Hacking PlatterSpeed for better diagnostics: case study with an old Rega

Short version: The PlatterSpeed app uses a 3150 or 1000 Hz signal from a 331/3 RPM test track to make plots of frequency variation with time (turntable speed variation). When a plot is saved, what’s written to disk is not the plot itself but an .xml file of the raw data that’s used by the app to construct the graphic. Taking this file into Excel and choosing the Fourier Analysis option under the Data | Data Analysis menu (presuming you’ve enabled the ‘Data Analysis’ add-in) performs FFT on the measurements to give you the results in the frequency domain. It’s a bit limited because the app’s sample rate is 10 Hz and FFT only gives valid results to half that, but it nonetheless has the potential to highlight the effects of any changes you make to bearing lubrication, platter weight, motor isolation, belt tension, etc., to diagnose the condition of your motor and platter bearings by illustrating half-speed whirl (precession of a bearing shaft at a rate typically 40-50% of the shaft’s rotation), and to reveal possible resonant vibrations of platter, plinth etc.

A recipe for formatting and analyzing analogous time-series data using Excel’s Fourier procedure is given here. More details are given in the following, but if you don’t want to read the interspersed stuff about my own trial-and-error, just have a quick look at the graphs and the conclusions before doing so. It’ll help to recall that 331/3 rpm = 0.55 Hz and that 300 rpm (Rega and various other synchronous motors at 60 Hz AC) = 5 Hz or that 250 rpm (same motors at 50 Hz AC) = 4.2 Hz.

Disclaimer: everything below refers to the Android version of the app. The iOS version has a ‘lowpass filtered’ display which might reflect a different algorithm than the Android ‘notch filtered’ one (displays I’ve seen of the two look different; @Marvey has previously noted this here), but I don’t know if things like data file format and measurement sample rate are different. If so, iOS version users might need to modify the process (which uses the unfiltered data only; the filtered data are not saved in the base files, they’re reconstructed by the app).

Long tedious version: As the PlatterSpeed screenshot below illustrates, the speed stability of my > 30 yr old Planar 3 wasn’t great.



The mean frequency looks odd here because I used a test record with a signal that’s nominally 1000 Hz but is actually, if the Feickert test track and another I have are to be believed, 1012-1013 Hz (nonetheless I used this one because my Feickert is horribly eccentric and the other is only 10 sec long). So yes, like most Regas absolute speed was too fast (even new ones – see e.g. here but, er, check the math for yourself…), but it’s really the poor speed stability I wanted to improve: presuming the “notch filter” gets rid of the worst of the 331/3 rpm variation (the peakiest parts of the yellow line) - which could include record eccentricity/warp as well as anything mechanical - +/- 0.24% for what’s left (red line) isn’t very good.

Side note: the ‘clipping’ of some of the yellow peaks is due to a combination of frequency being recorded to the nearest 1 Hz and the low sample rate. Together with the short record length – 30 sec test track – it leads to a bit of error in individual FFT results, but that’s partly overcome by averaging a few.

Transforming these data into the frequency domain gives a direct view of the frequency components contributing to the variation, allowing speculation as to their causes and possible cures. The graph below is an average of the 5 best of 6 readings (including the one above) taken in a single run - i.e., without moving the record w.r.t. the platter between measurements. It shows a big 331/3 rpm peak (0.55 Hz), a smaller one at twice that, and other possibly significant peaks at c. 2.4, 2.8 & 3.2 Hz.



The 0.55 Hz peak could have contributions from any eccentricity or a warp of the record and from anything in the drive train making its way to the platter. Hub bearing wear and sub-platter eccentricity could be suspected. Multiples of 0.55, within error - in this case the 1.1 and perhaps the 2.8 & 3.2 Hz peaks - could be record-related (i.e. warps). Wear in the motor bearings might show up as half-speed whirl at or a bit below 2.5 Hz, and similarly, hub bearing wear might show up around 0.25 Hz.

I first tried various combinations of cheap cures: different belt, different pulley, thrust bearing on the bottom of the motor, a more rigid (sorbothane-damped) motor mounting (the original motor was the old 110V Philips that dangled beneath the plinth on a rubber band). Average speed changed a bit with belt and pulley changes, but no change had significant effect on the speed variation. The minor peaks also remained, suggesting they weren’t due to belt or pulley issues.

Changing the load on a journal bearing can increase or reduce half-speed whirl. An increased load is often recommended to reduce it, but figuring that you never know and also wondering whether weighting/clamping might reduce any warp-related aspects of the measurements, I made a weight out of a 280 g practice hockey puck (Canadian, eh). Here’s the result:



A peak that’s likely half-speed whirl of the hub bearing appears around 0.25 Hz, and the 0.55 Hz peak is higher (off the graph at >1.2). These two phenomena suggest issues with the hub bearing. The small changes to the 1.1, 2.8 & 3.2 peaks are likely not significant, and given the test track is right at the outside of the record, if they were due to warps it wouldn’t be surprising that the puck didn’t suppress them. The 2.4 Hz (motor-bearing-related?) peak remains, as does the 1.4 Hz peak whose origin is a mystery to me but which appears in all the measurements.

Cleaning and re-lubricating the hub bearing and trying again with the puck eliminated the 0.25 Hz peak and reduced the 0.55 one to the no-puck level. Although this superficially cured that problem, it made me wonder what other components of the speed instability (‘notch’ deviation on the raw plots was still up there) bearing wear might have been contributing to. So in the interests of science, I bought a new hub/sub-platter assembly (GBP 22, kindly delivered by a travelling colleague). Results below:



I interpret the significant drop of the 0.55 Hz peak to show that the cause of much of the previous variation at that frequency was not record eccentricity but bearing wear and/or platter shape. The other possibly record-related peaks remain, as does the possible motor bearing peak (2.4 Hz). The improvement at 0.55 Hz is obvious on the raw plots too, absolute frequency variation being on average about half of what it was with the old assembly. Notch-filtered variation was improved, but still looks pretty wobbly; and annoyingly, average speed increased from formerly c. 0.75% too fast to c. 1.5% too fast.



To test if the minor peaks were record-related, the Feickert test track was used. The 0.55 Hz peak on this was off the chart at > 12 (> 14 before the new hub assembly), reflecting the much worse eccentricity of this record – which was bought from an eBay seller in Hong Kong, so I’m wondering if it was from a batch that didn’t pass QC and went out the back door of the pressing plant. On the original plots, the ‘raw’ curve oscillates over a range of about 40 Hz, whereas the record Marv used for his Classic 4 and 1Xpression measurements swings about 15 Hz.



The improved error and S/N (narrowing and raising of the peaks, respectively) result from a larger number of samples in these measurements: I ran each for 1 min, allowing 512 samples rather than the 256 that had to be used with the 30 sec long 1 kHz track. The peak just below 3.4 Hz might be a record multiple and correspond to the c. 3.2 one from the other record. The 1.7 and 2.25 ones are probably multiples. The 1.4, 2.4, and 2.8 ones seem to have disappeared but were present in Feickert runs done before changing the hub, so I’m not sure what to think about this; some record and some vibration effects, maybe?

Hearing that new 24V Rega motors aren’t too expensive and no doubt tiring of listening to test signals, Mrsdegraded offered an early Christmas present. I’d hoped to have received and installed it by now, but the retailer is having supply problems so I’ll save it for another post.

Learnings and conclusions
Doing FFT on PlatterSpeed measurements can better reveal hub and motor bearing issues and other low-frequency vibrations than the original plots – but it’s a shame the app’s sample rate isn’t higher so vibrations in the > 5 Hz range (including multiples of the motor speed) could be looked at.

Replacing my hub assembly made a difference. My bearing was initially badly lubricated and might have been worn; and/or the current RP3/RP6 subplatter (with the small raised lugs) is more circular than the old one (flat top). The speed increase with the new assembly (faster than the freshly-lubricated old one) is consistent with a contribution from the subplatter.

Having multiple test records is useful for cross-checking record-related variables, e.g. frequency (speed), eccentricity and warps. If your local second-hand shop doesn’t have one, encourage them not to turn down any that come in. And beware of new test records on ebay.

Looking forward to seeing what differences a new motor will make…

 

Forgot to mention probably the most important conclusion...

Whether or not you're going to do the FFT thing, take a few measurements and choose (a) representative one(s). Any spurious noise around 3150 or 1000 Hz, and spurious low-frequency vibrations, could be recorded - and because both the max and notch deviations shown in the plots are maxima, any such noise within the range of allowed recording frequencies but *outside* the range of speed variation of your TT or eccentricity of your record would show up as the max or min number.

The corollary to this is: be suspicious when you see any single plot or set of numbers that looks (or is claimed to be) bad.

 

Update with new motor

Rega 24v motor (GBP75 plus postage from Analogue Seduction in the UK; this is much cheaper than buying the full kit locally, but requires that you have or acquire a 60 Hz pulley and a mains to 24v AC transformer) finally arrived. I mounted it to the plinth with rubber bands and the same sorbothane gasket as I was using with the old motor rather than using the supplied sticky pad, figuring it would give better isolation. Had to use different belts, so the mount tension won't be quite the same as on the previously-measured old motor, but there wasn't a discernible difference with or without gasket on the old motor at these frequencies anyway.

The first noticeable thing was that absolute speed increased again. The new sub-platter brought it up to c. 1.5% too fast, and the addition of the new motor took it up to around 1.8% too fast (almost 34 RPM!). Adopting the internet's favourite trick of a wide rubber band around the sub-platter gave this result using the (very eccentric) Feikert record with 3150 Hz test track:



8.6 Hz or 0.27% too fast, about 33.4 RPM. Much better, but did it make the sub-platter too eccentric? Here's the best 5/6 result with my favoured test record as in the OP:



0.55 Hz variation is more than with the new sub-platter, probably because of the rubber band (I didn't check without), but less than with the old sub-platter. The 2.4 peak that I'd speculated was half-speed whirl of the motor bearing is still there, as are the 1.4, 2.8 and 3.2 peaks whose origins were more mysterious.

So at these frequencies, I'm not sure if the new motor has made a significant difference. However, when testing for noise by resting the stylus on the glass platter with the motor running (without the belt connected, obviously) and volume at max, things were much quieter than with the old one at audible frequencies, which is satisfying.

So in conclusion - this was an interesting exercise and did give some indication of the difference made by the sub-platter, but the limit to the frequency range that can be analyzed reduces the value of all the fooling around required. Back listening to music now...

 

AFAIK the speed thing is part of the Rega sound; they just run fast and that's the sound. It's kind of like tube distortion or EQ I suppose.

The Groovetracer subplatter products and upgrade belts should help immensely and are always highly praised.

I've tried the whole adjustable separated DC motor thing and in the end I just prefer my table to be one component rather than two accuracy be damned. Also within a small window of pitch deviation I can't perceive it unless it's direct A/B with a digital file playing simultaneously.

I do run two dead nuts accurate DD's and only one belt table these days though. The basic Thorens AC drive is stupid simple, you can't adjust anything other than the tension on the belt (either it's right or wrong basically) and if you're out of timing then you just need a new belt. They do sell these aluminum pulleys but I am not convinced they do anything better.

Checking platter speed by phone also has its challenges; the phone certainly weighs more than a record and is concentrated in one spot. The needle also drags on the record during playback which does not happen with a phone on the platter obviously. This matters more on belt suspension decks like Thorens rather than Rega I would think though.

 

The PlatterSpeed app I used works by recording the audio from a test tone on a record rather than sensing the speed by sitting on the platter, so it should reflect the 'in-use' speed characteristics of the TT.

Yep, Regas do run fast as a rule (the 60Hz versions maybe more than the 50?). Still, it's a bit annoying that the new sub-platter, while improving speed stability, doubled the absolute speed error, and the new motor, while eliminating noise, increased speed by that much again. The two together though were cheaper than the Groovetracer standard unit, and included the bearing sleeve, so I'm not complaining. And now I can amuse myself by rubber-band rolling

 

I'm sorry but unless you deal with the crap bearing everything else is a waste of time. I bought into the Rega b/s when they first started aka a Rega 2. I didn't learn and shelled out for a Rega 3.

Remove the ball bearing from the phosfer bronze cup and place it on something you know is flat. Just give the b/bearing a little shove and watch it roll around like a drunk 'as pissed as a newt'.

You may or may not know that there are various grades of b/bearings and you would have thought that Rega would have used the best - the difference in price, especially for quantity is not that great.

Before I wised up I shelled out for super-duper power supply for the Rega 3. Soon after as I posted on another thread I bought for my wife a well looked after Technics SL120 fitted with an SME 3009 arm and V15 - forget the arm and cartridge, it was the silence and speed stability that stunned me.

I have Rega 2 & 3 arms and can explain how to rewire them, how to set them up etc.

 

My Rega P2 always ran 1% fast no matter what I did until I enlarged the sub platter with some foam tape that gave it about 4mm larger in diameter (2mm all around).

This is one of the reasons I got rid of it. All Regas run fast and I think it os like this from the factory to make them sound more "lively". A psychoacoustic trick.

Same trick used by radio stations back,in the day. CD players fixed this until they had pro decks with speed adjust. My dad in radio knows this trick, but his stations never used it.

As @Mr.Sneis says, the Groovetracer, and there are others, that fix the over speed issue, but I read it makes the Regas sound more "dull", so YMMV. Mine sounded right finally, but still had speed stability issues. Flutter is a real issue.

At the end of the day, stepper motors are not designed for constant use or constant speed.

I finally went back to SS to a vintage Kenwood, but I think I am just going to pick up an SL-1200 and call it a day at some point.

 

What Kenwood do you have? You couldn't go wrong with a KD deck. The Technics sl 120/1200 are good but to shine they need lots of mods that cost money. With a KD - there's nothing to change to make it better just replace bog standard commercial PSU caps and rmost if not all the esistors, there are plenty but dirt cheap and the deck will outlive you and don't forget all the KDs have really good arms.

I bought a lightweight h/shell which should be better for the AT 33EV. Even with the original heavy aluminium h/shell the partly modded D/V sounds very good with the Kenwood and AT cartridge.

 

More necromancing, but it being my thread I can be Lazarus. So in the spirit of ‘Sheldon Cooper’s Fun with Flags’, I present another episode of ‘Biodegraded’s Listening to Excel’.

Short version:
For a synchronous AC turntable motor (24v Rega in this case), a dedicated power supply that generates its own AC signal improves speed stability over a simple step-down transformer supplied directly from the mains (probably - given some shortcomings in my data collection).

Long version:
I recently picked up a discounted demo unit of the now superseded Rega TTPSU mk2 (the one at the bottom) and wondered if I could improve on PlatterSpeed’s default data display to see what difference it made to speed stability. In previous posts I’d wanted to use the data to investigate the low-frequency behavior of my deck to see if platter bearing or motor issues could be revealed, so I did Fourier transforms on the datasets to break out the different frequency components. Now I’m interested in seeing how the speed varies over time, so I’ve been playing with the data to see if it can be analyzed in ways more illustrative than the app’s rather basic graph outputs.

The results are interesting, but a bit unsatisfactory for a couple of reasons. I haven’t compared in near real-time to see the difference between the right-now mains and the right-now power supply because the PS is a 50 Hz (UK-sourced) unit, so I’d have to change motor pulleys and spend some time getting them aligned right. The mains datasets I have are historic ones, actually from a different house. However, the power was likely to be cleaner there than in my current home because it was detached and I’m now in a multi-unit building. Also, there might be some inherent speed stability difference in the motor at the different speeds (250 rpm with the 50 Hz power supply vs 300 Hz with 60 Hz mains). For the same reason, it’s not straightforward to compare the FFT results with and without the power supply.

First, here’s the PlatterSpeed graphic output from a representative test of the mains-fed configuration after the last round of bearing, motor etc. replacements (same as in post #3 of this thread):



As mentioned in earlier posts the 7” Adjust+ test record is horrendously eccentric so the 0.55 Hz (33.3 rpm) cycle is wild, the 3150 Hz test signal changing in pitch by up to +/- 15 Hz. Feickert is no longer selling the 7” record for just this reason; most of them ended up this bad (they’re not making the PlatterSpeed app available on their site or at official app stores either, but you can find both the Android and iOS versions if you hunt). The notch-filtered curve (red) is better, but still shows a large amount of 0.55 Hz variation.

Now here’s a result using the TTPSU, obviously with a different motor pulley (larger 50 Hz version) and with the rubber band removed from the sub-platter. Note it’s a longer record, 124 sec compared to the mains one which is 63 sec:



The mean frequency is now lower than the input signal, about 0.15% below the target 3150 Hz as opposed to higher in the previous record (ha, sometimes stock Rega stuff doesn’t run fast). Taking the rest of the figures at face value though it doesn’t look like things have improved: both raw and notch-filtered deviations are higher than their counterparts from the mains-fed result. But with these figures being simple maxima and minima and with everything being swamped by the 33 rpm eccentricity, is that conclusion really valid; and might there be other variations hiding behind the saw-teeth of these datasets?

The data records were imported into Excel as previously, then plotted and various things were tried to see through the 0.55 Hz swing. Not knowing the details of how PlatterSpeed constructs its notch filter, I settled on a simple 5.5 sec (10 platter revolutions) moving average and then played around with fitting curves to that. The apparent long-period behavior seems to be fit pretty well by a 4th order polynomial. Here, with the frequency (y) axes zoomed to highlight the variations around the avg level, are the results with mains power vs TTPSU:





I took the last 60s of both records to minimize or eliminate the missing few seconds of moving average at the beginnings. Note that the graphed portions contain the largest and smallest +ve and -ve variations from the averages in both examples (compare these graphs with the screenshots).

Whether there’s much real variation going on at periods below 10 seconds or so is hard to judge because the 0.55 Hz variation still dominates the moving averages. At timescales of tens of seconds, however, the results are clearly different, with the mains dataset showing swings of a couple of Hz and the PSU dataset being very stable.

It would have been better to have longer records, especially for the mains data. I have a few other datasets with the Adjust+ record with older configurations of this turntable (including with the original 110v motor), and although shorter than the one displayed here, they seem to display the same behavior. And obviously it would have been preferable to have recorded the mains vs TTPSU behavior at the same time in the same room, but as I said earlier I’d have to fool around changing motor pulleys. Best of all would have been a better test record, without such extreme 33 rpm swings and with a long (2 min or more) test signal track at 3150 Hz (PlatterSpeed’s recording precision is 1 Hz, so 3150 is better than its other option of 1000 Hz). But nonetheless, I think analyzing the data this way has demonstrated a real difference and is more informative than the raw PlatterSpeed graphics.

Finally – how does it sound? Bass is tighter for sure, which is a big plus for me as I like a tighter more textured bass. It’s not up to the level of a direct-drive table I had in the system for a while, but it’s another addition to the improvement that resulted from a recent phono preamp change. Long sustained notes from voices, horns and some strings seem more stable and overall ‘cleaner’ as well but this effect being what I’d hoped to hear from better speed stability, at least some degree of my perception of it could be expectation bias.

So, worth the money, fun to play with, and convenient in that I now have electronic speed change for the handful of 45s in the collection. And because of that, I could continue bloody tweaking by adding one of those multi-belt pulleys…

But first I think I’ll buy more records

 

@Biodegraded, there is no difference in AC motors Rega puts in their tables. A stepper motor runs at the input frequency. So for the 60 Hz versions, they just change the pulley size. I found this out when I purchased a UK or 50Hz table myself (it was an NAD table actually that was OEM'd out by Rega). I did find a guy on eBay selling metal motor pullies out of stainless steel (opposed to plastic), and it did work. Michael Lim also sells them, but I did not get it from him.

But I am with you on the double or triple pulley for more torque. Lim even had one Rega with dual motors and triple pullies on each:

​

 

Yep, I know - the 24v motor I bought a couple of years back is a 'UK' (50 Hz) one. I already had a 60 Hz pulley, so all I did to adapt the motor for Canada was swap them over. The new PSU also being UK, though, generates 50 Hz so I had to change the pulley back again (for each speed the PSUs have both 50 & 60 Hz crystals, the only country-specific change being where a certain SMD goes on the board - photo at bottom of p2 here).

Having said that though, there might be small differences in the tuning of the '50 Hz' vs '60 Hz' motors: on the motor circuit boards there's a variable resistor in front of the phasing caps in each of the 33 & 45 rpm legs, which is presumably how the motors are "individually tuned for minimal vibration" (marketing). I imagine these are adjusted for correct phasing at different speeds, 250/341 rpm (= 33/45) at 50 Hz and 300/401 rpm at 60 Hz, depending on where they're going.

LOL. I have one of his standard pulleys; not sure ^that setup wouldn't just increase vibration though!

 

That is the best way to do it. This is one of the reasons I don't Ike stepper motors in a constant way like this. It is also why they use multiphase motors in industrial motors. I'm like you, I'd just prefer direct drive.