Lab Power Supply - The Lost Current

[John Heath]

I am sensing unknown variables as part of the problem. Measuring the voltage drop across a resistor relies on that resistor being what it says it is to know the current. 200 m ohms is not easy to measure. A hall effect DC current probe will measure the current in no uncertain terms and it would be a nice addition to your test equipment. My DC current probe cost 35 bucks. Cheap. Make sure it is a DC current probe not the other type that is limited to AC only.

 Failing this you can use your cell phone as a current probe to know how much current. It requires a little calibration as cell phones were not intended to be used this way. Google magnetic app for android and you will find lots of apps for free. Find a known to be true 1 amp source that can be trusted. Run 1 amp through a wire and find the sweet spot on the back of your cell phone where the magnetic reading is the highest. Mark that spot on your phone and mark what the magnetic strength is for 1 amp. That's it , you now own a free hall effect current probe calibrated to 1 amp and as a side benefit you can make a phone call.

[radoczi94]

I am sensing unknown variables as part of the problem. Measuring the voltage drop across a resistor relies on that resistor being what it says it is to know the current. 200 m ohms is not easy to measure. A hall effect DC current probe will measure the current in no uncertain terms and it would be a nice addition to your test equipment. My DC current probe cost 35 bucks. Cheap. Make sure it is a DC current probe not the other type that is limited to AC only.

The resistor was pretty close to the nominal value. It could not have vary that much, it was +-5% 300ppm and it was just a little bit above the ambient temperature. The problem was: I used inadequate equipment. The best theory is that there were some serious oscillation and it tricked the Wun-Hung-Lo multimeters,so they just showed an average.

I did not even knew, that such a thing exist, will do some research on it, thanks.

Just because a PSU has CC, doesn't really mean that it's short circuit protected. Even my Agilent U8002A will supply over 20 amps into a sub ohm load for 100us before it begins to sluggishly current limit.
I mocked up a CV/CC regulator to experiment with the CC response. The cause of the delayed CC response was because the output of the CC op-amp normally sits at close to its full + rail voltage. When the PSU is suddenly overloaded, it then takes a long time for the CC op-amp's output to slew down to the point where it takes control of the Base.
One of the reasons is that the loop compensation capacitor is usually connected directly between the op-amp's output and inverting input. I have connected the capacitor to the other side of the ORing diode so that the op-amp's output can slew at its maximum rate until the ORing diode conducts.
Because the LM358 that I have used doesn't slew that fast anyway, 0.3V/us, I have put a diode and LED in its feedback path that keeps its output at 2.2v so that it doesn't have far to swing down to take control of the Base. I am ordering some faster NE5532 op-amps to see if I can omit this extra complication.

Extra: I have used an LED for the CC ORing diode to give CC indication.

That design is almost identical to the one formed in my head. My concern is, that I want to use fast transistors, around 30MHz.The opamps should be order of magnitude slower to prevent loop oscillation without compensation. But the slew rate of the opamp should be high enough to be able to handle the switch on-off events, with an acceptable waveform on the output. Guess, I should learn to use a simulator software before I start soldering.

[John Heath]

Speaking of oscillations I noted a 10 uf condenser on the output of your power supply. Who ordered that? 10 uf in CC mode means much higher current than the setting for the CC mode in the first 1 m second. It defeats the purpose of current limiting for sudden current demands in the 1 m second range. That is fishy. Did they put it there as an after thought to compensate for an oscillation problem in CC mode? Try a 100 uf quality condenser and see if things improve. It's a long shot but you never know. If it does improve then the current limiting is only for average not sudden current demands. This is not good as one could burn out an IC with a sudden short even if the current limiter is set to 100 m amps. A raw 100 uf quality condenser will be more than happy to provide the destructive current to destroy this IC .

[xavier60]

That design is almost identical to the one formed in my head. My concern is, that I want to use fast transistors, around 30MHz.The opamps should be order of magnitude slower to prevent loop oscillation without compensation. But the slew rate of the opamp should be high enough to be able to handle the switch on-off events, with an acceptable waveform on the output. Guess, I should learn to use a simulator software before I start soldering.

Oscillation is a big concern. My experiment did oscillate for various reasons. One time it seemed to be Darlington that was the cause. That's why I put the 220 ohm in series with the Base.
One of my suppliers has some 150V 73A MOSFETS(PSMN020-150W) selling  cheap, so I bought some to experiment with in my regulator.

[C]

Many ways to stop or limit Oscillation.

Rate of change positive could be different from negative.

Multi-stage feedback.

Areas of circuit where frequency response is different.

Positive feedback to get large change with negative feedback for small change.

For an analog or you could have two. A min OR to control output & a MAX OR to prevent swing to rail.

Many ways other then just slow.

[jaycee]

Seen this circuit before. To be honest it's not a very good one, and those opamps are probably seeing way too much voltage across them!

Take a look at the schematic for the ELV 22532 power supply. A good approach used by many of the big manufacturers is to have a seperate "bias" supply for the opamps, which floats around the output. I've used a tiny little 9-0-9 transformer from an old bedside digital clock along with 7805/7905 regulators to create such a supply before. Things are vastly simpler and performance is much better with this sort of topology

[xavier60]

My idea of connecting the integrator capacitors to the other side of the ORing diodes was too unstable. I want the current limit fast acting yet stable. I'm experimenting with an idea that enables the feedback through the integrator capacitor  only after the CC op-amp has taken control.

[radoczi94]

Hi Everyone!

Sorry for the late reply, I had a critical exam last friday, studied my brain out. It was succesful, so rewarded myself with a scope,ordered a Rigol 1054Z. It should be here tomorrow.

On the weekend I was playing around with the PSU. I removed the CC opamp from the socket, and put on a 12V 10W chinese LED as a load. With this config, I measured 0,5A on the shunts and 1 amps on the output. So the CV opamp was oscillating (or something nasty like that). Then I replaced it (NE5534) with a 741, turned on, the result was 1 (or both) shorted 2N3055 and a smoking chinese LED. Doh. I just lost my temper, whacked out the industrial heatgun, desoldered everything from the pcb and threw it in the scrapbin.

Seen this circuit before. To be honest it's not a very good one, and those opamps are probably seeing way too much voltage across them!

Take a look at the schematic for the ELV 22532 power supply. A good approach used by many of the big manufacturers is to have a seperate "bias" supply for the opamps, which floats around the output. I've used a tiny little 9-0-9 transformer from an old bedside digital clock along with 7805/7905 regulators to create such a supply before. Things are vastly simpler and performance is much better with this sort of topology

I decided to build that supply on breadboard and play around with it,try to change small things and measure everything. What I want to do is replace the TIP14x transistors with 1 driver and 3 or 4 end transistors, use a more adequate reference than the 7805, and replace the opamp with a more skookum one. I will put together a transformer tap switch circuit too.

Many ways to stop or limit Oscillation.

Rate of change positive could be different from negative.

Multi-stage feedback.

Areas of circuit where frequency response is different.

Positive feedback to get large change with negative feedback for small change.

For an analog or you could have two. A min OR to control output & a MAX OR to prevent swing to rail.

Many ways other then just slow.

I will make some research on things like these, thanks. Will need to do some experiments to see what are the exact effects of these methods.

[C]

So you should now have a scope.

Think of the old time test equipment.
  Back then it was very poor, but the old timers built some fantastic stuff using poor. Very stable circuits that had power supplies in the ball park with a lot of noise.
A voltmeter or current meter was treated as a poor slow speed analog display that was not very accurate. A scope was treated as a higher speed video display, again not very accurate.

With every thing poor first high res voltmeters where voltage comparators. They built these by comparing resistors to build a voltage divider. They compared the voltage reference to a better voltage reference.
The result was a bunch of ratios. This resistor is 2x that resistor, this resistor matches that resistor..  The 2x was hard until you had two matching resistors and then very easy. As the comparator got better the degree of match got better and the resistor divider got better, but it still was all ratios.
Assuming that the D'Arsonval galvanomete was repeatable, and by using a Wheatstone bridge you could compare(voltage, current, resistance).

If you use your brain some you can work with uncalibrated test equipment using it more as a comparator. Good numbers are then just a way of passing information.

Old timers also used positive and negative feedback. The trick for this is balance. At static stage you want positive feedback countered by negative feedback.

Think of a switcher power supply, it is just on or off. What makes it work is Time( the rate of change). Good use of analog can act like switcher for large changes and analog for small changes.

If you think of an op amp output as a pot, then if you have a pot input you get a scaled pot output. But here you are forgetting that a op amp output is really two pot outputs. You only get a pot output if both output pots change correctly and you are still forgetting the output load.

Going a step further the op amp feed back is like a . seesaw( teeter-totter) and output is more like a set of springs.
In physical world a seesaw can bend and bounce. Load changes changes spring deflection.   

A min parts circuit has to work harder or be not as good as a many parts circuit. Many parts also has problems of it's own.The total result is what you want to be as good as possible.

A lot of old time transistor circuits would often use differential circuits. This signal goes positive while a second signal goes negative. This can be cheap while opening up options for control.

Really think about your control circuit.
When you have a big output current change, you have a lot of current change on 2N3055 base. Do you want the op amp handle all this change or would it be better if the circuit around the 2N3055 handled some of this change directly.
Do you want source change all handled by op amp or some handled by circuit around 2N3055.
Your output stage is based on current, think current.

[chris_leyson]

Stumbled across an interesting blog from Peter Oakes about building a bench supply https://www.element14.com/community/groups/test-and-measurement/blog/2014/09/15/the-modular-bench-power-system-the-essential-diy-build-for-every-ee-student-and-old-timer-alike

[C]

First, how much have you tested this with different loads? As a lab power supply I would think you would want to connect any load to this and know how well it will try to protect the load while trying to maintain it's settings.

Think of using a second copy of this as a load(shunt regulator) to test the first. The +30volt supply & 1000uf cap is your supply. Q2 & R5 is the regulator.   
By keeping both the same, you could use one to test the other. By making the same changes to both you improve the Power Supply use & Load use.

Might also want to think of some nice future additions and make it easy to add them.

One thing I have seen is using an internal known load to get more information about connected load.

An additional think to think about. Do you want only the load pulling output to 0 Volts & 0 Current output.
Really think about this, Say you are building/testing a circuit connected to this supply. You short something out in the circuit. Do you want all the cap's in test circuit dumping in to short or would you like the Lab Power Supply able to pull excess from test circuit. Right now you can source current, To make faster response to a lower output you would need a shunt across the output, Pull down in addition to existing pull up..

If you are going to trouble of creating a PCB, might be nice to be able to use the PCB as part of more test equipment of your test bench. 

If you plan ahead you can make ground loops less of a problem.


 

[xavier60]

There is the concern that some combination of load reactance will make it unstable. I use another large MOSFET driven by 15Hz to cycle various loads. Besides the function of Q1, the op-amps have been applied in a rather standard way, so I'm not expecting any problem that tweaking the compensation won't fix. 
I had no plans to make it sink current except for a bleed resistor across the output unless I find too much overshoot when it unloads.
Designs like that of Peter Oakes have the option of being controlled and monitored by a micro-controller. Every time I think about this, it adds another 4 op-amps, so I won't bother for now.

[radoczi94]

Did you tried to tie diodes between the fet's DS? That way the CC opamp could pull down the voltage from the capacity on the output, possibly from the connected circuit too. I don't know if it's a working idea or not, just a brainfart.

[xavier60]

Did you tried to tie diodes between the fet's DS? That way the CC opamp could pull down the voltage from the capacity on the output, possibly from the connected circuit too. I don't know if it's a working idea or not, just a brainfart.

If I understand properly, that would need another power MOSFET to force down the output. I'm trying to keep it as simple as practical.
Even though there looks like a lot of current overshoot, it is actually much better than other PSUs.  An even faster op-amp I'm expecting to reduce the overshoot further.

[radoczi94]

Did you tried to tie diodes between the fet's DS? That way the CC opamp could pull down the voltage from the capacity on the output, possibly from the connected circuit too. I don't know if it's a working idea or not, just a brainfart.

If I understand properly, that would need another power MOSFET to force down the output. I'm trying to keep it as simple as practical.
Even though there looks like a lot of current overshoot, it is actually much better than other PSUs.  An even faster op-amp I'm expecting to reduce the overshoot further.

No, I just tought about diodes, 2 or 3, they only open if there is a radical intervention by the CC opamp. Something like the on the schematic below. The 2 diodes on the bottom are the analog OR gate.

[xavier60]

The 2 diodes would clamp the Gate at -1.4 with respect to the Source. The Gate needs to be reduced quickly to below about 4v for the MOSFET I'm using.

[C]

You have to think of a load as being dynamic. At some point in time you will try to use your lab power supply to power almost anything

You should have no problem seeing a load switch from no current to max current and back.
Think of what this is doing to your current op amp.
The higher freq op amp change hides some of this problem.

Think of the states of your power supply.
You are designing for two states  CC & VC. You have a third state where output is not at set voltage or set current due to a load change. The time it takes for output to get to a set state is this third state. Most circuits handle less then set a lot better then greater then set.
You need to think about when you connect a circuit you are designing in the future. That circuit could have over voltage protect and/or over current protect circuit that is faster then this power supply. It could be more sensitive.
What will happen when the load makes a massive change on just a little overshoot? Protection will most likely go open circuit or dead short. Remember that this lab power supply has to protect it's self & the load.

So the higher speed op amp makes circuit respond faster. This also increases freq response range of power supply output. This adds greater chance for circuit to oscillate. 

Try testing you circuit with smaller value for C6. C6 is there to prevent the fast current changes from causing your power supply to oscillate.  T
Here smaller is better( less un-controlled dump into load).
Find where circuit starts oscillating and try to make osc freq higher in rest of circuit first. This will reduce output overshoot some. Then make circuit stable with min parts change.
While testing, if you are not using a lab power supply to supply your 30 Volt supply you might want to add some protection to Q2 to prevent magic smoke. A temporary resistor between Q2 & 1000uf could do this.
 
R1 should be the lowest value possible.


Need to remember that active circuits take time to change.

A good test point is to look at anode of D! & D3 and look at what is happening on cathodes.

[xavier60]

No, I just tought about diodes, 2 or 3, they only open if there is a radical intervention by the CC opamp. Something like the on the schematic below. The 2 diodes on the bottom are the analog OR gate.

I understand your idea now. After the MOSFET is turned off, the CC op-amp will help pull the output down, It is only 30ma though.
Actually the  op-amps can't pull the PSU output down directly because they are referenced to the the output rail.  Current needs to be sunk to the negative output rail.

[xavier60]

It did oscillate when I reduced C6 to 10uF. I found a choice of 2 possible fixes.
One is by simply reducing the Proportional response of the CV op-amp by reducing R3 to 33k and increasing  C2.
The other is by increasing the frequency response of the MOSFET by reducing the Gate resistor to 100 and also making the response of the CV op-amp only Integral by putting only a 100pF capacitor in its feedback path.
With C6 changed back to 100uF, the voltage dips by 0.25V with a 3A load and recovers in 5uS.

[C]

The lowest possible gate resistor is best here.  Remember that connecting a scope to gate changes things. It is only here to prevent ringing on the gate. You are trying to charge/discharge a gate cap and need a lot of peek current. Digital gate drivers peek current is in the amp's to get fast change.

Keep in mind that C6 is like a CV hack and CC harm. It is there to handle very high freq CV & CC changes. Smaller is better for CC while larger is better for CV. C6 also hides change.

Your CV circuit has two modes.
One is changing between two currents in CV mode.
Second is changing to/from CV to CC

Your CC circuit also has two modes.

You op amp datasheet should list specifications for large and small output changes.
Also of note is your op amp has a limited amount of output current.

In your testing look for differences.
In CV mode
1. lower current to higher current.
2. higher current to lower current.
Here you are looking at voltage change, rate of change and overshoot.
3. CV to CC where CC mode is small or large currents.
4. CC to CV where CC mode is small or large currents.
again for CC mode
and many more.
Good testing is a must.
DC up to freq's hidden by C6 and all possible dynamic load
changes.


Think of each part and what would happen if you larger or smaller and all the effects on the circuit.
For example it C6 was huge output changes are slower & at same time change is slowed so sense is slower..
At same time it's different for larger or smaller currents.
Your circuit can not compensate for what it can not see.
Huge also turns a short in to a spot welder.
And you have no perfect(matched) parts for a copy.

Think of using changed values or added parts to assist in making other parts of circuit function better.
For example changing R5 makes current sense larger or smaller & also effects CV mode with more or less change needed.

Some cap's are speed up and some are slow down, use with a lot of care.

Change I would thing about is using a quad comparator to drive CC & CV leds. This should leave more current for driving Q2 and provide output test points that reduce effect on control circuits.

One problem area is mode shift.
In CV mode, CC wants more current. You could have one op amp at rail before change and require a huge output change for transition.

[C]

The output of the CC op-amp starts slewing down when it senses that set current has been reached.

This is why I suggested you look at the anode & cathodes of D1 & D3
You have a limited rate of change on the outputs of the op amps. In addition more time is usually required for op amp's output to come off it's rails.

You need to test both CC and CV mode before making a change. What happens to one should happen to the other.
CV mode to current op amp.
CC mode to voltage op amp.
When you find a problem for one make same change for other.

With your current connections, C6 is hiding current changes from R5. In addition R5 voltage drop changes is effecting Q2 directly.
moving C6 to Q2 source would let R5 see more of the load current change & let CV control see need to change.

Think of what circuit is doing.
A negative change of Q2 gate is output increase.
A negative change on R20 is output increase.

Swapping current sense makes a negative change of R5 and output increase.

With all working in same direction simpler control.

You have D1 & D3 setting min and controlling Q2
Think of another analog OR that sets ready to take control max.
Set voltage or D1 anode
Set current or D3 anode

Remember to keep testing with lower values of C6 as this will let you see problem areas better.

[xavier60]

actually it is stable with 47uF for C6, I neglected to update the schematic.
There is a delay of the op-amp's output initially dropping from sitting at full voltage. It is responsible for the remaining current overshoot.
I have never seen any designs with the shunt exposed directly to the load.  Ill see what happens.

[xavier60]

Moving C6 to the other side of the shunt allows the shunt to see the capacitor's discharge current but nothing can be done about it anyway.
I mainly set out to solve the problem that many PSUs must have, where the short circuit current is unlimited for some time.  My Agilent U8002A will supply the full set voltage into a sub ohm load, tens of amps, for 100us  before the CC loop begins to respond.

[C]

Moving C6 to the other side of the shunt allows the shunt to see the capacitor's discharge current but nothing can be done about it anyway.
I mainly set out to solve the problem that many PSUs must have, where the short circuit current is unlimited for some time.  My Agilent U8002A will supply the full set voltage into a sub ohm load, tens of amps, for 100us  before the CC loop begins to respond.

Something can be done, decrease size of C6 to decrease current dump.
Increase control speed to reduce 100us.
The more you reduce R1 the faster you can change.
The more current you can pump in/out Q2 gate the faster you can change.
If you have control then you can add functions.
The faster you can make circuit respond the less time output is uncontrolled.
Faster also reduces overshoot.
Think about it. you can drive a hot rod on the street with proper control. A bad driver will end up in a tree, while a good driver has power to get out of way.
The difference is how good is the control.


One of the big problems for a power supply is the unknown load.
I have seen power supplies that add a scaled internal voltage load & a scaled internal current load not effected by output load to get better control.

Need to keep in mind total result & how each part effects local area & total.

for example
C3 has two effects, removes noise & slows changes.
A cap across R20 does same for output.
A cap across R21 adds output noise & speeds changes.
There will be matching places for CC side.
Using both an AC divider.
You can use separate dividers and mix two at input to op amp. 
Remember that when you add a cap that the effect is different at DC then at high frequency.
Try to have few caps and really think about it before adding one.

Right now you are running op amps open loop with total circuit controlling gain. You have option of some local resistor feedback which would slow all changes.

Note that you can use many feed back paths. Op amps are commonly used to combine two audio sources for example.
Might think of three inputs to op amp, scaled output, reference input and local feedback that gets canceled or over ridden to keep output off rails.

Old timers often used positive & negative feedback.
If you have 1v positive feedback and 1v negative feedback. A set of matching resistors in series = 0 feedback.

Remember that mode change is by conduction curve of diodes. Diodes have capacitance.
Diodes only pull one way. Watch for difference in over shoot vs under shoot.

Your LED's are adding error to output that is different between low and high. I would think of using a quad comparator for leds.

When thinking of a change, really look at the many ways to accomplish same thing.

[C]

You are using
NE5532p
http://www.ti.com/lit/ds/symlink/ne5532.pdf

Q2 = PSMN020-150W
https://media.digikey.com/pdf/Data%20Sheets/NXP%20PDFs/PSMN020-150W_2.pdf

D1 & D3 =_____

Here is my thinking
Q2 gate is a voltage control device, current is rate of change.

The  Q2 gate-source slows rate of change. And with out proper control can lead to ringing.

D1 & D3 have non linear conduction curves but only effect positive  output change.
NE5532p is a voltage output device.
From data sheet output short circuit current is 10-60 ma.
The current limit of NE5532p limits positive output change
A part of this current goes to over coming diode pull up resistor.
diode pull up is source of negative output change output & Q2 gate.

To get fast you need to change fast but must remember time delays of Q2 gate-source cap & NE5532p delay.

Now look at
CC op amp in CV mode or CV op amp in CC mode.
Both are swinging full range which takes time.
If you can build a window around these, then you could shorten the swing from a little above current output needs to control of output.

You can connect two more diodes below D1 & D3
and build a max analog OR
D1 & D3 are a min analog OR of CV/CC to control output.
New Max analog OR can be used to limit the swings of the op amp's

Really try to keep both CV and CC op amps the same circuit with only difference being difference of output & reference.

Also note that you can regulate down to 0.
With R15 connected as shown in last the rate of change will get slower the closer to 0 you get.
Think of dropping the output and catching it with the op amps.
You have option of using a constant current or a better third op amp to do this function.

You could add a resistor between op amp output and diode to be able to sense op amp output current.

You can window a control signal and slide the window.
You can also window the window and gain even more control.