In Reply to: Re: I disagree with Roger Sanders' claim posted by bwaslo on January 5, 2007 at 08:39:08:
David Hafler tried to prove his DH-200 power amp was "perfect", by showing it could achieve a -60 dB "null" in the midband using a similar kind of technique, but unfortunately for him, no one else would agree that his amp was "perfect".The problem has multiple aspects. The largest one in my opinon is that most folks won't agree on what is an adequate amount of null or how far down the residual error signal has to be before something is adjudged "perfect". Some very well respected folks that used to post here on Prop Heads insisted that as long as error signals were more than approx. -40 dB down, they would be inaudible. However, the anecdotal evidence from the DH-200 seems to show that -60 dB is not enough (I have personally auditioned the DH-200 under controlled conditions, it was not "perfect", not even close; yet casual examination using a input/output comparison circuit would typically reveal an I/O error level below -60 dB across most of the midrange, falling to only about -45 to -50 at the frequency extremes).
Then there is the issue of any transit delay in the DUT, or any other form of phase shift or frequency roll-off that causes a phase shift anywhere near the audio band. With all but a few types of wideband line level electronics, these kinds of problems and issues can crop-up.
Some folks advocate allowing correction for all of the above, some take the opposite tack, do not allow ANY correction. Yet any DUT that involves such issues would have greatly varying results on a null type test, depending on what was permitted or accepted as 'allowable'.My own experience with such issues, is that a certain amount of pre-filtering should be allowed, as long as it does not start to enter into the audio band proper. Time delay issues are still a stumbling block, but my take on these are that correcting for them is OK if the situation is not real-time sound reinforcement, that is, playback of previously recorded material can have time delays allowed for.
Finally, based on my experience in the recording studio, error signals need to be BELOW -90 dB in order to even start to be considered as being in the realm of inaudible, and that it may be necessary for certain types of signal abberations to be below -100 to -110 dB before they could be said truly to be audibly insignificant. Additionally, most sound cards are NOT rated for any sort of jitter performance, and the tests that have been done by certain objectivists that claim to be jitter tests of sound cards, are not exactly the last word in that department.
The problem now is that if you look at what most computer based sound cards systems can do, merely altering how the input/output cables are placed, how long they are, exactly where they are at any given time, can cause changes in the I/O error detection at these levels and higher.
Part of this seems to be due to the huge amount of hash generated by most computer systems, and leaked into the sound card IO.Some of it seems related to a particular computer system's software details, leading me to speculate that different operating systems: DOS versus Windows, as well as different methods of accessing the sound system - ASIO versus WinCrappyXXX, etc. all affect the timing, repeatability and cleanliness of the overall performance of the sound card system.
If we effectively can not look reliably below -80 to -90 dB using a PC based measurement system without huge variations in the error residue, then the more subtle aspects of these issues will completely escape us and remain unknown. Worse yet, some fool will use such a system to demonstrate 'perfection' to -90 dB (just for that day, on that PC, with that sound card, for that particular button push), and declare sonic equality, end of story, we can all go home now. Of course, it would be a mistake, but that person would not be very receptive to the actual facts or truth of the matter, nor would they be likely to take into account not using simple purely resitive load resistors instead of actual audio components, nor would they be likely to be exposing the DUT to actual sound playback levels comnesurate with those in the home (or profesional) environment, thus a lack of typical operational vibration stimulus would be missing, and so on and so on.
There is a lot more to this "simple" idea than meeets the eye at first, but so many folks are all to willing to overlook everything but a single basic measurement criteria, and we are back to the banality of a THD measurement number faux pas.
Jon Risch
This post is made possible by the generous support of people like you and our sponsors:
Follow Ups
- Same problem all along - Jon Risch 19:16:40 01/07/07 (2)
- Re: Same problem all along - bwaslo 08:39:24 01/10/07 (1)
- Minus 140dB is measurable. - cheap-Jack 09:34:14 01/10/07 (0)