Also beware of common mode distortion. Unity gain buffers are the worst offenders since both input pins carry identical signals - so it is operating only in common mode. The 5532, an otherwise excellent (if ancient) opamp is particularly prone to this. As is the TL072 (measurements by Self).
This is something that never appears in datasheets. No amount of decoupling makes any difference to this effect.
Newer generation devices like the LM4562 are almost immune to this effect (measurements by Self), and even more recent ones like the OPA1611/1612 probably are too.
Those latter ones also give 1.1nV/rootHz (so getting into AD797 territory) and a 1/f knee at 10Hz.
Good to see you here Ricardo BTW!
Craig
This is something that never appears in datasheets. No amount of decoupling makes any difference to this effect.
Newer generation devices like the LM4562 are almost immune to this effect (measurements by Self), and even more recent ones like the OPA1611/1612 probably are too.
Those latter ones also give 1.1nV/rootHz (so getting into AD797 territory) and a 1/f knee at 10Hz.
Good to see you here Ricardo BTW!
Craig
About star grounding. Can you explain this further, separate coupling capacitor's ground from signal ground is not a good idea I be leave. So connect the coupling capacitors and the signal ground references for each op amp together first and then connect all op amps stages individually to a ground star?
From what I read (here for example) the ground path from the decoupling capacitor to the load must be as short as possible. This is to help with transient currents. So if the load at the output of the op amp should have a short path to the decoupling capacitors ground, I assume it's the same for ground reference for the resistors of an normal non inverting op amp circuit for example. Not sure how important this is for signal level audio circuitry but I still haven's found anything real that says the ground from decouling caps should be separate. Nor have i seen any good PCB examples of this. Designing a PCB with several op amps and then run separate grounds for the decoupling caps would resulsts in very long ground path, which I'm sure is NOT a good idea.
I wrote an Engineering Note at Calrec on this circa 1979 when it wasn't well known. IIRC, TL072 was much worse than 5532. Can't remember if I had any suggestions to alleviate this or alternate methods if a Unity Gain buffer was required.
Measurements on single devices, like Self's, don't show how OPAs perform in real life like Kingston's thread. LM4562 has a particularly nasty lockout mechanism which appears on single rail applications and also when LF filters (with large caps) overload. Proaudiodesignforum had a huge thread on this. I think he brought it up at one of the TI or old NS forums where this was acknowledged. 1ppm THD is usually less important than the OPA crapping out completely at times.
Good to see you here too Craig 🙂
Last edited:
This is a HUGE subject and I could write a whole book on it for high quality audio. ONE of the biggest problems is that decoupling dumps the SEWAGE from the power lines onto your nice clean signal earth which is your 'ground reference'.
This is to minimize inductance. I think I say in the Kingston thread
- Think what happens to the sewage your decoupling caps are dumping
- Think what happens to your currents
- Think what happens if your Ground/Earth lines/planes have resistance and inductance (cos they do)
Kingston reports that the biggest improvement came when he used electrolyics at each package AND a separate earth .. which is what I expected.
Usually, the 'load' in small signal circuitry is small; eg 100k input resistor on the next stage. This is not the case for your 'main outputs' but for these you can usually treat them like a power amp with Star Earth.
Where you can't assume this, is 'low noise filters at LF' which have large caps and correspondingly large currents which HAVE to be referenced to Clean/Signal earth. For these, look up Heirarchical Earthing for Tube mixers/preamps which often have a single earth bus/line and da old gurus thought very carefully about decoupling for each stage and also how signal currents would affect the overall signal and ground integrity.
And don't think separate audio & decoupling ground planes will solve all your problems. When designing a 'clean audio ground plane' (you don't just flood it), you work out where and how you want your currents to flow and ensure the flood doesn't let the current go via a different route. And don't have ANY ground or power plane under your OPAs (strictly under the nets which form your 'input loop') as these put unwanted capacitance to 'earth'
I've only touched on the subject. Your TI link on Mixed Signal stuff covers some of it but leaves out some important points.
Well this about grounding sure is a huge subject and maybe should be discussed somewhere else. But that I'm interested in is if the decoupling caps dumps "sewage" into the ground or not. Read the kingston thread btw, very interesting.
Agree, "just flood it" was something I just to do then I was younger. But no ground planes under the op amps or the input nets? This was new for me, but I guess it is to not introduce noise into the signal, for the same reason (TI datasheet I think) says that power track at the PCB should be routed as far away from the inputs of the op amp. Whih I do btw, but never thought of the ground plane.
Guess the effect would be greater for surface mount op amp compared to DIL-package in a socket for example, because of the distance to PCB.
True, but what if we want to drive lower resistance pots? 1k for example, it's much lower than 100k and most OP amps can drive loads down to 600 ohms easily. Reason I ask is cause I just design a circuit using an op amp to drive a 1k pot.
Used a ground plane and option B for decoupling caps for this design btw. And ground plane under the op amp ... damn
Agree, "just flood it" was something I just to do then I was younger. But no ground planes under the op amps or the input nets? This was new for me, but I guess it is to not introduce noise into the signal, for the same reason (TI datasheet I think) says that power track at the PCB should be routed as far away from the inputs of the op amp. Whih I do btw, but never thought of the ground plane.
Guess the effect would be greater for surface mount op amp compared to DIL-package in a socket for example, because of the distance to PCB.
True, but what if we want to drive lower resistance pots? 1k for example, it's much lower than 100k and most OP amps can drive loads down to 600 ohms easily. Reason I ask is cause I just design a circuit using an op amp to drive a 1k pot.
Used a ground plane and option B for decoupling caps for this design btw. And ground plane under the op amp ... damn
Interesting what you say about the LM4562. I populated a whole active crossover with them, and had real problems with it - presumably via the filter mechanism you point out. Fortunately I'd designed and built it socketed, and ended unplugging the lot and putting 5532's in.
I'm a member of the ProAudio forum - I'll check for the 4562 thread.
The FET input OPA2134 also worked well in the crossover, but higher voltage noise than the bipolar 5532's.
The only problem I know of with the AD797 is how twitchy it is to make it stable and not hoot at low RF. Decoupling is very important for that device, as is the position and capacitor types for the distortion reduction feature. I've steered clear thus far.
There are several issues but the main one is you don't want capacitance from the -ve i/p to ground (or power)
That's another case where you have to look at 'Heirarchical Earthing' as in Jurassic Tube mixers and preamps.
I've got a document, LNprimer, in my MicBuilders Files where I list most of links I know on this Grounding ****. You have to join.
NJM4562 has many of the supa dupa characteristics of LM 4562 but without its problems. It's a completely different topology ... the Raytheon 4136 which was the first OPA, I found to be 'undetectable' in DBLTs ... though you had to be careful with levels. NJM have produced a huge family of OPAs with many different specs with this topology. People sneer at them cos they are so prevalent but some are SOTA for certain applications. eg NJM2068 is now my favourite OPA for a MM preamp. Biggest caveat is their high Ib (like 5532) ... and that it's PNP i/p which is the opposite of 5532 and other good OPAs
'Only' ??!! 😱 This was the gist of my correspondence with Guru Wurcer 🙂
Alas obsolete now. From the datasheet it looks like a fine product. Why are most useful audio devices falling off the perch like a Norwegian Blue?
For one thing, I can't get past the nuisance of 100n + 100n in the 'B' config most likely being 97n + 104n or something like that. Very messy.
Do they have to be closely matched though? If the intent is to stabilise the amp and keep junk away from the power lines, is 'close enough, good enough'?
1% tolerance C0G/NP0 capacitors are readily available anyway, so it would be easy enough to get matching values if it is important.
I just wonder how critical it is to have them closely matched.
1% tolerance C0G/NP0 capacitors are readily available anyway, so it would be easy enough to get matching values if it is important.
I just wonder how critical it is to have them closely matched.
Don't think they have to be match. Option B is what most datasheet suggests and usually they suggests values between 0,01 and 1,0 uF or so.
Kingston shows the MOST benefit is using Electrolytics in 'B'. Get your 1% matched Electrolytics from 'Ye Olde Unobtainium Shoppe'. Once you have these in place, you can use anything else in addition and they won't worsen the performance .. They won't make things better either but who cares 🙂
You jest, but I'm also suspicious of what happens when there's actual ripple on the lines (which the capacitors are trying to take care of).
If there are audio signals loaded to ground, or PSU ripple, or audio going up through the op-amp's/s' rails (etc.), ceramic capacitors are going to be harmonic generators as the voltage across them varies. Except for C0G, maybe, but 2 x 100n per op-amp package wouldn't be a viable commercial option anyway.
With unmatched values there's an additional concern (probably unfounded but worth checking) that common mode noise will be unbalanced and will leak to ground. My preferred way of doing things is usually with single-ended rails and virtual grounds all referenced to one side, precisely to avoid "solvable problems" like mains hum. Otherwise, common-mode hum from leakage capacitance in transformers can be hard to fix.
*I forgot to add to the list the likely high-frequency garbage generated by the 'better' op-amps as they operate their output stage in ultra-low THD class b.
If there are audio signals loaded to ground, or PSU ripple, or audio going up through the op-amp's/s' rails (etc.), ceramic capacitors are going to be harmonic generators as the voltage across them varies. Except for C0G, maybe, but 2 x 100n per op-amp package wouldn't be a viable commercial option anyway.
With unmatched values there's an additional concern (probably unfounded but worth checking) that common mode noise will be unbalanced and will leak to ground. My preferred way of doing things is usually with single-ended rails and virtual grounds all referenced to one side, precisely to avoid "solvable problems" like mains hum. Otherwise, common-mode hum from leakage capacitance in transformers can be hard to fix.
*I forgot to add to the list the likely high-frequency garbage generated by the 'better' op-amps as they operate their output stage in ultra-low THD class b.
Last edited:
One of the problems with separating the grounds is you can create unnecessary loop area which will radiate garbage so you really have to think about this as well. I like the idea of a power ground plane and then a separate signal ground track overlaid on top of that, but of course we don’t always have the luxury of multi-layer boards.
Seems it’s a case of balancing common impedance coupling with radiating loops - they have conflicting requirements sometimes.
Seems it’s a case of balancing common impedance coupling with radiating loops - they have conflicting requirements sometimes.
I just follow this guide for opamp based pcb layout's from ti's e2e blog.
Do check additional resources and comments on the same page.
- Make the connections to the inverting pin as short as possible.
- Place decoupling capacitors as close to the supply pins as possible.
- If using multiple decoupling capacitors, place the smallest decoupling capacitor closest to the supply pin.
- Do not place vias between decoupling capacitors and supply pins.
- Make traces as wide as possible.
- Do not route traces with 90-degree angles.
- Pour at least one solid ground plane.
- Do not sacrifice good layout to label a component with silkscreen.
Do check additional resources and comments on the same page.
Good link @krevilplays . Interesting it does no mention that ground planes should ne be placed under under IC and/or input nets. But someones mentions it in the comment at that page. Wonder how important this is, at least for audio use.
Recommendation is because of parasitic capacitances forming between tracks and ground plane. They are easily 3 – 10 pF. Main issue is parasitic capacitance effect on negative feedback network, where it causes gain peak at high frequency (MHz range) and can provide conditions for circuit oscillation.
Ok, so is this mainly a problem then designing HF circuits in the Mhz range. Or could this also be a problem for audio circuits, for example making them unstable and oscillate?
It is not a problem for audio circuits with reasonably limited frequency response. But, most high-quality operational amplifiers used for audio, have bandwidth from 10 – 50 MHz or even more. So, we still have to take care.
I’ve seen here comments from diy-ers that don’t use oscilloscope or measure anything, but they design audio circuits with opamps and claim that their circuit in latest version not having ground planes around opamp, sounds better than previous with ground planes. Probably it was a case of circuit oscillating at high frequency, which affected sound.
I use opamps with ground planes around opamp, covering input, output and feedback network, but taking care to minimize parasitic capacitances with very narrow tracks and using low impedance feedback network. That way, I can get both excellent performance and immunity to electromagnetic interference from transformers or environment.
I’ve seen here comments from diy-ers that don’t use oscilloscope or measure anything, but they design audio circuits with opamps and claim that their circuit in latest version not having ground planes around opamp, sounds better than previous with ground planes. Probably it was a case of circuit oscillating at high frequency, which affected sound.
I use opamps with ground planes around opamp, covering input, output and feedback network, but taking care to minimize parasitic capacitances with very narrow tracks and using low impedance feedback network. That way, I can get both excellent performance and immunity to electromagnetic interference from transformers or environment.