janneman said:
How do you account for the non-stationarity of the recording environment? I assume digitally comparing a CD ripped w/wo a green pen yields nothing, or maybe we should not go there.
No, just looking for a few good engineers, who will learn what we already know about the subject, before they come up with their own opinions.
scott wurcer said:
It would be easy to compare rips from the same CD (not painted or painted) and see if there is a difference. If there isn't any than it's probably green pen that makes difference.
john curl said:
Well yes that's what I also would think in your place. But if you actually think about it and listen to it, it sounds like a noisy version of the original, clearly understandable. Now, non-linear distortion causes harmonics, so I would think those harmonics would not sound like the original tracks, would they?
You should take a listen yourself John, see what you think.
Jan Didden
Well, no process is perfect 🙂. But he does mention that the EQ is acquired using a log-swept sine. When doing such an EQ measurement, the distortion components come out at negative time in the impulse response, so windowing in the time domain can be used to remove the effect of nonlinear distortion on the frequency response calibration.
From the manual:
Here is an article that describes how the swept sine technique helps reject the influence of distortion on impulse response measurements.
From the manual:
Here is an article that describes how the swept sine technique helps reject the influence of distortion on impulse response measurements.
vuki said:
Done that. Linux has a diff tool never got a single bit difference until some sandpaper was applied.
On reconsideration, I think that you are right, Jan, in THIS specific measurement. What I was saying is that, without difference testing, you would hear the ceramic non-linear distortion first, as the ear is more sensitive to it. It apparently takes a lot of DA to be really obvious, but I would not be surprised if the ceramic had .1-4% DA distortion, depending on loading and test signal.
scott wurcer said:
Sorry, I didn't mean "rips" made from PC CDROM, but tracks made from audio CD player dig. or analog output.
andy_c said:
Andy I'm not sure what you're driving at? The windowing and harmonics attenuation is only applied to the EQ file, not to the outcome of the differecing itself?
Jan Didden
john curl said:
Yes, Bill did mention that his test file was with 100k load to the ceramic, and that with a 1 meg load the distortion was much, much less.
Jan Didden
janneman said:
Yes, and that's because it's very hard to define a consistent and simple metric that the software can work on to minimize.
To John's caps matching, 0.001% means roughly 16bit. I wonder if 25 years ago there were any A/D converters to measure beyond 16bit the "linear distortion" effects. And if these are, in fact, audible at all. For the purpose of calculating the SPL, 16bit ref. 1V mens roughly 10uV, or 1.25*10^-11 W into 8ohm, or a SPL of -218dB
andy_c said:
This is interesting. The reason for my comment was that Greiner claims (sorry lost the references) that you can't count on speaker/room/mic stationarity better than -50dB or so.
However, it never completely goes away. Is a .1% deviation from ideal, OK with you? How about 10 such caps in series?
?
I consider my reply to you appropriate, and polite. Also, I used Scott Wurcer's AD524 op amp as the difference device. Please take it up to him. Digital processing was not part of the test procedure. The test signal was BAND LIMITED, both low and high, to remove secondary problems with the IC or subsequent measuring equipment.
janneman said:
The EQ file corrects for linear differences in the two paths.
Yet one or both of these paths may be somewhat nonlinear, so the question arises as to the integrity of the EQ file (which implicitly assumes a purely linear circuit) being corrupted by nonlinear behavior of that path in the characterization process. The technique used in the EQ measurement rejects the nonlinear behavior (not all such techniques do!) and should therefore be very accurate, assuming high-quality sound card, etc.
Edmond Stuart said:
I've never denied the existence of listener bias...that's why I said 'Will it alter my perception of the sound (in either way)?' above; note the parenthetical clause at the end. I'm just plain tired of people claiming--or implying--that they're free of bias simply because they read meters; I'm trying to level the playing field a bit.
Given that the majority of the distortion in current systems occurs in the speakers, I'd have to disagree with your disappointing/dull/not exciting remark. ESLs, for instance, typically have low measured distortion and I find them to be quite nice to listen to, particularly in the midrange. The fact that I don't own ESLs comes down to the head-in-a-vice problems (and no, the curved ones don't impress me), not because I don't like the sound.
This could easily turn into a long post if I followed this thought to its logical conclusion, so I will simply note in passing that ESLs achieve their low distortion without feedback, which I think is a contributing factor to the subjective sense of ease many report when listening to them. To really push THD levels in an amplifier requires NFB after a certain point, and that has unfortunate consequences for the sound.
I've said it before, but these things tend to get lost--I think servo subwoofer systems are very promising. If you're faced with 5-8% THD at 30Hz, then by all means begin by lowering the distortion before resorting to fine-tuning tweaks like magic dots on the sides of the cabinets. Fix the biggest problem first. Hammer it. Pound it. Beat it silly. That sort of number is unacceptable. Yes, yes, yes...I know...the average consumer doesn't have the wherewithal to modify a subwoofer system to include a servo, but this is DIY and we can do anything we please.
In an ideal world, listeners and measurers would team up to design gear, but measurers have a deep-seated mistrust of listeners and would rather go it alone. The converse is true also, of course, but not to the same degree. Most listener/builders wouldn't mind measuring things like THD, but buying an HP distortion analyzer costs more money than they can afford. Face it, they aren't cheap, even used. I've got an HP, and would love an Audio Precision, but Santa Claus didn't bring me one, and I've run out of uncles; none of them died rich, leaving me a fortune. Bummer.
Grey
andy_c said:
I was refering to uses of Diffmaker that have a speaker in the path. I can't see how there would be much use for equalization in a good D/A -> A/D chain.
- Status
- Not open for further replies.