I have a list of 28 Opamps and I need your experience.

[Avelino Sampaio]

Hi!

This year I decided to invest $100.00 in Opamps. See my list in image 1.



Now I need your experience to define, in order of priority, the best ones for each project (image 2). I would like you to indicate me three per project. I will try to specify what I want to achieve in each project. Let's start with the PSU.

Floating PSU (image 3).

* Vmax=18v / Imax=2A / CV and CC
* CV with full priority, getting the best possible transient response.
* A protection transistor will be inserted in the project, to respond faster if the output is shorted.
* Less noise possible.

What are the top three Opamps on my list for this project? I would very much like you to justify the reasons for your choice, so that I can acquire better knowledge.

[MasterT]

Circuits is non functional, +-5V is not enough to control 19V

[ledtester]

The "ground" of the +/- 5V rails seems to be connected to the power supply output -- where it says "18V".

[MasterT]

If so, than +3V is virtually -15V - way below -5V power line.

[Avelino Sampaio]

Nesse caso, +3V é praticamente -15V - muito abaixo da linha de alimentação de -5V.

Thanks. I changed the design. I hope I got it right.

[Kleinstein]

The power supply part would likely need some extra compensation, unless one uses a pretty slow OP-amp (LM358 may be slow enough).

If precision is not top priority the RC4558 should be a reasonable pick - probably good enough. The NE5532 may also work but is kind of overkill. If good precision for the current is wanted, the OP27 or ADA4075  (about similar to the OP27) may be an option.
Noise wise (CV mode) the reference can be more a factor than the OP-amps. For the amps part the noise may matter, but usually the amps mode is more like lower priority.

[Avelino Sampaio]

The power supply part would likely need some extra compensation, unless one uses a pretty slow OP-amp (LM358 may be slow enough).

The drawing I presented is very incomplete, being only a small representation. Yes, there will be compensation. The use of the LM358, for being slow, would force me to increase the value of the output capacitor, to compensate in the transient response? And wouldn't that affect the CC answer?

If precision is not top priority the RC4558 should be a reasonable pick - probably good enough.

When you refer to accuracy, are you referring to Offset voltage? Is there anything else that influences accuracy?

The NE5532 may also work but is kind of overkill.

What does exaggerated mean? An SR=9v/us ? BW=10MB ?

I would like to know what is the line that divides the Opamps, considered slow, moderate and fast. For example, is a sr up to 2 considered slow for this project? Between 2 and 7 is moderate ?

If good precision for the current is wanted, the OP27 or ADA4075  (about similar to the OP27) may be an option.

Would you choose different Opamps in this project of mine? For example, a faster one for CV and a slower one for CC. Does this have any foundation?

Noise wise (CV mode) the reference can be more a factor than the OP-amps. For the amps part the noise may matter, but usually the amps mode is more like lower priority.

Understood.

[blackdog]

Bom dia Avelino Sampaio, 

I have searched extensively for opamp that can be properly applied to control loops for linear power supplies.

These are the specifications I have emphasized:
Minimum 12V Supply Voltage
Slew rate >10V/uSec
Wide phase margin > 75 degrees at normal load capacitance
Low noise
Fet inputs if posible

Candidates
No1
The best opamp found so far is the ADA4625-1.
This one gets up to about a 90 degree pase margin.
The dual version of this opamp has a clearly worse pase margin and common mode input capacity.

No2
Another of Analog Devices is the ADA4522-1 this also has a pretty good phase margin with a normal capecative load at the output, this is around 75 degrees.
This opamp has a number of other advantages which are:
Auto zero
Low bias currents up to 50 Celsius.
High supply voltage possible, up to 55V.
Low 1/f noise
What is unfortunate is this low -PSRR, but with good decoupling this can be controlled
Also, the phase margin is dependent on the supply voltage.
The phase margin in the datasheet of this opamp is given at 50pF load of the opamp output.
With a good design, this can be significantly less, which is why I indicate 75 Degrees.

Kind regards,
Bram

[David Hess]

Of the ones on your list that I am familiar with, the LM358 is perfectly adequate for this and its low bandwidth will simplify frequency compensation.

Using a faster operational amplifier will improve performance only up to the point where additional frequency compensation must be added to make it stable.  The output resistance from the transistors and output capacitance are usually what limits performance, and some designs minimize the output capacitance for better performance but this requires careful consideration of the frequency compensation.

HP made a lot of power supplies which use the same idea of floating low voltage rails for the control circuits.

Another of Analog Devices is the ADA4522-1

...

What is unfortunate is this low -PSRR, but with good decoupling this can be controlled

Isn't the PSRR 150dB or higher?  A lower negative PSRR is normal but common.

Usually the recovery time of a chopper stabilized amplifier would preclude it from this application, but the ADA4522-1 recovers pretty fast.  Some recover extremely slowly.

[Avelino Sampaio]

Bram:

thank you very much for the nominations but it has to be a nomination for my list. I do not intend to invest in other opamps.

A curiosity - in your NA-01 project you used the ADA4077 (sr=1.2v/Us , Bw=4M , Margin of 55%). A modest SR, compared to its indication. This makes me very confused.

David Hess:

OP27, nominated by Kleinstein, seems to be a good candidate as well. It already has a high compensation (image 1) and a good phase margin (image 2). (sr=2.8v and BW=8M). What do you think ?

[David Hess]

OP27, nominated by Kleinstein, seems to be a good candidate as well. It already has a high compensation (image 1) and a good phase margin (image 2). (sr=2.8v and BW=8M). What do you think ?

I used an LT1007, which was Linear Technology's version of an OP27, in my last power supply for low noise, but it required extra frequency compensation for good transient response.  In retrospect a 1 MHz LT1001 (OP07 equivalent), or an even slower part, would have been just as good or maybe better.  The extra frequency compensation needed by a faster part lowers the bandwidth and slew rate leaving performance mostly back where it would have been with a slower part.

Most precision operational amplifiers, including the OP07, OP27, LT1001, and LT1007, have a limited input differential voltage range which can be a problem because the current or voltage error amplifier which is not currently active will have a voltage applied across its input.  The current into the inputs needs to be limited to prevent damage, and the effective short across the inputs needs to not affect circuit operation.

[blackdog]

Hi Avelino Sampaio,

Power supplies look easy but are definitely not.

The schematic you're talking about dates back at least 10 years, when those fancy modern opamps weren't available.

The ADA4077 is a beautiful opamp and the 1.2V/uSec is certainly fast enough for a good LAB power supply.
You do need a fast and good Power section if you use the ADA4077 as shown in my schematic.

The reason I started experimenting with the ADA4077 is because in the first datasheet that came out a much wider phase margin was mentioned...
But of course I could not really stick to that, because on that datasheet it said "provisional".
In the end, the fast Power section showed in my schematic gives a good performance.

This just as I indicated to you, that this also with an LM324 opamp posible with a good fast power section. :-)
It doesn't always make sense to figure everything out optimally, you also have to take your application into account.
I got the nicest and most wideband low Ri of my design with 2x the 2SA1943 + 2SC2911 and then with two low inductance resistors for the current sense resistors(0.1 Ohm), short wiring and tuned output network direct over output connectors of the power supply.

For an impression of how good that power section can be, below is a picture of the impedance measurement.
"F" is the double compound power section, with 2x 2SA1943 and de filter in de picture.


The output impedance is certainly a magnitude better than I have been able to find from other LAB power supplies.
However, very good types have come out in recent years from some IC manufacturers with very low noise and wideband low Ri.
But of course those are not LAB power supplies.

In a private email to Avelino Sampaio I had already explained that that very nice low Ri that is possible with my design, disappears like snow in the sun due to long and/or bad cabling!
Why then still design it so that the Ri is very low....
You can as example then e.g. connect multiple loads to the power supply without them affecting each other.

So it really depends on your application and how well you are going to make it.

Not to mention, it also depends on your knowledge and measurement equipment and measurement setup how well you're going to get it.
The optimal result is pretty hard to get right.
You can compare it a bit to building a good 10V Voltage Reference, pretty simple in schematic, but oh so hard to keep it within 1PPM between 20 and 30 Celsius and load variations.

Some tips for the builders of linear power supplies.
2N3055 type power transistors are too slow.
Most old types Darlingtons are too slow.

These above type transistors cause the phase margin to become too small for a relative fast responding current and voltage loop.
One then often solves this by e.g. making the currrent loop much too slow, don't do that!
Use good and not China junk power transistors.
And certainly they can make good components in China, but through ebay and aliexpress this is almost impossible to find them.

Use good Motorola audio transistors (if you can still get them original) like the MJ21193 - PNP MJ21194 - NPN series.
These have a Ft of around 7-Mhz.
This makes the internal cabling in the power supply less critical for oscilations.

The 2SC5200/2SA1943 are of course also good(Fairchild and On Semi make them with differetnt parnumbers),
but because of their high Ft they can give problems, usually this can be solved by a capacitor at the input of the power section that compensates the impedance of the longer wiring to the buffer capacitor.

Most of the Sanken power transistor you can also use.
If you can order at Mouser or another company, look at the NJL1302DG and th NJL3281DG these ar "Thermaltrek" power transistors from ON-Semi.

These transistors have an extra diode in the housing intended for temperature sensing of audio amplifiers.
For power supplies you can use them for driving the fan control.

Short side jump
I use these transistors (four pieces) to heat an oven and the diodes drive the opamp for the temperature control, which works very well!

Conclusion
Especially for beginners, it is better not to try to get the most out of it.
Use medium speed power transistors and as I have already shown you can make a very good power supply with a NE5534A (YES! the "A" type) that is low noise and fast enough to respond.
With the advantage that you can then also compensate the opamp as needed.
I use at least five different types of power supplies, one sites fits all is not really possible.

When building a linear power supply, you can learn a lot.
Remember that every piece of wire has resistance and inductance.
So also on the wires of the capacitors, or the wires that are e.g. on induction low resistors.
You will notice this for example when you start connecting the sens wires and the network that comes over the terminals of the power supply.

This is not of course incomplete, but keeping it reasonably legible is for is hard enough for this "Dyslexic Monky" <= Thats Me! *Grin*

Kind regards,
Bram

[David Hess]

Some tips for the builders of linear power supplies.
2N3055 type power transistors are too slow.
Most old types Darlingtons are too slow.

It depends on the application.  For a power supply with 50+ microfarads of aluminum electrolytic output capacitance per amp, they are fast enough.  There are a lot of regulated power supplies built with transistors slower than the modern 2N3055.

Use good Motorola audio transistors (if you can still get them original) like the MJ21193 - PNP MJ21194 - NPN series.
These have a Ft of around 7-Mhz.
This makes the internal cabling in the power supply less critical for oscilations.

Modern audio power transistors are a great choice.  The original ring emitter transistors like the DH44/DH45 series work well for even higher performance but are faster so more attention has to be paid to local oscillations.

[blackdog]

Hi  David Hess,

Psttt, don't tell anyone, but one of my treasured power supplies here on my workbench (Dutch DELTA D030-1) has 2N3055 transistors.... 

The DH44/DH45 series transistors are nicely fast, but also less robust, I did tests with four of them and they didn't stay alive a few times in my design.
When working out the schematic even better, they will probably be good to use.

But as you point out, the speed of these transistors makes it difficult to keep them HF stable.

Kind regards,
Bram

[Avelino Sampaio]

I used an LT1007, which was Linear Technology's version of an OP27, in my last power supply for low noise, but it required extra frequency compensation for good transient response.  In retrospect a 1 MHz LT1001 (OP07 equivalent), or an even slower part, would have been just as good or maybe better.  The extra frequency compensation needed by a faster part lowers the bandwidth and slew rate leaving performance mostly back where it would have been with a slower part.

It was very enlightening. Thanks

Use medium speed power transistors and as I have already shown you can make a very good power supply with a NE5534A (YES! the "A" type) that is low noise and fast enough to respond.
With the advantage that you can then also compensate the opamp as needed.

I don't have the "A". And as for the ADA4075, what do you think? Fast, accurate and has a very interesting phase margin.

https://datasheet.lcsc.com/lcsc/1809171712_Analog-Devices-ADA4075-2ARZ-R7_C143913.pdf

[David Hess]

Psttt, don't tell anyone, but one of my treasured power supplies here on my workbench (Dutch DELTA D030-1) has 2N3055 transistors.... 

Several power supplies that I have rebuilt use 2N3771 (40V, 30A, 150W) power transistors which only have a specified minimum Ft of 0.2 MHz instead of the 2.0 MHz of a typical 2N3055.  My kit variable power supply on my bench is a 723/2N3055.

The DH44/DH45 series transistors are nicely fast, but also less robust, I did tests with four of them and they didn't stay alive a few times in my design.

Their lower secondary breakdown safe-operating-area definitely needs to be taken into account.

But as you point out, the speed of these transistors makes it difficult to keep them HF stable.

I would say less forgiving rather than difficult.  We get away with a lot with big slow bipolar power transistors.

[Avelino Sampaio]

Hi!

Well, I've completed my list of opamps, candidates for use in linear PSU projects.

Here's a video I found on YT, explaining in a simple way, the importance of maintaining a phase margin and a gain margin, to achieve good stability.

[Avelino Sampaio]

HI

It is time to list the possible candidates for the design of an electronic load, with the following specifications:

* Powered by a 9v battery
* Maximum voltage over the Opamp of 5v.
* The mosfet used will be the IRLZ44N
* Imax = 5A
* Pmax = 40w
* There will be power and current monitoring, using a PIC
* Maximum DUT voltage of 30v
* 500Hz input from my generator, for transient testing.

Below is a reduced sketch of the circuit I intend to use.

Which opamps from my list would you use for this project and for what reasons?

[blackdog]

Hi Avelino Sampaio, 

I have some comments on your Active Load circuit.
After studying your circuit, I preliminary come to the conclusion that the OPA364 is the best opamp for driving the IRLZ44.

The IRLZ44 is a Logic level MOSFet and these types have a very large Gate capacitance and your chosen type is that 3300pF.
Da is not an easy impedance to drive, especially for many opamps that work with low voltages and are RR I/O.
The OP188 and 189 are not very useful because of this.

But the OPA364 seems to be specially made for this kind of applications.
It can drive large capacities without generating.
Those large capacities will of course give more overshoot if you would not configure the circuit properly.
Your Opamp in your schematic already has a 100 Ohm resistor at the gate, that is at least a good starting point.

Some things that require attention.

You are using a measuring resistor of 0.1 Ohm, this requires careful design of the circuit board, this with respect to the copper resistance and its inductances.

Second point, the OPA364 is not well specified regarding DC offset voltage, which can be up to 2.5mV! for the types with the worst specifications.
You can inject a small current for the offset into the input through a properly scaled resistor and trimpotmeter.

Because in your circuit there is 100mV across your sense resistor at 1-Ampere.
1mA gives 0.1mV across the sense resistor.
Only by the offset of your opamp can you in the low values say 24mA wrong.

Bias compensation is not necessary for the OP364 because the bias currents are quite low.
So choose the feedback circuit then a little higher inpedances, make the 1K resistor 4K7 and the 1nF 220pF as the starting value with which you will test.
That makes it easier for the opamp, then has some more control current available for driving the gate impedance.

Your modulation way is a nice find with the PC817B.
But pay attention to the value in the datasheet, I don't know how fast you want the modulation to be.
The switching times depend on the impedance that will drive the output of the opto.

One more note about the modulation, it is On/Off in your circuit.
When I use one of my Active Loads it is almost always 10% minimum to 100%.
This can then still be done in your circuit by using an external minimum load.
But the opamp and Mosfet are not always fast enough to start quickly from "0", this is a point to consider.

The IRLZ44 and similar MOSFets have a side effect that ultimately the variation of the voltage at the gate is not very large
and the opamp output therefore does not have to make a large swing, except when switched off completely.

An impression below what kind of puls I usually use, with an Active Load,


.
Now when I want to test with fast pulses on a power supply output, I use my Jim Williams Dynamic Load
This has a bandwidth of a few MHz.
Also look where I am measuring, note the scope probe, it is connected to the point where the sense wires are connected to the connector.
And I do this without wiring, because even when twisting the wiring, the induction is still too high.
I want to test the circuit and not the wiring, that's why I measure on the sense connections at the back of the connectors.

For larger currents, it is better to use a piece of RG58 coax to measure, which then goes to your scope input.
Keep the coax perpendicular to the wires tha has high currents, otherwise you get coupling between the power supply wiring and the coax!


Kind regards,
Bram

[Avelino Sampaio]

Bram

I was quite enthusiastic about your statement and so I decided to put together a quick rehearsal on a breadboard. I started with the MCP6002 and the results seem to be very positive.

Analyzing the graph on the left, in image 1, I could see that the worst scenario is due to a low Vds. So I set the DUT to 3v and reduced the integrator capacitor to 330p. The result is that it started to oscillate from 1A load. I raised the capacitor value to 560p and it started to oscillate from 3A of load. I returned with the 1nF capacitor and was able to reach 5A without any oscillation. I tested it with higher DUT voltages and everything remained very stable.

I performed transient tests from two sources. The first test was on my DIY power supply, designed with the LM723 and a sziklai output (bd139 +2SA1943). I tested it with the short cables I have and the result was that the phase margin seems to be around 45º. The second test was on a switched power supply and the result points to a margin above 30º.

As for the test with the OPA364, I'll have to postpone it, because I bought it with the SOT23 package and I don't currently have the appropriate socket for the test on the breadboard. I'm going to test it today with the OPA376 and I'll bring the results.

[Avelino Sampaio]

Bram

I performed several tests to verify the stability, messing with different values of the integrator capacitor and the feedback resistor. The OPA376 proved to be a great candidate for my low voltage project. The transient result, from the switched source, was the same as the one I got with the MCP6002, see in the images.

[Avelino Sampaio]

Bram

I managed to test with OPA364 and the result was also very good. With that I complete my table, with the options for my future project of an Electronic Load.

[Avelino Sampaio]

Hey

Images 1, 2 and 3 - I have a project done for a long time, with the objective of carrying out comparative tests with my multimeters. My budget was very limited and I made it with very common components. In order to achieve better Vref stability, I raised the temperature of the LM723, stabilizing its casing at 40ºc. I used current sources for the best possible stability for the LM336. It was a simple project but it works perfectly.

Image 4 - I intend to build a new project, expanding its functionalities. Below is an outline of what I intend to build and I would like your experience here, to choose the opamps from my list that best suit the project (U1 and U2).

[Avelino Sampaio]

Hi

After researching a lot here on the forum, I conclude that the good opamp for this project of mine will be OPA188. And it costs half the price of the OPA277. I have doubts about using INA157 in this project and whether it will work in single source.

The precision requirements of my project will be modest. My Brymen BM869S will serve as a parameter.

[Avelino Sampaio]

Hi

To clear my doubt about the operation of the INA157 in simple font, I performed a quick test and it worked well..

1 - Output INA157
2 - Output REF5050
3 - Output OPA188
4 - List update