Good opamp for that1646

[abbey road d enfer]

I'm also trying to find more information about correct bypassing caps. Some people suggest only a cap between the rails, some other suggest one cap to ground for each rail and some other suggest both options. Go figure...

It's been demonstrated often enough that rail-to-rail bypassing only is not adequate.
Rail to ground should be the primary mode of bypassing.
There are two factors in bypassing: stability, which is usually covered by using the typical 0.1uF ceramic caps, and rail sagging, which depends on how power is distributed to the rails.
When the ouput of the regulators is directly distributed to the rails, sagging is minimal, but when power is distributed via low-value resistors (usually 10-47 ohms), it is essential to place large value (electrolytics) close to the opamps that have to deliver significant current. the common point of these bypass caps should be grounded as close as possible to the load return point.

 

[warpie]

It's been demonstrated often enough that rail-to-rail bypassing only is not adequate.
Rail to ground should be the primary mode of bypassing.
There are two factors in bypassing: stability, which is usually covered by using the typical 0.1uF ceramic caps, and rail sagging, which depends on how power is distributed to the rails.
When the ouput of the regulators is directly distributed to the rails, sagging is minimal, but when power is distributed via low-value resistors (usually 10-47 ohms), it is essential to place large value (electrolytics) close to the opamps that have to deliver significant current. the common point of these bypass caps should be grounded as close as possible to the load return point.

Thank you abbey. That makes sense.

This is what got me confused (from Doug's book).

The 5532- and 5534-type opamps require adequate supply-decoupling if they are to remain stable; otherwise they appear to be subject to some sort of internal oscillation that degrades linearity without being visible on a normal oscilloscope. The essential requirement is that the +ve and -ve rails should be decoupled with a 100 nF capacitor between them, at a distance of not more than a few millimetres from the opamp; normally, one such capacitor is fitted per package as close to it as possible. It is not necessary and often not desirable to have two capacitors going to ground; every capacitor between a supply rail and ground carries the risk of injecting rail noise into the ground.

 

[Newmarket]

OK, so you also suggest leaving R2, C1 out. Interesting. As I said I've seen it it many circuits (including some Neve) and that's why I used them.

Anyway, I'll give a try on the breadboard and probably come back with more questions

I'm also trying to find more information about correct bypassing caps. Some people suggest only a cap between the rails, some other suggest one cap to ground for each rail and some other suggest both options. Go figure...

BTW, thank you all for your help! Much appreciated.

Well different designers have different ideas about what is needed in a particular application. In detail, I wouldn't use it in your circuit. But if I were laying out a pcb and there was room I would have it on the schematic and have a footprint for it on the pcb. You can then either Zero-Ohm link it or fit a resistor as you see fit. And I would likely also allow for a resistor from - to 0V to allow easy conversion into a non-inverting amplifier. So I'd also allow for the cap' across the feedback resistor and also one across the resistor from - to 0V. And if I was fairly certain that it would mainly be used as a buffer I'd also put a track directly from output to - in an accessible position. So by default it would just be a unity gain buffer but if I wanted to use any of the optional components I'd be able to cut that track.

Interesting question re decoupling caps. Douglas self is adamant that 100n rail to rail is preferred:

"The 5532 and 5534 type op-amps require adequate supply decoupling if they are to remain stable, otherwise they appear to be subject to some sort of internal oscillation that degrades linearity without being visible on a normal oscilloscope. The essential requirement is that the positive and negative rails should be decoupled with a 100 nF capacitor between them, at a distance of not more than a few millimeters from the op-amp; normally one such capacitor is fitted per package as close to it as possible.

It is not necessary, and often not desirable, to have two capacitors going to ground; every capacitor between a supply rail and ground carries the risk of injecting rail noise into the ground."

but I've read what Abbey has posted here (and previously) on this and that makes sense to me from a load/voltage distribution point of view.
I'd say allow for both options - esp as I think you said this was basically for educational purposes

 

[Newmarket]

Thank you abbey. That makes sense.

This is what got me confused (from Doug's book).

The 5532- and 5534-type opamps require adequate supply-decoupling if they are to remain stable; otherwise they appear to be subject to some sort of internal oscillation that degrades linearity without being visible on a normal oscilloscope. The essential requirement is that the +ve and -ve rails should be decoupled with a 100 nF capacitor between them, at a distance of not more than a few millimetres from the opamp; normally, one such capacitor is fitted per package as close to it as possible. It is not necessary and often not desirable to have two capacitors going to ground; every capacitor between a supply rail and ground carries the risk of injecting rail noise into the ground.

Ah - we crossed posts using the same extract there

 

[abbey road d enfer]

This is what got me confused (from Doug's book).

The 5532- and 5534-type opamps require adequate supply-decoupling if they are to remain stable; otherwise they appear to be subject to some sort of internal oscillation that degrades linearity without being visible on a normal oscilloscope. The essential requirement is that the +ve and -ve rails should be decoupled with a 100 nF capacitor between them, at a distance of not more than a few millimetres from the opamp; normally, one such capacitor is fitted per package as close to it as possible. It is not necessary and often not desirable to have two capacitors going to ground; every capacitor between a supply rail and ground carries the risk of injecting rail noise into the ground.

I tend to disagree with Doug in that respect. It's true that either option (one rail-to-rail or two rail-to-ground) is necessary for taming tendency to oscillation, not only with 5532/34, but also with many others opamps, particularly the faster-than-light ones.
But where I differ is regarding this: "every capacitor between a supply rail and ground carries the risk of injecting rail noise into the ground." Rail noise doesn't come out of the blue. It comes from stages pumping current that then goes throuh a load and returns to ground.
If the load current is returned exactly where the bypass caps are grounded, the current loop is contained and Kirchoff makes sure there is no resulting voltage.
Having the return current returning anywhere else than the bypass ground is bound to develop parasitic voltages between "grounds", which results in longitudinal noise.
Using rail-to-rail decoupling places the point where the opamps draw current from at the PSU ingress, and the load current return somewhere else, which results in parasitics.
QED.

 

[JohnRoberts]

Adding an extra cap rail to rail shouldn't hurt, but rail to ground seems pretty necessary.

+1 to what Abbey said.

JR

 

[Bo Deadly]

Interesting question re decoupling caps. Douglas self is adamant that 100n rail to rail is preferred:

"The 5532 and 5534 type op-amps require adequate supply decoupling if they are to remain stable, otherwise they appear to be subject to some sort of internal oscillation that degrades linearity without being visible on a normal oscilloscope. The essential requirement is that the positive and negative rails should be decoupled with a 100 nF capacitor between them, at a distance of not more than a few millimeters from the op-amp; normally one such capacitor is fitted per package as close to it as possible.

He seems to be making an observation about the 5532 and 5534 specifically having an atypical oscillation issue. He is not describing a method for bypassing op amps in general. The only way to really judge when and where bypass caps should be used is to understand the specific layout and circuit. For example, if there is no output going in to / out of local 0V, then you will not have significant currents that must go to a distant bypass cap. In this case, you might get away without bypass caps and adding them does potentially provide a path for coupling noise. But a lot of modern op amps are the high speed variety and usually there is some kind of connection between the output and 0V even if it is just a small cap. At high enough frequency, maybe parasitic capacitance to 0V could be enough to cause instability. So the general DIY rule these days is to just bypass caps between each rail to 0V. But in between these two scenarios, you might find, as D. Self seems to have, that a rail-to-rail cap actually resolves some instability about that specific amp without rail-to-0V caps.

 

[warpie]

OK, so I decided to add the option to be able to trim the level by approximatey +/- 1dB.

Does this look correct (i.e resistor values, cap values, etc...)?

Also, since it's not a unity gain buffer anymore the offset at the 5534 output will be higher. Is a DC blocking cap needed at the input of the 1646 in this case?

Finally, I'm a bit concern that the 2k trimmer (at the feedback) might introduce some noise. I think I had a similar problem ages ago but can't remember very well.

Once again, thank you all for your help!

 

[abbey road d enfer]

/*OK, so I decided to add the option to be able to trim the level by approximatey +/- 1dB.

Does this look correct (i.e resistor values, cap values, etc...)?

Yes. I just wonder why you power the 5534 for +/*15 and the 1646 from +/-18...

Also, since it's not a unity gain buffer anymore the offset at the 5534 output will be higher.

Will it?

Is a DC blocking cap needed at the input of the 1646 in this case?

No.

Finally, I'm a bit concern that the 2k trimmer (at the feedback) might introduce some noise. I think I had a similar problem ages ago but can't remember very well.

A trimmer should not create extra noise, except if it has accumulated dirt, and then it can be cleaned up.

 

[warpie]

Yes. I just wonder why you power the 5534 for +/*15 and the 1646 from +/-18...

Ideally I'd like to power both chips with +/-18V but I'm afraid that the 5534 will run too hot and shorten its life.

OK, so it looks good. Cool.

 

[Newmarket]

He seems to be making an observation about the 5532 and 5534 specifically having an atypical oscillation issue. He is not describing a method for bypassing op amps in general.

Yeah - I was commenting in the context of 5534 as that is what the OP was proposing to use. It's not quite clear whether Self is saying anything about use with other opamps here but focussing this extract on 5532 / 5534 as it is / has been so widely used - especially by himself.

The only way to really judge when and where bypass caps should be used is to understand the specific layout and circuit. For example, if the output is to local 0V, then you will not have significant currents that must go to a distant bypass cap. In this case, you might get away without bypass caps and adding them does potentially provide a path for coupling noise.

I say let's not make life more complicated than necessary Just always have bypass caps
Agree that the criticality depends on the physical system / layout etc. A small interface/balancing box is very different to even a modestly sized desk.

 

[Newmarket]

Ideally I'd like to power both chips with +/-18V but I'm afraid that the 5534 will run too hot and shorten its life.

OK, so it looks good. Cool.

Just to note that the 'Trim' is 0 - +2.2dB relative to th input signal. But yes its '+/-1dB' relative to mid point of the trimmer pot.
Also (and I know it jas been stated before) the THAT output chip will give you a +6dB gain into either a balanced or unbalanced output.
Just to make sure you're aware.
On the voltages above - is there any real advantage in the THAT chip at +/-18V rather than +/-15V (I know the datasheet specs' are at +/-18V). Personally I'd probably go with +/-15V all round or preferably use LM317/337 as required to configure the rails to +/-17V.

 

[FIX]

When ever I run any op-amp and unity, I use a 10K or 20K from out to inv in, and a small cap if I need unity never with the out to inv. I do this because the op-amp sounds a lot better. As far as ruining the "low freq CMMR" I wouldn't worry about that too much unless you are in a hostile environment and need it. You are running in unbalanced anyway. Since the 1646 has +6 dB of gain, if you need to drop the input, use input pads, but never put any R between the amp and the 1646. If you want to make the schematic even simpler, use a 1256 as the input amp. You won't need any external parts except .1u on the power rails. Make sure you do that right next to the 1646, as they sing like birds if you don't. (I made that mistake on an ADR console I built for TODD-AO) You can play with the 1256 resistor configurations to get -6, unity or +6 by flipping them around, balanced or unbalanced. I use them everywhere and they sound fantastic.

 

[warpie]

Just to note that the 'Trim' is 0 - +2.2dB relative to th input signal. But yes its '+/-1dB' relative to mid point of the trimmer pot.
Also (and I know it jas been stated before) the THAT output chip will give you a +6dB gain into either a balanced or unbalanced output.
Just to make sure you're aware.
On the voltages above - is there any real advantage in the THAT chip at +/-18V rather than +/-15V (I know the datasheet specs' are at +/-18V). Personally I'd probably go with +/-15V all round or preferably use LM317/337 as required to configure the rails to +/-17V.

Is there any better solution for +/-1dB trimming?

The circuit is powered by +/-28V so I was thinking of crudely using a resistor to drop the value down to +/-18V instead of using separate regulators. Is this very bad practise? So, I thought that dropping the voltage by 10V is better than dropping it by 13V (from +/-28V down to +/-15)

I realise that the 1646 has 6dB gain but I still need to be able to adjust the exact value by maximum +/-1dB.

When ever I run any op-amp and unity, I use a 10K or 20K from out to inv in, and a small cap if I need unity never with the out to inv. I do this because the op-amp sounds a lot better. As far as ruining the "low freq CMMR" I wouldn't worry about that too much unless you are in a hostile environment and need it. You are running in unbalanced anyway. Since the 1646 has +6 dB of gain, if you need to drop the input, use input pads, but never put any R between the amp and the 1646. If you want to make the schematic even simpler, use a 1256 as the input amp. You won't need any external parts except .1u on the power rails. Make sure you do that right next to the 1646, as they sing like birds if you don't. (I made that mistake on an ADR console I built for TODD-AO) You can play with the 1256 resistor configurations to get -6, unity or +6 by flipping them around, balanced or unbalanced. I use them everywhere and they sound fantastic.

Thanks for chiming in Paul. So, are you suggesting changing the trimmer to 10k and R3 to 33k or thereabouts?

and a small cap if I need unity never with the out to inv.

Not sure I understand this

Isn't it a bit overkill using a 1256 in this scenario? What would be the benefit?

 

[abbey road d enfer]

Just to note that the 'Trim' is 0 - +2.2dB relative to th input signal. But yes its '+/-1dB' relative to mid point of the trimmer pot.

The input signal is attenuated by 1dB, so that makes the range -1/+1.

 

[Ike Zimbel]

Ideally I'd like to power both chips with +/-18V but I'm afraid that the 5534 will run too hot and shorten its life.

That is absolutely not a concern. 5534 will be totally fine at +/-18v.

 

[JohnRoberts]

When ever I run any op-amp and unity, I use a 10K or 20K from out to inv in, and a small cap if I need unity never with the out to inv. I do this because the op-amp sounds a lot better. As far as ruining the "low freq CMMR" I wouldn't worry about that too much unless you are in a hostile environment and need it. You are running in unbalanced anyway. Since the 1646 has +6 dB of gain, if you need to drop the input, use input pads, but never put any R between the amp and the 1646. If you want to make the schematic even simpler, use a 1256 as the input amp. You won't need any external parts except .1u on the power rails. Make sure you do that right next to the 1646, as they sing like birds if you don't. (I made that mistake on an ADR console I built for TODD-AO) You can play with the 1256 resistor configurations to get -6, unity or +6 by flipping them around, balanced or unbalanced. I use them everywhere and they sound fantastic.

I don't know about sounds better but there are subtle stability concerns related to stray pin capacitance etc.

For example an op amp that is not unity gain stable and should not tolerate a feedback capacitor (100% NF) can actually have stability margin improved by a adding a small (several pF) feedback cap. Imagine a small C to ground at the - input forming an RC network with the feedback resistor, this RC introduce phase lag detracting from stability. A small feedback capacitor adds some phase lead to cancel out that phase lag. In the example of 5534 you want the two capacitors to maintain the 3:1 noise gain ratio (resulting in a very tiny feedback cap). Low resistance feedback networks will tune these poles well above the audio band.

This is likely TMI so feel free to disregard.

JR

 

[Newmarket]

I don't know about sounds better but there are subtle stability concerns related to stray pin capacitance etc.

For example an op amp that is not unity gain stable and should not tolerate a feedback capacitor (100% NF) can actually have stability margin improved by a adding a small (several pF) feedback cap. Imagine a small C to ground at the - input forming an RC network with the feedback resistor, this RC introduce phase lag detracting from stability. A small feedback capacitor adds some phase lead to cancel out that phase lag. In the example of 5534 you want the two capacitors to maintain the 3:1 noise gain ratio (resulting in a very tiny feedback cap). Low resistance feedback networks will tune these poles well above the audio band.

This is likely TMI so feel free to disregard.

JR

Sounds (errr...not literally ) quite marginal to me. Would be good to have some schematic to avoid misunderstanding what you mean.
But there is some confusion in the whole (interesting) discussion over whether we are talking generally or specifically wrt *5534 OpAmp.
And while I don't know what "sounds better" means objectively, I will note that technically some OpAmps do 'require' a resistance in the feedback path of a non-inverting unity gain buffer. LT1028 comes to mind. If I have the wrong device there I'll advance the mitigation that, in the words of a popular comedian - "I've done no research". Not actually true but I'm relying on memory

 

[warpie]

Would be good to have some schematic to avoid misunderstanding what you mean.

Yes that would really help. I've read the post something like 3-4 times and I'm still not sure if I understood.

 

[JohnRoberts]

Sounds (errr...not literally ) quite marginal to me. Would be good to have some schematic to avoid misunderstanding what you mean.
But there is some confusion in the whole (interesting) discussion over whether we are talking generally or specifically wrt *5534 OpAmp.
And while I don't know what "sounds better" means objectively, I will note that technically some OpAmps do 'require' a resistance in the feedback path of a non-inverting unity gain buffer. LT1028 comes to mind. If I have the wrong device there I'll advance the mitigation that, in the words of a popular comedian - "I've done no research". Not actually true but I'm relying on memory

Don't hold your breath for a schematic. As I offered this is probably TMI.

It is the nature of web discussions for threads to veer and ramble.

Back several decades ago when op amps were lower performance we routinely tried to squeeze out the last drop of performance with trickery... don't open the can of worms about alternate compensation schemes.

I repeat feel free to ignore this...

JR