opamps and local decoupling of rails, some questions

[JohnRoberts]

I've a 100Mhz scope with 10x probes, Rigol DS1052E that got universal praise at eevblog upgraded to 100Mhz by their recommendation. What size resistor should there be in series with the opamp output and probe? I've been measuring the output pin directly of course, didn't know any better...

As far as me playing around with exotic opamps that began with blind swapping, it seems I'm finally asking the right questions. I also suspect it was the better decoupling implementation that gave me these improvements, little to do with opamps themselves. 0.001% THD is still several zeroes away from what the better opamps are capable of in ideal environments. Perhaps after this thread I'll gravitate towards new designs where I will achieve that.

Capacitive loading at the inverting input. Cancel with a feedback capacitor; for general audio work, I'd size the time constant with the corresponding feedback resistor for about 300 kHz (this assumes a unity-gain stable opamp).

Is this the compensation/integration cap parallel to the feedback resistor? How do I calculate the time constant and the corresponding cap size?

freq= 1/(2x pi x R x C)  or 1/(6.28 x R  C)
solving for C
C= (1/freq) /(6.28xR)

note: as Sam mentioned this is only applicable for unity-gain stable opamps, since it could actually cause issues with decompensated opamps. In theory you can add a cap across a decomp opamp feedback R also but that cap needs to be smaller than the cap to ground at - input in the proper ratio to satisfy the minimum gain requirement to insure stability, so generally not worth the effort, unless you have unusually large input pin capacitance.

Probing an output pin with a scope probe is generally not a problem unless the circuit is really marginal. The x10 probe has less capacitance. A small resistor (couple hundred ohms) in series should be adequate. As Ricardo mentioned too large of a resistor could form a LPF pole with the probe capacitance and hide very HF oscillation.

JR

 

[Samuel Groner]

LM/NJM4562

These are completely different opamps (even if implied otherwise in the article you linked, and the numbers suggest otherwise).

What size resistor should there be in series with the opamp output and probe?

100r is fine, usually. Most 10x probe have an input capacitance of 10 pF or less, so the BW limit is above 100 MHz. Particularly for prototype PCBs, I often "design in" these series resistors together with a test point.

Probing an output pin with a scope probe is generally not a problem unless the circuit is really marginal.

For a NE5532 or slower, yes. But not necessarily for an ADA4898-2 or anything else in this class.

In some circumstances (which I've encountered more frequently than one'd think) the added loading of the probe can make an amplifier stop oscillating. To the unwary this can be even more confusing than the other case (where the probe caused oscillation)...

Samuel

 

[ricardo]

What size resistor should there be in series with the opamp output and probe?

The 10x setting on your probe introduces an appropriate resistor to reduce loading AND compensation so the response remains flat to 100MHz

Using a separate resistor also reduces the loading but mucks up the RF response so you may not see the oscillation.

 

[Kingston]

The best THD chain of this project is now 0.0008% (aux1 or aux3 to their respective group). Sure it's also the shortest chain, but I'm happy. Generally from a single channel through master bus and its line drivers I get 0.001-0.002% (most common chain).

The opamp that was most easy to handle, didn't ever oscillate, and gave lowest THD was OPA2132. LME49860 was another one with slightly higher THD, no oscillation, and lowest noise floor in all cases (lower than OPA2132). Surprisingly LT1358 got me nearly equal results (nearly equal to LME49860). NE5532 has been unstable in certain spots, still trying to figure out what's causing it.

What's most important is that I can no longer tell the difference between opamps when comparing channel sets. I used to be able to pick them easily: "the fluffy bass OPA2132", "nervously sharp LT1358" and "ultra-accurate LME49860" etc. sounds. But then, I was listening to 0.1% THD. Looks like I've been mostly hearing PSU and dirty power distribution artifacts. What a ridiculously complex interactive network it must have been with every single opamp of the mixer talking to everyone else through the rails. No more, not ever again will I make this mistake. I can also lay to rest this blind swapping madness and concentrate on design and data.

Some of the stranger opamps I'm using are still oscillating around 25-50Mhz 2-12mV. I will probably revert back to stable ones, I'm not making a radar just yet.

 

[ricardo]

What's most important is that I can no longer tell the difference between opamps when comparing channel sets. I used to be able to pick them easily: "the fluffy bass OPA2132", "nervously sharp LT1358" and "ultra-accurate LME49860" etc. sounds. But then, I was listening to 0.1% THD. Looks like I've been mostly hearing PSU and dirty power distribution artifacts.

And probably the effect of oscillation which may not be immediately visible too.  I think you can see why some people report chalk & cheese results with OPA rolling  :

Baxandall tells a story of trying to make a power amplifier stable by tacking on ceramic capacitors.  He couldn't see the oscillation cos it was an internal loop which changed as soon as he probed it .. like Sam mentions.  He was getting very confusing results until he realised that the 1" or so leads on his tacked on ceramics were just the right inductance for the 200+MHz or so stuff he thought he was getting.

Trying to cure oscillation is extremely frustrating so old fogeys (like me) tend to take a belt & braces attitude.

Kingston, this has been a long thread but I feel I have learnt as much as I have pontificated.  Many thanks and well done.  Can you tell us which of the

stranger opamps I'm using are still oscillating around 25-50Mhz 2-12mV.

?

 

[Kingston]

I still can't be absolutely positive that these oscillations are not a measurement error. Strangely they all have now "stabilised" into 26Mhz 5mV. This is what I measure anywhere that "oscillates", with any of the opamps now, which is awfully strange and seems more like an artifact of my rig. It could have something to do with the fact I'm using DIP adapters for these SOIC8 devices. Some of the opamps have thermal pads (not connected), and it could be they aren't too happy about these adapters in general. Path to decoupling has more joints than it should (SOIC8 opamp -> adapter -> adapters legs -> opamp sockets -> circuit/decoupling).

ADA4627-1, ADA4898-2 and AD8599 are the ones oscillating (and in adapters). I tried AD8512 also, but it had too much distortion in any of the tasks of this mixer. OPA827 and OPA2211 are the only ones in adapters and still happy. I probably won't be using these adapters anymore. For this project it was just a way to exhaust my opamp sample collection.

Although my decoupling is now up to snuff, the rest of the endeavors are still far less scientific than I'd like them to be. The boards are so thickly populated and p2p modded that some measurements are simply impossible. For my next project I will make absolutely sure I can make conclusive measurements, now that I have a hunch how they should be done. Thanks everyone that helped, this has been the best learning project I've ever done.

 

[Samuel Groner]

If there's just one oscillation frequency it could be that there's just one opamp instable, and the oscillation signal is just distributed through the board. Often the oscillation also appears on the supply rail, where it is easily conducted to other ICs.

Fast opamps don't like sockets; you can migate the stability effect somewhat by installing the feedback capacitor right on the IC, rather than the PCB. This keeps the HF feedback path short.

Samuel

 

[Kingston]

I'm always about to close the lid on this, but then optimiser in me takes over. several things happened this week. Firstly, TLE2072 is a great opamp. I was hunting oscillations and rediscovered this opamp that had sounded "harsh" in previous projects. It sounds exactly like every other better quality opamp: like nothing at all. Most importantly, it doesn't oscillate (in my environment).

Then I also remembered I had these PCB's stocked away for some future project:

http://www.sjostromaudio.com/pages/hifi-projects/36-hifi-projects/97-jsr03-sjoestroem-super-regulator (and its negative version)

They're Jung Super Regulators http://tangentsoft.net/elec/opamp-linreg.html

I built a PSU based on them just to see what would happen, a simple thing, transformer -> rectifier -> some CRC -> these super regulators. But there's voltage/gnd sense so the PSU can forget there's a 2-3 meter cable before the current reaches destination. The original PSU (that I also had built) was LM317/LM337 based thing with really big caps before and after the regulators. Jung regulators don't need that much capacitance to stay clean. As far as I know they provide purest DC of any regulators.

Result? All the last (minimal) oscillations I had been fighting are completely gone! Everything is just a flat line, nothing there for the oscilloscope to see. Also, some channels had noise that I had thought to be from the opamps themselves (because not all signal chains had that type of noise). Not so, this new supercharged PSU completely cleaned noise floor, strange (negligible) hash at mid frequencies upwards. THD measurements stayed the same as I had them before (now around 0.001 for most unity gain channel-to-bus-to-lineout chains).

I'm using this mixer mostly as a line level unit so when I measure these unity gain chains, the noise of any chain is now *below* RME HDSP I/O noise (crosstalk is worse than RME, but that can't be helped.). I will post some pictures later. The specs were not bad before, but now, holy crap! PSU quality matters more than I thought it would. It's not like the old PSU was bad, and I could not measure any noise from its outputs. But looks like a simple LM317 solution isn't always enough. Perhaps the voltage sense matters as well. The old PSU had some 0.3ohm resistance in the cable that was additionally filtered at the mixer end. Could be said the old PSU was a bit "soft" in comparison to the new one that is technically perfect.

 

[living sounds]

Kingston, it's extremely interesting that you ended up with exactly the same op amps in your console as I did in mine. I've been modding it for three years and am finally happy with the sound, and ended up with OPA2132/132, OPA827 and LME49720 (metal can). ;-) For the EQs I prefer the sound of the 5532.
I tried several ways of decoupling and found that 100nf rail-to-rail sounded best. I've also removed all caps from the signal path and made several changes to the servo and other circuit parts with the help of the good people here.

I'm curious about the PSU, I'm still using the standard one with a few upgrades, maybe the discrete regulators could improve performance even more... hm... ;-)

 

[Kingston]

Kingston, it's extremely interesting that you ended up with exactly the same op amps in your console as I did in mine. I've been modding it for three years and am finally happy with the sound, and ended up with OPA2132/132, OPA827 and LME49720 (metal can). ;-) For the EQs I prefer the sound of the 5532.
I tried several ways of decoupling and found that 100nf rail-to-rail sounded best. I've also removed all caps from the signal path and made several changes to the servo and other circuit parts with the help of the good people here.

You're reading this thread completely wrong.  The modifications I've done create a unit that does not have a sound. That is my most important conclusion from all these experiments. The only reason I'm still swapping opamps is to see how they react in this technically robust environment, and whether I can match the published specs of even their datasheets. For example, I can swap three OPA2132's and TLE2072 in a chain, and cannot measure or hear any difference. If there is some, they fall under the capabilities of my measurement rig. They have been provided a very optimal environment, and since there is no distortion from numerous PSU and decoupling related artifacts, they act the same.

"100nf rail-to-rail sounded best" is also a meaningless statement with regards to the many theoretical and practical tests on this thread. It does not even begin to cover the complex decoupling scenarios discussed here. My conclusion again is that proper decoupling does not have a sound. If it does, you're doing something wrong or decoupling is inadequate.

[edit]

...and whatever I may have said about opamp sound in the past I have to retract. I know I've posted a lot about opamp swapping and even reviewed their "sound". It's unfortunate to whoever searches this forum and lands on them. But to my defense, I had no idea what I was doing and causing with all that blind swapping. I know now I was listening to opamps trying to deal with bad decoupling, interacting with each other and other PSU artifacts. There are big differences in that area that can actually be heard. But that's a completely unpredictable sector that should be avoided.

 

[living sounds]

"No sound" is what I'm looking for (except in the EQ), what exactly did you end up with?

 

[Kingston]

"No sound" is what I'm looking for (except in the EQ), what exactly did you end up with?

An incredible amount of tedious decoupling, damping and signal vs. decoupling ground isolation additions, modifications and re-routings throughout the mixer. And finally the supercharged PSU.

 

[JohnRoberts]

Congrats, "straight wire with gain" has been a design goal for a long time (for some of us).

JR 

 

[jackies]

And finally the supercharged PSU.

So, have you used a single Jung regulator for the entire mixer, or like, one for each strip or something? I imagine it has somewhat limited current capacity?

 

[Kingston]

And finally the supercharged PSU.

So, have you used a single Jung regulator for the entire mixer, or like, one for each strip or something? I imagine it has somewhat limited current capacity?

There's a super regulator for each rail, two in total for the whole mixer. Their current capability is actually limited by the LM317/337(1.5A) used as a pre-regulator (see their use in Jung 2000 Regulator here: http://tangentsoft.net/elec/opamp-linreg.html). The pass transistor - depending on your choice - could actually do far more than that.

This mixer eats 17V ~900mA each rail and the PSU pre-regulator/pass transistor pairs already need significant heat sinking.

You could use LM338 for the positive rail to bring the regulation all the way to 5A, but there's no equivalent for the negative rail as far as I know.

 

[living sounds]

Kingston, have you looked at the op amps in other gear as well? Like the PRR 176 for example? I wonder if the differences there, not only in the signal path but also in the side chain, will dissappear as well with all the issues addressed.

Also, are the transparent ones only the modern high speed low distortion op amps or does this apply to older ones like the TL072 or 5532 as well (obviously in places where they don't have to "work" much)?

 

[Kingston]

Kingston, have you looked at the op amps in other gear as well? Like the PRR 176 for example? I wonder if the differences there, not only in the signal path but also in the side chain, will dissappear as well with all the issues addressed.

Also, are the transparent ones only the modern high speed low distortion op amps or does this apply to older ones like the TL072 or 5532 as well (obviously in places where they don't have to "work" much)?

Yes, this project has me thinking of improvements in other gear. Removing distortion is a great equalizer of sound. Whether a complete lack of distortion is desirable is another thing (less than 0.001%THD in my opinion). Certainly opamp isolation from each other could be improved in almost every single project present on prodigy pro. THD and noise will go down measurably. I've been wondering how it isn't much more common knowledge that the usual 100nF decoupling present everywhere doesn't necessarily help, or even do much anything. Opamps seem to like damping and their own little personal electrolytics along with them.

And while I haven't used TL072 anywhere recently, I found NE5532 remarkably transparent. Just like the academics here are always saying, it's good enough for just about anything when implemented correctly.

 

[living sounds]

Thanks! How important is the seperate drain ground and where does it connect to the common ground / audio ground? This would be a major complication in my console.

Stupid question: Does the order of electrolytic series / resistor to ground make a difference, in other words, should the resistor or the cap be closer to the opamp leg / the ground?

 

[Kingston]

Thanks! How important is the seperate drain ground and where does it connect to the common ground / audio ground? This would be a major complication in my console.

The most dirty ground "owned" by the decoupling caps separated from the reference ground seemed very significant. This trick completely removed any trace of 50hz hum and its multiples from noise floor. I cannot answer how this works in other consoles, or if it's even necessary there. I made the connection in each channel strip/group/master main gnd connection point (there is only one) where it connects to chassis. In this mixer chassis is used as the "ground bar". It's like hierarchical ground, but there are two ground buses per strip. Someone experienced might be able to implement a perfectly noiseless layout where there's just one bus. I wouldn't know how.

Stupid question: Does the order of electrolytic series / resistor to ground make a difference, in other words, should the resistor or the cap be closer to the opamp leg / the ground?

I didn't actually do the electrolytic+resistor trick, but in series connection the order doesn't matter.

 

[ricardo]

Does the order of electrolytic series / resistor to ground make a difference, in other words, should the resistor or the cap be closer to the opamp leg / the ground?

The most important and useful resistor is the ESR of the electrolytic.  Bigger caps have smaller resistors but in fact this is exactly what's needed to damp any propensity for oscillation & thd.

Just put electrolytics between each rails and dirty earth near each OPA.  Then you can add any other Golden Pinnae caps wherever you want.  The ESR of the electrolytics provide the correct damping for everything else.

The separate dirty earth is essential.  Exactly where the earths join is also important.  In most cases a 'star earthing point' works well.  If you have many modules plugged into mother rack you may need to re-think this.

The most useful advice is to

• think where that nasty stuff you are decoupling is going.
• think what happens if your earths have resistance & inductance (cos they do)