Not exactly. You seem to be thinking of things like white noise, pink noise, etc. which can be produced by filtering random noise.
Barkhausen noise and quantizing noise are not as random, or random in the same way.
For example, noise observed at the output of an audio amplifier is present whenever the amplifier is turned on, even in the absence of an input signal.
Barkausen and quantizing noise require some kind of signal to excite them into being. They could be viewed as parasitic to the input signal, and thus in some ways they are closer to distortion than something like white noise is.
Hmmm. I just played one from my laptop. You should be able to download them, or play them with the dropbox streaming player. Is anyone else having a problem?
I see you need to learn to read, this comment makes it obvious you are just trolling.
Scott Wurcer said:
Duu.uuh! The 48k '3120pt' RIAA FIR is available from https://linearaudio.net/downloads
Here's my first attempts. It's using stuff which isn't ideal as it only has 2048pt FFT blocks and Guru Wurcer's FIR is actually 2361pt (not 3120).
The software was for doing other stuff where the 2048pt FFT blocks dun matter. Frequency resolution is 48k/2048 = 23.4375Hz
The vertical scale is 50dB and the resolution is 0.2dB. If you print it out, you can hold it up against the light with plots from a B&K 2307 chart recorder and compare response curves as God intended. 50dB pot on 2307 has 0.2dB resolution 😀
I've got stuff with phase, better frequency & amplitude resolution, printouts bla bla .. but dis beach bum will need some (a lot?) of time to dig them out. Scott bemoans the loss of Visual C++. I do my serious stuff with the Compact version of TurboC on my last remaining XP computer. If I don't move into the 21st century before it dies, I'll be stumped 🙁
_____________________________
3_3.gif is SAI applied to Scott's FIR with 3 feedback and 3 feedforward coeffs. I'm a bit suspicious of his FIR as it appears to have some hash which you see as a thickening of the curve above 3kHz. 3_3 is clearly more than 0.2dB out below 800Hz
4_4.gif, 4 feedback & 4 feedforward, is out by 0.2dB
5_5.gif is out by 0.2dB at 90, 130, 320 & 670Hz
More coeffs are within 0.2dB ie no change on this display when overlaid
3 feedback & 3 feedforward is the 'same' computational load as 2 x 1st order bi-quads .. ie Scott's 2 x bi-quad RIAA solution is superior ... but as I said, I wouldn't use SAI for anything for which I had a s=jw function.
I expected rather worse results from SAI.
I might do a bit more work with more amplitude & frequency resolution, a 'correct' large RIAA target .. but don't hold your breathe. It does show you can us it in the manner I suggested ... combine the cartridge EQ with RIAA in a single small arbitrary IIR.
You can of course, do it with a 2K size FIR too like Scott's RIAA
Attachments
Last edited:
I sent you more stuff, you need to be careful when shifting FFT sizes around. As I pointed out in the article I used a 48K point FFT, when you use the filter on data sampled at 48K it just works, in fact exact to micro-dB.
Thanks for dis Scott. I'll pre10 2 reed them in da next few days.
I know about that Scott.
But I sorta expected the 'errors' to be at LF rather than HF. 😕
Milkshake, please understand that the comparison of digital vs analog or the exploration of the merits of these two technologies, is not in the interests of this thread.
Insisting on such postings here will be intentional violation of the Forum Rule #2 diyAudio Rules
If you are genuinely interested in such issues, search for material and answers in a relevant thread. There were many in the past, a few are closed now, one of the reasons is that the subject has been beaten to death.
George
Ricardo, please try to use the established 26 letter communication code.
We are all of some age here, any unnecessary difficulty added only …
😀
George
You might try focusing attention on the guitar chord distortion sound, and try to ignore everything else, especially the percussion. Is the guitar distortion rough, smooth, creamy, chimey, grating, what? Don't listen to the music, or the pitch, or the song. The idea to start out is just listen to the texture of the sound of one instrument.
Cymbals are another thing that often shows some easy to detect difference if attention is focused on how they sound in terms of texture in different frequency ranges. Not so much the initial ping, but as they ring and decay after being struck.
For me at least, once I find something in the sound that's easy to notice as different if I focus my attention on it, then I start noticing how all the other sounds are affected in a similar way.
Once attention is focused on the distortion it can sort of leap out at you, and it can be like, "how did I ever miss that before?"
Last edited:
Nope, back and forth, every time maybe there's something it ends up in both, still no difference. If blind 50/50 random choice.
I don't know what to suggest then. I can tell you it's very hard for me to notice a difference on my laptop through the tiny speakers. But, its fairly easy listening on my main reproduction system... that is, if I focus on the distortion and not on the music, of course.
Could be there are some files around here of other professional ADCs that are more distorted than either I posted so far. Give me awhile and I'll see what I can find.
BTW, of the two I posted, Bob Katz rated one an A+ and I would agree with him that's probably about right. The other one has very good specs, but its not an A+ for sure.
There are now three more wav files, including possibly better and worse quality than the ones I posted before. Some of these are starting to sound almost like undithered. Cymbal chick sounds pretty bad on one. Here's the link again: https://www.dropbox.com/sh/9r5kiiptr00seub/AAC5zEynkg-ZmriNSkQxFPaEa?dl=0
Delving deeper into the thread from proaudio I posted an interesting snippet related to the side topic of declicking. Wayne tends to convert the stereo into M+S then declicks that before re-creating L+R. I though about his reasons for that and it tends to make sense. Dirt or damage would normally create a vertical motion so getting the de-clicker to work on that seperately might result in a better fix. Certainly worth a try. A long way in my future at the moment though.
Mark: Now I have the living room back and my Koss electrostats I do intend to give these a go. Just need to get an evening where the tinnitus isn't messing things up.
I'll try, but I assume there is more to a turnkey A/D product than a converter chip.
Well you would have to arrange an ABX or preferably an ABC test with your 2 samples.
AB testing is meaningless in this context. I pontificate slightly on this in Wayne's forum that Hans contributed to and Bill pointed out.
But I AM interested in which ADCs were used and how the files were made. Can you send me a PM with this info?
One of the ADCs is probably poor or faulty if they can be reliably distinguished. But the first step is to check if they can be distinguished and this requires ABX or ABC DBLTs.
There are digital recorders today with zillion bits & zillion GHz sampling .. which have a practical S/N about that of a Dolby B cassette recorder. These are hardly suitable for assessing 'digital' vs 'analogue'.
Which of course begs the question why you haven't included an analogue tape recorder and a direct cut disc recording in your line up 🙂
Sorry George.
Last edited:
The files were made using complete ADCs marketed for pro audio use. They range in price from several hundred dollars to over ten thousand dollars.
I was glad to see he agrees with my assessment that declicking before equalization makes more sense. I was surprised he didn't mention that forensic audio restoration article I found, it was pretty old but full of some very good material.
- Status
- Not open for further replies.