[WorkInProgress] LM324/358 based, 0-20V, 0-1A, cc/cv PSU

[debininja]

My

used a LM723 with foldback current limiting, which is not very accurate. It's a good "beginner PSU" because the voltage output is pretty stable. The current limit, however, drifts around by a large margin, depending on ambient temperature, heatsink temperature, and so on. By no means is the LM723 a "lab grade PSU." It's time to move on to something better...

======================================================
As a first time op-amp based PSU builder, I'm keeping my goals modest.
Target specifications:
* Input voltage max: 25V
* Output voltage: 20V max
* Output current: 1A max @ 20V
* Output power: 20W max
* Must use commonly found hobbyist parts. No using fancy shmancy ICs.
======================================================

EDIT: Settled for Klein's simulated schematic. Transferring over to KiCAD. 

[Kleinstein]

It is a good idea to limit the voltage and current for a first try.

The current sensing would not work with an LM358. It violates the common mode range, as it will not work near its positive supply. The regulator is using a kind of emitter follower output stage - so it would need a really fast current limiting or a secondary current limit (the LM395 should provide that). The LM358 is already on the slow side for this.

The original circuit also used the compensation pins to speed up the reaction of the current limit. So it is a little tricky to replace the upper OP in this circuit.

The TIP120 is a little unusual choice. TO220 transistors are not that well suited for high power. Two TIP140 or two 2N3055 (or TIP3055) and a BD139 would be the more obvious choice. A transistor usually fails with a short - so if one of them fails the whole circuit would fail. 40 W for a tip 120 is also rather optimistic.

I don't think the part around Q4 would work well.

The reference part for the voltage should be a little different. It is Ok to replace the refs with more modern ones. For the voltage I would consider the LM329. Usually the TL431 and like work best with just a little more than there minimum current, so more like 1-1.5 mA.

[debininja]

Damn, that's a bummer. I was hoping there was a decent schematic for a cc/cv opamp based PSU using commonly found opamps and discrete components.

Seems there's very few designs using common opamps. Even this from the 1970s uses the LM301.

Maybe I should start with the basics as a newcomer. In other words, build a constant current source (can be easily done with a LM358 + mosfet + stable voltage reference + potentiometer), and then add in another opamp for the constant voltage source.

What do you think? Sounds like a good plan?

[edavid]

(Well, honestly speaking, I'd rather pay $60 for those GPC-3020 CC/CV PSUs on sale, but I'm broke at the moment).

You can get an eBay power supply for $36.  Does that change the equation?

http://www.ebay.com/itm/3-1-2-LED-Digital-30V-3A-Adjustable-Variable-DC-Power-Supply-Regulated-Lab-Cable-/162551854197

[debininja]

I'd much rather buy a used $70 GPC-3020 linear PSU than some no-name garbage, switching PSU with horrible decoupling caps and the bare minimal, cheapest components from China that explodes (seriously, some guy's PSU caught fire, and it was one of those models). Even if a PSU uses analog meters, I don't find it a hassle. Better safe than sorry.

I actually made my first SMPS a week ago, after taking almost a month. It was a challenging project for sure, my very first SMPS and very first serious project. It's not the most efficient thing, nor is it pretty, but it does work and meets all my set goals (FBT, secondary isolated via optocoupler, 5V regulated output, 1A continuous current).

Anyways, it's a learning experience for me with opamps. I'm not afraid to make mistakes, so long as they are calculated risks. Building a cc/cv is a big challenge, but I'm sure I can get it done and learn a lot in the process.

[Kleinstein]

It is possible to build a lab supply based on standard OPs. The LM358 might not be the best choice due to its low maximum supply voltage and not so good speed. However it works - I also have an old supply base on the LM324 (quad equivalent of LM358).

There are mainly two options:
1) The classical floating regulator, with a second transformer to power the regulator circuit.
This one is flexible and relatively easy to use with digital control / display as the extra transformer can also power the display / control circuit.
2) The classical emitter follower circuit, similar to the circuit shown here. Usually one has the shunt at the low side. It is usually limited to something like 20-30 V due to the OPs supply and the better versions need a negative supply too. Here the output capacitor can be smaller, but current regulations tends to be slower in response.  For a low end supply up to about 20 V this circuit is also a little easier, and easier to understand.

There is a parallel thread on a supply of the 1st type:
https://www.eevblog.com/forum/projects/help-with-psu-design/
The shown circuit has the basic features, but could be simplified in some areas.

[debininja]

Perfect, I'll go with the first type using two transformers.

Although, it doesn't have to be a separate transformer, right?
An auxillary winding on the same xfmr would be electrically isolated, so it should work just as well, correct?

(And regarding that circuit you linked. I noticed there are output caps on the output. Seems like a terrible idea to do so for a cc/cv PSU, as those capacitors can output dozens or hundreds of times the set current once a load is placed, being limited by the ESR and load resistance).

If you have a schematic for the LM324 PSU that you created, I'd appreciate it if you could upload it here.

[edavid]

I'd much rather buy a used $70 GPC-3020 linear PSU than some no-name garbage, switching PSU with horrible decoupling caps and the bare minimal, cheapest components from China that explodes (seriously, some guy's PSU caught fire, and it was one of those models). Even if a PSU uses analog meters, I don't find it a hassle. Better safe than sorry.

The eBay supply I cited is a linear supply.  I don't see why it would be any more likely to catch fire than a home-made one.

[debininja]

Sure it's "linear," but it has a SMPS output stage with questionable filtering capabilities and high ripple (even though they claim it's 10mV, I highly doubt it --100mV, maybe). It weighs ~3 pounds. The linear series pass stage will attenuate the noise, sure, but it wont be much with the amount of puny little capacitors and inductors they can fit in something that size. Instead of paying $36 on that sort of SMPS garbage, I'd much rather pay double for the GPC-3020.

The GPC-3020 available on eBay for $60-80, 30V, 2A, parallelable or series-able (to 60V, 2A), and weigh 18 pounds. Look at its impressive datasheet and you have a clear winner in terms of what the better quality PSU is.


I'm aware it's economically unwise to create a DIY power supply, because, well, it won't quite meet the performance of quality, factory made devices. With $20/hr wages, you can buy a quality PSU for 3-4 hours of work. Besides, building an op-amp based PSU is a good learning experience, and that's my main goal--experience and knowledge.

EDIT: Found a post with a schematic for a lab cc/cv PSU using very common parts: https://www.eevblog.com/forum/testgear/are-cheap-ebay-power-supply-worth-the-money/msg248693/#msg248693

Attached schematic for those curious.

EDIT2: edavid, those are all SMPS based cc/cv lab PSUs you're linking There's nothing wrong with them, but what is a dead giveaway is the cost. A SMPS based lab PSU will need a lot of proper filtering for decent noise levels. At $36, I know for a fact that I can't find anything decent when buying those, new. A 10kuF, major branded capacitor costs $5 (Mouser), some TO-247 power transistors ($3 each), opamps, precision resistors, etc etc, it'll come out to at least $30 for the base parts if lucky. Much better off buying used linear PSUs for long term guarantee that the thing wont expode or pop the capacitors.

[edavid]

Sure it's "linear," but it has a SMPS output stage with questionable filtering capabilities and high ripple (even though they claim it's 10mV, I highly doubt it --100mV, maybe). It weighs ~3 pounds. The linear series pass stage will attenuate the noise, sure, but it wont be much with the amount of puny little capacitors and inductors they can fit in something that size. Instead of paying $36 on that sort of SMPS garbage, I'd much rather pay double for the GPC-3020.

Hmm, how about this one: http://www.ebay.com/itm/0-30V-0-3A-3-Digits-Variable-Adjustable-Digital-Regulated-DC-Power-Supply-W9Y8-/192073872220

[floobydust]

I like the aux. floating supply approach the best- you get almost no V losses in the pass-transistors and can use ordinary (non rail-rail) op-amps, they have to slew less.
Disadvantages I find are the output overvoltage spike on power up/down, extra transformer and Vregs.

Most of the lab bench PSU's from Asia use the old-school low cost dual op-amps:

LM358 $0.55
LF353 $0.79
TL082 $1.18

LT1013 $5.91


I'd rather spend $5 of the budget on decent power transistors like 2SC5200's instead of high-performance op-amps.

[aries1470]

@ OP - debininja,

Can you also comment on what Voltage range and Current you are looking for.
Just curious, and what your usage will be.

Then I can search more specifically for you.

Does it need to follow that old design, and if yes, why?

Will you be using a large transformer - linear, or do you prefer SMPS?
If SMPS, what convertion target of efficiancy are you looking for? Will 60% -75% do, or do you need 80%+

What is your total budget for the project so realistic goals can be set.

Sorry for all these questions, but they will help identify many things

[debininja]

I listed those in my first post.

* +25V rail max, -3V rail min
(unless there is a way to get a single supply based cc/cv design. I couldn't find too many of those around the net)
Oh look, here's one. And this! I love this thing! I think it'll make a good first cc/cv project!

* 20V max output @ 3A max current (60Watts. A big heatsink from an old Pentium 4 PC should work just fine).

* I would prefer a modular design. If I can hook it up to, say, a 24V laptop charger (SMPS) that can output 3.5Amps, that's fine. I can add in external filtering if required. I would rather not build another SMPS at this point, so my preference is for pre-existing SMPS that can be found around the house (inkjet printers have nice and juicy SMPS that output 30V or so, and I have one that outputs 32V @ 1.6A. I can definitely use that by modifying the TL431 voltage divider present within to output 25V).

--> I also have a MOT pre-wound (24turns_CT_24turns) and ready, which outputs 36V-0V-36V rectified at the moment (I can unwind some turns to keep it at 25V instead--I'm keeping some headroom in case the mains power surges by +10%, that way LM358s wont get burned. The filtration stage consists of a common mode choke and 6600uF combined capacitance (will do fine for 3Amps max).

* It doesn't have to be an old design. I'm just looking for something that can be built with common, inexpensive, discrete parts--parts that are surely in stock by 90% of hobbyists: LM358s, LM324s, TIP120s/TIP127s, TL431s, 12V and 5.1V zeners (non-precision), 2N3904s and 2N3906s, IRF640Ns, IRFP250Ms, 1N4148s, 1N4007s, various TO-220 and TO-247 schottky diodes.

--> I'm willing to spend $10 OOP in parts beyond what I listed above, which I have at the moment. I spent $30 on the SMPS build from last week, so I can't be blowing more money around at this point. The life of an EE student on a budget is not an easy one

[aries1470]

Hi,

I found a "simple" and hopefully easy for you to source, or should have most components handy, power supply to make your own Oh, and it ain't too fancy either
I came across a few, but here is the 1st on the list:

1. 30v-4a-adjustable-bench-power-supply

I will be updating and adding to this list in due time for future websearches too.

from underneath here will be all the new additions:

2. Simple commercial bench top power supply, the CSI3005X5, found here. Just needs a X-Former, and no, you don't need that fancy front end, and now more people have a stepping stone to do more google searches for it, without the need for the fancy LCD front-end, LED Front ends work great too

[David Hess]

This designs takes advantage of two unusual properties of the LM301A making it difficult to replace.  Offhand, I know of no modern equivalents to the LM301A.

1. The LM301A has a common mode input voltage range which includes its positive supply making it useful in high side current sense applications like here.

2. In this particularly design, the compensation pin of the LM301A is used for clamping to improve switching between constant current and constant voltage mode.  Offhand, I know of no available modern parts which provide this capability except maybe the LT1008 and while it would make an excellent error amplifier except perhaps for speed, it's common mode range does not include the positive supply.

[floobydust]

... 1. 30v-4a-adjustable-bench-power-supply ...

This is a chinese design kit sells for $6 on ebay. No way it can reliably output 4A with a single pass-transistor, see SOA curve.

All are missing the protection diode from Vout to Vfilter cap.
That Mini Bench PSU has Vout going directly to op-amp X4. Not a good idea.
Making a lab PSU bullet-proof, a lost art. Sigh

[debininja]

Hello everyone. I ended up finding this gem of a project. I transferred over the schematic to KiCAD and modifed many of the parts. See my first post here for all the hacks I put in.

Please let me know what you think. I think it'll be a nice little PSU to start off with and work my way up.

[Kleinstein]

The MiniBenchPsu does not look that promising.  Having the shunt at the high side is making things more complicated than necessary. It can be bad for stability too and it would need an amplifier that work up to the positive rail. The way of using a differential amplifier, to bring down the common mode voltage introduces quite some errors and needs accurate resistors. So it is more like a collection of points not to do.

The compensation input of the LM301 is used to speed up reaction of the current limiting. There are other ways to achieve this and having the shunt on the high side has more disadvantages than just needing the OP to work at the upper rail. Usually it is not a good ideal anyway.

Possibly having a spike on turn on is nothing specific to the floating regulator type. It can happen with other circuits too and there is a way to prevent this. It might be even a little easier with the floating regulator.

The commercial Mastec circuit can make a good bases for a supply with floating regulator. It is not perfect, but also not that bad. The compensation used is very simple, maybe even oversimplified, but this is possible to fix / improve with a few small caps and resistors, but quite some thoughts behind them.  One can also add sense input to that circuit if needed. With only 20 V and 1 A one can simplify or leave out transformer tap switching. For just two taps (e.g. a transformer with center tap) one can also use an elegant electronic way of using two raw voltages. There are mainly two downsides with the floating regulator: it needs the second supply (e.g. extra transformer or isolated winding) and the output capacitor might need to be a little larger.

The simple circuit without a floating regulator usually has the shunt on the low side. One can get away without a negative auxiliary supply, if single supply OPs like the LM324 are used. However without a negative supply there is no way of making a kind of minium load current sink and this limits the performance. With a transformer as a power source one can use a simple charge pump for the extra negative voltage. This is not a big deal, as one only needs something like 10-20 mA. With something like a laptop supply to start with, this is more difficult, but one could still use a kind of small switched mode converter or even just a diode in series for the negative side (so loosing some 0.7 V) - just -0.6 V can be enough to make a current sink. In some cases the extra negative supply can help to reduce the drop out - so there is not that much lost by the diode.

While it is possible to use just an LM324 type OP, it might be a good idea to use a slightly better (faster and maybe more accurate) OPs especially for the CC mode loop in the non floating design. Often one starts with a kind of generic OPs in the design and later one decides which OPs would fit best.

[mikerj]

That has a significant flaw for a bench supply, look how much capacitance is on the output.  If you connect a circuit to the output expecting the current to be limited to the value set by the pot, then you will be disappointed.

[Kleinstein]

Many lab-supplies have a capacitance in the 500 µF range at the output. This is not ideal in some cases, but this what you have to expect. With less capacitance the regulator circuit needs to be rather fast or there will be significant voltage drop and overshoot on load transients. A fast regulator can get surprisingly tricky, as suddenly parasitic inductance of just a few cm of wire or an capacitor can get important.

One can build a regulator without an output capacitor, but this would be not a normal lab supply, but a rather special one with deficits at other points, like a high no load power dissipation and not so good regulation.

The output capacitance of the shown circuit (now in inital post) is not the main problem. The main problem is that current limiting is so slow, that chances are high that the circuit would not survive a dead short when at a higher voltage. So even without a physical capacitor at the output, there can be huge current spike. This potential current spike is a problem with most circuits based on a emitter follower output stage, and even worse with a source follower stage (which is not commonly used anyway). Chances are also high that the CC mode might oscillate under some conditions and it will show significant ripple.

A simple regulator with the shunt on the low side would about look like the attached files.
The OPs use in the simulation are slightly faster (1MHz) than the LM324 though. So more like an OPA170. The simulation shows a  10 mA -  1 A  - 10 mA transient test.

[alm]

Output capacitance is just something to be aware of, in my opinion. So do not connect a LED with the power supply set to 20 V, 10 mA. Something like 500 uF is typical for linear supplies with up to 1 A output current, but higher current supplies can easily be much higher. Yes, this energy gets dumped into the circuit in the case of a sudden short. A solution is to set the voltage limit to 0 V before connecting the load, so the output cap is discharged. A downprogrammer might also alleviate the problem.

Specialty high-speed power supplies without significant output capacitance, in addition to extra complexity, also tend to be more finicky about reactance on the output. They are not necessarily suitable as general-purpose power supply.

[ZeTeX]

So how slow is too slow?
Ultra fast PSU can be advantage and disadvantage, in terms of usability and the effort of designing correctly. some people like them, some not.
Slow PSU is pretty much horrible for everyone, the current limiting is useless.
what is the sweet spot? something that will not take too much theory to design and make it stable but not something that is too slow to be useable.
The circuit posted by Kleinstein takes about 142us to switch from CV to CC:

is that fast enough? is that slow enough? would most people be happy with that kind of speed? will the circuits under test that are connected to the PSU survive or be damaged in case of current-limiting? is it worth to go to the efforts to design a faster one for a beginner?

[T3sl4co1l]

That's not a current limit, that's an oscillator.

Tim

[debininja]

Everyone here is just shooting me down with their fancy shmancy, microvolts precision, super quiet PSU output requirements. As a hobbyist on a budget, and as a first time cc/cv PSU builder, I don't care about that sort of precision and accuracy (as long as it's better than the LM723). I just need something to work, and work decently, following the K.I.S.S. principle.

People expect too much from people without an extensive engineering background. The massive income that comes from this sort of profession tends to spoil many, and I think that's the trouble with an electronics forum. There are too many here with extremely precise equipment, costing thousands, or tens of thousands of dollars, who have become spoiled from the performance of their equipment.

Here's another one I found, and it seems to fit the bill for me perfectly.
According to one of the commenters:

Hi DIYFAN, I built your version 3 PSU (2nd schematic). I used the LM358 op-amp and it worked quite well. Output voltage can only go up to 27 volts (from 44 volts of unloaded rectified DC from the transformer) maybe because I used an 7809 regulator for the reference (I could not find a 6.2 Volt Zener from my junkbox). Current limiting was lightning fast to not burn a little green LED connected directly to the output terminals at full output voltage (27 volts) at minimum current limit. Really loved it. I have yet to find a case for it though. I hope i can put it together soon as this will be my first proper lab PSU.

I'll have to change out the reference voltage and gain setting resistors, but I think I'll go with this schematic (current sense is on the low side).

[ZeTeX]

Everyone here is just shooting me down with their fancy shmancy, microvolts precision, super quiet PSU output requirements. As a hobbyist on a budget, and as a first time cc/cv PSU builder, I don't care about that sort of precision and accuracy (as long as it's better than the LM723). I just need something to work, and work decently, following the K.I.S.S. principle.

People expect too much from people without an extensive engineering background. The massive income that comes from this sort of profession tends to spoil many, and I think that's the trouble with an electronics forum. There are too many here with extremely precise equipment, costing thousands, or tens of thousands of dollars, who have become spoiled from the performance of their equipment.

Here's another one I found, and it seems to fit the bill for me perfectly.
According to one of the commenters:

Hi DIYFAN, I built your version 3 PSU (2nd schematic). I used the LM358 op-amp and it worked quite well. Output voltage can only go up to 27 volts (from 44 volts of unloaded rectified DC from the transformer) maybe because I used an 7809 regulator for the reference (I could not find a 6.2 Volt Zener from my junkbox). Current limiting was lightning fast to not burn a little green LED connected directly to the output terminals at full output voltage (27 volts) at minimum current limit. Really loved it. I have yet to find a case for it though. I hope i can put it together soon as this will be my first proper lab PSU.

I'll have to change out the reference voltage and gain setting resistors, but I think I'll go with this schematic (current sense is on the low side).

I'm sorry but your response is childish.
are you trying to learn from this project? are you doing it just to get over it because you must?
nobody is going against you, and nobody throws microvolt precision at you. and to be real everybody likes Ferrari over fiat, even if they don't need to accelerate to 100km in 3 seconds every day.
Current limited PSUs are special in that to make sure then are stable and don't oscillate under any load that is not a pure resistor, they require you some knowledge of control theory. and it's complicated to be real, that's why you have people here helping you giving you a proper schematic or knowledge, because there are many designs on the web that are just bad and will not work you.


Nobody is spoiled, nobody has ultra expensive equipment just because they can, and nobody started to begin a hobbyist with those kinds of equipment.