• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Wow and Flutter measurements like you've never seen before! Maybe...

Joined
Jun 2, 2024
Messages
12
Likes
33
Location
Italia
Introduction
I’ve had a Vernier force sensor for educational use for a couple of years, acceleration and gyroscope. I didn’t have time to play with it until a couple of weeks ago, it connects via Bluetooth and can collect up to 1000 samples per second, here, acquiring the data, 5 minutes of data therefore 300,000 samples, on turntables with different technology and in different state of conservation I was able to make the most precise measurement of wow and flutter that I happened to be able to do and see.

Measurement was made on the turntable alone, so it does not include imperfections in reading and printing of the test disc and resonances of arm/cartridge system, the sensor was first balanced, the data is saved in CSV with its native app, I first examined them with LTspice, a program that I always use and that I know, then with a couple of Python scripts that I specifically created.

Most interesting things, which I was able to see are:
  • spectrum of speed variation, with a resolution, at least by me, never seen
  • polar diagrams of speed variation

Have been tested:
  • Lenco L75 revised and lubricated.
  • Sansui XP-99 revised and lubricated.
  • Sony PS X555 ES NOT revised in the engine part.
  • Revox B790 revised and lubricated.
Who will win? I’ll tell you right away they are already in order of performance, from the first to the last.

In conclusion

Complete maintenance is also important for DD turntables, it doesn't matter what the service maual says, nor the fact that they lock the correct speed by turning on the "quartz locked" light. If you want the declared performances you have to disasseble, clean, lubricate, carry out service calibrations and measure before and after the treatment, to verify the effectiveness of the interventions.

I’m sorry I don’t have a belt drive turntable at hand, I have some but they have been stopped for a long time, I would be curious to measure a high-end one (mine are not), I think it would win at the test bench.

I now understand those who use the Lenco platter by putting the mechanics on a new base and changing the tonearm.

The DDs I measured are more precise in nominal speed but dirty the frequency spectrum more by “vobulating” the signal in some way. Is it audible? What is the acceptable limit? I don’t know, and I can’t make a direct comparison between these turntables since they have different arm and cartridge, and then it would only be a subjective opinion strongly influenced by my preferences and beliefs.

The W&F weighting curve was made several decades ago, when technologically outdated devices were available, I am not entirely convinced that the dirt in high frequency is to be considered less important than that at 4Hz, the maximum sensitivity according to the IEC386. If so, the jitter of digital devices would have no importance, and it doesn’t seem correct to me.

It is possible, if I can I will try to do it, that these measurements can also be obtained with the gyroscope of the mobile phone, I will try sooner or later.

1717360758050.png

The sensor https://www.vernier.com/product/go-direct-force-and-acceleration-sensor/

1717360862389.png

Lenco wins above all, its fundamental at 0.555Hz is placed at -70dB the highest in this test, but it is the one that matters and disturbs the least, consider that a 25 micron decentralized disc is enough to obtain the same "degradation". Close to 0Hz the noise floor rises, it is the speed which, being mechanical, is not very precise, but still within 2 points per thousand. Above 6Hz you see absolutely nothing.
Note the dynamic range of the measurement, between the 0dB (I normalized to 33 and 1/3) of the continuous rotation and the noise floor there are 110dB!

1717360915769.png

Sansui and Lenco overlapped for direct comparison. This one is clean too, DD four coils effect is dominant, can’t understand what happens at half the freq rotation of the platter.

1717360965128.png

The Revox has several defects, motor cogging stands out, there are no adjustments, trimmers to compensate for the asymmetries of the two driving circuits, nor to compensate for drift and differences that the Hall sensors could have accumulated over the years, hfe differences of power transistors.

1717361018390.png

I widen to 120Hz because the Revox, the only one tested, has a component at -90dB 111Hz, I guess the unshielded tachometer cables as the root cause. Not measurable with the means of the time, the German is the one with the most spartan electronics, the Japanese were much more cutting edge.

1717361063875.png

The Sony has never been serviced and it shows. It probably has an imperceptible ticking, an impulsive braking on the lap that the control immediately puts back on its feet. More precise RPM. Ticking evidence in polar diagram.

1717361166906.png

The results of the script that calculates weighted W&F and not of Lenco and Sansui, the Sansui was also measurable at the time via the tachymetric freq, the Lenco was not, it has performances much much higher than those declared.

1717361197025.png

The crowding of harmonics (however very low value) of the Sony that is slightly defective, the cogging at the frequencies that weigh more for the Revox. The Sony can be improved with the right maintenance, probably the Revox too by adding trimmers to balance the two motor control branches, that will be for another time. Who knows how much that spurious at 111.11Hz weighs, it's a kind of periodic jitter.

1717361241696.png

The gray circle labeled 1.000 (covered by the graph) is the speed normalized to 33+1/3, the inner one a thousandth less, the outer one 1/1000 more. Graphs are obtained by processing the average of 50 rotations.
The repeated acceleration and braking of the Revox are due to the geometry of the motor and the asymmetries of the driving circuit.
The Lenco simply brakes and accelerates smoothly during the turn. Would re-milling the bottom of the platter also remove this "defect"? Also note the slight positive deviation from the nominal speed.

1717361275427.png

The variations of the Sansui are softer and at low frequency, the 4 lobes are the 4 coils of the stator.

1717361296964.png

The Sony would win hands down over other DDs if I serviced it.

1717361339926.png

Brief recap.
 
^^^ More than my brain can handle in one sitting -- but this looks great!
I shall peruse anon. :)
I saw "Sony" and looked a little harder. I know nothing about this particular (later -- from my perspective) model, but I do know that the early Sony DDs do have a certain... shall we say... cachet with even hard-core idler and belt devotees. The design of the early Sony DD motor was quite different (not that I have much understanding of this!) than the Matsushita/Technics DD of roughly the same vintage. Supposedly, the speed stability of the Sony is (much?) better than the Technics in some way that pushes some of the vinylistas' buttons. ;)

All I can say is I do use a Sony 2251 and it is good -- it might (???) be interesting to measure. Ever get to northern New England, @Luca Sorrentino ? ;) (I'm guessing no, unfortunately)

 
The unit you are using likely has vibration suppression inherent in the 6 axis accelerometer which precludes its use for this application. I went down this hole with accelerometer/gyroscopes for turntable W&F and it wasn't successful.

What particular IMU chip does it use?
 
The unit you are using likely has vibration suppression inherent in the 6 axis accelerometer

Good point.

I use the gyroscope Z axis. I have no evidence of filtering, below a plot of the sensor standing still.
Turntables W&F measured by the FG signal are pretty spot-on with my measurement, spectrum and polar plots seems coherent with the technology of the device under test.
Unfortunately I find no info on what chip is inside and I'm not curious to disassemble it :)

1717409539777.png
1717409596723.png
 
The design of the early Sony DD motor was quite different
Your unit looks gorgeous, congratulations! I wold like to measure it but I'm in Italy, is not possible right now. I'm a Sony fan too, I have a couple of PS X800, and a PS X600 too, "biotracer" tonearms, I was fascinated by that technology.
 
Is the unit weight balanced when applied on the spindle? (There are a number of Shaknspin measurements on this site as well)
 
I will check, thank you. My feeling is that shakenspin plots have less resolution.
500samples per second for 8 seconds
For better clarity I often use a 20hz low pass filter to remove noise in time series plot (4th order Butterworth). I think this gives a resolution of 500/4000=0.125 Hz?
 
Last edited:
500samples per second for 8 seconds
For better clarity I often use a 20hz low pass filter to remove noise in time series plot (4th order Butterworth). I think this gives a resolution of 500/4000=0.125 Hz
Definitely resolution of my experiment is far higher, 1Ksample per second during 5 minutes, 3.3mHz -> 0.0033333333Hz no filtering.
However for polar plots I had to average 50 revolutions to obtain something acceptable and remove noise
 
Definitely resolution of my experiment is far higher, 1Ksample per second during 5 minutes, 3.3mHz -> 0.0033333333Hz no filtering.
However for polar plots I had to average 50 revolutions to obtain something acceptable and remove noise
Do you have max and min for the polars? Average can hide variations.
 
Definitely resolution of my experiment is far higher, 1Ksample per second during 5 minutes, 3.3mHz -> 0.0033333333Hz no filtering.
However for polar plots I had to average 50 revolutions to obtain something acceptable and remove noise
Ok but 0.0033 resolution is not needed or a relevant measure of quality. It is like sayin the patients temperature is 38.67352683 C vs 38.7 it does not result in a different or better diagnosis...

By the way, here is the Shaknspin data UNfiltered , filtered with 4th order Butterworth low-pass at 30 Hz and 10 Hz.. Technics 1200G
1717439561315.png


FFT unfiltered data. 2048 sample length( Excel limitation with 4000 data points) gives 0.25Hz resolution, adding some extra data allows for higher resolution using all data points and higher resolution (4096 samples)
1717439680962.png

1717441157289.png
 
Last edited:
@Balle Clorin What is the actual IMU chip in the Shake'n'spin? The data sheet will tell all.

The 'data' captured is most likely in the noise/error/minimum floor of the A/D converters in the IMU and all you are doing when filtering is looking for a 'trend', not an absolute indicator of true wow or flutter.

Minute turntable variations in rotational velocity are at the absolute bottom end of the most sensitive ranges of these chips- they aren't designed or capable of accuracy or repeatability in measurements- they are mass produced for phones, tablets and drones, not precision turntables.
 
Last edited:
Ok but 0.0033 resolution is not needed or a relevant measure of quality.
Maybe your right, but it's free, I put my sensor on and simply capture with my PC 5min of data.

1717480992153.png

This graph brings no info al all, apart that was acquire @45rpm.

No need of filtering if data quantity is big enough to make a detailed FFT, with high resolution I can capture any detail @ low freq, mechanical tt slightly change speed here and there and in five minutes I can document it.

1717481429125.png
i.e. with this resolution I can clearly see spikes detached form the noise floor that are perfectly measurable, that 111.11Hz is not detectable by shakenspin at least in that way.

1717481591303.png

output results of automatically measure speed variation, this are dB RMS referred to 33+1/3.

index.php

In this one I barely detect periodicity of speed variation but absolutely not it's shape. Magnitude seems not negligible instead.

1717482780637.png

for example this polar plot is the average of 50 turns -> 90seconds of acquisition
 
The 'data' captured is most likely in the noise/error/minimum floor of the A/D converters in the IMU and all you are doing when filtering is looking for a 'trend', not an absolute indicator of true wow or flutter.
Yes, if you look @ the raw data without any calculation, noise floor of the sensor dominates, not for lack of resolution I think.
1717483179961.png


but just make the FFT of the whole 300.000 data points (without windowing in my experiment) reveals a lot of interesting things that are perfectly measurable, repeatable and "turntable technology" correlated.
1717483503516.png



A very nice example of a slightly defective Sony tt, an imperceptible ticking dominates the spectrum at very low level. I never serviced that motor.
1717484239960.png
 
Last edited:
Do you have max and min for the polars? Average can hide variations.
Yes you're right, but in time domain random noise of the sensor dominates peak to peak value of the signal and brings to incorrect results.

Averaging can hide something but brigs to readable and repeatable results, coherent with mechanical status and technology type of the tt under test.
 
Interesting for sure.

The fact that the go-direct unit is also picking up 120Hz mains means it could also be picking up the actual motor coil currents/switching and even the sensor (FG) coils/heads, along with whatever the drive circuitry is doing. So, how much of the 'data' is actual rotational data and how much contribution there is from electrical fields/magnetic coupling etc is unknown.

You should align and setup the Sony as per the service manual, replace any out of spec components, service and lubricate the motor bearing/thrust assembly and retest at some stage. What model Sony is it?
 
What model Sony is it?
Is a PS X555ES, at that time I've just replaced a faulty hall sensor in the tonearm.
is also picking up 120Hz
Nope, in Italy mains is 50Hz, no magnetic fields from the main transformer are involved, no evidence of fringing flux from whatever, just rotation speed in rad/sec and random noise of the sensor.

How bare data arrives.
1717489084089.png
 
@Balle Clorin What is the actual IMU chip in the Shake'n'spin? The data sheet will tell all.

The 'data' captured is most likely in the noise/error/minimum floor of the A/D converters in the IMU and all you are doing when filtering is looking for a 'trend', not an absolute indicator of true wow or flutter.

Minute turntable variations in rotational velocity are at the absolute bottom end of the most sensitive ranges of these chips- they aren't designed or capable of accuracy or repeatability in measurements- they are mass produced for phones, tablets and drones, not precision turntables.
The data in this thread shows that the sensor is good enough for the task at hand. It may not detect some very small random deviations, but everything periodic is clearly visible in the plots.

The lowest Allan Deviation of MEMS gyros is typically on the order of magnitude of the rotational velocity of earth (0.004 °/s) and their noise is strongly dominated by white noise in the short term (~100 s). That means a longer measurement time results in big sensitivity gains up to that point. The performance of these little sensors is super impressive.
 
Back
Top Bottom