Does Op-amp Rolling Work?

[Buckeye Amps]

The Marantz and Mcintosh gear are part of my collection which friends in professional circles come to enjoy and reminisce about 'the good old days'.

Nothing wrong with that at all! Keep enjoying it.

 

[Jose Jimenez]

Nothing wrong with that at all! Keep enjoying it.

Thank you.

 

[Cbdb2]

I'm also pointing out there is no way to know for sure, because the only real way to know is to carry out blind testing, and that is not possible with this type of modification.

Actualy its easy to blind test stereo gear by doing mods only on one channel first and listening. But nobody does that because they think a mod always improves things.

 

[Jose Jimenez]

I believe!

 

[Cbdb2]

I replaced the LM201 in the front end with newer Burr Brown/TI devices and the apparent difference was noticeable.

LM201 has a slew rate of only .5V/us which is only good for about a 4 volt output at 20khz. Depending on the application, that might cause some slew limiting distortion, which can be nasty. Not all opamps are built the same, there designed for different applications. Its the job of the designer to choose the best one for the application. This means most other opamps are not the best for that application. And thats why opamp rolling is almost always a waste of time. Using your ears to pick opamps is also a waste of time.

 

[Cbdb2]

That really is half a thumbs up! Perhaps not everyone. The Model 15 and 16s are two separate mono chassis amplifiers and that is how it was done, while friends wer over and enjoying some Buffalo Trace. Over the years you learn to pick up what is important, what the ear considers good and not all the specsmanship.

Please dont modify other peoples posts. Dont understand what your saying. Did you mod one amp and compare it to the other?

 

[notinuse]

Wow, a currently 51 page message thread about psychoacoustics.

 

[SnakeLord]

Using your ears to pick opamps is also a waste of time.

You might even go nuts while doing it!

Not joking, most of those people have no idea about the absurdity of what they are actually doing and activly try to find/imagine those alleged audible diffences because a bunch of snake oil sellers and fanatic believers keep the snowball rolling.
If you do the swapping long enough, you will eventually lose track and those op-amps will just start to sound all the same.

Btw. here are some nice test results made by e3 on their A5 amp showing how some of the widley used op-amps are performing on the device. You can see that the OPA1656/1612 are performing best in this specific test and application while the NE5532 is clearly beaten in THD+N and power reserves before clipping sets in.

While the noise floor/distortion might be ignored as inaudible at those levels anyways, the values of the output power seem not to be insignificant - i see at least >10W more power comming from the OPAs, so in that case its around +10% clean power gain only from swapping the "right" op-amps.

 

[solderdude]

That is assuming one is driving the amp to clipping levels and even then that is only a 0.4dB difference in max loudness of peaks in music.
Not many people will compare the 'sound' of amplifiers near or at the clipping points anyway.
To get the idea of a device sounding a bit louder than another device you need a few dB more power, not just half a dB.

While most op-amps perform very similar in many aspects (when in a low-gain circuit) op-amps do differ more near their clipping points.
This has to do with how close the output (and in some cases input) the signal voltage can come to the internal voltage rails of that op-amp.

This aspect certainly can NOT be responsible for the (by the brain 'imagined') differences in sound aspects that are said to always occur when the amp is not anywhere near their clipping points.

 

[antcollinet]

op-amps do differ more near their clipping points.

But surely the clipping of a power amp has (should have if properly designed) nothing to do with the clipping point of any op amp in the circuit - but is caused by the power amp running out of volts or amps from the power supply. It seems particularly unlikely that an op amp circuit clipping point (which will be defined by the op amp power supply) is going to be anywhere near coincident with the power amp clipping point.

To be honest I'd be surprised if the difference in clipping point shown in that chart was caused by any characteristic of the particular op amps used.


Edit: Having thought about it more - the increased distortion of some of the op amps (visible between about 10W and 100W can increase the peak voltage of the waveform, to cause earlier clipping. It is illustrative that those with higher distortion clip at lower power levels.

Though it is also interesting that there seem to be two distinct clipping points which don't match the different levels of distortion variation. Which might indicate slight test setup variations between two groups of tests.

 

[solderdude]

Good point.

If the rail voltage of the op-amps were the limiting factor than the amp would not be able to put out more power at a higher than 38V supply voltage.

Given that the knee is sharp at only 2 different points it is more likely that 2 different power supplies were used during the test or the power supply decided to output a lower voltage having become too warm after many test runs at max power ?

When the TPA3255 is used in PFFB mode one is more reliant on the gain of the pre-amp opamps as that needs to have a higher output voltage than 7.8Vpp

 

[MaxwellsEq]

You might even go nuts while doing it!

Not joking, most of those people have no idea about the absurdity of what they are actually doing and activly try to find/imagine those alleged audible diffences because a bunch of snake oil sellers and fanatic believers keep the snowball rolling.
If you do the swapping long enough, you will eventually lose track and those op-amps will just start to sound all the same.

Btw. here are some nice test results made by e3 on their A5 amp showing how some of the widley used op-amps are performing on the device. You can see that the OPA1656/1612 are performing best in this specific test and application while the NE5532 is clearly beaten in THD+N and power reserves before clipping sets in.

View attachment 524009

While the noise floor/distortion might be ignored as inaudible at those levels anyways, the values of the output power seem not to be insignificant - i see at least >10W more power comming from the OPAs, so in that case its around +10% clean power gain only from swapping the "right" op-amps.

This is not entirely surprising.

I think everyone who has covered op-amp theory at college or designed real-world circuits with op-amps in, will confirm that different types of op-amps work differently...

... Which is precisely why I advise AGAINST op-amp rolling. Without precision instruments you don't know what you are getting (including increased oscillation, noise or distortion).

The graphs show some unsurprising differences when the signal approaches the rail voltages. The differences are not sufficient to make one implementation sound louder, but we know tiny variations in level, which don't sound louder, can make one version sound more exciting than the others. Only at clipping, however - at lower levels the differences are too small to matter.

 

[SnakeLord]

Not many people will compare the 'sound' of amplifiers near or at the clipping points anyway.
To get the idea of a device sounding a bit louder than another device you need a few dB more power, not just half a dB.

Im not talking about sound improvments as in "moister bass" or "bigger soundstage", its not that simple in this case.

There is clearly a total power performance gain thru the op-amps which in the end will be measured and land on the spec sheet of the device and this is an important one!

It does matter when you compare device A with ~95W and device B with ~110W of total clean power output at similar price point. Considering the fact that all the feauters are similar, as a customer you will go for the better performing one 99% of the time.
So in case of the A5 with higher performing OPA16xx you actually get an "better" device compared to the NE5532s.

 

[SnakeLord]

Though it is also interesting that there seem to be two distinct clipping points which don't match the different levels of distortion variation. Which might indicate slight test setup variations between two groups of tests.

That would be interesting to investigate and try to replicate e3 A5 data shown on the graph.

Unfortunately i do not have the needed tools nor the knowledge.

 

[solderdude]

Im not talking about sound improvments as in "moister bass" or "bigger soundstage", its not that simple in this case.

There is clearly a total power performance gain thru the op-amps which in the end will be measured and land on the spec sheet of the device and this is an important one!

It does matter when you compare device A with ~95W and device B with ~110W of total clean power output at similar price point. Considering the fact that all the feauters are similar, as a customer you will go for the better performing one 99% of the time.
So in case of the A5 with higher performing OPA16xx you actually get an "better" device compared to the NE5532s.

No one can hear a difference in peak loudness of 0.4dB so the measured difference (which may not even be caused by the op-amps !) is moot in real life IF the small shift in clipping point is indeed caused by the op-amp.

The question here is whether or not the measured differences are caused by the op-amp or that something else was the cause of the observed drop/increase in max output power and not the actual op-amp swap.
This would have to be re-checked by only swapping a 'higher power measurement' op-amp and 'lower power measurement op-amp' back and forth at least 2 times to see what the cause was.
For now the shift in measurement is suspect and may or may not be op-amp dependent.

 

[Cbdb2]

It would be strange (a bad design) if the little extra output (rail to rail) of the OPA is the cause of the distortion difference. A proper design would have the OA running at a few volts for max amplifier output, not maxing out the OA.

 

[solderdude]

That could only happen if the +/- voltage rail for the input op-amps would not be regulated and simply a percentage of the power supply voltage AND there would not be enough headroom when op-amps were used that should be rail-to-rail out.
This is unlikely.
The most plausible explanation is that some other aspect is responsible for the difference in clipping point.
The ONLY way to find out is to repeat the test with one 'low clip point' op=amp and one of the 'higher clip point' op-amps and test that twice.
I don't think 3E is going to repeat the test with the 3251 design (all 3255 now).

FWIW the test Amir did between the 5532 and 994Enh-Ticha showed the exact same clipping points (as one would expect).
I reckon the E3 test was incorrect in some way or that amp was poorly designed (unlikely).

 

[SnakeLord]

Found something interesting here:

2. Op-amp working voltage should meet 36V(+/-18V) maximum,34V Minimum.
7. Op-amp supply voltage equal or lower 30V, a dedicated setting(J30) is MUST to set supply voltage to 30V, refer to below picture.

x5532x seem to have lower supply voltage of 30V(+/-15V) compared to OPA15xx 36V(+/-18) which would make the rule 7. apply.
No idea if this would cause the clipping behaviour. Experts?

Btw. for example Bursons V7s are also operating at 30V(+/-15V) which only shows that you just shouldnt randomly swap electrical parts of an finished circuit.

 

[amirm]

Experts

The buffer circuit with op amp runs at much lower voltage that output rail. So that is not an issue at all.

 

[SnakeLord]

The buffer circuit with op amp runs at much lower voltage that output rail. So that is not an issue at all.

So are there any other plausible reasons for the slightly better results of the OPA15xx besides some change of the testing conditions possible?