Home Propeller Head Plaza

Technical and scientific discussion of amps, cables and other topics.

Re: Even if the software were perfect ....

Tom wrote:
"In the case of say 80 dB of resolvable level, lets remember that while one might be able to detect some things >80 dB down, that is like hearing something which is .01% of comparative Voltage level, one hundred million to one in power.. "

I realize that -80 and -90 dBFS are way down there, but first let's put them in proper perspective.

On a 16 bit capable digital recording/playback system, you have approx. 96 dB of theoretical dynamic range available. You can record the average signal levels at 0 dBFS, you have to leave headroom for transients, or looked at another way, you have to allow for the crest factor of the signal. Live music is one of the most unpredictable signal sources, and the crest factor can be enormous compared to common test signals. Thus, it was common practice to place the VU meter "0 dB" for digital audio recording systems at approx. -26 dBFS, thus -96 dBFS becomes -70 db VU.

Now with a known signal source, such as a prerecorded bit of music, or an electrical recording of music while monitoring a capacitor or cable, you have a much better idea of the maximum signal level, and can push the digital system near it's recording limits, but good instrumentation practice will tend to keep the signal level some -6 dBFS to -10 dBFS down from the absolute maximum of 0 dBFS. Obviously, if you clip, the difference test will have a lot of false garbage and would be invalidated altogether.

Given that most sound cards that are 16 bit are NOT accurate down to 16 bits, this is not a trivial thing, we are now down to a situation not that far from what I was writing about, a loss of the laast bit's worth of accuracy, and not going all the way to the MSB by keeping the levels down, we are now down to a 14 bit recording system. To expect that to be able to capture the subtle details of an analog system, is just not being reasonable at all.

Remember, when I was talking about the dither algorithms? They all live at -90 dBFS and below, but they ALL sound diferent. For the typical soundcard recording the output from a digital audio workstation, and toggling baclk and forth betwen the various dither choices, it is entirely likely that the subtle differences would be entirely lost, that the difference file would only have very low level noises in it, along with some of the soundcards own colorations.

You mentioned a microscope, but trying to test high performance audio components with a $200 sound card is like trying to check the accuracy of a well built and executed Lab-grade microscope, with a magnifying glass from the dime store. Not going to be very revealing.

Tom wrote:
"For the person who is sure they hear large differences between say two speaker cables, “if” those differences are electrical in nature, they would also be audible as a comparatively large difference between one end of the cable and another.
Same for amplifiers etc, what seems like an obvious audible difference with music is not produced by having two identical signals."

Actually, I was under the impression that the test method was to involve recording cable A's output, and cable B's output, and comparing them to each other. Same for the power amps, rather than trying to compare input to output for one device. As you noted, trying to interpret the I/O difference is going to be futile and frustrating, while comparing two different outputs is at least at the same signal level, and involves the same interface issues.

BTW, I will not be going to NSCA this year, sorry, I would have liked to meet you and chat.

Jon Risch


This post is made possible by the generous support of people like you and our sponsors:
  Kimber Kable  


Follow Ups Full Thread
Follow Ups
  • Re: Even if the software were perfect .... - Jon Risch 16:37:48 03/12/07 (0)


You can not post to an archived thread.