Op amps listening test with 96kHz/24bit files

as a follow-up to previous test

http://www.diyaudio.com/forums/everything-else/308514-another-big-opamp-listening-test.html

technically described in http://pmacura.cz/anothertest.htm

we have had a new test with the same op amps, but this time the source material was in 96kHz/24bit resolution, as requested by several participants of the first test. The test has been published in diyaudio as

http://www.diyaudio.com/forums/everything-else/309098-hires-96-24-listening-test-opamps.html

and the test files could be downloaded from http://pmacura.cz/test_hires.zip

The second difference from the previous test, in addition to hires music samples used, was higher voltage level at the input of the op amps under test, now it was 0dBFS = 2Vrms. So the opamps under test were exposed to higher signal slew rate, both for the reason of wider music bandwidth and higher signal amplitude.

1. Music sample

The music sample spectrum was extended well above 20kHz, as one can see from spectrum analysis

Fig.1. Spectrum analysis of the music sample used

2. Op amps under test

The opamps were tested in the same circuit as during the previous test. The recorded files were assigned to opamps as:

rr - LM4562

ss - OPA2134 first take

tt - MA1458

uu - TL072

vv - OPA2134 second take

One can see that the same OPA2134 was used in 2 sequentially recorded files.

3. Measurements

The biggest measurable difference between the opamps used was, this time, in THD at high frequencies. This was also for the reason of higher signal amplitude used, closer to the slew rate limit of the worst opamp, MA1458. Let's see spectra of 10kHz output at 1.4Vrms measured with the opamps in the D/A ==> A/D chain used. 10kHz signal with 2V amplitude (1.4Vrms) has maximum dv/dt of 0.125V/us. This is close to MA1458 slew rate, which is about 0.5V/us.

LM4562 and OPA2134 THD 10kHz just copied the spectrum of the D/A ==> A/D chain, so it was unmeasurable. But it is shown here for the record.

Fig.2. 10kHz THD of the chain with LM4562

Fig.3. 10kHz THD of the chain with OPA2134

Fig.4. 10kHz THD of the chain with TL072. We can see that distortion is rising.

Fig.5. 10kHz THD of the chain with MA1458. Distortion is about 15-20x higher that with the other opamps

4. Poll results - preferred file

rr - LM4562 - 1 vote

ss - OPA2134 - 2 votes

tt - MA1458 - 2 votes

uu - TL072 - 9 votes

vv - OPA2134 - 1 vote

7 people voted for "I can not hear a difference". So, the clear winner in the poll is the old TL072, and the best performers LM4562 and OPA2134 got the least votes, similarly with the old 1458. It was an interesting test, but in my opinion it did not tell much about op amp "sound", which is most probably non-existent as long as the part specs are not exceeded.

5. Preferences that arrived by Personal Messages

7 people sent me their opinion by PM, I appreciate it very much. From most of those messages it was possible to arrange the preferences in descending order, so I tried to make a chart:

We can see that there is no agreement in preferences of individual voters, with one exception, the 1458 never won and never totally lost. The others oscillated between "best" and "worst" ranking.

6. Conclusion

Based on the preferences that arrived in PMs and on the fact that no member was able to support his preference by a positive result in the ABX test, my personal conclusion is that there was no verifiable sound difference between the files in this test. This is at least interesting.