Another big opamp listening test

[PMA]

Did you read Samuel's original, it's free. IMO his common mode test is unrealistic (for bi-polars) and is akin to exaggerating an effect in order to measure it..

Agreed at 1000%, Scott.

 

[PMA]

Hey Karl, you remember seventies, right? I think you would be about my age (60+). In the seventies, when I started with my professional career, the usual audio link level was about 200mV (not 2V as now). So, even that bloody 1458 was fast enough to make it. Only it was a bit too noisy, and it still is. So, the key is to use the part inside the range where it does not go out of specs. That's engineering 😀

 

[ErnieM]

Interesting. I have a studio channel strip designed with LME49860's in critical spots. Wonder if they do listening test or design by measurements.

 

[Mooly]

I do remember DIN standards so fair comment re 'engineering' and yes, at 55 I'm catching you up Pavel 😉

 

[PMA]

OK, the next test would be in true hires, 96/24, frequencies well above 20kHz, same opamps, renamed, 2Vrms. Is it OK?

 

[Mooly]

That would be great Pavel

I also know how time consuming organising these things can be, so many thanks for doing this.

 

[Mooly]

One question.

In your circuit diagram I note that the output is DC coupled. Does the varying offset from each device affect the AD process ?

 

[PMA]

One question.

In your circuit diagram I note that the output is DC coupled. Does the varying offset from each device affect the AD process ?

No, because the A/D input is AC coupled. And the only one with some DC output was MA1458, another 3 opamps were <=2mV.

 

[Mooly]

Thanks, I just wondered on that one.

 

[PMA]

Karl, I am receiving questions if it would be possible to hide votes in the poll from public, before the poll is closed. I do not know, but you, as a moderator, do you have an idea? Shall I send a PM?

 

[maty tinman]

Pavel, maybe is a good idea Lady stardust (David Bowie, 1972), from a very good vinyl rip. In three of my four rips this theme has DR15.

foobar2000 1.3.9 / Dynamic Range Meter 1.1.1
log date: 2017-04-28 20:44:35

--------------------------------------------------------------------------------
Analyzed: David Bowie / Ziggy Stardust {UK, MainMan}
--------------------------------------------------------------------------------

DR Peak RMS Duration Track
--------------------------------------------------------------------------------
DR11 -0.26 dB -15.93 dB 4:43 ?-A1 Five years
DR13 -0.31 dB -17.52 dB 3:35 ?-A2 Soul love
DR13 -0.80 dB -16.42 dB 4:43 ?-A3 Moonage daydream
DR11 -0.27 dB -15.09 dB 4:14 ?-A4 Starman
DR12 -2.11 dB -19.42 dB 2:57 ?-A5 It ain't easy
DR15 -0.55 dB -18.06 dB 3:22 ?-B1 Lady Stardust
DR13 -0.77 dB -15.74 dB 2:47 ?-B2 Star
DR13 -0.73 dB -15.31 dB 2:41 ?-B3 Hang on to yourself
DR13 -1.23 dB -16.79 dB 3:14 ?-B4 Ziggy Stardust
DR14 -0.65 dB -15.85 dB 3:30 ?-B5 Suffragette City
DR12 -0.34 dB -17.38 dB 2:58 ?-B6 Rock 'N' Roll suicide
--------------------------------------------------------------------------------

Number of tracks: 11
Official DR value: DR13

Samplerate: 96000 Hz
Channels: 2
Bits per sample: 24
Bitrate: 2835 kbps
Codec: FLAC
================================================================================

 

[PMA]

Hello Maty,

sorry, but not a vinyl rip. Technically, any vinyl rip is a disaster and you can always tell it with headphones - the way the guys mostly listen. Look at my link to Fremer's recent test of tonearm cables. Vinyl rip is just good for 12-bits dithered, anything more is wasting web space. I will probably use Linn 96/24 sample. I am not a tube/vinyl guy and to me this combination is a worst possible sound regarding sound accuracy. And quality, for me, equals accuracy.

 

[stocktrader200]

my 2 cents : OPA2134 and LF353 for input impedances > 10k ohm
LME4562 is great for inputs less then 10k ohm
LF353 load > 3.3k ohm

Are any of these on the list?

PMA , if your Vinyl is only good for 12 bits then it is very dusty or your preamp or cartridge is shot.
good Vinyl sounds as good as DVD ( 24bit 96khz) with a bit more noise

 

[billshurv]

Another priceless. I needed cheering up.

 

[Markw4]

Pavel, In addition to those things you said so far about plans for the next test, I would like to offer a few items to maybe consider:

1. It would be nice if there was somewhere people could privately send listening impressions prior to the end of the poll. Then maybe people would be less inclined to blab early. In addition, if somebody does blab, then there is a record of who described listening impressions prior to the blabing. That way there should not be suspicions that the earlier reports were biased by hearing the blab. And the impressions that were received can be revealed at the end, along with the times tamp when they were received. I think this type of information might be important because it could turn out that listening impressions correlate better with measured opamp distortion than poll votes do. Especially, if people are asked to vote for the one they prefer, which could be the most distorted one, depending on source material and other factors.

2. Since I think a few people remarked that cymbal tails often tend to be revealing, source material that offers some clear, high quality, out in the open, cymbal hits may help listeners differentiate in what may be a very difficult test. No sense in handicapping listeners, if we want to see the best they can do.

 

[billshurv]

People will always blab sadly. They will also load the files into a DAW to compare and some will outright cheat to try and show their listening prowess. Is the nature of the internet.

 

[Markw4]

People will always blab sadly. They will also load the files into a DAW to compare and some will outright cheat to try and show their listening prowess. Is the nature of the internet.

That may be true, but there may be some ways to help identify such people. And also ways to help discourage bad behavior. Won't go into any more details on that.

Fact is, most people are mostly honest, and only slightly dishonest. Or as Dan Ariely describes it, most people cheat, but only a little, not so much that they can't easily reconcile it as being okay. I'm talking about things like driving 5 mph over the speed limit, or saying, "no, you don't look fat in that outfit." Whatever. We think such minor cheats are usually okay, and we aren't doing anything bad.

In addition, most people won't go to a lot of trouble to cheat, in part because they may feel more and more wrong about it the harder they have to work at it. Or maybe it could be described by saying, if one has to go to a lot of trouble to cheat, it gets harder to convince one's self it's okay, and that it doesn't really amount to true, full-blown cheating.

Seeing how this is the internet, all we can do is try to do is the best we reasonably can to run a good experiment, and if cheaters want to spoil things, I guess they can do that.

Doesn't mean we shouldn't try to run a well designed test though. It's good to get some experience making the best test designs one can.

Also, if you look at what we have after this last poll, we don't seem to have a lot of cheaters bragging about listening prowess. It doesn't seem that way to me, at least. In addition, I'm not sure if some people have all the know how to do sophisticated cheating, although others surely do.

What we do seem to have is people from both sides saying, "see, it's just like I said all along. I was right."

So, I think I would like to worry more about reducing opportunities for ambiguity in interpreting the results next time. Then if it looks like cheating is getting to be a bigger issue, maybe have to work some more on that then.

 

[jcx]

more complicated but I believe some have setup automatic file randomization - each request gets new randomized labels and a random id number

only the host knows the keys - one person's aa may be another's cc

 

[billshurv]

Oh I think these tests are great and really help dispel some audio myths. Karl's 'copper or banana' test in particular! But we did have someone here who used to post 10/10 on foobar abx tests. He was rumbled when he got 10/10 on 2 bit identical files. Oddly he doesn't post here much now. But anything that helps people open their ears and both listen and think about what ACTUALLY matters with audio is really positive in my book.

 

[PMA]

Oh I think these tests are great and really help dispel some audio myths. Karl's 'copper or banana' test in particular! But we did have someone here who used to post 10/10 on foobar abx tests. He was rumbled when he got 10/10 on 2 bit identical files. Oddly he doesn't post here much now. But anything that helps people open their ears and both listen and think about what ACTUALLY matters with audio is really positive in my book.

With the new foobar ABX plugin, cheating like this is close to impossible, though it was possible in the old version.

The process is>

- you make a test
- save signed foobar ABX result as a txt file
- verify your result in foobar ABX log signature online
- post here the ABX result not like a screenshot, but as a file as follows

foo_abx 2.0.2 report
foobar2000 v1.3.7
2017-06-13 13:38:57

File A: aa.wav
SHA1: 81e0cc3b0146df3562d4421385098251239b3789
File B: bb.wav
SHA1: 4da5d9045285da07488f69c24cda9187be2882c9

Output:
WASAPI (push) : Speakers (Cambridge Audio USB Audio 1.0), 24-bit
Crossfading: NO

13:38:57 : Test started.
13:39:28 : 01/01
13:39:40 : 02/02
13:40:03 : 03/03
13:40:44 : 04/04
13:40:54 : 04/05
13:41:16 : 05/06
13:41:30 : 06/07
13:41:39 : 06/08
13:41:39 : Test finished.

 ---------- 
Total: 6/8
Probability that you were guessing: 14.5%

 -- signature -- 
26994466d1ad7e78f19f720799029ed6752abf23

- anyone can copy your result (the whole text file) and paste it into box at
foobar2000 ABX Log Signature Tool

and push "Verify signature" button to see if the result is valid. The result is protected by SHA codes and the unique signature.