16/44.1 digital recording (on vinyl): source of ultrasonics

[DF96]

At 7.5ips you won't be able to record much beyond 20kHz, and even then only when everything is set up really well. A good CD player should not be sending anything out beyond about 20kHz, so anything above 20kHz on replay
(a) should not be there
(b) probably comes from tape distortion etc.
My bet is that you would see significantly less ultrasonics than an LP.

 

[lcsaszar]

Tape bias oscillator residual or beat byproducts perhaps?

 

[Pano]

and sometimes even the line scan frequency of the mixing desk automation controller

I often see the video 15.75 kHz on opera recordings. Maybe from the controller as you say, but I suspect also from video monitors in the mixing suite or truck.

 

[scott wurcer]

I never really bought the stylus/cart./lathe "mistracking" hypothesis.

Then take a test LP with high level 1/3 octave noise bands and any FFT software and see for yourself. There's even a warning on the Telac Omnidisk that the low frequency crackle on the 5k and up bands is "normal". It's very instructive to just watch the bands go up in frequency and see the upper 2nd's and 3rd's track and then the lower IM artifacts grow from the low end.

 

[luckythedog]

There's even a warning on the Telac Omnidisk that the low frequency crackle on the 5k and up bands is "normal".

Increased surface-noise-like crackle/pop on tests which are difficult to track, even accidently difficult such as f sweeps, is a very interesting loose end.

Pulling on that loose end is very revealing about the nature and origins of surface noise, IMO.

Nothing to do with the OP though, it has a flicker noise like profile 1/f.

LD

 

[scott wurcer]

Increased surface-noise-like crackle/pop on tests which are difficult to track, even accidently difficult such as f sweeps, is a very interesting loose end.
LD

Sorry I meant Telarc, IMO the low frequency noise does include a lot of inner IMD of the third octave bands as well as some of the these other effects.

 

[luckythedog]

Sorry I meant Telarc, IMO the low frequency noise does include a lot of inner IMD of the third octave bands as well as some of the these other effects.

The random crackle-pop audible in the background of hf monotone tests is what I mean, and prob what Telarc refer to. I doubt that's IMD and besides it shows up in monotone tests too.

LD

 

[scott wurcer]

The random crackle-pop audible in the background of hf monotone tests is what I mean, and prob what Telarc refer to. I doubt that's IMD and besides it shows up in monotone tests too.

If you get the f1+f2 where does the f1-f2 go?

 

[luckythedog]

If you get the f1+f2 where does the f1-f2 go?

Only if there is an f2, which there isn't for a monotone...………..

...........and besides, IMD doesn't yield random crackle-pop type noise with a flicker distribution in the time domain.

LD

 

[scott wurcer]

Sorry I misunderstood your point, I was only pointing out how easy it was to demonstrate high IMD in vinyl playback. I don't recall even remotely the same amount of "crackle" on spot frequencies or sweeps on the same LP.

 

[gabdx]

actually the easiest way to see once and for all which is best between CD and vinyl is to slow down a vinyl playback by 5% and compare the same CD track slowed down by 5%.

I can hear CD losing lot of information at only 1%, for vinyl the sound changes but there is still plenty of info and detail at 2% slow down.

 

[xx3stksm]

It's very easy to slow down a vinyl playback without unintentional degradation. But how do you slow down CD playback in the digital domain? I would say the way to slow down CD may dominate the result. Please explain the method you use to compare vinyl and CD.

 

[lcsaszar]

Replacing the master clock/XTAL 11.2896 MHz to 10 MHz?

 

[gabdx]

In the free program, Audacity, change tempo, which mimics a vinyl playing slower,
affecting both pitch and playback time.

*it could slow down better than youtube, because I think* the way it is done is by upsampling and extrapolation, then down sampling, but it works

 

[hollowman]

If anyone reading this has access to a calibrated, decent open-reel deck (and some good-quality tape), might I suggest an experiment:
Feed the analog output of a CD player to the input (line in) of the open-reel deck. Record any CD at 7.5, 15 or 30 ips.
Now play the tape back on same deck whilst recording it in the digital domain . You can use any digital recorder you wish including your PC sound-card.
As per OP, re-run the vlogger's experiment.

My bet: you'll see pretty much the same as the LP.

Haven't tried it ... but I'm gonna take back my bet.

Here's why (and I did mention this in one of the earlier post in this thread).

The explanation may be this simple:

The tape recorder/player is band-limited. Let's say, 18khz is the highest freq. it can record/playback.

We record a very high-freq sound ... for the sake of argument, let's invent a piano key that's centered at 16khz. When this key is hit, its string will produce 16k and overtones (harmonics). The tape recorder will attenuate any signal over 18khz, so most of the overtones will be "cut off" in the recording.

When it's time to cut the record, the tape deck (with that 16k piano key audio signal) is connected to the lathe and we begin cutting. Although the lathe is also bandlimited (say 18k again), this is an electronic filter (in the lathe's preamp); the cutting stylus acts like a tuning fork and can vibrate to its mechanical limit (so it may very well vibrate at overtones well past the electronic filter cutoff).

Ditto case at the playback end (at the record player): when the stylus meets the groove with the 18k signal, the stylus may very well (naturally) ring at overtones. And this is reproduced in the generated electrical (audio) signal.
This "phenomenon" is seen in a lot of vinyl record spectrograms, as in the case of the prev. noted:

Image ref:
Analysis of Vinyl Frequency Content

 

[Bill Coltrane]

In the free program, Audacity, change tempo, which mimics a vinyl playing slower,
affecting both pitch and playback time.

*it could slow down better than youtube, because I think* the way it is done is by upsampling and extrapolation, then down sampling, but it works

That Audacity algorithm can cause audible artefacts...



An other thing:
If a redbook recording has significant >22kHz content, something has gone seriously wrong with that recording. Df96 pointed that out correctly.

 

[rsdio]

First of all, there are so many variables in this YouTube test that I can't even take it seriously.

When I try to digitize vinyl (for listening or analysis), I get rid of the analog RIAA filter and try to digitize the signal as close to the vinyl as possible. Even still, I pick up noises from switching powers supplies and digital equipment in the room. It's incredibly difficult to measure just what's on the vinyl and nothing else.

This YouTube guy has the signal going through lots of extra hifi gear, speakers, the air, a microphone, a consumer ADC, and free spectrum analysis software. There're too many mistakes there to take this seriously.

To answer the OP's question more directly: It's entirely possible that all of the content above 20 kHz is actually due to the spectrum analyzer software - AND NOTHING ELSE!


I've written spectrum analysis software, and the first thing I noticed is that 32-bit processing has so much noise that it's quite misleading when you're trying to analyze the quality of frequencies above 20 kHz. This is true even if you feed in a perfect digital sine wave. There might appear to be noise across the spectrum, but you know it's not actually there if the source is guaranteed to be a pure sine wave. I suspect that the spectrum analysis software that this guy is using doesn't actually have the resolution needed for the task.

The result of my lessons was that I rewrote my FFT code to use 64-bit math, and now my frequency plot goes down below -288 dBFS (I'm not sure how far it goes, as I've never extended the window below that).

When this level of accuracy, it's actually possible to see the difference between a standard 16-bit CD and one that has been mastered at 24-bit and dithered properly to 16-bit with noise shaping. The standard 16-bit CD has an obvious noise floor centered around -96 dBFS, the theoretical limit. The noise bounces around, slightly above and below the -96 dB reticle, but it's obvious that you're seeing 16-bit noise. In comparison, a carefully dithered 24-bit recording shows a noise floor at -144 dBFS, even though it's being played from a 16-bit digital data source. This is not visible with the 32-bit FFT, or at least not in all cases. The 24-bit in 16-bit recording does show some part of the noise floor that's well above -144 dBFS, but that's expected because it's how noise shaping works.

Another thing that happens is that any discontinuity in the digitized signal, such as a square wave, spike, click or other transient, ends up producing in infinite spectrum of harmonics.

Admittedly, a lot of the potential spectrum artifacts that I've described above would only appear briefly on the display. The YouTube video seems to show a fairly constant signal above 20 kHz, although you do see brief patterns with repeating shapes. Those patterns are certainly artifacts that are not actually recorded on the vinyl, even though the vinyl did originate from digital sources. Most likely, the bulk of the content above 20 kHz in that display is due to processing resolution limitations, not much else.

 

[hollowman]

Huh????????????????

Huh????????????????

 

[DF96]

Scrape a piece of rock against a piece of plastic and you will get noise of various types at various frequencies, including ultrasonics. This will show in a spectrum analyser. The noise is generated by the vinyl playback process, not put there by the recording chain. There may in addition be a little noise or distortion from the recording chain, but even this will mostly come from the cutting process not the digital audio which fed the cutter.