• Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Does Op-amp Rolling Work?

Rate this article on opamp rolling:

  • 1. Terrible. Didn't learn anything

    Votes: 9 3.4%
  • 2. Kind of useful but I am still not convinced

    Votes: 17 6.5%
  • 3. I learned some and agree with conclusions

    Votes: 53 20.3%
  • 4. Wonderful to have data and proof that such "upgrades" don't work

    Votes: 182 69.7%

  • Total voters
    261
However, the biggest change I heard from the new op-amps was harder hitting impulses, particularly from kick drums. No matter how much A/Bing I do, I come to the same conclusion: the new op-amps slam harder. If this is happening, the two explanations I have are a change in FR in mid-bass frequencies—which seems unlikely based on Amir's measurements—or a faster slew rate.
I think this is one of those cases where intuition has little to do with electronic reality. "Dynamics" or "transients" as musical phenomena have nothing to do with slew rate, which, in the unlikely event of it affecting the audio band (unless we're talking about some absolutely decrepit opamp like a ua741), will be caused by high-amplitude upper treble or ultrasonic frequencies, where you find the sparkle of cymbals rather than the thump of a bass drum or snap of a snare. Music generally has little energy at these frequencies, which further limits the practical importance of slew rate. When you are slew rate limited, the amp is overloaded and distorting, which with a complex signal will sound like distorted intermodulation crud rather than lack of "slam".
 
Last edited:
And about random OPA rolling and impact on FR, I found 5min to overslay 12 of the 90+ measurements I performed some time ago (with different combinations of OPA):

View attachment 437692

That was not in an amplifier, I must say, but in a low price DAC (OPA rolling possible in the I/V and LPF slots).
Bandwidth was not flat, stock, but none of the OPAamps had an influence on it (plot at 1k).

This is the missing piece in the tests.
This trace debunks claims about EQ-like changes.

1742532245841.png
 
However, the biggest change I heard from the new op-amps was harder hitting impulses, particularly from kick drums. No matter how much A/Bing I do, I come to the same conclusion: the new op-amps slam harder. If this is happening, the two explanations I have are a change in FR in mid-bass frequencies—which seems unlikely based on Amir's measurements—or a faster slew rate.
Slam is in the 150Hz range (as you already mentioned) and can't change. There is no 'mechanism' for that and would be very measurable.
The slew-rate of all the mentioned op-amps (the gain will be max. 3x or so) is many, many times higher than that of the output stage so the output stage is the limiting factor in slew rate. For op-amps in low gain mode the slew-rate is much higher than that of the output circuit.

The key here, most likely, is the test method used (knowing what amp is in there) and or the volume being changed.
When an op-amp in a circuit is changed out the gain will NOT change as it is determined by the components around the op-amp which have not changed.

Before some design designer pops in saying that isn't exactly true as (s)he knows the gain technically can change ever so minutely (hard to measure even) when one swaps out an op-amp with very low input current like a (MOS)FET input LTP and plug in a BJT input op-amp, with a higher input current, that will siphon away a very, very small amount of the signal resulting in a very, very small level difference.
 
Last edited:
No disrespect intended, but if this is 'My personal, subjective conclusion' from 'purely subjective and impressionistic findings', why should that be of any interest to any if us?

Jan
That's OK. As far as I know subjective impressions aren't banned on this site. I know some people would prefer that they were; if that's your position, take it up with the moderators. Meanwhile I will continue to occasionally post a personal impression and you may continue to totally ignore them.
 
"Dynamics" is a general audiophile term. It has no precise meaning.

Dynamic range is measured. That is the difference between quiet and loud.

Frequency response is measured. That shows the bandwidth of the system as to make sure it covers audio band.


Psychological. The impact of instruments is in the content itself. I have heard the same thing but once I confirmed nothing had changed, that perception disappeared. Or when I matched levels.
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?
 
Slam is in the 150Hz range (as you already mentioned) and can't change. There is no 'mechanism' for that and would be very measurable.
The slew-rate of all the mentioned op-amps (the gain will be max. 3x or so) is many, many times higher than that of the output stage so the output stage is the limiting factor in slew rate. For op-amps in low gain mode the slew-rate is much higher than that of the output circuit.

The key here, most likely, is the test method used (knowing what amp is in there) and or the volume being changed.
When an op-amp in a circuit is changed out the gain will NOT change as it is determined by the components around the op-amp which have not changed.

Before some design designer pops in saying that isn't exactly true as (s)he knows the gain technically can change ever so minutely (hard to measure even) when one swaps out an op-amp with very low input current like a (MOS)FET input LTP and plug in a BJT input op-amp, with a higher input current, that will siphon away a very, very small amount of the signal resulting in a very, very small level difference.
Thank you and to ask/ponder (and considering @seedragon impressions), with Slam occuring in the 150Hz range and to consider Harmonic decay where let's say 37.5hz (and it's Harmonics) decaying to 70hz then 150hz and so on and a Speaker can Authoritatively produce 37.5hz (within the room where it is placed). Could there be (measured) differences produced, re opamp swaping, around 37.5hz (and it's Harmonics)? If yes, can you clarify and if no, well No implys that the resulting Harmonic decay (from the Speaker) will be unchanged, won't it?

Edit:
Note: Below 100Hz we can see electrolytic capacitor distortion in the E1DA Cosmos ADC. It is no myth, it exists.
I have just noticed this posted by @pma, below. This suggests that there could definitely be a consideration to consider with what is posted above, doesn't it? and is that correct?
 
Last edited:
And THD vs. frequency and frequency responses.

Topping D10s LM1458-LME49720_FR.png


Topping D10s LM1458-LME49720.png

Note: Below 100Hz we can see electrolytic capacitor distortion in the E1DA Cosmos ADC. It is no myth, it exists.

Just to clarify for others, the LME49720 is the stock opamp in this DAC.


JSmith
Best to keep it and make no changes, from engineering point of view. Unqualified personnel should do no changes, no opamp swapping.
 
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?
To overheat an op-amp it would need to be providing quite a lot of current (so low load impedance and high voltage rails).
So, when using op-amps as output stages (driving headphones for instance) there certainly is a lot of (very measurable) and when pushed to one or more of its limits audible differences between op-amps. The 'Chu Moy' amps are a good example.
The discrete op-amps usually can provide a lot more current (power) and can certainly have audible benefits (due to higher power output) in situations where low impedance headphones are driven directly from an op-amp compared to regular DIP-8 op-amps that usually have current limited outputs. On top of that often the positive and negative rail even differ resulting in asymmetric distortion (even order).

That said, in the discussed amplifier there is no 'heavy load', it is just a simple buffer or few x gain circuit with a very low load so in this amp the answer is... that property of op-amps is not an issue.

Thank you and to ask/ponder (and considering @seedragon impressions), with Slam occuring in the 150Hz range and to consider Harmonic decay where let's say 37.5hz (and it's Harmonics) decaying to 70hz then 150hz and so on and a Speaker can Authoritatively produce 37.5hz (within the room where it is placed). Could there be (measured) differences produced, re opamp swapping, around 37.5hz (and it's Harmonics)? If yes, can you clarify and if no, well No implys that the resulting Harmonic decay (from the Speaker) will be unchanged, won't it?
The plots from Amir as well as those from PMA for the suitable op-amps show no issues at low nor high frequencies nor would any of those levels relate to 'more slam'.
More slam is perceived at either louder levels, but changing an op-amp does NOT change levels (are you reading this Danny ?) or increased 150Hz region. That would need an increase of more than 1dB, more like 3-5dB.
This (1dB increase in level for harmonics) would require like 30% distortion in the lowest frequencies or even more, not 0.001% and also not 0.1% so the answer is no won't be harmonic distortion.
Speakers can reach that but op-amps won't. Not even the uA741 will do this in this amp nor will ceramic or electrolytic coupling caps ever reach values higher than that of the output stage in the Douk.

Most likely scenarios are a difference in volume and/or because of 'knowing' what one listens to.

Unless ... there truly is 'magic' or 'unknown signals' or 'things we don't know yet that matters', or 'things we cannot measure yet', or 'we haven't measured all aspects.
Think the lack of frequency response plots which undoubtedly would not have changed anyway but as it has not been reported shown lets people speculate that there might be audible changes there.
 
Last edited:
To overheat an op-amp it would need to be providing quite a lot of current (so low load impedance and high voltage rails).
So, when using op-amps as output stages (driving headphones for instance) there certainly is a lot of (very measurable) and when pushed to one or more of its limits audible differences between op-amps. The 'Chu Moy' amps are a good example.
The discrete op-amps usually can provide a lot more current (power) and can certainly have audible benefits (due to higher power output).

That said, in the discussed amplifier there is no 'heavy load', it is just a simple buffer or few x gain circuit with a very low load so in this amp the answer is... that property of op-amps is not an issue.
Thank you for the enlightening words! My last straw of hope now is that the longer legs might provide greater thermal distance from the amplifier when it's under heavy load.
 
Once again, I am saying nothing about audibility. However, if it is legal to have a SINAD chart competition in this forum, with the audience waiting for new winners, is it illegal to show that opamps, depending on the application, do make a difference as well and may contribute to those SINAD charts wars?? What is the difference? The same approach.
I like your measurement approach, but sometimes I lose the thread of what you are trying to say and show.

I'm not interested in measurements of Op Amps in isolation, or any individual components. Do you have measurements of Op Amps that are commonly in use today, in an amplifier under normal operating conditions and demonstrating a material change to the output?

I'm cheerfully convinced that changing Op Amps is a futile activity (perhaps not for phono stages) but you seem keen to demonstrate otherwise. It's always cool to see evidence - but I don't get what you are showing.
 
Above you have measurements of several opamps in the Topping D10s DAC. Exactly what you ask for. The only opamp from that list in the post #549 that should never be used is the 1458. The others are standard dual opamps that have been used by nowadays. D10s has a socket - there is a question why Topping did that. You either can read something from the plots or not. As this forum is called Audio Science, I am showing the hard data. I am not interested in personal opinions other than supported by data and own work of the posters.
 
Above you have measurements of several opamps in the Topping D10s DAC. Exactly what you ask for. The only opamp from that list in the post #549 that should never be used is the 1458. The others are standard dual opamps that have been used by nowadays. D10s has a socket - there is a question why Topping did that. You either can read something from the plots or not. As this forum is called Audio Science, I am showing the hard data. I am not interested in personal opinions other than supported by data and own work of the posters.
That's great, thanks - no point in changing Op Amps, there are measurable differences but not audible and the stock Op Amp is the best of the bunch anyway.
Cheers :)
 
I am showing the hard data. I am not interested in personal opinions other than supported by data and own work of the posters.

Thank you for sharing! Much appreciated.
 
That's great, thanks - no point in changing Op Amps, there are measurable differences but not audible and the stock Op Amp is the best of the bunch anyway.
Cheers :)
Tell that to a hard core subjectivist and you'll get an answer like:
You can't measure everything you can hear.
Or this; the reason why blind test X with participants Y didn't show any audible differences is because they are too poorly trained (no golden ears), too poor listening equipment (speakers/headphones), that they had a bias to NOT hear any difference so they didn't hear any difference and so on.

Edit:
To be clear, I do not take the hard core subjectivist view.
 
Confirmation bias can cut both ways :)
 
Back
Top Bottom