AES Paper Digest: Do Audio Op-amps Sound Different?

[jan.didden]

Yes. I thought Stuart Yaniger's article in Linear Audio 'Testing, one, two, three...' was interesting as he used the opamps in a more usual role as buffer or gain-of-one driver. IIRC his threshold of detecting the opamps in the circuit was 5 or 6 in series. Double blind test although only one man's result, but a good example of how to design this type of test.

 

[restorer-john]

It would be fun to grind the numbers and edges off a or so dozen opamp pairs, paint the tops white and label them with just A,A, B,B, C,C pairs (where only one person knows the true identity) and send them around to 'op-amp' rollers to see what conclusions (if any) they came up with. They could plug them into their CD/DAC buffer stages (sockets of course), compare notes etc and then, down the track, reveal what the identity of each opamp was.

 

[Cosmik]

It would be fun to grind the numbers and edges off a or so dozen opamp pairs, paint the tops white and label them with just A,A, B,B, C,C pairs (where only one person knows the true identity) and send them around to 'op-amp' rollers to see what conclusions (if any) they came up with. They could plug them into their CD/DAC buffer stages (sockets of course), compare notes etc and then, down the track, reveal what the identity of each opamp was.

But you are asking amateur bodgers in uncontrolled circumstances to handle integrated circuits.

ESD events not only reduce assembly yields, but can also produce device damage that goes undetected by factory testing, and later, is the cause of a latent failure. These devices with latent ESD defects are called walking wounded because they have been degraded, but not destroyed, by ESD. This occurs when an ESD pulse is not sufficiently strong to destroy a device, but nevertheless causes damage. Often, the device suffers junction degradation through increased leakage or a decreased reverse breakdown, but the device continues to function and is still within datasheet limits. A device can be subjected to numerous weak ESD pulses, with each successive pulse further degrading a device until, finally, there is a catastrophic failure. There is no known practical way to screen for walking wounded devices. To avoid this type of damage, devices must be given continuous ESD protection...

http://www.ti.com/lit/an/ssya010a/ssya010a.pdf

So you are trying to detect minuscule differences using humans instead of test equipment, and on top of that possibly introducing unquantifiable *real* differences by damaging the devices in the process. Audiophilia is silly.

Edit: In fact... this could be an issue in *any* test using 'boutique' audiophile equipment. We can be pretty certain that ESD protection will be non-existent on many products. When it is plugged and unplugged, is cumulative damage being done? And as we saw in another thread, a certain brand of headphone amplifier damages headphones at power off and on. For all you statisticians and listening test aficionados, do your perfect experiments take into account the potential for degradation of the hardware due to simple bad design?

(I once had to review a circuit that was intended to work in a particularly demanding industry, and despite its expensive connectors and machined-from-the-solid enclosure, its input went to a CMOS logic gate input with *no protection* whatsoever. No ESD protection, no overvoltage protection, not even a series resistor or pull-up/down resistor, no EMC components; just a wire straight from the connector to the chip. And they hadn't even used the Schmitt trigger version of the logic gate. The designer had a highly professional demeanour, was highly respected and highly paid.)

 

[Wombat]

It would be fun to grind the numbers and edges off a or so dozen opamp pairs, paint the tops white and label them with just A,A, B,B, C,C pairs (where only one person knows the true identity) and send them around to 'op-amp' rollers to see what conclusions (if any) they came up with. They could plug them into their CD/DAC buffer stages (sockets of course), compare notes etc and then, down the track, reveal what the identity of each opamp was.

Similar to tube-rolling which can be an expensive and amateurish alternative to bias adjustment-not to mention the unnecessary depletion and price gouging of NOS tubes. Audiophool cred. stuff generally-look what I can do, I out-thunk the designer. 'I know SFA but look(hear) what I done'.

 

[Wombat]

It would be fun to grind the numbers and edges off a or so dozen opamp pairs, paint the tops white and label them with just A,A, B,B, C,C pairs (where only one person knows the true identity) and send them around to 'op-amp' rollers to see what conclusions (if any) they came up with. They could plug them into their CD/DAC buffer stages (sockets of course), compare notes etc and then, down the track, reveal what the identity of each opamp was.

Isn't that what boutique sellers do, the number reprinting, that is?

I wonder how many 'swappers' buy ICs on Ebay and unknowingly are commenting on the perceived differences of fakes?

 

[Arnold Krueger]

A simple learning exercise in test methodology and paper presentation, I feel.

Unfortunately, not the best that we can do today.

 

[Krunok]

I have noticed that more than a few DAC/preamp manufacturers are making claims that discrete DAC/preamp stage (after the DAC chip itself) is sounding significantly better comparing when made with OPAs. Specs don't really justify this thesis but they are still very good, so I find it really hard to believe that it makes any difference in sound. What do you guys think?


Example from Audio-GD: http://www.audio-gd.com/Pro/dac/NFB1.38/NFB1.38EN.htm

Simple and short signal transfer is best:
The NFB-1 applies the newest ESS high-end Sabre32 DA chip ES9038Pro which can support up to 32Bit/384KHz input .
ES9038Pro in voltage output model , there is -12DB THD+N worst , to be achieve the best performance of ES9038Pro , must work in current output model .
In some other ES9038Pro built in DAC design , ES9038Pro output signal have to through 3 step OPAs implement analogy signal process .
The NFB-1 had different design , applies Non-feedback ACSS design. The ES9038Pro feed to the ACSS circuit by balanced model without any OPAs .
The ACSS circuit naturally working in current model and have design difference current input , it can transfer the ES9038Pro current signal to output and implement the I/V conversion in current model in one ACSS amp . And this ACSS amp only have one current gain stage , the ES9038Pro output signal only through one stage can arrive the output stages .
The shortest signal journey can reproduce the best detail and dynamic, and almost without sound coloration , It can show the real performance of the ES9038Pro .
So our ACSS design is the best combo with ES9038Pro achieve the best performance .






Example from L.K.S. Audio MH-DA004: https://www.shenzhenaudio.com/l-k-s...bu-for-dop-usb-i2s-optical-audio-decoder.html

4. L.K.S Audio three years ago to proceed with the development of discrete circuits used to replace the traditional op amp chip, because the traditional op amp chip is subject to the chip size and SOC integrated circuit of some innate factors, the composition of the internal circuit of the monomer parts Performance is very limited, can withstand the heat power is very low, basically all of the integrated operational amplifier chip quiescent current is in a very low state (less than 10mA), even if the enthusiast of the senior opa627 gold seal version is the case, Processing ES9018 / 28/38 more and more audio output current (ES9018 / 28 in the mono mode output current is about 30mA or more, while the ES9038 is more than 100mA), so the overall quality of performance can not meet the demanding needs of enthusiasts The The output of the sub-output circuit, the use of twin JFET field effect transistor input, the last pole 3A power tube output, you can easily meet the ES9038pro high current output processing, and after three years of accumulated over 100 circuit parameters tuning, sound Performance is extremely natural, Chiang Kai-shek, high and low frequency of the ductility is excellent.

5. With a high standard of analog output circuit, In order to meet the best performance of this circuit, in particular the design of the same high-grade A (A) power circuit, constant current source SHUNT circuit mode design, ON Semiconductor high power Adjust the tube, and use Jensen new electrolysis, with the appropriate output capacitor and decoupling capacitor, natural sound, transparent, very balanced and very musical appeal.

 

[jan.didden]

You often see statements like this: "The shortest signal journey can reproduce the best detail and dynamic, and almost without sound coloration ".

This is in my view a somewhat naïve view. Firstly, in the limit, a single device is shortest and fastest, but will almost guaranteed color your sound due to distortions, possibly noise, frequency response deviations. When you take that one device and want to make it more "detail and dynamic, and almost without sound coloration", you need to add devices.....

Secondly, what is the value of a short signal journey at one specific place after a journey through dozens of stages (mostly opamps!) in the recording and mixing console, the tape recording or ADC-CD recorder, the DAC etcetera?

I can well believe that a discrete stage sounds different, under some circumstances, to a competently engineered opamp stage. But different does not mean better by whatever metric.

I have been privy to home recordings, directly through a home-brew opamp mic stage directly into a portable digital recorder. THAT, my friends, is what I would call a short signal journey and what gives you incredible detail, dynamics and uncolored sound.

 

[Krunok]

You often see statements like this: "The shortest signal journey can reproduce the best detail and dynamic, and almost without sound coloration ".

This is in my view a somewhat naïve view. Firstly, in the limit, a single device is shortest and fastest, but will almost guaranteed color your sound due to distortions, possibly noise, frequency response deviations. When you take that one device and want to make it more "detail and dynamic, and almost without sound coloration", you need to add devices.....

Secondly, what is the value of a short signal journey at one specific place after a journey through dozens of stages (mostly opamps!) in the recording and mixing console, the tape recording or ADC-CD recorder, the DAC etcetera?

I fully agree with this, except that I would add that IMHO during the music production stage the main source of the distortion is microphone, not the electronics.

I can well believe that a discrete stage sounds different, under some circumstances, to a competently engineered opamp stage. But different does not mean better by whatever metric.

You think our ears could tell the difference between this Audio-GD DAC and say, Topping D10 which has exactly the same THD of 0.0008% as this "discrete" DAC? I'm really not sure but I think they would sound equally well.

I have been privy to home recordings, directly through a home-brew opamp mic stage directly into a portable digital recorder. THAT, my friends, is what I would call a short signal journey and what gives you incredible detail, dynamics and uncolored sound.

As I said, as long as you're using high quality mic I'm sure your recordings will sound very well. Especially if the room you're taking the recordings in is well isolated from the outside noises.

 

[Arnold Krueger]

I have noticed that more than a few DAC/preamp manufacturers are making claims that discrete DAC/preamp stage (after the DAC chip itself) is sounding significantly better comparing when made with OPAs. Specs don't really justify this thesis but they are still very good, so I find it really hard to believe that it makes any difference in sound. What do you guys think?


Example from Audio-GD: http://www.audio-gd.com/Pro/dac/NFB1.38/NFB1.38EN.htm

Simple and short signal transfer is best:
The NFB-1 applies the newest ESS high-end Sabre32 DA chip ES9038Pro which can support up to 32Bit/384KHz input .
ES9038Pro in voltage output model , there is -12DB THD+N worst , to be achieve the best performance of ES9038Pro , must work in current output model .
In some other ES9038Pro built in DAC design , ES9038Pro output signal have to through 3 step OPAs implement analogy signal process .
The NFB-1 had different design , applies Non-feedback ACSS design. The ES9038Pro feed to the ACSS circuit by balanced model without any OPAs .
The ACSS circuit naturally working in current model and have design difference current input , it can transfer the ES9038Pro current signal to output and implement the I/V conversion in current model in one ACSS amp . And this ACSS amp only have one current gain stage , the ES9038Pro output signal only through one stage can arrive the output stages .
The shortest signal journey can reproduce the best detail and dynamic, and almost without sound coloration , It can show the real performance of the ES9038Pro .
So our ACSS design is the best combo with ES9038Pro achieve the best performance .






Example from L.K.S. Audio MH-DA004: https://www.shenzhenaudio.com/l-k-s...bu-for-dop-usb-i2s-optical-audio-decoder.html

4. L.K.S Audio three years ago to proceed with the development of discrete circuits used to replace the traditional op amp chip, because the traditional op amp chip is subject to the chip size and SOC integrated circuit of some innate factors, the composition of the internal circuit of the monomer parts Performance is very limited, can withstand the heat power is very low, basically all of the integrated operational amplifier chip quiescent current is in a very low state (less than 10mA), even if the enthusiast of the senior opa627 gold seal version is the case, Processing ES9018 / 28/38 more and more audio output current (ES9018 / 28 in the mono mode output current is about 30mA or more, while the ES9038 is more than 100mA), so the overall quality of performance can not meet the demanding needs of enthusiasts The The output of the sub-output circuit, the use of twin JFET field effect transistor input, the last pole 3A power tube output, you can easily meet the ES9038pro high current output processing, and after three years of accumulated over 100 circuit parameters tuning, sound Performance is extremely natural, Chiang Kai-shek, high and low frequency of the ductility is excellent.

5. With a high standard of analog output circuit, In order to meet the best performance of this circuit, in particular the design of the same high-grade A (A) power circuit, constant current source SHUNT circuit mode design, ON Semiconductor high power Adjust the tube, and use Jensen new electrolysis, with the appropriate output capacitor and decoupling capacitor, natural sound, transparent, very balanced and very musical appeal.

Pretty much 100% BS if your standard is a reliably audible difference in a proper DBT. The DACs in the first generation CD players were almost sonically transpareent, but by 1990 sonically transparent DACs were about a dime a dozen, figure of speech. Sorry, but you did
ask!

 

[Krunok]

Hahaha Yep, I did ask.

So it seems our cheap DAC champion (D10) for whose promotion is Topping continuously paying Amir huge monthly sums doesn't need the promotion at all, as all these expensive discrete "short signal path" DACs can't anyhow beat it in a proper DBT.

 

[oivavoi]

Just for the record, the Swedish Audio Society were able to identify a dac as recent as the Oppo Sonica Dac in a DBT about a year ago. But not with music signals though, AFAIR, only with clicking/ticking sounds. So the claim that no dacs have been spotted in DBTs is not correct, methinks. But they were not able to identify the rather cheap Yamaha WXC-50 in a similar test.

 

[Krunok]

Just for the record, the Swedish Audio Society were able to identify a dac as recent as the Oppo Sonica Dac in a DBT about a year ago. But not with music signals though, AFAIR, only with clicking/ticking sounds. So the claim that no dacs have been spotted in DBTs is not correct, methinks. But they were not able to identify the rather cheap Yamaha WXC-50 in a similar test.

Because Yamaha was not making clicks, right?

Well, if you're saying they were able to identify Sonica DAC ONLY by the clicking noises that kind of still proves Arnold's statement, at least in my eyes.

 

[Arnold Krueger]

Just for the record, the Swedish Audio Society were able to identify a dac as recent as the Oppo Sonica Dac in a DBT about a year ago. But not with music signals though, AFAIR, only with clicking/ticking sounds. So the claim that no dacs have been spotted in DBTs is not correct, methinks. But they were not able to identify the rather cheap Yamaha WXC-50 in a similar test.

If there is a record of those tests, where is it?

Google does not seem to be able to find it!

Can't even find a trace of "Swedish Audio Society". Maybe someone is not taking this very seriously?


I feel silly pointing out something so seemingly obvious that the listening test needs to be based on commercial recordings, not made-up test signals. They are little more relevant than turning up the gain by 60 dB to hear some artifact.

 

[oivavoi]

If there is a record of those tests, where is it?

Google does not seem to be able to find it!

Can't even find a trace of "Swedish Audio Society". Maybe someone is not taking this very seriously?

I feel silly pointing out something so seemingly obvious that the listening test needs to be based on commercial recordings, not made-up test signals. They are little more relevant than turning up the gain by 60 dB to hear some artifact.

Here is their website: http://www.lts.a.se/

It's formally called Ljudtekniska sällskapet, which you could translate as "The society for audio technology" (I've previously translated it as the Swedish Audio Engineering Soceity, but I don't think it's a good translation).

Their measurement reports and listening tests are documented in their printed magazine in Swedish, which is not peer-reviewed, and unfortunately not available online either. But all the testing is run by this guy, who does acoustic and psychoacoustic research for a living: https://www.kth.se/profile/ssg/

In short: There is no reason to rubbish these tests.

They also tested the dac in the Sonos connect, btw, and were able to identify it when playing music because of some frequency changes in the bass department. And I believe they have been able to identify some other dacs as well. I only began reading their magazine last year, so I don't have access to their previous testing.

On the other hand they recently ran measurements of the dacs in the most common smart phones on the market, and found that most of them had measurements which were on a par with quite a lot of "audiophile" dacs.

 

[Krunok]

Speaking of the THD (and IM distortion) - what is actually percieved as a noticeable distortion? Let's assume I have integrated amp which measures 0.08% of THD and IM distortion over the hearing freq spectrum. Would I be able to tell the difference if I swap it with the integrated amp that has say 0,008% of THD and IM distortion?

 

[oivavoi]

Speaking of the THD (and IM distortion) - what is actually percieved as a noticeable distortion? Let's assume I have integrated amp which measures 0.08% of THD and IM distortion over the hearing freq spectrum. Would I be able to tell the difference if I swap it with the integrated amp that has say 0,008% of THD and IM distortion?

My intuitive hunch: you would never notice it or think about while listening to music if someone exchanged them without you knowing, but you could maybe be able to spot the difference in a DBT if you trained for the test and knew exactly what to listen for.

 

[Krunok]

My intuitive hunch: you would never notice it or think about while listening to music if someone exchanged them without you knowing, but you could maybe be able to spot the difference in a DBT if you trained for the test and knew exactly what to listen for.

Maybe.. But frankly, I think I wouldn't. Because I'm 54 and I can notice the difference between mp3 320kbps and FLAC in a DBT only on my favorite songs when I know exactly what to look for. And I think the difference between 320kbps mp3 and flac is larger than the difference in distortion between 0.08% and 0008% .

 

[RayDunzl]

0.08% converted to decibels is -61.9dB - Within the possibly audible range, depending on the signal and circumstances.
0.008% converted to decibels is -81.9dB - You decide.

http://www.sengpielaudio.com/calculator-thd.htm

Shoutometer guestimation:
-60dB is the difference of someone shouting at you from a meter and from a kilometer
-80dB distances the shouter to about 9 kilometers

 

[Krunok]

0.08% converted to decibels is -61.9dB - Within the possibly audible range, depending on the signal and circumstances.
0.008% converted to decibels is -81.9dB - You decide.

http://www.sengpielaudio.com/calculator-thd.htm

Shoutometer guestimation:
-60dB is the difference of someone shouting at you from a meter and from a kilometer
-80dB distances the shouter to about 9 kilometers

Hahaha

I know how to convert THD % to dB but this shoutometer comparison is really giving things a new perspective! Does it have something to do with the real shouting scenario in the otherwise quiet space (mountain for example) or your figures were just for fun?