Joseph K said:
George,
Agree to that wholeheartedly. When increasing those caps, you reach a point of diminishing returns where the negative side effects get larger than the benefits.
Thanks very much for the interesting measurements. There's much in there that needs more than a casual glance. Although there was some criticism on Thorsten's sims , I think by and large he used realistic values.
One question: as far as I saw, you didn't try out having the "snubber" at the large cap, with the wiring between that combo and the (bypassed) "chip"?
Jan Didden
janneman said:
If you are skeptical about whether a modification could make a difference, you might not hear a real difference when comparing that mod to a reference because you are biased against the possibility of difference.
So from this you see that a blind test is not more reliable than a sighted test.
Related question: Why do the results of sighted tests often go against expectation? If expectation is strong enough to override our senses it seems this should never occur.
jeff mai said:
Jeff,
Yes, that is a concern. I have no ready answer to that. But I don't think it is black-and-white in that sense. These are concerns for ANY test, blind or sighted, which have to be dealt with or accepted as an uncertainty. But, what is CERTAIN, is that if you do a sighted test, the odds are stacked against you overwhelmingly big time.
jeff mai said:
The first sentence: One possibility is that expectations are based on things like price, looks, design, your own involvement, marketing claims, while blind tests are designed to isolate these issues and bring the actual sound quality to the forefront. So you have two judgements based on widely different attributes. It would actually be a coincidence if both would lead to the same preference.
Your last sentence: I don't understand that. Could you explain what you mean here?
Jan Didden
Not easy to know but my 5 year daughter knows it! She is listening to a song with lyrics of "hello" in at least 20 languages including "göteboska" = "haj på daj du" dialect around Göteborg.Upupa Epops said:Little off topic question : What is " Konnichiwa " and " Sayonara " ?
Pavel, it's in japanese, hello and goodbye.
jeff mai said:
Just a technical point of logic, please do not read anything else into this:
It's true that if one person cannot detect a difference, or has a bias against finding a difference it will not prove that there is no difference. However, if one person in a crowd can reliably discern a difference, then it can be said that an audible difference does exist (audible to some, at least), even if most others cannot.
Sheldon
janneman said:
Let me clarify as it appears you misunderstand both sentences. Why does a sighted test often result in a conclusion one does not expect? I've been witness to a number of listening tests where the outcome has resulted in rather large disappointment of an individual with a lot at stake. You'd think the self-deception mechanism would be hard at work here, keeping persons with a vested interest happy.
Joseph K said:
Excellent posts! Thank you!
Addressing the quote above: use a choke input power supply and/or full wave rectification instead of a full wave bridge. I intend to try both soon in a chip amp (it's standard practice in my valve amps), but need to source the chokes before I can try the latter option. I'll probably have some locally wound.
In the setup Joseph K used could you say that the added stray inductance is ~1nH per mm, count it twice since there is inductance in the ground plane as well.
The total inductance seemed to increase from 24nH to slightly above 100nH with a distance of 2inches ~ 50mm.
I like to have approximate values for everything.
It would fit approximately with the estimated inductance increase in my simplified test setup.
i wanna have a network analyzer as well.
Janneman: You claim that an audio circuit does not generate signals above audio range. Fine i can buy that. The thing not discussed though is for digital circuits and class D amplifiers. There this could be a big concern. A common clock requency in a DAC is 2.8MHz? This would create a resonant circuit with the popular decoupling 100nF and a ~30nH cap/stray inductance but this is maybe off topic in a chipamp area.
The total inductance seemed to increase from 24nH to slightly above 100nH with a distance of 2inches ~ 50mm.
I like to have approximate values for everything.
It would fit approximately with the estimated inductance increase in my simplified test setup.
i wanna have a network analyzer as well. Janneman: You claim that an audio circuit does not generate signals above audio range. Fine i can buy that. The thing not discussed though is for digital circuits and class D amplifiers. There this could be a big concern. A common clock requency in a DAC is 2.8MHz? This would create a resonant circuit with the popular decoupling 100nF and a ~30nH cap/stray inductance but this is maybe off topic in a chipamp area.
jeff mai said:
No, I think what you described is what I would expect. In a well prepared blind test, the self-deception isn't active; after all, there are no clues to trigger a perception one way or another. The listener is convicted to try to make sense of what he/she hears only.
Now, since the effect will cause a bias towards his/her own equipment or his/her own preference, there are several possibilities. Suppose there is an audible difference, then the biased preference can either be in line with the blind preference, or not. If there is no audible difference, I would expect a 50-50 score. In either case, globally, you have a 50-50 chance to get it right.
I would hazard the guess that the listener would be quite disappointed at anything less than a 90% score, since he/she is so strongly *convinced* that he/she can clearly hear the superiority of their own stuff. So there's great scope for disappointments, yes.
Jan Didden
jeff mai said:
Yes, but note that the heavvy current pulses discussed in your quoted post are the result of the very large capacitor directly following the rectifier. The choke doesn't do too much here. Or do you propose to skip the first cap? That wouldn't work, you need *some* initial cap as reservoir otherwise your rectified voltage will be much, much smaller than expected.
Jan Didden
hjelm said:
Hjelm,
Yes you are right, of course. But what would the effect be of supply ripple on the DAC chips? I don't know. I notice that in the things I read about it, must solutions involve distributed low-L capacitance at all the supply pins. I understand that some hi speed chips even have several supply pins, internally connected to different parts of the circuit, just for decoupling purposes.
Don't forget that we tend to think about supply as providing current to a load, but it really is a circulating current through the chip and through the supply. Local bypassing then short-circuits that circulating current, so that L of the supply lines cannot ring.
That's what I asked George before: you can have ringing at the supply but if you use a low-L cap at the pins of that GC, who cares what happens at the supply, as long as the chip pins are clean?
Jan Didden
I follow this topic more and more, as it is getting interesting since the postings from Joseph K (thank you very much, Joseph!).
Thats exactly my question at this point: what do you recommend with a detached psu, as I use it in my mini-gainclone?
Franz
BTW: I accept filters in the psu-path, but never in the signal path. I am sure, high frequencies are important for good sound reproduction. I can hear bats, at least, when more than one is present: the differences of the frequencies are in the audible area.
Thats exactly my question at this point: what do you recommend with a detached psu, as I use it in my mini-gainclone?
Franz
BTW: I accept filters in the psu-path, but never in the signal path. I am sure, high frequencies are important for good sound reproduction. I can hear bats, at least, when more than one is present: the differences of the frequencies are in the audible area.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.