Speed Stability Analysis
To fully characterize turntable speed stability, in a most straight forward way, one definitely needs a test record with a standard wow&flutter 3.15kHz test tone. Then, there is a healthy selection of apps and programs to extract meaningful information on-the-go. But what do you do when you are stuck with HFN test record, which only has 300Hz track ? And there isn’t enough time to order or borrow anything else ? Well, that’s when you get creative. There is quite a bit of information to bee extracted even in this case. Let’s capture this 20 sec. track and look at plain spectrum first, but now in high 0.046Hz resolution. All the speed variation will show up there as inter-modulation in form of side-bands near the single 300Hz tone.
Spectrum is totally dominated by the eccentricity of a test record (0.55Hz, 1.1Hz, 2.2Hz etc.) and idler well irregularities (2.55Hz, 5.1Hz, 7.65Hz etc.). Otherwise it looks quite good, with sharp fundamental peak. Now, let’s see what happens if we put record just 0.5mm off center. HFN record center hole measures 7.4mm, which is little out of spec. (IEC98-1987 standard 7.24 ±0.09mm). Dual’s centering pin is 7.0mm, and it’s plastic one, so let’s assume a play of ±0.1mm. This gives us “amazing” opportunity to create up to 0.5mm eccentricity.
Gone is our stable peak, and now we have massive side-bands, spaced by 0.55Hz from our fundamental tone. Using relative levels of these side bands, we can calculate, that our tone is deviating ~1.5Hz around fundamental, or 3Hz peak-to-peak. But that’s a really awkward way to look at things.
We already know, that our cartridge is velocity sensitive device, and this velocity is proportional to groove amplitude and angular frequency:
Here we are talking about groove velocity of stylus so let’s call it lateral velocity VL. We consider only the case of constant amplitude static frequency tone. Then groove amplitude is always constant too.
But our groove lives on a record, which is also spinning and has it’s own velocity.
Let’s call this record velocity tangential VT. It is proportional to groove radius and angular frequency of a record.
Tangential velocity is always decreasing, because radius of the groove is shrinking when record is being played. Now comes the important part. Because LP record is constant velocity, ratio between tangential and lateral frequencies is always the same!
Then it becomes obvious, that ratio between tangential and lateral velocities is only function of groove radius.
It means that every variation in tangential speed, because of radius or tangential frequency change, will produce proportional increase in lateral velocity. Groove amplitude can’t change, it’s already there pressed as is, so only thing that can change is frequency of our tone. Now that we established this, our problem becomes quite trivial. Our track lies 96mm from record center. It’s tangential frequency is:
Then it’s tangential velocity:
When we have center hole 0.5mm off center, on one side of the record radius becomes 96.5mm, and tangential velocity increases to:
On the other side of the record, radius becomes 95.5mm, and tangential velocity decreases to:
That’s a ratio of:
Which from 300Hz is:
From another perspective, that’s a wow of 0.5%.
So visual tone sharpness inspection can give us crude stability estimation, but that’s about it. If we want more meaningful and accurate information, we have to go down the rabbit hole.
Software Demodulation of Wow&Flutter
In order to make accurate analysis of a turntable speed deviation, we must acquire spectrum of wow&flutter. For this purpose, let’s treat our 300Hz tone as carrier, then speed deviation will be our modulated signal.
It’s really not that difficult concept to get your head around, even for less technically inclined average person. Although most of audio gear reviewers think different it seems, as there is little to none published results, based on this methodology.
Let’s leave AM modulation aside this time, and concentrate only on FM modulation. Performing frequency demodulation on this FM tone, will give us speed deviation in time domain. Running frequency analysis on this waveform, will give us spectrum of wow&flutter.
Actually this is how all commercial wow&flutter meters work, they just rectify demodulated signal and show it’s value on analog meter, instead of producing it’s spectrum. Here is a functional diagram of industry standard B&K 6203 meter for reference.
There is a variety of software demodulation schemes and implementations. However, most of them will always involve some amount of code crunching, be it MatLab or LabView or something else. Luckily for us, there is thriving community of amateur radio enthusiasts online. These folks have quite a few tools for software signal demodulation, most of them command line only, but there are few with nice GUI’s. Most user friendly, probably is Spectrum Lab from DL4YHF. Let’s give it a try with our previously captured 300Hz signal. Here it’s spectrum.
Our main limitation is straight obvious here. We have only about 100Hz of usable bandwidth, as there are spurs of modulating signal already at 236Hz. This will give us speed deviation results only to 50Hz or so. That’s why at least 1kHz carrier signal should be used.
Program Settings and Workflow
Let’s not make this step-by-step “how to”, so here is a couple of screenshots with all relevant program settings for those playing at home.
Start the sound thread and adjust gain for left filter output amplifier, so that demodulated signal, at adder L5, had a level of -40dB. Waveform on the scope should confirm that we have ±1% full scale amplitude. Record the waveform and save it for later processing.
Now let’s feed our previously recorded 300Hz track. We have to start audio thread and start recording, then open audio file. This is a bit awkward, but there is no way around, as demodulator will produce digital noise without data and triggering will not work. This part is a little bit buggy I’m afraid, but that’s not a problem, as we can edit these later.
It goes without saying, that it’s fully possible to analyze speed stability with this setup on-the-go. Just enable ADC and hook up your phono preamp to the sound card input.
Open file in your favorite editor, and cut all noise from before-after. Now we’ve got ourselves a detailed turntable speed deviation in time domain.
Interpreting the Results
After importing processed wave files to editing software and making sure, that vertical scale is linear and shows percentage of full scale, it should look something like this.
Synthesized 1% calibration signal
Dual 1218 speed deviation (with off centered record)
Dual 1218 speed deviation (with centered record)
Let’s remind ourselves what all the terminology actually means.
| “Drift” – Frequency modulation in the range below approximately 0.5 Hz that may be perceived as a steady or slow change in pitch. |
| “Wow” – Frequency modulation in the range of approximately 0.5 to 10Hz that may be perceived as a fluctuation in pitch of a tone. |
| “Flutter” – Frequency modulation in the range of approximately 10 to 100 Hz that may be perceived as a roughening of the sound quality. |
| “Scrape Flutter” – Frequency modulation in the range above approximately 100 Hz that may be perceived as a noise added to a reproduced sound that is not present when the sound is absent. |
So what we have plotted here is turntables unweighted wow&flutter part of speed deviation, which is 0.2% when record is centered, and jumps to 0.5% with just 0.5mm record eccentricity. We haven’t removed DC anywhere in processing chain, so there is also drift information embedded, but it’s just not enough time samples to get a good resolution. Also flutter is limited to 50Hz only, as our bandwidth limitation states. Let’s run spectral analysis on these turntable speed waveforms.
record being off center. As was the case in a noise analysis, here again, spectrum is the only useful specification, when defining turntable speed deviation and it’s components.
Applying Weighting to the Results
In order to obtain an objective indication of the audibility of the measured wow&flutter, industry decided to apply a weighting filters. Almost all flutter measurement standards (DIN 45507, IEC 386, CCIR 409 and IEEE 193), and their corresponding weighting filters, are centered at 4Hz. Listening test have shown that this modulation frequency is most audible.
You can download exact values of this filter here. It was used to FIR filter 16 sec. of centered record speed deviation. Here is resulting waveform and spectra.
We see a drop of wow&flutter to 0.12% and now our spectrum is dominated by idler wheel irregularities. That’s very close to the factory specification of 0.09%. Let’s try to push things a little bit further by turning this idler wheel on lathe and making sure it’s round and aligned.
Now that’s an improvement! We’ve achieved idler harmonics drop of ~10dB and if I could be bothered to center record better, we’ve been hitting weighted wow&flutter figure of 0.04% easily. Also notice general decrease in flutter noise floor after repair, or I should say – upgrade.
Why Should I Care ?
You may ask yourself after reading last couple lengthy analysis sections. All that rubbish you have measured is so low in audio band, that you can’t possibly hear it anyway? That’s the problem – you actually can!
It’s a long known fact, that wow and flutter (low frequency modulation) folds up in audio range as sidebands on audible tones. Turntable is inherently non linear mechanical system, so whenever you have two tones played (or excited as resonance) there will be inter-modulation distortion present, which is product of frequency and/or amplitude modulation. In any case, it represents it self as sidebands on higher frequency component and is clearly audible. The threshold of this audibility could be argued, but I would say 0.1% is a safe bet. Here is a standard 3.15kHz test tone with decreasing 4Hz (most audible) modulation. Listen and find out for yourself.