OP Amp power supply question

[Frankentronics]

Greetings,

I would like to build the following power supply...



...which is described here http://www.joebrown.org.uk/wp/?p=7294.

I see that the OP Amp gets +8V from the 7808 and also gets -22V straight from the rectifier.

Questions:

1 - Does that mean that it will be pulling only -8V from the available -22V?

2 - Wouldn't it be better to just install another 7908 regulator and feed -8V to the OP Amp?

Thanks...

[StillTrying]

"1 - Does that mean that it will be pulling only -8V from the available -22V?"

? No, the op amp output can go lower because of the -22V supply.

"2 - Wouldn't it be better to just install another 7908 regulator and feed -8V to the OP Amp?"

The output of the opamp & 337ADJ pin has to go from about +0.5V to -14V, so supply rails of +8 and -22V would be about right.

U3 could be made a bit faster and stable with feedback components, rather than just a 10u on it's output.

[Frankentronics]

Thank you very much for your super fast reply. I'm glad I asked.

U3 could be made a bit faster and stable with feedback components, rather than just a 10u on it's output.

The recommended OP Amp is LF351.

Is this done by adding a resistor and cap in parallel from the output back to the +input?

Thanks...

[David Hess]

The forum ate my post but I kept a backup copy.  Why is this even necessary?

1 - Does that mean that it will be pulling only -8V from the available -22V?

Huh?  No, with the 78L08 positive voltage regulator, the operational amplifier will be operating from a regulated +8 volts and an unregulated -22 volts which is no problem.  Of course you need a 36 volt minimum regulator for this to work.

The +8 volt supply is really a little high and could be lowered.  The Tektronix PS503A power supply uses similarly asymmetrical voltages for its operational amplifiers but only 5 volts and 78L05s are much easier to find than 78L08s.  I have seen 3 volts used in some circuits.

2 - Wouldn't it be better to just install another 7908 regulator and feed -8V to the OP Amp?

The operational amplifier has to drive the adjust pin of the LM337 down to the -ve output voltage (plus 1.25 volts) which is considerably more negative than -8 volts.

The recommended OP Amp is LF351.

There are lots of parts which would work including the LF351, LF355, LF356, TL031, TL051, TL061, TL071, TL081, 741, 301A, etc.  Practically any 36 or higher volt part is suitable.  I have used the LT1007, a slightly more modern OP-27, in a similar circuit for low noise but that does not apply here.

Usually a slow operational amplifier should be used but the output resistance of the operational amplifier combined with C6 implements dominant pole compensation preventing oscillation.  In that respect, a 301A with suitable external compensation would be a much better choice but any part will work.

[Frankentronics]

There are lots of parts which would work including the LF351, LF355, LF356, TL031, TL051, TL061, TL071, TL081, 741, 301A, etc.  Practically any 36 or higher volt part is suitable.  I have used the LT1007, a slightly more modern OP-27, in a similar circuit for low noise but that does not apply here.

Usually a slow operational amplifier should be used but the output resistance of the operational amplifier combined with C6 implements dominant pole compensation preventing oscillation.  In that respect, a 301A with suitable external compensation would be a much better choice but any part will work.

Thank you for this information. I couldn't find a through hole LF351, so now you gave me plenty of choices. This list will be helpful.

I am now Googling what it mean "...with suitable external compensation" but it's a bit much for my brain to process as I don't have any real background in electronics. I think I understand that compensation makes it more stable and that it is done by connecting C6 in a feedback loop back to the positive input of the OP amp. Correct? And if so, would the value of C6 change depending on which OP amp is used?

Thanks...

[danadak]

A pretty good ap note on compensating V OpAmps -


http://www.ti.com/lit/an/sloa020a/sloa020a.pdf


Regards, Dana.

[David Hess]

Thank you for this information. I couldn't find a through hole LF351, so now you gave me plenty of choices. This list will be helpful.

It may be that they selected the LF351 because it is particularly suitable to the form of frequency compensation they used because it has a controlled output resistance.

I am now Googling what it mean "...with suitable external compensation" but it's a bit much for my brain to process as I don't have any real background in electronics. I think I understand that compensation makes it more stable and that it is done by connecting C6 in a feedback loop back to the positive input of the OP amp. Correct? And if so, would the value of C6 change depending on which OP amp is used?

Usually a much lower value of capacitance would be placed between the output and inverting input (1) and I am not sure why they did not do that here.  A larger capacitor from the output to ground achieves the same objective of making the feedback loop stable but at the expense of performance which was not needed in this case anyway because an integrated regulator was used as a pass element instead of a transistor.

The value of the capacitor between the output and inverting input does not change but slower operational amplifiers would not require the external feedback capacitor at all.  In that respect, the LF351 strikes me as an odd choice in this application.  Below is an example of the same type of tracking regulator circuit made with an LM301 with its standard 30pF compensation capacitor and no feedback capacitor which a faster operational amplifier would require.  A 741 operational amplifier would be a drop in replacement for the LM301 in this circuit.

(1) The value of this capacitance is inversely proportional to the parallel resistance of R2 and R3.

[Frankentronics]

First of all, than you both for this useful information.

A pretty good ap note on compensating V OpAmps -

http://www.ti.com/lit/an/sloa020a/sloa020a.pdf...

Thank you, Dana. I opened the PDF but as I'm sure you can understand it is going to take me time to study and understand all this. I have no formal training in electronics. I'm more of a broad range DIY enthusiast and as such I have limits. I also have a big priority in my life, which is raising a small child that is very demanding (and a lot of fun) and I haven't had much spare time in the past few years. But thanks again. I will try to study that document, as time permits.

...It may be that they selected the LF351 because it is particularly suitable to the form of frequency compensation they used because it has a controlled output resistance.

...Usually a much lower value of capacitance would be placed between the output and inverting input (1) and I am not sure why they did not do that here.  A larger capacitor from the output to ground achieves the same objective of making the feedback loop stable but at the expense of performance which was not needed in this case anyway because an integrated regulator was used as a pass element instead of a transistor.

The value of the capacitor between the output and inverting input does not change but slower operational amplifiers would not require the external feedback capacitor at all.  In that respect, the LF351 strikes me as an odd choice in this application.  Below is an example of the same type of tracking regulator circuit made with an LM301 with its standard 30pF compensation capacitor and no feedback capacitor which a faster operational amplifier would require.  A 741 operational amplifier would be a drop in replacement for the LM301 in this circuit.

(1) The value of this capacitance is inversely proportional to the parallel resistance of R2 and R3.

This is all very helpful. Thank you.

I think in my case it makes sense to buy 2 or 3 different IC's and experiment a bit. I am not designing a circuit to sell a product. This is just a project for me to take my mind off of all the daily obligations and have some fund DIY-ing. So, I'll do just that and experiment a bit. I'm sure I'll end up with a power supply that will be just fine for my future experimental needs. And I'm also using all the other parts from a salvage bin, so buying a few IC's won't break my bank.

Again, many thanks to all that took the time to share your knowledge. I'll order the IC's, start breadboarding and if I run into any issues I guess I'll have to ask for more help.

Thanks...

[StillTrying]

If you really wanted the outputs to go below +/- 1.25V, you could quite easily produce a -1.25V from the -22V to allow the outputs to go very close to +/- 0V.

[Frankentronics]

If you really wanted the outputs to go below +/- 1.25V, you could quite easily produce a -1.25V from the -22V to allow the outputs to go very close to +/- 0V.

Yes, I was thinking about that, too.

I saw another circuit, described here http://www.nutsvolts.com/magazine/article/a_test_bench_power_supply, that uses a zener.



I'm in the process of figuring out how to incorporate that trick in the circuit that I want to build and am also searching for the appropriate zener. But the Mouser part number #511-TS821A1Z listed on that page does not return any results at Mouser. When I search DigiKey for "Zener 1.25" I only find surface mount zeners with forward voltage 1.25V @ 200mA. However,  I thought I would have to use a zener with reverse voltage of 1.25V, but DigiKey does not have zeners with this spec.

So, I guess I don't really know what I should be looking for. Is it a zener with forward voltage 1.25 or reverse voltage 1.25?

Thanks...

[rdl]

Don't use a Zener, use a voltage reference such as the LM385.

[Frankentronics]

Don't use a Zener, use a voltage reference such as the LM385.

Oh, I see.

Would the AZ432AZTR-E1 be a better choice, as it has output voltage exactly 1.25V?

Thanks...

[Zero999]

Is a regulated power supply really necessary? Check the supply voltage rejection ratio of the op-amp you're using. It might be so high, a regulated supply is unnecessary, as long as large enough filter capacitors are used.

[rdl]

Don't use a Zener, use a voltage reference such as the LM385.

Oh, I see.

Would the AZ432AZTR-E1 be a better choice, as it has output voltage exactly 1.25V?

Thanks...

Having an exact output voltage is probably not that important. All that's being done is pulling the LM317 adjust pin down to zero by offsetting its internal reference voltage, and the LM317 reference voltage is not likely to be exactly 1.25 volts.

[StillTrying]

The easiest way to produce a stable -1.25V would be to use a low power LM337.

"Is a regulated power supply really necessary?"

The positive 8V/5V supply is just to save needing a 50V opamp.

[David Hess]

The easiest way to produce a stable -1.25V would be to use a low power LM337.

The various easy ways to generate a +/-1.25 volt reference are all about the same cost:

1. TO-220 LM317 and LM337
2. TO-92 LM317LZ and LM337LZ
3. LM385-1.2

More complex designs use a separate operational amplifier and reference to control the LM317 and LM337 regulators which are used as protected pass elements so +/-1.25 volt references are not required.

[Zero999]

The easiest way to produce a stable -1.25V would be to use a low power LM337.

"Is a regulated power supply really necessary?"

The positive 8V/5V supply is just to save needing a 50V opamp.

I think I misunderstood the purpose of this thread/circuit. I thought it was about a power supply for an op-amp circuit, rather than an op-amp based general purpose power supply.

Quite often a PSU for op-amps doesn't need a regulator, because the op-amps themselves reject changes in power supply voltage and stop it from changing the output.

As far as using an op-amp as a part of a regulated PSU is concerned: that's common too, although care needs to be taken with phase compensation, otherwise it will oscillated. In this case, C6 on the output of U3 might cause oscillation, as op-amps often oscillate when driving a capacitive load.

[Frankentronics]

I think I misunderstood the purpose of this thread/circuit. I thought it was about a power supply for an op-amp circuit, rather than an op-amp based general purpose power supply...

Actually, you were correct. I started this post by asking about the power supplying the OP Amp. Then the thread took a life of its own...

In any case, I guess there's many solutions for the reference voltage. Again, perhaps in my case it's just best to order 2 or 3 different components and have fun experimenting. For me, it's all about learning and experimenting, at this point.

Thanks...

[David Hess]

As far as using an op-amp as a part of a regulated PSU is concerned: that's common too, although care needs to be taken with phase compensation, otherwise it will oscillated. In this case, C6 on the output of U3 might cause oscillation, as op-amps often oscillate when driving a capacitive load.

I explained that in my earlier post.  The large value of C6 shunting the output of U3 *is* the frequency compensation in this circuit.  Without C6, the fast LF351 will oscillate because of the phase delay produced by the regulator.  The output resistance of U3 combined with the large value of C6 reduces the gain below 1 before the total feedback reaches 360 degrees.  This is a completely legitimate way to implement frequency compensation (dominant pole compensation) but it has the cost of reducing bandwidth severely which is acceptable in this case because the integrated regulator itself has such a low output impedance and low adjustment pin to output pin drift; after all, by itself it is a regulator with its own feedback loop.

The alternatives are to use a much smaller value of capacitance between the output of U3 and its inverting input or to use a slower operational amplifier so no external compensation is required.  These would be appropriate choices for a higher performance design.

[Zero999]

As far as using an op-amp as a part of a regulated PSU is concerned: that's common too, although care needs to be taken with phase compensation, otherwise it will oscillated. In this case, C6 on the output of U3 might cause oscillation, as op-amps often oscillate when driving a capacitive load.

I explained that in my earlier post.  The large value of C6 shunting the output of U3 *is* the frequency compensation in this circuit.  Without C6, the fast LF351 will oscillate because of the phase delay produced by the regulator.  The output resistance of U3 combined with the large value of C6 reduces the gain below 1 before the total feedback reaches 360 degrees.  This is a completely legitimate way to implement frequency compensation (dominant pole compensation) but it has the cost of reducing bandwidth severely which is acceptable in this case because the integrated regulator itself has such a low output impedance and low adjustment pin to output pin drift; after all, by itself it is a regulator with its own feedback loop.

The alternatives are to use a much smaller value of capacitance between the output of U3 and its inverting input or to use a slower operational amplifier so no external compensation is required.  These would be appropriate choices for a higher performance design.

Well with C6 being that high, it probably won't oscillate at all, as the bandwidth will be far too low.

I agree about using a slower op-amp. I've built a similar circuit to this before, with the µA741 and it was perfectly stable. I tried applying a load transient, at different supply/output voltages/currents and couldn't get it to oscillate.  The negative linear regulator acts as a follower, so doesn't add  any gain which would make it unstable. The delay may be a problem. I haven't done any tests to confirm whether it would oscillate, when used with a faster op-amp.

I think I misunderstood the purpose of this thread/circuit. I thought it was about a power supply for an op-amp circuit, rather than an op-amp based general purpose power supply...

Actually, you were correct. I started this post by asking about the power supplying the OP Amp. Then the thread took a life of its own...

In any case, I guess there's many solutions for the reference voltage. Again, perhaps in my case it's just best to order 2 or 3 different components and have fun experimenting. For me, it's all about learning and experimenting, at this point.

Thanks...

Please post the op-amp circuit you want to power. It it's something like a low gain, high level, audio amplifier, then it probably doesn't need a regulated supply. If the voltage doesn't need to be changed, then there's no need for any potentiometers. Select fixed resistor values, to give the appropriate output voltage or fixed linear regulators, such as the LM78(L)xx (positive) or LM79(L)xx (negative), where L is for low current if you don't need more than 100mA and xx is the output voltage, so LM7812 and LM7912 would give +/-12V at up to 1A and LM78L09 and LM79L09 would give +/-9V at up to 100mA.

[David Hess]

Well with C6 being that high, it probably won't oscillate at all, as the bandwidth will be far too low.

That is the intention.  Sometimes this type of compensation is used to present a low output impedance at high frequencies which the operational amplifier itself is not capable of.

I agree about using a slower op-amp. I've built a similar circuit to this before, with the µA741 and it was perfectly stable. I tried applying a load transient, at different supply/output voltages/currents and couldn't get it to oscillate.  The negative linear regulator acts as a follower, so doesn't add  any gain which would make it unstable. The delay may be a problem. I haven't done any tests to confirm whether it would oscillate, when used with a faster op-amp.

Negative linear integrated regulators almost always use NPN output stages so they actually have voltage gain and problematical stability.  That is why they *must* have a bulk output decoupling capacitor for stability and may have restrictions on ESR.  Low dropout regulators of all kinds have the same issue.  This may explain why the unusual frequency compensation method was used here; it would be easier to get working.

Please post the op-amp circuit you want to power. It it's something like a low gain, high level, audio amplifier, then it probably doesn't need a regulated supply. If the voltage doesn't need to be changed, then there's no need for any potentiometers. Select fixed resistor values, to give the appropriate output voltage or fixed linear regulators, such as the LM78(L)xx (positive) or LM79(L)xx (negative), where L is for low current if you don't need more than 100mA and xx is the output voltage, so LM7812 and LM7912 would give +/-12V at up to 1A and LM78L09 and LM79L09 would give +/-9V at up to 100mA.

There are also designs where the positive and negative integrated regulators are cross coupled to produce a variable tracking output supply.  Shown below is one example.

[Frankentronics]

Please post the op-amp circuit you want to power. It it's something like a low gain, high level, audio amplifier, then it probably doesn't need a regulated supply. If the voltage doesn't need to be changed, then there's no need for any potentiometers. Select fixed resistor values, to give the appropriate output voltage or fixed linear regulators, such as the LM78(L)xx (positive) or LM79(L)xx (negative), where L is for low current if you don't need more than 100mA and xx is the output voltage, so LM7812 and LM7912 would give +/-12V at up to 1A and LM78L09 and LM79L09 would give +/-9V at up to 100mA.

At this time I am only building a bench power supply. I Googled through many schematics and settled on the one I posted above, because it matched a very nice salvaged transformer that I have and want to use. I initially just wanted some clarification so I can order components from DigiKey.

I am new to electronics but I believe it makes difference what will be the intended use for a bench power supply.

In my case, I want to experiment a lot with audio amplifier circuits. When I Google audio amp schematics I find that some require a +/- power supply, so this is one reason I settled on this schematic. Of course, I will experiment with circuits that require different voltages, which is why I'd like to have the adjustable option.

I feel like I've already learned a lot more that I was expecting from this thread and I really appreciate all of you are taking the time to explain things to me.

Perhaps my next step should be to revise the above schematic, based on everything I've learned in the past few days, and post my revised schematic here. But I can't work fast, so I need a little time to do this. And I also don't have any software, so I'll have to draw by hand.

Thanks...

[Frankentronics]

OK, this took less time than I thought.

So, here is the schematic. I'm sure I made mistakes.



Thanks...

[David Hess]

That looks about right.  C8 should go between +5 volts and ground and does not need to be nearly that large; 10uF would be better.  C7 and D6 should go between the positive output and ground.  U3 usually requires a 30pF capacitor between pins 1 and 8 but might work without it in this case.

[StillTrying]

I think C3 should also go to GND rather than the -1.25V.

At 0V/0V output, unfortunately C3 and C6 get 1.25V in reverse, but I think most electrolytics won't mind that too much.

Connect the pot's wiper and open end together - slightly safer when the wiper becomes noisy.

[Zero999]

Here's the circuit I made a long time ago, while I was still a student. It doesn't have any compensation capacitor and it's perfectly stable. There are a couple of other things about the design which are marginal and I wouldn't built is exactly like this again, but frequency compensation isn't one of them.

[Frankentronics]

Thanks again.

I made some revisions (all shown in red).

...C8 should go between +5 volts and ground and does not need to be nearly that large; 10uF would be better.

I think you misread my handwriting. That was a 470n cap, not 470u. So, I left it 470n. That value is OK, right?

C7 and D6 should go between the positive output and ground.

I fixed that.

U3 usually requires a 30pF capacitor between pins 1 and 8 but might work without it in this case.

I made that revision, too. Could you please check if my pin numbers are correct? Thanks...

[David Hess]

Here's the circuit I made a long time ago, while I was still a student. It doesn't have any compensation capacitor and it's perfectly stable. There are a couple of other things about the design which are marginal and I wouldn't built is exactly like this again, but frequency compensation isn't one of them.

The 741 has the same relatively slow performance as a 301A with its standard 30pF compensation capacitor so further compensation is not needed in this circuit.  The precision version of this circuit might use a slow 308 or OP-07 also without further compensation.  The LF351 shown in the original circuit is much faster so requires the 10uF output capacitor or something else.  When I used the fast OP-27 or LT1007 in a precision low noise design, the external compensation became pretty complex.

Usually it is not really necessary for the output to go all the way down to zero volts; most loads are not going to draw much current at +/-1.25 volts.

[David Hess]

I think you misread my handwriting. That was a 470n cap, not 470u. So, I left it 470n. That value is OK, right?

Yes, that will be fine.

U3 usually requires a 30pF capacitor between pins 1 and 8 but might work without it in this case.

I made that revision, too. Could you please check if my pin numbers are correct? Thanks...

Those are correct; you can always check the LM301A datasheet.  The 30pF capacitor between pins 1 and 8 makes the LM301A duplicate the performance of a 741 which has this capacitor built in.

StillTrying brought up a good point about the voltage across C6 going to negative -1.25 volts.  In this circuit with the LM301A and its 30pF compensation capacitor, C6 should not even be needed.  If the negative side is unstable without C6, then the value of C11 can be increased to gain the same effect which is something that an operational amplifier with a built in compensation capacitor like the 741 cannot do.

I have a funny story about the 741 and 301A involving power supplies.  Tektronix used the 741 in their PS500 series power supplies but I have a very late PS503A which came stock from the factory with 301As replacing the 741s and the needed 30pF compensation capacitors soldered to the back of the board.  I have no idea why this was done and this change does not show up in any of the detailed service documentation they provided.  Maybe they ran out of 741s that day?

[Frankentronics]

Wow! I already got a lot more from this post than I expected. I should end up with a very decent power supply thanks to all your help.

I made yet another revised schematic.

1 - C6 is now dashed, in case a different OP Amp is used.

2 - I added fine and coarse output voltage adjustments, based on another schematic I've seen on the internet. Are those values correct?



Also, thanks for the PDF (it came in as I was posting my previous reply).

[StillTrying]

I wouldn't think course and fine pots were really worth the bother on a 0-15V supply.

I've often thought that for a 5k pot, you could use a dual 10k with the tracks in parallel to increase the stability and resolution, I bought a dual pot to test the theory a while ago, but never got a round tuit, as usual.

[David Hess]

I prefer fine and coarse single turn potentiometers to 10-turn potentiometers unless extra stability is required; wirewound 10-turn potentiometers have better absolute and ratiometric stability.  In most applications, just a coarse single turn potentiometer is sufficient.

[Frankentronics]

I've often thought that for a 5k pot, you could use a dual 10k with the tracks in parallel to increase the stability and resolution...

That sounds like a neat idea, also to keep in mind for any future reference.

...I bought a dual pot to test the theory a while ago, but never got a round tuit, as usual.

Sounds like you and I have something in common. I got quite a few project bins filled with parts I never got around assembling.

[Frankentronics]

First of all, I want to thank everyone for taking the time to reply to my questions. This was all very helpful.

I ordered the components and will be posting some photos when I put it all together. I'll be experimenting with different pots to see the effects, but I guess I'll settle on a 4.7K multi turn pot that I found at a very good price.

One last question (I think important) if you don't mind.

Should my output ground be connected to the same ground as the ground from the AC outlet? In other words, are the transformer center tap, the transformer core and body, the enclosure, the heat sinks and the output ground all connected?

Thanks...

[Richard Crowley]

Should my output ground be connected to the same ground as the ground from the AC outlet? In other words, are the transformer center tap, the transformer core and body, the enclosure,

Typically, no.  If you examine the "common practice" you will note that  the "ground"/0V node is NOT connected to mains green-wire safety ground.  It is typical to find the mains ground connected to a green terminal on the power supply front panel so that the user has the option to connect circuit ground to mains ground and ultimately to the crust of the planet if it is appropriate for the use-case.

the heat sinks and the output ground all connected?

NO! Pay very careful attention to where the heat-sink tab is connected in the regulator components! The data sheet shows that the LM317 tab is connected to the output node. So it should most certainly be isolated from ground.  And the LM337 tab is connected to the input node.  There are some fixed, positive voltage regulators whose heat-sink tabs are connected to ground, but most other regulator chips must have their heat sink tabs isolated from circuit ground.

[Zero999]

As mentioned above, it's better not to connect the 0V rail to mains ground.

As far as the heat sink is concerned: an insulating tab should be used, otherwise the two devices should be on separate heat sinks, which are not connected to anything else.

[Frankentronics]

Got it! Thank you both for taking the time to explain.