Does Op-amp Rolling Work?

[T&T]

I have 2 Fosi ZA3 using it mono with xlr, I don't touch 1 of the ZA3 and change opamps on the 2 ZA3, level Macht both volume, same song over and over and over, turn the volume down on one ZA3 and up in the other ZA3 and vice-versa...There is always 1 opamps that sound better, worse or similar.

 

[trevor_bartram]

I have to admit, I swapped a NE5532 for a LME49720HA. I was happy when I was able to see a higher slew rate on my test board with the oscilloscope. Unfortunately, I couldn't hear a noticeable difference. However, the LME has a thermally healthier TO-99 housing. Are there any measurements that prove perceptible impairments in overheated opamps?

In the past (Bob Widlar @ Nat Semi etc) paid close attention to chip layout to minimize variation over ambient temperature (-55C to 125C) to enable military grade performance.

 

[trevor_bartram]

Before rolling opamps please check that your alternate opamp is NOT a current feedback type, as a listed feature on the datasheet. Those types are sensitive to circuit values & may not work well in a conventional opamp circuit.

 

[boristau]

When I post my last video on review of Douk A5 amplifier, I mentioned that rolling (changing) op-amp ICs in there is fruitless. I got have a dozen comments under that video in youtube asking why so I thought I experiment again with the A5. Note that I have done the same testing with both DACs and Power Amplifiers and found the change to not make a difference. But let's see if the results are different this time.

View attachment 436787
The Douk A5 makes this job both easy and hard. It is easy because you can just lift the plexiglass and replace the op-amps. The difficulty was replacing the standard TI NE5532P opamp with the discrete Sonic Imagery Labs as it was too larger to fit in there. I pushed the adjacent caps more than I was comfortable with to get it to fit. Fortunately it worked.

There is a massive cost difference with the stock NE5532P costing US 57 cents in single quantity compared to minimum of $49 I found for the Sonic Imagery 994Enh-Ticha dual opamp. You would need two of them for stereo operation which would represent a premium that matches that of the amplifier itself! Here is a close up shot the 994Enh-Ticha:

As you see in the above picture, i decided to replace the right channel (Ch 2 below) and left left channel (Ch 1) the same. That way we can compare the two channels simultaneously under the same environmental situation. Amplifiers are temperature sensitive and shutting down to replace the opamp and powering back up would have created another variable. Alas, there is also channel to channel variations so the testing is not 100% exact but very close as you see below.

Opamp Rolling Measurements
Here is our usual dashboard:

View attachment 436792
As we see the performance is the same with SINAD which sums noise and distortion. This is of course at one power level so let's sweep the input voltage and measure at all power levels up to clipping:
View attachment 436793
There is the tiniest gap between the two but that may just be variations between channels. Even if it weren't so, it is a miniscule difference.

Maybe the differences becomes more visible if we use other frequencies than 1 kHz represented above:
View attachment 436795

I have zoomed into this graph to make differences larger. Dashed line is the Sonic Imagery discrete op-amp. We see that both at 15 kHz and at less than 500 Hz, the discrete amplifier is actually worse! But again, that could be variations between channels.

I wanted to investigate that a bit more so ran a couple of FFTs at both 100 Hz and 1 kHz:
View attachment 436797




View attachment 436798
The profile of distortion changes but not the high-order message that any difference is relegated to high order harmonics that are at or below threshold of hearing.

Discussion And Conclusions
It is natural to assume that the much more expensive, larger and fancier hand-made opamp IC would do better. All is not as it seems. An integrated circuit (IC) benefits from high precision components and even components that cannot be instantiated using discrete parts. Path lengths are also shorter allowing for better optimization of the design. Mass production using automated systems follows up by sharply reducing its cost.

On the other hand ICs can be subject to thermal coupling where rise in temperature in one part of the IC can negatively impact the performance of the rest of the IC. This doesn't apply here because the opamp is used at the front-end of the amplifier that is not attempting to produce power (only acts as a buffer and/or gain stage). Importantly, there is feedback that is used to correct the non-linearities in the op-amp. This correction highly linearizes both discrete and integrated op-amps as to almost erase any signature of the original part. This is why we don't see much difference in measurements.

People report improvements in sound and with it justify the upgrade. As members of this forum well know, such listening tests are improper. Testing must be controlled to exclude all extraneous (non audible) factors. When done, the measurements powerfully predict no audible difference. Indeed, I only know of one research paper that dug into sonic differences in op-amps and that only happened when the opamp was vastly overdriven.

Finally, I am not saying that all opamps are the same. There are countless ones for a reason. But unless you have instrumentation such as I am using, you have no prayer of knowing if a change improves anything. Or worse yet, made things worse. Here are the results form the DAC test:

There we do see a bit of differentiation but not enough to bother with any of this.

Net, net, leave the engineering to well, engineers! But a performant audio product and use it as is. Don't risk damaging and spend money on something that has essentially no chance of doing you any good.

Video version available as well:

------------
As always, questions, comments, recommendations, etc. are welcome.

Any donations are much appreciated using: https://www.audiosciencereview.com/forum/index.php?threads/how-to-support-audio-science-review.8150/

With all due respect Amir, no Op-Amp can improve mediocre amplifier.

TI NE5532PTHD + N at 1 kHz (typ) 0.002%
https://www.ti.com/product/NE5532?u...MIuonBtuGajAMVvEdHAR0cihacEAAYASAAEgKB8PD_BwE
994Enh-Ticha Dual Matched Discrete Operational Amplifier
Distortion+Noise at 1 kHz 0.0003%
https://www.sonicimagerylabs.com/pr...e_HD_OpAmp/994Enh_DiscreteOpAmp_Datasheet.pdf

I tried many Class D amps with and without Op-Amps. Ended up with 8 Purifi 1ET9040BA and 8 Hypex NCOREx NCx500 amplifiers all with 994Enh-Ticha and can not peel my ears from the systems.

Regards,
Boris

 

[DanielT]

With all due respect Amir, no Op-Amp can improve mediocre amplifier.

TI NE5532PTHD + N at 1 kHz (typ) 0.002%
https://www.ti.com/product/NE5532?u...MIuonBtuGajAMVvEdHAR0cihacEAAYASAAEgKB8PD_BwE
994Enh-Ticha Dual Matched Discrete Operational Amplifier
Distortion+Noise at 1 kHz 0.0003%
https://www.sonicimagerylabs.com/pr...e_HD_OpAmp/994Enh_DiscreteOpAmp_Datasheet.pdf

I tried many Class D amps with and without Op-Amps. Ended up with 8 Purifi 1ET9040BA and 8 Hypex NCOREx NCx500 amplifiers all with 994Enh-Ticha and can not peel my ears from the systems.

Regards,
Boris

Which you listen to and compare between on a pair of speakers that have...0.5% distortion?

 

[pma]

It is the forgotten TIM/SID distortion that makes the 1458 so horrible. Never use an opamp with slew rate below 5V/us for audio, at any position. D10s photo with the socket is shown below.

A textbook image with the LME49720.

 

[DaneelOlivaw]

It is the forgotten TIM/SID distortion that makes the 1458 so horrible. Never use an opamp with slew rate below 5V/us for audio, at any position. D10s photo with the socket is shown below.

View attachment 438000

View attachment 438001
A textbook image with the LME49720.

View attachment 438002

I found a slew rate as low as 0.5 V/μs for the LM1458, 9 V/μs for the NE5532, 20 V/μs for the LME49720, and 27 V/μs for the OPA1612.
Can you check the NE5532 in the same way?
I dont think so but - perhaps swapping the NE5532 for a relatively inexpensive LME49720 was not too bad.

 

[Cbdb2]

I found a slew rate as low as 0.5 V/μs for the LM1458, 9 V/μs for the NE5532, 20 V/μs for the LME49720, and 27 V/μs for the OPA1612.
Can you check the NE5532 in the same way?
I dont think so but - perhaps swapping the NE5532 for a relatively inexpensive LME49720 was not too bad.

The NE5532 has a sufficient slew rate for audio, higher rates dont make a difference. Slew rate has nothing to do with normal linear operation if it is sufficient.

 

[DaneelOlivaw]

The NE5532 has a sufficient slew rate for audio, higher rates dont make a difference. Slew rate has nothing to do with normal linear operation if it is sufficient.

I learned today 0.5V is not sufficient and 9V is more than enough. So we agree that 5V is minimum?

 

[solderdude]

Whether or not it is enough also depends on the output voltage

 

[Cbdb2]

Whether or not it is enough also depends on the output voltage

Just have to use the max rail voltage the opamp can handle and youll be safe.

 

[Cbdb2]

I learned today 0.5V is not sufficient and 9V is more than enough. So we agree that 5V is minimum?

Slew Rate Calculator

This calculator computes the minimum slew rate necessary to output a certain voltage at a certain frequency, calculates the maximum output frequency an op amp can deal with at a given voltage, and calculates the maximum voltage that an op amp can output at a certain frequency.

www.learningaboutelectronics.com

Because slew rate is a measure of input overload, its not in regular operation, you dont want to be anywhere near it. The rule of thumb is to calculate the slew rate required (for 15v rms and 20khz is 1.9v/us) and multiply by 5, so an opamp that has a slew rate of 10v/us can be used for any audio application without slewing.
The TI data sheet has the NE5534 SR as 13v/us, more than enough.

Most opamps slew rate is determined by the compensation capacitor, if this is an external cap the slew rate will change with the capacitance

 

[amirm]

With all due respect Amir, no Op-Amp can improve mediocre amplifier.

This amplifier outperforms more than half of all amplifiers I have tested so it is not mediocre at all from performance point of view.

That aside, this is the point of the video/article. People who were interested in this amplifier were shocked when I said there was no point to swap opamps. And highly suggested the discrete opamps as the target. So I did the testing to show there is no reason to do that.

As for your choice with higher power amps, I tested those as well and found little sensitivity to choice of discrete opamps. See: https://www.audiosciencereview.com/...-opamp-review-sonic-imagery-vs-sparkos.10325/

 

[Salt]

ASR a it's best again

 

[Salt]

The discussion in short, we are on page 29:
OP amps do their job if a fitting one is selected for their job.
If not, there may be some intention for tuning/sounding/voicing inteded. Only in this case it might be advisable to try another OPA to achieve better results/transparancy.

 

[wwenze]

Confirmation bias can cut both ways

That's why we need to fight against it by proper testing both ways

 

[Tranquility Bass]

OPA1656 specs below. What's not to like ? Do any of these discrete varieties offer better bang for the buck ?

Also, has anyone checked out the input bias current on some of these discrete parts ? No engineer worth their salt would ever design in or offer as a replacement a silicon part with input bias currents as high as 5uA or more on some of these discretes ! Typical input bias current for an LM4562 is 10nA ! uA741 had a typical input bias current of 80 nA ! And of course the FET input parts are orders of magnitude lower again !

What exactly are the issues that these discrete opamps are trying to solve in their silicon chip counterparts ?

https://www.ti.com/lit/ds/symlink/opa1656.pdf

 

[milosz]

So, you're saying that opamps don't matter, so if my audio gear had IC709 opamps in it, it would sound just as good as if it had Sparkos SS3602's? Hmmm. I'm not so sure about that.

 

[Tranquility Bass]

So, you're saying that opamps don't matter, so if my audio gear had IC709 opamps in it, it would sound just as good as if it had Sparkos SS3602's? Hmmm. I'm not so sure about that.

No I am saying why would you spend 20-50x as much on something that on balance appears to be inferior and is impractical in terms of size constraints ?

What problems on an existing chip opamp are you trying to solve with this device ?

 

[milosz]

No I am saying why would you spend 20-50x as much on something that on balance appears to be inferior and is impractical in terms of size constraints ?

What problems on an existing chip opamp are you trying to solve with this device ?

Even lower noise, lower distortion....