Source of Fremer's "1 arc second" claim?


In the latest TAS April 2025, page 34, Fremer reviews some Technics TT, and repeats his claim that "listeners in blind tests could hear arc second speed shifts". where one revolution ~1.3 million arc seconds. Anybody have any idea where this is coming from?

Basic math will make you wonder whether any listener can hear a difference between chamber a' = 440.00000 Hz and 440.00004 Hz, rounding the 1.3M to an even 1M. When tuning my violins, I can hear 2–3 cent difference, where 800 cents = 1 octave = doubling of frequency. At 2 cents, that is over 1 full Herz difference. Even playing a cord with tones at 1 Hz difference will result in an oscillation at 1 Hz, i.e. peak to peak 1 second. For easy math, assume even a 0.00005 difference, which would lead to an oscillation with frequency of 20,000 seconds = 33 minutes. Good luck hearing that. 

"Golden Ears" being able to hear ten times better than a normal human, why not. But 20K better? We are off by several orders of magnitude. Maybe I don't understand that he is talking about, but I consider it complete BS.

Maybe it has to do with consistency (accuracy vs. precision), but then the a different unit needs to be used that includes time in the denominator. But even then the math/physics don't add up.

If anybody can provide any insights, LMK. Thanks!

The alternative is rather unflattering for Mr. Fremer.

oberoniaomnia

Showing 4 responses by richardkrebs

Gentlepeople

Michael's arc second comment probably came from a conversation I had with him about the speed sensing architecture in the OMA K3 turntable. The design sends a little over 1.3 million pulses, counts, to the motor controller every revolution. Slightly more than one count per arc second. This equates to a little more than 728,000 pulses per second at 33.33 RPM. The speed sensor assembly, like everything that exists, is not perfect. This means that  the time spaces between the pulses are not absolutely equal, even if the motor is running at a perfectly constant speed. The controller will sense these differences and signal the motor to correct a speed error that does not exist. The motor may or may not be able to accelerate or decellerate the platter in time to correct this non existent error but it will receive a current change none the less. 

The people who make the motor controller designed it to control systems that require much more speed precision than used in a TT. They also are well aware of speed sensor count errors. To mitigate this problem, the controller has a selectable function that activates a rolling average of the counts. With a few key strokes this can be set to 1, no averaging and 2 up to 8 count averaging. The controller also adds math to this function by giving more weight to the first count and little less for the second and so on. This function acts to smooth over the sense errors, resulting in more stable (smooth) and accurate rotation. It also acts to soften, smooth, the reaction, correction, of actual speed errors. When programming the controller, our listening panel trailed this function and found that it did indeed make an audible difference. The person making the program change did not tell the panel what change had been made, he only asked....better, worse, same?  Since it was software driven we could easily toggle it on/of or change the averaging number. In this way we could confirm in near real time what we were hearing. We also had a policy of revisiting the programming some weeks later to ensure that we hadn't make mistakes.

We settled on a rolling average count of 4.

When calibrating the speed to 33.33, 45 and 78 we simply adjusted the count command per second. Once close to the required speed we set about  incrementing or decrementing by 1 count per second. This meant that when targeting the required speed, we were changing the speed up or down by arc seconds per time. For a laugh, we listened to these changes, no one could detect a difference and no one was surprised. 

@pryso 

I designed the OMA K3 and K5 turntables. Jonathan then used an industrial designer for the aesthetics. 

My company provides the machined rotating parts and the K5 chassis. The cast iron chassis is made in the States. Both TTs are assembled by OMA. 

The "We" refers to a brilliant mechatronic engineer who helped to get the drive up and running and then a panel of enthusiastic friends who helped finesse the program.

Cheers 

@oberoniaomnia

Yes, it is an interesting trade off. We did not need to go to the max averaging available, so I'm happy with the hardware design choices I made. .  The change from no averaging to averaging of 2 counts was astonishing, we all heard the effect. But we could hear no difference going higher than 4. This implies that data noise is low. 

To be clear, as you point out, this tuning is not a speed change of the equivalent of one arc second/ rev.

However it was a change to the granularity of the feedback/controller/motor loop in the order of one arc second. 

Cheers.

Guys

I think that we need to cut MF some slack.  In my conversation with him, I said something to the effect that a change in controller motor command in the region of one arc second resulted in an audible change in the sound. While you could argue that there is a difference in changing the speed by the equivalent of one arc second per revolution to changing the speed sensor averaging by one arc second, which is of course true.... The fact remains that we heard a difference by making a change that is microscopic. This was my key point.

In this hobby, I think that we are too fast to ridicule. We are all, I hope, striving to achieve excellence of sound in our homes. It is not a competition. Why can't we accept that and considering the big picture, pardon semantic errors that don't really matter. 

Cheers.