Just going into some more details with the OPA1656: with an average 3.5nV/rtHz, over 80Khz, gives 0.99uV, or a THD+N of -131dB below 3.5Volt, just because the THD with -150dB vanishes in the noise.
This fully conforms to the specs.
Taking a 500Khz BW, increases this THD+N figure to -123dB.
Using two OPA1656's will increase noise to -120dB as compared to the -110dB that the INA851 specifies.
The OPA stage will need at least a gain of 10, not to let the THD be worsened by the following OPA1633.
And matching the overall CMRR will need the use of 0.1% 25ppm resistors around the OPA1633 plus a trimpot for tuning CMRR at LF plus a capacitor trimmer for tuning CMRR at HF.
Curious to see the @JohnC124s comment.
This fully conforms to the specs.
Taking a 500Khz BW, increases this THD+N figure to -123dB.
Using two OPA1656's will increase noise to -120dB as compared to the -110dB that the INA851 specifies.
The OPA stage will need at least a gain of 10, not to let the THD be worsened by the following OPA1633.
And matching the overall CMRR will need the use of 0.1% 25ppm resistors around the OPA1633 plus a trimpot for tuning CMRR at LF plus a capacitor trimmer for tuning CMRR at HF.
Curious to see the @JohnC124s comment.
Just an idea:
If the whole thing has some (almost) fixed gain then the two-opamp with CM servo implementation of an FDA would come to mind. See I/V of ES9039pro for the general idea, though it is for inverting config.
But it can be modified for non-inverting gain high-Z input, with gains between 2 and 10 or so (otherwise input CM range will suffer severely, or CMRR). There is no subtractor circuit, and there is only need for precision for the two equal CM signal mixing resistors. The servo's noise and distortion appears only common-mode in the output, therefore the input amps dominate the picture.
From servo amp's POV, it's a composite with slaves in parallel. Compensation can be made two-pole (at least for a fixed gain) to increase available feedback for audio range for best CMRR.
Hi Claude
I just stumbled upon this.
Was it you that used 10mf Silmic II with 10mf X7R parralleled as supply decoupling for the OPA1656?
It's not me, but that's not a bad start. I'd use 100uF electrolytic || 10 uF X7R at the opamp pins right after a 0.22 ohm resistor if you're using something fancy like LT3081/LT3091 as regulators since they don't care about capacitive load. If the regs are something like 7815/7915 or LM317/LM337 then I'd make the resistor 10 to 47 ohms to isolate the regulators from the bypass ceramics because the negative regulators oscillate with a naked ceramic capacitor load. Of course bypass the regs with a reverse-biased diode so you don't smoke them if the power goes out and there's still charge on the caps.
Now why bypass electrolytics with a biggish ceramic? Electrolytic capacitors have nontrivial self-inductance, on the order of 100 nanoHenry, or 0.1 microHenry, which with the usual 0.1 uF ceramic makes a dandy parallel resonant circuit. Paradoxically the lower the equivalent series resistance (ESR) on the electro, the higher Q of the resonance. On the other hand, 10 uF in parallel with 0.1 uH needs only 0.1 ohm ESR to damp out the resonant tank.
The 0.22 ohm resistor is there to damp out the resonance created by the power supply trace to the opamp bypass caps, about 2-5 nH per centimetre, so with a 10 cm trace you can get something around 20 to 50 nH. This would be a series resonance with the ceramic bypass capacitor, amplifying power supply noise at the bypass capacitor. The trace would have its own resistance but usually too little to matter, hence the 0.22 ohms in front of the 10 uF X7R.
I suspect much of the different opamps' "sound" is a reflection of how well they deal with power supply source impedance and RF interference, but RF is outside scope here. Someone who knows opamps better than I do could weigh in with an opinion of whether internal nodes could be susceptible to instability when faced with high power supply impedances at greater than 100 kHz.
Now why bypass electrolytics with a biggish ceramic? Electrolytic capacitors have nontrivial self-inductance, on the order of 100 nanoHenry, or 0.1 microHenry, which with the usual 0.1 uF ceramic makes a dandy parallel resonant circuit. Paradoxically the lower the equivalent series resistance (ESR) on the electro, the higher Q of the resonance. On the other hand, 10 uF in parallel with 0.1 uH needs only 0.1 ohm ESR to damp out the resonant tank.
The 0.22 ohm resistor is there to damp out the resonance created by the power supply trace to the opamp bypass caps, about 2-5 nH per centimetre, so with a 10 cm trace you can get something around 20 to 50 nH. This would be a series resonance with the ceramic bypass capacitor, amplifying power supply noise at the bypass capacitor. The trace would have its own resistance but usually too little to matter, hence the 0.22 ohms in front of the 10 uF X7R.
I suspect much of the different opamps' "sound" is a reflection of how well they deal with power supply source impedance and RF interference, but RF is outside scope here. Someone who knows opamps better than I do could weigh in with an opinion of whether internal nodes could be susceptible to instability when faced with high power supply impedances at greater than 100 kHz.
Last edited:
Late to the party, but: this is typical of such measurements. Most audio analyzers like AP have bandwidth limiting filters at 80kHz, with 500kHz being the unfiltered bandwidth. Hence the numbers.
Many manu's can't resist limiting the measurement bandwidth to improve the numbers, although it can be considered reasonble to look to the extended audio bandwidth only and not all the way to lightspeed.
Jan
The difference is we had to add quite a lot of caps directly at the OPA legs in order to correct what seemed to our ears a non perfect OPA PS from the manufacturer. Remember this is a Chinese 70E amp and no one knows if the parts inside are genuine etc.
We found in that very specific case, with the offboard board mount and that OPA, plus the fact that the power amp overall could benefit from some tonal tweaks IMHO, that the 3 caps we mounted directly on the legs, on top of what he had already tweaked re OPA and overall PS, sounded best to our ears and closest to what std power amps from the market sounded.
That is very specific and tweaked to our ears. I have other set ups where the OPA PS is more garanteed (short paths, usual decoupling bypasses) and no such things are required. Not saying they can't be benefical nor sound better, I didn't bother going beyond.
Fact is this is an excellent OPA with the usual OPA PS following the pec sheet recommendation, and indeed it does suffer, perhaps more that some other OPAs (?) with an non adequate OPA PS IME. In fact, apart from decoupling at the legs, OPA PS isn't really something hat should be discussed around this OPA as more system dependent.
MFG
Claude
We found in that very specific case, with the offboard board mount and that OPA, plus the fact that the power amp overall could benefit from some tonal tweaks IMHO, that the 3 caps we mounted directly on the legs, on top of what he had already tweaked re OPA and overall PS, sounded best to our ears and closest to what std power amps from the market sounded.
That is very specific and tweaked to our ears. I have other set ups where the OPA PS is more garanteed (short paths, usual decoupling bypasses) and no such things are required. Not saying they can't be benefical nor sound better, I didn't bother going beyond.
Fact is this is an excellent OPA with the usual OPA PS following the pec sheet recommendation, and indeed it does suffer, perhaps more that some other OPAs (?) with an non adequate OPA PS IME. In fact, apart from decoupling at the legs, OPA PS isn't really something hat should be discussed around this OPA as more system dependent.
MFG
Claude
I am a bit puzzled now. Aren't the various areas where the OPA1656 can be used a part of what this thread is about?
I was really hoping I could spare a big chunk of the trial and error time to implement this excellent OPA by picking on your guy's brains.
What is the suggested local decoupling scheme when using Jan's super regs? This is for I/V on a Dac.
Would those same requirements also pertain to 2x LT3045 in series?
What is the suggested local decoupling scheme when using Jan's super regs? This is for I/V on a Dac.
Would those same requirements also pertain to 2x LT3045 in series?
Radian, what I am trying to say is that the OPA1656 works perfectly well IF you follow the recommendations of its spec sheet. That goes as far as basic PS decoupling.
You may already know that there are different types of OPAs, with different entry stage - JFETs, CMOS, Bipolars... all have various capabilities, and then there are of course much more variables re OPAs, that's why blind op amp rolling is nonsense - One has really to consider why the OPA is there for and what kind of circuits are feeding it and where it will go after the OPA. It is impossible to discuss that on an OPA thread as really circuit related.
For the same reasons, possible PS flows (or say particularities) are circuit / unit and not OPA dependant and no one can predict beyond some basic decoupling what caps in addition to the existing unit's OPA PS should be added and where IMHO.
I hope this helps
Claude
You may already know that there are different types of OPAs, with different entry stage - JFETs, CMOS, Bipolars... all have various capabilities, and then there are of course much more variables re OPAs, that's why blind op amp rolling is nonsense - One has really to consider why the OPA is there for and what kind of circuits are feeding it and where it will go after the OPA. It is impossible to discuss that on an OPA thread as really circuit related.
For the same reasons, possible PS flows (or say particularities) are circuit / unit and not OPA dependant and no one can predict beyond some basic decoupling what caps in addition to the existing unit's OPA PS should be added and where IMHO.
I hope this helps
Claude
Thanks DSP-Geek,
That is valuable information.
I wonder why my LT3045 regs do not like higher capacitance. It is much better with values below 10uF.
This was observed on the analog supplys of my various AKM Dacs.
And the best sounding for me is the opa 1656 on this amp https://nl.aliexpress.com/item/1005....47.a50779d2C3ZGsD&gatewayAdapt=glo2nld...I'm waiting for the nex 2 opa 1656 for the A07 and the A07 max.
Great info thank you Mark.
I the past I led others do the fiddling and copied maybe to much to save time. I haven't touched my scope more than twice the last couple of years. Realizing the complexity of the subject maybe I need to dust the thing off and collect my own data.
•