Bench CC/CV PSU Based on Daves uSupply (Not Anymore)

[iMo]

It simulates ok with an LM358. The current limit is not precise with I_limit<100mA, though. Bigger shunt (ie 1ohm) makes it better, for I_limit>=50mA it works precise then.
PS: a small negative V- for the opamps would help as well, imho.

With a bigger shunt you get higher voltage on it, therefore you have to decrease the dividers ratio in the current sense diff amplif to be still within the DAC range.

Also add some resistance in series with the output capacitors (ie 100-200mOhm) in the simulation to get more realistic ILoad currents when switching cv->cc.

[jaycee]

PS: a small negative V- for the opamps would help as well, imho.

A small inverting regulator could do that. It also opens the choice of opamps up greatly. -5V would probably be enough. My own supply based on this topology uses an auxilliary transformer winding to get +5/-5V (around the output) to run the opamps, and uses TL082's.

Also add some resistance in series with the output capacitors (ie 100-200mOhm) in the simulation to get more realistic ILoad currents when switching cv->cc.

The capacitors on the output already include ESR - LTSPICE allows you to specify this. In my latest version of the files, you will see I specified a Panasonic FM series capacitor - the values for ripple current and ESR come from the datasheet.

[Kleinstein]

Even the LM358 might like a little negative supply. No need for -5 V, but something like -0.5 V or -1 V would already be enough.

Another part that would like a slight negative supply would be a minimum load, as a constant current load usually needs at least some -100 mV.  A minimum load of some 10 mA could avoid the rather nonlinear and slow part of the power stage.

The OPs initial idea was to start with an (old)  laptop supply, so no easy choice for a 2 nd winding.

[PeterZ]

There is even a dedicated chip to generate a small negative voltage (-0.232V) and allow the OAs outputs go down to zero:

LM7705 Low Noise Negative Bias Generator
https://www.ti.com/product/LM7705

[iMo]

Ok, I see the capacitors models.. So no additional resistors needed in the simulation.
Btw, why not to wire the constant current load at the output (after the shunt)? It creates an offset on the shunt, that helps to go a little bit lower with the lowest I_Limits, and you can compensate for the current offset in software.
PS: the simulation shows -10mA at -3V for the 4x LM358.

[Kleinstein]

The minimum current setting could be due to the LM324 like nearly all OPs can not get it's output all the way to the output when there is a load at the output. Just a little of negative supply should do the trick. Alternatively one could add an offset to the divider, so that zero current could be shifted to some 10 or 50 mV as needed. No need to have the offset at the shunt.

Without a negative supply the constant current load would also not work well, when the output voltage is low. The simulation already includes a constant load before the shunt, so it would not interfere with the current measurement.

A LMP7705 or similar could be an option, though a little tricky to solder (VSSOP 8 case). As there is likely enough voltage to spare, I would consider a diode (like 1N5403) for the negative side current to drop something like 0.6 V from the negative side.  Usually the laptop supplies are isolated, and thus no need to have the negative side of the input to be also the negative side (ground) of the output.

[iMo]

A small modification of small_psu 50mV-20V/2.5A..
Added "Green LED" current sources and with junkbox transistors.
Needs a negative supply as well, -5V for example, or similar.
Simulates stable with min 10uF at the output.
PS: Q3 requires a heatsink as well..

[KC0PPH]

I do not have a Laptop supply to use yet, but I was thinking I could easily get one in the 18V range.

I can use a diode to create a negative voltage rail for the op-amps. This should be ok without causing any issues to the rest of the circuit. I need to still figure out how to use LTSpice so I can play before I build. I have parts on order to breadboard this up.

As for OpAmp selection if a 324 works well I would like to use it as I have a bunch laying around.

It will also be interesting to see how this likes being fed from a homebrew switchmode PSU as I will keep the voltage supply to this right above the cutoff threshold to reduce heating.

[iMo]

I get almost identical results with LM324. Of course, it is a simulation only..

[Kleinstein]

A LM324/LM358 should work though relatively slow.  My usual bench supply is build similar with an LM324, though with the low side shunt, as only single channel.  As most of the  OPs have a standard pinout there is a chance to use a different OP for better accuracy, higher speed (could be a factor for the current regulation).

For the current sources I would not use green LEDs, as quite some voltage is lost. The more usual choice would be a red LED.  One could still use a divider to reduce the voltage drop even further - so that some -0.5 V would be sufficient.

[jaycee]

I'm glad this has given everyone food for thought

One worthwhile change for better transition response is to use a better transistor than the TIP35C. Audio output transistors work very well for this. The 2SD1047 as used in the Korad supply works very well, as does ONSemi's NJW0281, or the 2SC5200. Dont try getting those from eBay though.

[KC0PPH]

What about replacing Q1 and Q2 with a TIP-120?

[jaycee]

What about replacing Q1 and Q2 with a TIP-120?

It'd work for testing I expect but a TO-220 case would have limited dissipation.

[KC0PPH]

I plan on using something in a TO-220 case most likely as my max current will be 1A, and 95% of use case will be sub 200ma i would guess. I do have some 2n3055 but dont want to use them for this project due to my low current requirements.

Also the dissipation should be not all that much as i will have a tracking pre regulator to keep the voltage low.

[jaycee]

OK just bear in mind you need to factor in capacitive/inductive loads into your dissipation requirements... I would probably use two TO220 devices in parallel, but doing that with darlingtons could give you issues. You'll also need emitter resistors to make them share load

[Kleinstein]

Starting with 18-20 V and 1 A maximum current the maximum power would be slightly below 20 W. This is about the practical limit for the TO220 case. At this power level one may not need a pre-regulator and may get away without a fan.

With relatively slow OPs there is no real need for a really fast transistor. The question of a fast transistor would be something if it's about getting the output capacitor really small (e.g. 10 µF range). The speed of the current regulation is more limited by the speed of the OPs.

[KC0PPH]

The TIP-120 lists a max current of 5A, so absolute outside max current of my PSU of 1A gives me a decent bit of headroom.

If I am reading the datasheet correctly it also lists 60W of dissipation. With an absolute max voltage drop across the transistor of around 4V this gives me 4W.

I honestly do not know how to calculate the capacitive or inductive but would be hard to believe that it can increase my dissipation from 4W to 60W. I am thinking about this in terms of Power Factor (Im an industrial guy) Regardless I will put the fixed regulators (5V, 3V3, 2V5, 1V8) on heatsinks and any transistors in this design that might get a bit toasty.

Just by guessing how far am I pushing my luck if I only use 1 TIP-120? What If I lower my current down to say 800ma?

[KC0PPH]

Starting with 18-20 V and 1 A maximum current the maximum power would be slightly below 20 W. This is about the practical limit for the TO220 case. At this power level one may not need a pre-regulator and may get away without a fan.

With relatively slow OPs there is no real need for a really fast transistor. The question of a fast transistor would be something if it's about getting the output capacitor really small (e.g. 10 µF range). The speed of the current regulation is more limited by the speed of the OPs.

I do want to pre-regulate - if for no other reason other than to dabble in a switching reg. Depending on performance I dont plan to keep more than about a 4V drop across the transistor.

I may look into some different OPs that are DIP so I can swap them easily (I plan on socketing them for easy tinkering).

In my design ill leave in the 2nd transistor to supply base current to the pass element if I decide that a Darlington is not a good option. Hopefully I win one of those new Keysight scopes and take some nice picrures of what I am doing.

 

[Kleinstein]

The power ratings of transistors are more theoretical numbers for 25 C case temperature. So the more practical limit is often something like half of that (except for some modern FETs with unrealistic theoretical numbers to start with).
The current limit is not a problem, its more the question on how much heat sink is used. If the heat sink is not that large it may help to have a temperature sensor to turn of if too hot.

With a linear regulator a capacitive  or inductive load would not change the worst case heat load. The worst case would be a short anyway.
Capacitive and inductive loads are the more difficult cases for the loop stability.

With a pre-regulator a highly variable load (like the infamous file test) could increase the worst case loss, though usually still lower than without a pre-regulator.

[iMo]

I've tried with LT1363, 2SC2078+2SC5200, 47pF comp, 1uF ceramic output cap. It simulates ok, and it is faster, indeed.

[not1xor1]

I get almost identical results with LM324. Of course, it is a simulation only..

here is a simpler design which should also work with LM358/LM324
https://www.eevblog.com/forum/beginners/lm324-power-supply-with-variable-voltage-and-current/msg2255262/#msg2255262

BTW it would be better to use 20k/100k resistors for the current control.
Those should be 0.1% or selected 1% resistors with a 100 ohm multiturn trimmer for balancing and cancelling the quiescent current.
I've simulated more complete circuits, I'm not posting other images/asc files for lack of time.
I may do that later if anybody is interested.

[jaycee]

here is a simpler design which should also work with LM358/LM324

It's pretty much the same as what I proposed, but instead of having separate sense and error amps, the sense/error function is combined.
This will indeed work pretty nicely, but one of the good things about having separate sense amps is that you have appropriate signals for feeding to ADC's for display.

[not1xor1]

here is a simpler design which should also work with LM358/LM324

It's pretty much the same as what I proposed, but instead of having separate sense and error amps, the sense/error function is combined.
This will indeed work pretty nicely, but one of the good things about having separate sense amps is that you have appropriate signals for feeding to ADC's for display.

Your design (with IMO contributions) works better than I expected (at least in simulations) although a bit slower than other solutions.
As soon as I've some spare time I'll make more tests in order to compare the (just simulated) performances of various solutions including the floating (HP/Harrison's) design.

[iMo]

I've been playing with both versions a bit. Added FastCC (see a different Mike_Mike's PSU thread on that).
As an example here is the response to a "short" (from 20V) for jaycee's and not1xor1's version with identical params in compensation, parts, settings and load, LT1022A used.
Measured at the 0.33ohm 3055 emitter resistor (FastCC).

Added V(out) dropping from 20V.

Added "shorting" into 10mH, measured at the Shunt resistor.

[jaycee]

imo, how are you doing these tests? can i see the LTSPICE file please ?