PSU student version à la Peter Oakes

[info]

Hello guys! I stumbled upon Peter Oakes' DIY PSU on element14's website and eventually on his YouTube channel. I thought this would be a good project to work on as it's a nice opportunity to learn how stuff works hands on. I watched all the video series and I'm confident I can get something along those lines working. Checked my parts bin and found a couple of op-amps and the other passives required to build the PSU. Before going all-in and etching/ordering a PCB, I decided to test on the breadboard what I had. Here is my version:

Then the suggested basic psu

and then the final iteration of the psu.


I have deviated a bit from the suggested layout by:
- using no differential measurement (using a 1R sense resistor) as I don't have a precision op-amp.
- not using a voltage reference: I have a couple of TL431s around and can replace them to the I&V-set pots.
Otherwise:
- the zener diode is a 6V zener
- V+ is filtered DC input, for now from a power brick but later I will be using a 24V transformer.
- I'm aiming for 0-3A and 0-15V ( eventually 0-24V) on the output
- precision is not my topmost priority. This is going to be a proof of concept and if it works, V2..etc will aim for precision and/or accuracy and also possibly digital control.

So now I have everything on the breadboard but if I float the "analogue" ground on the V_OUT rail, it just doesn't work: zero volts on output, zero milliamps. But when I do join both "analogue" and V_RAW ground, the circuit works as intended but the current goes only to about 4mV on the output even if I turn the pots all the way down.

Now I actually don't know why the "analogue" ground floated doesn't work but with common ground all works. Any hints or gotchas (a.ka. traps for the young players ) I missed?


p.s. The images Basic+PSU.png and PSU+MkII+RPO+fixed.png are all properties of Peter Oakes hosted on element14.com.

[danadak]

Suspicious that TLE2142 is not a RRIO opamp. So it inputs
don't CM to its rails, and its output to neither rail.


Regards, Dana.

[info]

Just had a look at the datasheet and it's true. Is that really relevant here? I mean, according to my probing (simple DMM, no scope yet) on the inputs/outputs of the opamp, there is no sign of either of the outputs driven/input hard to either rail.

[paulca]

This is why I hate op amp datasheets.  They always lead with headline info which suggests one thing, then in the details they pull a fast one and tell you the real story is completely different.

The datasheet says:

Input voltage range, VI (any input)  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VCC+ to VCC-  -0.3 V

So it's 0.3V under it's negative rail and fully to the upper rail?

Nope.  It then says:
V Common (@5V) -0.3V to 3.2V.  So not VCC+ at all, not even bloody close!  Grrr.

Still at least it includes it's lower rail.  I think.

So it tells you that it will accept Vcc+ on it's inputs, but then in the small print bascially says it won't function as an opamp (common mode) with Vcc+ on it's inputs.

They are like car salesman I swear!

[IanMacdonald]

The original is a a strange design if you don't mind me saying so, and I can forsee several reasons why it might have reliability problems.

As mentioned the opamp input swing to supply rails is only going to work with certain types designed for this.   

I don't understand the reasoning behind the gate drive arrangement through a zener and 10k resistor. With the FET acting as a source follower, surely that will limit the max output voltage to well under 12v?

Besides, if both opamp outputs are at 12v then all of D1, D2 and D3 will be non-conducting and you will have a 'floating' FET gate whose voltage will vary at random between the zener voltage and 12v with any static doing the rounds. That can't be good.

A TO220 device is going to be mighty sweaty handling 5A with an input voltage of 15. That's going to be at least 75W worst case dissipation.

The IRFP064 would be a better bet but it would still be outside of SOA if it was on a 30V input, 5A output. 15V, should be OK.

There doesn't seem to be any FET gate protection. I'd be inclined to have a gate to source zener of 20v max.

[Kleinstein]

The differential amplifier for voltage sensing can be tricky. One might have to adjust the resistors for common mode gain to make sure not to get a negative output resistance. With a negative output resistance chances are high that one would get instability with a highly capacitive load.

Anyway it would be a good idea to check the design in a simulation run. The two 100 nF speed up caps for the voltage regulator night not be that suitable. It might need extra diodes to clamp the voltage during fast transients. For the simulation the ESR of the output capacitor can be an important factor. For the real world the choice of capacitor type (normal vs low ESR) can be important too - it might take a combination of both for best operation.

The MOSFET (IRFP064) is not such a greate choice for linear operation. It might be Ok for 15 V, but I am not sure about that.  A better, though still not perfect choice would be an IRFP250 - it is common low cost choice in old audio amps. No guarantied for linear operation by some manufacturers, but chances are it works - though 3 A at 30  V would still be tough. The 10 K in series to the gate is rather high - this would not allow for a fast regulation - less resistance here should be better.

Anyway the MOSFET would likely need a source resistor to make it behave more linear, when currents get higher. Without it, the increasing  transconductance with higher current makes it difficult to get stable regulation. A minimum current (extra load) could help to reduce this nonlinearity and thus improve regulation at low currents. There is a way to have a constant extra bias current flow from the +12 V to the negative output.

For a high power supply (even 3 A at 24 V) it would be a good idea to have a kind of transformer tap switching - this reduces the power loss and thus makes is easier to get away without having 2 or more MOSFETs in parallel, which is not that simple. There is a relatively simple way to use a center taped transformer and a second power MOSFET for this.

[hli]

Since "capacitive load" was mentioned, I'm having a question regarding these OpAmp supplies.
When you have a larger capacity at the output, and the set voltage gets reduced by a fairly large amount (lets say from 12V down to 5V), then the OpAmp tries to reduce the output voltage by reducing the base voltage (or the gate voltage in this case) of the pass transistor. Since the output voltage will not go down fast enough, the OpAmp will probably reduce the base/gate voltage down to 0. But that would mean that the base/gate (at 0V) is negative with respect to the emitter/source (which is still at 12V or so). The typical limit for NPN transistors is 6 to 7V, for FETs its typically higher (the IRFP250 has 20V). So would this mean that the transistor would be destroyed, or does the base / gate resistor limit the current to safe values?
(The same problem would occur when you connect a voltage source to the output which is higher than the voltage set point, e.g. a battery)
If it is a problem, whats a good solution?

[Kleinstein]

In the given circuit the gate voltage is not a problem as it is limited to the +-12 V range of the auxiliary supply.
This usually also applies to other versions of the LDO / floating regulator type circuit.

A simple way to prevent to negative a gate / base voltage is a diode in parallel. Instead of driving the base very negative the OP will than draw current from the output and thus with it's limited current try to bring the output.

[hli]

Thanks. Yes, as long as the output voltage is lower than the breakdown voltage of the pass transistor this is not a problem. But I was asking in a more general sense.
I just simulated this in LTSpice to see what happens. Your solution has the nice advantage that it will also discharge the output capacitance faster.

[mikerj]

This is why I hate op amp datasheets.  They always lead with headline info which suggests one thing, then in the details they pull a fast one and tell you the real story is completely different.

The datasheet says:

Input voltage range, VI (any input)  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VCC+ to VCC-  -0.3 V

That specification is listed under "Absolute maximum ratings", and right underneath is the following note:

"Stresses beyond those listed under absolute maximum ratings may cause permanent damage to the device. These are stress ratings only, and functional operation of the device at these or any other conditions beyond those indicated under recommended operating conditions is not implied."

[info]

Wow, I never thought the choice of an op-amp could make or break your project/bank. I was looking around mouser for amps that are RRIO and can be driven by ±12V and  most are out of my budget. BTW, the mosfet in my rendition is just a placeholder from my eagle library as I could easily substitute it with any other mosfet. I guess I'll have to clamp down my requirements regarding the current for now, i.e. 0-24V and 0-2A. The transformer I'm using has 6 Taps, 1*±12V @1A each, 1*24V@5A and 1*±5V @0.5A. For my needs that's enough as I'm learning as I go along. Any pointers on a guide to deciphering op-amps' datasheets?

So I guess the "floating" aux ground on the output rail is used by the differential readings he makes for current+voltage which I'm not using in my version. I will try his version with 1% resistors and see if that makes a difference. In his videos, he uses 5% resistors and the current/voltage control works as stated.

[Kleinstein]

The floating regulator usually does not need Rail to Rail OPs, especially not with a +-12 V supply for the OPs. However the compensation with the LDO type circuit can be a little tricky. So it needs those small (e.g. nF range) caps to get the right AC response. The LM324 is a rather slow OP, but it might not be slow enough to work without extra compensation. Especially the case of capacitive load is not that easy.

With the given transformer it is mainly the 24 V winding to give a high current. However without a center tap, it could be limited by the maximum power dissipation of the MOSFET. So the transformer might allow for 2.5-3 A, the limited capability of the MOSFET might limit the safe current limit to a lower values (e.g. 1-2 A). For the floating supply of the regulator one could use either the 2x5 V or the 2x12 V windings.

[peteroakes]

As the design has the op-amp power supplies tracking the output supply (op-amp supply is floating and Analogue Zero it tied to power output voltage rail), it will never get to be close to the op-amp supply rails, also as the op-amp only turns off the FET and anything below a couple of volts will have the FET completly off and as the op-amp output goes high it allows the fet gate voltage to rise rather than actually driving it, the output also does not need to be rail to rail.
The other benefit of this arrangement is that the op-amp selection never needs to worry about the actual RAW voltage for the power supply, a +- 12V is all that is needed, no need for a 60V op-amp for instance.

[peteroakes]

The rails driving the FET are floating on the output of the FET so as the voltage of the PSU is increased, so op-amp floats up with it, so as long as your fet can be turned hard on with say 8V, all will work well.
The reason for this arrangement allows for many controls to turn off the fet, if all of them are ok (No Current limit, No Disable, Voltage is not high enough yet, the GATE will be pulled higher through the zener and resistor turning it on harder, as soon as any limit is reached, that section can then proceed to pull down the volts effectively turning off the fet by that required to balance things out again. Nothing except the zener can drive the fet on
if you think about it, the last thing you need is the current limit turning down the fet to limit the current, then have the voltage control say... wow, I need to drive harder as im not were I am supposed to be

On Element 14 right now I and some friends are working through built circuits refining the design, also we have added a micro controller DAC and ADC to automate the whole thing, if there is anything we can try for you, please let me know and ill do what I can
Regards
Peter Oakes
https://www.youtube.com/thebreadboardca

[paulca]

This is why I hate op amp datasheets.  They always lead with headline info which suggests one thing, then in the details they pull a fast one and tell you the real story is completely different.

The datasheet says:

Input voltage range, VI (any input)  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VCC+ to VCC-  -0.3 V

That specification is listed under "Absolute maximum ratings", and right underneath is the following note:

"Stresses beyond those listed under absolute maximum ratings may cause permanent damage to the device. These are stress ratings only, and functional operation of the device at these or any other conditions beyond those indicated under recommended operating conditions is not implied."

Yes.  Hmm.  I also discovered today that it won't go to 0V on it's output.  70mV was as low as I could get it to go tonight.  Pain in the proverbial.  When I looked at the data sheet - if I am reading it right - the "low level voltage" at 25*C for a 5V supply and uA load is 75mV typical, 125mV max.  With 15mA load and higher operating temp that can rise as far as 1.7V!

I am running it on 12V, so I'm probably lucky to only get 70mV.

[peteroakes]

Which circuit are you testing, The basic or the full blown diff inputs more complex one, and do you just have the Op-Amp part running right now.
The reason I ask is that the MOSFET wont be turning on with just a few hundred millivolts on its gate so the output should be zero, the beginner basic power supply did have issues with not getting all the way to zero but really, is that an issue is your trying to power something

[info]

The floating regulator usually does not need Rail to Rail OPs, especially not with a +-12 V supply for the OPs. However the compensation with the LDO type circuit can be a little tricky. So it needs those small (e.g. nF range) caps to get the right AC response. The LM324 is a rather slow OP, but it might not be slow enough to work without extra compensation. Especially the case of capacitive load is not that easy.

With the given transformer it is mainly the 24 V winding to give a high current. However without a center tap, it could be limited by the maximum power dissipation of the MOSFET. So the transformer might allow for 2.5-3 A, the limited capability of the MOSFET might limit the safe current limit to a lower values (e.g. 1-2 A). For the floating supply of the regulator one could use either the 2x5 V or the 2x12 V windings.

Gotcha. Will redo the circuit first on breadboard with the differential inputs and see if that gives a result. I'm using the 2x12V for the op-amps and the 5V is going to be used to power a fan in my final version. As I said before, I'm going to shoot for a lower current output just because I don't have a big enough heatsink for the FET and also because I don't really need all that juice for now. I'm learning as I go along and anything that works, even 0-1A would be fine for me.

The rails driving the FET are floating on the output of the FET so as the voltage of the PSU is increased, so op-amp floats up with it, so as long as your fet can be turned hard on with say 8V, all will work well.........
.................
On Element 14 right now I and some friends are working through built circuits refining the design, also we have added a micro controller DAC and ADC to automate the whole thing, if there is anything we can try for you, please let me know and ill do what I can
Regards
Peter Oakes
https://www.youtube.com/thebreadboardca

  The man himself. Thanks for your clarification. I guess I butchered your circuit out of proportion and that's why it's not working without them differential inputs. Going over to Element14 to check out your progress now.

 

[paulca]

Which circuit are you testing, The basic or the full blown diff inputs more complex one, and do you just have the Op-Amp part running right now.
The reason I ask is that the MOSFET wont be turning on with just a few hundred millivolts on its gate so the output should be zero, the beginner basic power supply did have issues with not getting all the way to zero but really, is that an issue is your trying to power something

Hi Peter, sorry I was referring to the TLE2142 generally, rather than the PSU circuit in question.  I am using it in your original DC load circuit.  It's not a big deal, but currently the circuit won't go to 0mA.  I'm fairly sure it used to, but the TLE2142 now bottoms out at around 70mV, even those the control voltage from the pot divider goes down to 7mV.