• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. 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!

Yes – opamps swapping may make a measurable difference in a preamp

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,591
Likes
10,727
Location
Prague
Yes – opamps swapping may make a measurable difference in a preamp

With my new measuring setup as described in
I have decided to make some measurements of opamps swapping influence to a link preamp parameters. The circuit chosen is my SE to balanced converter, it has single ended input, Alps volume potentiometer and a balanced output with 51 ohm output impedance of each +OUT/-OUT output terminals.
The preamp/converter was driven from Topping D10s by a 75 cm long single-ended link cable (RG-59) and it was loaded by balanced input of E1DA Cosmos ADC with 500 ohm balanced input impedance (or 250 ohm each +IN and -IN vs. common ground).
The opamps were tested at positions IC2/IC3 (or IC102/IC103 in the second channel) and the opamps tested were OPA134 with a JFET input and LM6171 with a bipolar (BJT) input.
These two parts are quite considerably different, regarding input current noise, input bias current, output current, speed, and load drive capability.
I do not want to get much into details but let's note that the much higher input bias current of LM6171 creates voltage drop across R4 (R104) and thus there is a permanent DC voltage at the preamp output, in our case 260mV at balanced output. On the other hand, LM6171 has much better load drive capability.

Preamp schematics is as follows.

SYM2_swapping_sch_sm.png


And this is how the sample looks like.

SYM2_module.JPG


Measurements

1. Measurements of THD and THD+N (SINAD) at 1kHz at 2.54V output


SYM2_E1DA_1k_2.54V.png


SYM2_LM6171_E1DA_1k_2.54V.png


SYM2_swapping_chart.png

We can see that with LM6171 we get both better THD and SINAD at 1kHz, though the differences are not very big

2. Next plots show THD and THD+N vs. output voltage with both opamps

SYM2_OPA134_E1DA_1k_thdnlevel.png


SYM2_LM6171_E1DA_1k_thdnlevel.png



3. Measurements of THD vs. frequency at 1.8V

SYM2_swapping_thdfreq.png


We can see that the main measurable difference is in distortion at higher frequencies, LM6171 has lower HF distortion than OPA134.

Conclusion

Before any opamp swapping, we need to study the circuit where the opamps are placed, to study the datasheets and check if the opamps are suitable for the considered circuit. Here, LM6171 brings much lower HF distortion, but adds quite high DC output offset due to its high input bias current.
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,591
Likes
10,727
Location
Prague
Or swapping may cause ringing or even oscillation! You need to have a high frequency O-scope and know how to use it before you start swapping.
I have it, 3 different oscilloscopes :).
 

Gregss

Active Member
Joined
Jan 26, 2021
Messages
160
Likes
208
I have it, 3 different oscilloscopes :).
Hello,

I have used the LM6171 op amps in projects over several years. Only once had a problem and it was easy to fix. Seems this is a pretty high speed amp and it can oscillate fairly easily. Putting a COG capacitor directly soldered across the top of the chip with the leads as short as possible, decoupled the power supply and eliminated any trace of the oscillation.

Personally like the results of using this amp. FWIW.

Regards,
Greg
 
  • Like
Reactions: pma

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
A measurable difference may not not result in an audible difference, which is the usual claim made by opamp swappers and most reported differences seem to be in areas where there is no direct measurement for.

And yes, all opamps do measure differently. As output stages in headphone amps substantial differences can occur (A symmetric) current limiting and increased distortion etc.
-90dB at 15kHz, for instance as there will never be a 0dB stimulus at that frequency in practice and is not something I would ever worry about.
Also 250/500 ohm loads are unrealistically low for something that drives a power amp or pre-amp.

Does 0.0003% or 0.00015% at 1kHz make an audible difference ?

Of course, inside a power amp we don't really know what the actual load is to an opamp in a circuit. Could be anything even largely capacitive.
Power rail decoupling, feedback resistance values, Miller capacitances, load decoupling, load current, gain, offset, input reference resistance, mimimal gain all play a role in proper operation of an opamp.
In many cases there are no real problems but sometimes swapping opamps can give all sorts of problems.

But yes opamps DO make measurable differences even when correctly applied. I don't think anyone thinks otherwise.
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,591
Likes
10,727
Location
Prague
A measurable difference may not not result in an audible difference, which is the usual claim made by opamp swappers and most reported differences seem to be in areas where there is no direct measurement for.
However, it has not been my claim (that audible difference). One can see that opamp swapping may make the same measurable difference as is seen between 2 different DACs and I do not think that anybody has suggested that DAC measurements would be pointless. It is the same case, just another components. I am sure that average reader is not aware of the fact that changing opamps may bring same difference in distortion as changing DACs. It is just the topic that is not discussed so often here.
 

DVDdoug

Major Contributor
Joined
May 27, 2021
Messages
2,916
Likes
3,831
However, it has not been my claim (that audible difference)
Then, why waste time & money?

I could build a preamp that goes into the MHz range and it's no better (for audio) than one that just covers the audio range. And with flat frequency response in the audio range, once noise & distortion are inaudible there can be no improvement in sound quality.
 

Gorgonzola

Major Contributor
Forum Donor
Joined
Jan 27, 2021
Messages
1,016
Likes
1,398
Location
Southern Ontario
People here at ASR will certainly scoff at subjective impressions, nevertheless I'll report on my trial of several different op amps in the VTV buffer of my VTV Purifi Stereo amplifier. I auditioned:
  1. Sparkos (discrete) SS3602
  2. OPA1612
  3. OPA627
  4. Burson (discrete) V6 Vivid
It seem to me that there were subtle differences in sound. FWIW I preferred the Sparkos SS3602. Again FWIW, this is the harmonic spectrum self-reported by Sparkos ...

SS3602_THD.jpg
 

solderdude

Grand Contributor
Joined
Jul 21, 2018
Messages
15,891
Likes
35,912
Location
The Neitherlands
Blind tested with 2 amps and the only difference being the usage of different opamps or a switch of opamps and subjective evaluation knowing what opamp is in there ?
Measurements of both situations (or rather all 4 mentioned) with at least the quality of measurements PMA did.
Measurements of the entire amp incl. dummy load ?

Posting a plot made by a vendor in not similar conditions isn't real proof.
I am sure all 4 opamps are excellent. I am sure all measure differently and have stronger and weaker points.

However, it has not been my claim (that audible difference)

I know... but the test fuels the ones claiming audible differences where the differences are far below any audible thresholds (as is the case with most DACs that have high SINAD and proper reconstruction filters, low noise and proper behavior.

I do not think that anybody has suggested that DAC measurements would be pointless.

No measurements are pointless as long as they are done correctly and the data is interpreted correctly. Even when measurements are beyond audible limits they are still worthwhile otherwise we do not know the technical performance.

I am sure that average reader is not aware of the fact that changing opamps may bring same difference in distortion as changing DACs. It is just the topic that is not discussed so often here.

That can be the case if parts around that opamp are capable of higher technical performance. When they are not its pointless. Just as pointless as changing an opamp in a totally flawed DAC design (think jitter/noise/poor design) is not going to make it a better DAC.

It also can be the other way around. One can make performance worse by changing an opamp. That may well even be audible and preferred while technical performance is worse.

Safe to say changing opamps changes performance of the device it is used in. Can be for better or for worse or of no consequence. Could be audible or not (tested blind) or even no audible in reality but as the late Peter W Belt's wife said:
It does not make technical changes but alters the perception of the listener.
 
Last edited:

MaxwellsEq

Major Contributor
Joined
Aug 18, 2020
Messages
1,627
Likes
2,422
I would expect swapping op-amps to result in a measurable difference in almost any circuit that has an appreciable frequency response. Manufacturer's data sheets confirm that different op-amps perform differently.
 

restorer-john

Grand Contributor
Joined
Mar 1, 2018
Messages
12,579
Likes
38,273
Location
Gold Coast, Queensland, Australia
Also 250/500 ohm loads are unrealistically low for something that drives a power amp or pre-amp.

Exactly. The total for each leg as laid out above is 300R (51R+250R).

The OPA-134 has a 600R drive rating and the LM6171 has a specified 100R rating. So I would expect the THD to rise for the 134, as it does, when it is driven at half its rated load.

The preamp would be driving a power amplifier with an input impedance of many k ohms in a typical scenario.
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,591
Likes
10,727
Location
Prague
The OPA-134 has a 600R drive rating and the LM6171 has a specified 100R rating. So I would expect the THD to rise for the 134, as it does, when it is driven at half its rated load.

You are certainly right John, but many opamps do not raise distortion as long as they work inside their rated current output. Like LME49720 in my Topping D10s output buffer, which has rated output current up to +/-23mA, though OPA134 has higher rated output current up to +/-35mA. Interesting, right? It is the output stage design and loopgain plot that do define distortion at higher current. In my setup, the output current was a mere +/-7.2mA per one opamp, so nothing is exceeded.

So this is how LME49720 works when loaded by 250 ohm (+ 20ohm output resistor), at +/-9.96mA output current. Please take into account that voltage is measured across the load resistor (behind the board series resistor), not at the opamp output.

Please also take into account that LME49720 has also "only" 600 ohm drive rating. This is for the swing reason, 10Vrms output into 600 ohm makes +/-23.5mA.

D10s_E1DA_0dBFS_1k_loop_262k.png

You may see that it is the general design of the opamp circuit that makes the difference. BTW, LME49720 is nowhere specified to work with 250 ohm load, however it is the output current that makes the limit, you just get lower output voltage swing. We did not need a high swing here.

Another interesting point, nowhere in the LM6171 datasheet one can see what is the distortion at audio frequencies, they just show THD+N plots from 10kHz up to 15MHz, and the part below 300kHz is of course buried in noise, due to very high measurement BW.
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,591
Likes
10,727
Location
Prague
FWIW I preferred the Sparkos SS3602. Again FWIW, this is the harmonic spectrum self-reported by Sparkos ...

I cannot see definition of measurement conditions - the load impedance. As such, the plots at the Sparkos site have mainly marketing importance. The measurement conditions must be carefully described, otherwise the plots are pointless.
 

restorer-john

Grand Contributor
Joined
Mar 1, 2018
Messages
12,579
Likes
38,273
Location
Gold Coast, Queensland, Australia
Another interesting point, nowhere in the LM6171 datasheet one can see what is the distortion at audio frequencies, they just show THD+N plots from 10kHz up to 15MHz, and the part below 300kHz is of course buried in noise, due to very high measurement BW.

Yes, I noticed the lack of distortion ratings.

I'd like to see the classic NE-5532/4AN if you get a chance.

Are you running the the Cosmos and the Topping simultaneously with one as an recording and the other as a playback device?
 

Mnyb

Major Contributor
Forum Donor
Joined
Aug 14, 2019
Messages
2,632
Likes
3,592
Location
Sweden, Västerås
Yes if one is as competent as pma , go ahead choose the proper op for the application after careful study of the circuit. and do this.

This is usually not the typical context this is done in :) i would say this is very atypical .

It is right to to point out that is analog circuits after the DAC chips in most products and these to have an effect on the total results, in bad cases more effect than the DAC chip in for example AVR's

Laypersons should not be suckered into any component swapping schemes at all .

Example I'm a bit skeptical to class-D amp sales sites where one can pick the kind of bufferstage one wants ? That's a service i expect the amp designer to do for me and i expect it to perform transparently etc and i expect them to competently choose one design.
When i buy a car no one ask which compression ratio I'm interested in or ratios between the gears ?
 

eliash

Senior Member
Joined
May 29, 2019
Messages
407
Likes
209
Location
Bavaria, near lake Ammersee
Yes – opamps swapping may make a measurable difference in a preamp

With my new measuring setup as described in
I have decided to make some measurements of opamps swapping influence to a link preamp parameters. The circuit chosen is my SE to balanced converter, it has single ended input, Alps volume potentiometer and a balanced output with 51 ohm output impedance of each +OUT/-OUT output terminals.
The preamp/converter was driven from Topping D10s by a 75 cm long single-ended link cable (RG-59) and it was loaded by balanced input of E1DA Cosmos ADC with 500 ohm balanced input impedance (or 250 ohm each +IN and -IN vs. common ground).
The opamps were tested at positions IC2/IC3 (or IC102/IC103 in the second channel) and the opamps tested were OPA134 with a JFET input and LM6171 with a bipolar (BJT) input.
These two parts are quite considerably different, regarding input current noise, input bias current, output current, speed, and load drive capability.
I do not want to get much into details but let's note that the much higher input bias current of LM6171 creates voltage drop across R4 (R104) and thus there is a permanent DC voltage at the preamp output, in our case 260mV at balanced output. On the other hand, LM6171 has much better load drive capability.

Preamp schematics is as follows.

View attachment 212199

And this is how the sample looks like.

View attachment 212200

Measurements

1. Measurements of THD and THD+N (SINAD) at 1kHz at 2.54V output


View attachment 212201

View attachment 212202

View attachment 212203
We can see that with LM6171 we get both better THD and SINAD at 1kHz, though the differences are not very big

2. Next plots show THD and THD+N vs. output voltage with both opamps

View attachment 212204

View attachment 212205


3. Measurements of THD vs. frequency at 1.8V

View attachment 212206

We can see that the main measurable difference is in distortion at higher frequencies, LM6171 has lower HF distortion than OPA134.

Conclusion

Before any opamp swapping, we need to study the circuit where the opamps are placed, to study the datasheets and check if the opamps are suitable for the considered circuit. Here, LM6171 brings much lower HF distortion, but adds quite high DC output offset due to its high input bias current.
Interesting, did a similar exchange in my T+A P1260R preamp, where they installed a crude video amp as output stage. This one had so much distortion (at low levels, poor open loop gain as well), I was convinced to perceive it on an absolute level. Anyway T+A had already spend some OPA211 in the first gain stage, so I installed OPA211 as well in the output stage. Being aware of the input current and 10MOhms bias resistor, I reactivated the abandoned layout traces for a DC control loop (reutilising one OPA2134 from the as well upgraded XLR output board). Dimensioning had become very slow, because of the 10M bias resistor and the coupling cap pole to maintain stability and unwanted noise suppression. Takes some 40s until the last 100uV DC-offset are gone on one of the OPA211 outputs, stabilising at around 3-4V control loop output voltage (output of one OPA2134 half). Btw. distorted impression gone...yeah I know, no measurements made...

before/original, w AD825 (~76dB openloop gain only), left piggyback boards, non populated DC/offset control loop traces below

1655039574835.png


afterwards/modified w OPA211 signal (~120dB open loop gain; left piggyback boards; pls. note brown 22pF input cap needed for stability below) , OPA2134 (dual OPA134) DC/offset control loops (left bottom)

1655039458318.png
 
Last edited:

Geert

Major Contributor
Joined
Mar 20, 2020
Messages
1,936
Likes
3,516
Given the results of the measurements and the feedback about load, maybe change the tread title to "No, opamp swapping makes no audible difference in a preamp"? That they can measure differently is no surprise.
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,591
Likes
10,727
Location
Prague
Given the results of the measurements and the feedback about load, maybe change the tread title to "No, opamp swapping makes no audible difference in a preamp"? That they can measure differently is no surprise.
No, because in such case we could give the same attribute to almost any low level component test like preamps and DACs and would say there are no audible differences. But, such statement would need to be supported by facts and evidence of such statement, like statistically demonstrable results of controlled DBT listening tests - do we have it? This is the difference from my thread title, which is supported by indisputable facts. And, the opamps under test are not operated out of specs. There is no parameter like "lowest allowed minimum load" in the datasheets, there is a parameter of maximum output current, and I was at about of 1/3 of this parameter.
 

Geert

Major Contributor
Joined
Mar 20, 2020
Messages
1,936
Likes
3,516
No, because in such case we could give the same attribute to almost any low level component test like preamps and DACs and would say there are no audible differences. But, such statement would need to be supported by facts and evidence of such statement, like statistically demonstrable results of controlled DBT listening tests - do we have it?

The evidence you provided via your measurements actually supports the statement. I think it's generally accepted at ASR that the distortion and noise numbers you came up with are below hearing thresholds, making different components audible 'transparent' as we call it. You might think some numbers are on the edge, but that doesn't warrant a title that looks like click bait to me. But you do as you please of course.

And, the opamps under test are not operated out of specs

Not out of spec's, but out of common use cases. Interesting to engineers, but not to the end user.
 

Gorgonzola

Major Contributor
Forum Donor
Joined
Jan 27, 2021
Messages
1,016
Likes
1,398
Location
Southern Ontario
No, because in such case we could give the same attribute to almost any low level component test like preamps and DACs and would say there are no audible differences. But, such statement would need to be supported by facts and evidence of such statement, like statistically demonstrable results of controlled DBT listening tests - do we have it? ...
The evidence you provided via your measurements actually supports the statement. I think it's generally accepted at ASR that the distortion and noise numbers you came up with are below hearing thresholds, making different components audible 'transparent' as we call it. You might think some numbers are on the edge, but that doesn't warrant a title that looks like click bait to me. But you do as you please of course.

Nevertheless I'm glad I can listen and choose what I prefer. Granted I won't get any 'Likes' for saying so.

The thing about real science is that it isn't deaf to what human beings report. ABX and hearing threshold tests aren't telling the whole story because they don't reflect the often very consistent subjective listening results from people who listen rather that just read papers and measure.

And part of the problem too is that so many of those on the high-horse of "science" aren't scientists but merely engineers.

Kudos by the way to Amir for more consistently reporting his subjective impressions at least of speakers.
 
Top Bottom