:
https://www.youtube.com/channel/UCKEVTN-ZlbM6vWJMQ_ZjatQ/videos
he has a series about stability and he explains it very well.

[Kleinstein]

The circuit shown, does not look right. There is likely something fundamentally wrong (e.g. mixed up GND symbols). There usually needs to a kind of analog minimum function with something like a diode or two.  So either. A second point it that a MOSFET power stage really profits from a resistor in series with source as it limits nonlinearity. That type of circuit usually also needs a larger capacitor at the output - more like a few 100 µF, at least for the start.

The make the circuit initially stable to start loop gain calculations one could make an intentional very slow dominant pole compensation. So make the compensating capacitor rather large (e.g. 1 µF) even if this means starting very slow. This usually needs an idea where to add compensation, but this is usually required anyway. As a rule of thumb it is a good idea to have compensation where the circuit is slowest, but this might not apply to all OP circuits.

With the current circuit it is hard to tell where compensation is supposed to be - I can't even get a good guess one which OP should do the CV and which the CC mode loop. Usually one could start with just the CV loop or only the CC loop, but it is a good idea to know where the other loop will be.

[pitagoras]

The circuit shown, does not look right. There is likely something fundamentally wrong (e.g. mixed up GND symbols). There usually needs to a kind of analog minimum function with something like a diode or two.  So either. A second point it that a MOSFET power stage really profits from a resistor in series with source as it limits nonlinearity. That type of circuit usually also needs a larger capacitor at the output - more like a few 100 µF, at least for the start.

Thank you for your comments.
You're right, the use of two symbols for ground is useless in this case, but I was kind of recreating the HP circuit with many edits. In the simulation they work as a single  ground.
This is only the CV part, the CC part is similar (senses the shunt and has another diode -like D1- to perform the minimum). But I was staring with the CV part to understand how it works.
The original HP circuit does not have a series resistor for the mosfet but it does have another mosfet (which apparently does a "rough" regulation). I'll try to understand that part too.
Maybe the circuit seems wrong because the AC source I inserted to perform the gain loop analysis. But without it, it is working ok (in spice!) and regulates the output according to 4 * CV input.
I'll come back with the circuit including CC.
Thanks

[Kleinstein]

With this only for the CV part, this circuit makes sense. U5 is just an inverter for the reference voltage and U4 is a difference amplifier, needed for sense terminals.

The compensation would than be mainly at U3.  With just a capacitor in FB around U3 the circuit should be stable.

It might need some realistic ESR at C9. At least for the final analysis this can be important.

R10 is likely way to large. 100 Ohms in series to the gate make a lot of sense.

The power MOSFET can be nice at low power. However with normal parts the power handling capability is limited and it is difficult to use more MOSFETs in parallel. Load sharing is more difficult than with BJTs. The circuit could work essentially unchanged with a NPN Darlington instead of the MOSFET.

[David Hess]

All "good" circuits I see having remote sensing use analog controls (PS-501, 503, PD-2020) or the ones having digital control use the differential opamp to measure the shunt drop voltage.

The PS501 and PS503 design does not do it but if the variable resistance which would control the current is fixed, then the current control signal can be referenced to common by level shifting it to the high side sense amplifier using a transimpedance (current) output.  On the PS501, this fixed current is generated from the positive supply by Q60 but it could just as easily be variable and referenced to common or any other voltage as long as enough compliance exists to support the full output voltage range.

[blackdog]

Hi Kleinstein, :-)


The pre-regulator is switching OFF  type (stop charging the buffer capacitor, when the de voltage is high enough)
Everi new period it start charging again en switch of the charging if de dropout voltage is reached, simple and elegant.
It is NOT a switching pre regulator.

And more inportant it is Low Noise, this is a realy low noise LAB Power Supply <5uV 22KHz bandwith @ 5-Ampere.
And the Ri is extreme low at high frequency.

The 10 turn potmeter is fine for setting the voltage, i do not need to set this powersupply to 1mV precision.

Everitime i hear the same remarks about the Sizlaki output stages on several fora, its like i here a parrot talking, as a designer you have to do your job, make it stable!
Look at R13 and R14, these resistors keep it calm, and powersupply (with good specs) is not a DC circuit, you have to treat is as a "low HF" circuit, say 10MHz 

Kind regarts,
Blackdog

[minion]

The pre-regulator is switching OFF  type (stop charging the buffer capacitor, when the de voltage is high enough)
Everi new period it start charging again en switch of the charging if de dropout voltage is reached, simple and elegant.
It is NOT a switching pre regulator.

That is true, but the problem is, that because the voltage on the capacitor will be low, the current through transformer when you are charging it will be high, higher than without this circuit. So, even if the average current is relatively low, the RMS won't be, because of the high current spikes. As a result, you have to "oversize" the transformer or limit the current with lover voltages.

But limiting the current with lover voltages is exactly the thing that you wanted to avoid with this circuit in the first place. You wanted to limit power dissipation with low voltages and high currents.

That is how I see it.