A Few Turntable Measurements using the RPM Android App

As is often the case, measurement is the issue. RPM is a continuous quantity only in the abstract. In all practical applications, it is an average. Maybe over a fraction of a millisecond, but that still is a far cry from continuous.

So the major issue, variations on the 10 - 20 KHz level, may be averaged out and not be addressed at all. Unless the specs indicate the time span over which the average is taken, the results don't mean much. As one of the greats from the glory days of audio once told me, "The best audio analyzer available is hardwired to your brain." (Jonas Miller)

I like the Nottingham sound too, which is why I built my air bearing with a 45Kg platter and a 1.8 watt motor.

@kevemaher  

With all due respect, keve, I think that you are putting the cart before the horse. Before one aggregates observations, one must make observations. The difficulty with making speed observations is that for practical purposes they must be averages: distance travelled divided by elapsed time. This is not an instantaneous quantity.

Even aggregating a finite number of observations implies discrete, not continuous. That means sampling frequency.

This is not a moot point because modern controllers tend to correct speed at about 10KHz. If it were easy, they would correct more frequently. I doubt that an app is better; in any case, it's important to quantify the limitations of the observations; in this case the granularity  imposed by sampling frequency.

Then one could aggregate the data into moments, from which the central moments of mean, standard deviation, skewness, kurtosis, etc. could be calculated, although perhaps a spectral or Fourier analysis would be more revealing.

There is no problem with the data arising from a process which is not perfectly random - it is data. The point at which randomness enters the picture is in identifying which inferences are valid, and which are not.

Yes, I think that speed stability is a problem. The question then is, "Which frequency?" I strongly suspect that most of the audible differences between turntables arise from small speed variations in the lower kilohertz range, which we hear as  brilliance or sibilance. It is obvious to suspect speed instability of causing this - but at greater frequencies than are usually measured. For example, bearing noise affecting speed.

By way of evidence, a common criticism of Nottingham turntables (high mass, low torque) is that they are 'dark', that is, without brilliance. I prefer the term 'smooth'. Also, that is how the bearing noise from my 1.8W motor manifests itself. Yes, it's repeatable.

As you point out, "Measuring performance parameters of instruments is crucial to product development." But unfortunately, associating numbers to phenomena is one of the most difficult human activities - doing it right, that is. It is therefore of cardinal importance to report the limitations of the measurement before reporting the results of the analysis. In my opinion.

Thanks for the discussion, @kevemaher .

@kevemaher I agree that wow, flutter, and rumble measurements are quantifiable and repeatable and agreed upon. I think, however, that this is not the whole story and that a 50KHz sampling rate would give us the data needed for a more robust analysis.

I think that both of you make good points.

For 40 years digital has claimed perfection, while getting better every year. It is still too bright for my ears, but an entire generation has been brought up to think that brighter is better. I have noticed that every major improvement to my system, US cleaning, reflex record clamp, Koetsu, air bearings, every one removed copious amounts of high frequency, yet after a few hours, it was obvious that the sound was closer to the concert hall.

So I concluded that what I was after was ’smooth’.

Raul talks about there being bright in a bad way and bright in a good way, which I also think is quite right. Mijostyn talks about this too.

Might both of you be referring to the leading and trailing edges of the signal? Electronically, this performance is dictated by odd-number harmonics (square wave Fourier Series). Unfortunately, distortion of the odd harmonics is notorious for being offensively bright.

I think that you are both onto something.

@kevemaher 

I refer you to a talk by Richard Krebs, and the considerations that went into his latest turntable design.

If you want to hear about higher order moments and their importance, though not in formal statistical terms, and the sampling frequencies required, please do take a look.

https://www.youtube.com/watch?v=smib6OJm-T8&t=2065s

Also the limits of the ear in the time domain (Fourier theory).