If you use the "normal" formula (RPM=freq*60/pole pairs) that should yield 3000/16=187.5. Speculating that 4x mismatch from "normal" yields half the expected speed (normal: 3000/4=750) half of which=375RPM which is what you were getting.
I lost my eyes searching for decent description of bldc operation and how to calculate no of poles... Still nada...
Today I run few more of the last tests before starting to finally assembly the TT.
Motor is equipped with new pulley and assembled on the mounting, I run one more speed variation test by playing 1kHz test signals from ortofon test record and capturing signal "dance" in RTA window of REW.
Same is done in post # 62 earlier, but then with stock plastic V-shaped pulley and o-ring belt.
This is new capture:
I got 8.4 Hz dancing window, or +-0.42% , might be slightly better but almost the same as earlier. Could be if someone is doing this that stock plastic pulley can be kept and TT run with O-ring, saving machining cost and probably very little difference.
Then I connected original Linn Sondek motor with its original pulley. Since this is 110V motor and I don't have such amp (did not want to bother with trafos now) , I run it from mains 50Hz + capacitor as originally intended. Speed is reduced to almost half in my set up now (Sondek is run against much smaller sub platter) , 1kHz becomes 540Hz at test record. For speed oscillation testing quite not relevant.
With same method of measurement, speed oscillation window was 5 Hz, or +- 0.463 % (base F is 540Hz, not 1kHz any more).
Ortofon LP has 2x 1kHz tracks per channel per side. I run 2 tracks of left channel side 2 by both motors. Figures above are from "better" track.
Funny thing is that on another 1kHz track, both motors showed about +- 0.6% speed deviation...
More and more I'm sure that I'm actually measuring quality of test record, not of TT.
If anyone else has this LP and is willing to conduct same test on her/his TT, I would like to see result.
And final comment on Linn motor (Airpax and Premotec productions I think) in comparison with BLDC capstan motor I have chosen.
At beginning I did not like that I can clearly feel poles ripple while rotating by fingers.
It does run quite smooth, but still observable vibrations comparing to VCR motor
Much lower torque
Requires 110 VRMS 2 phase versus 3 VRMS 3 phase.
Motor is equipped with new pulley and assembled on the mounting, I run one more speed variation test by playing 1kHz test signals from ortofon test record and capturing signal "dance" in RTA window of REW.
Same is done in post # 62 earlier, but then with stock plastic V-shaped pulley and o-ring belt.
This is new capture:
I got 8.4 Hz dancing window, or +-0.42% , might be slightly better but almost the same as earlier. Could be if someone is doing this that stock plastic pulley can be kept and TT run with O-ring, saving machining cost and probably very little difference.
Then I connected original Linn Sondek motor with its original pulley. Since this is 110V motor and I don't have such amp (did not want to bother with trafos now) , I run it from mains 50Hz + capacitor as originally intended. Speed is reduced to almost half in my set up now (Sondek is run against much smaller sub platter) , 1kHz becomes 540Hz at test record. For speed oscillation testing quite not relevant.
With same method of measurement, speed oscillation window was 5 Hz, or +- 0.463 % (base F is 540Hz, not 1kHz any more).
Ortofon LP has 2x 1kHz tracks per channel per side. I run 2 tracks of left channel side 2 by both motors. Figures above are from "better" track.
Funny thing is that on another 1kHz track, both motors showed about +- 0.6% speed deviation...
More and more I'm sure that I'm actually measuring quality of test record, not of TT.
If anyone else has this LP and is willing to conduct same test on her/his TT, I would like to see result.
And final comment on Linn motor (Airpax and Premotec productions I think) in comparison with BLDC capstan motor I have chosen.
At beginning I did not like that I can clearly feel poles ripple while rotating by fingers.
It does run quite smooth, but still observable vibrations comparing to VCR motor
Much lower torque
Requires 110 VRMS 2 phase versus 3 VRMS 3 phase.
I've never been a fan of the 2 phase motors by Airpax, Premotec, Hurst and others, they have significant cogging because of the metal pole pieces. A lot of the BLDC motors have metal stator cores as well, but their construction is different from the 2 phase types and reduces the cogging you are seeing. The Sony motor (#4 in post #50) is coreless so it should have the lowest cogging of any of the motors; they used to refer to this motor as a BSL type (Brush and Slotless) which is another way of saying coreless. If it has sufficient torque, it may be the best choice.
This Sony motor, inside view in post #118, is nice, runs very smooth on 750 RPM@ 50hz which is very useful, and has huge torque (this one I cant actually stop by fingers).
However it has observable noise when running most likely from its ball bearings, I could machine bushings and insert, at cost....
Besides this is coming from Sony HDW-M2000P, HD Video recorder with 4 digital audio channels, costing over 50,000 USD when actual 20 y ago. Except me being lucky to get one broken, nobody else will manage to get motor from such expensive and rare studio machine, not easily for sure.
I used cheapest motor from Philips VCR so far as it is very suitable and as such, or similar, available to all from scrapped VCR's.
Do you think it would be worth investing (at least for myself) another about 100 bucks in new bushings, pulley and housing to be turned?
Studied a bit more on this Sony motor , indeed this seems preferable design for DD TT's, so cant be bad.
Also checked specs for Digital Betacam , tape speed should be 101.5 mm/s. Since shaft driving pinch roller is 5mm diameter, that means motor was running at about 380 RPM that would call for 25Hz or about.
Run it also at that speed, still it is much more vibrations observable than with cheap Phillips, this time I can better locate them to winding carrying plate, not bearings.
I cant measure torque ripple (yet 😉 ), but I can feel vibrations . If I use it I would like to design something or to reduce it or to dump it. At the moment its tested with pure sinewave from Supaspin, if I manage I will try to utilize its original drive with hall sensors, could be it makes it run less vibrant (for note, these vibrations are not hurricane, just more than best motors so far)
One slightly off subject and idea for Lenco club:
These capstan motors are actually idler drivers; they all directly drive pinch roller in tape machine. The shaft housing is opened to accommodate idler, and shaft itself is precise for the same reason. Might be someone finds use for it as idler driver too.
Also checked specs for Digital Betacam , tape speed should be 101.5 mm/s. Since shaft driving pinch roller is 5mm diameter, that means motor was running at about 380 RPM that would call for 25Hz or about.
Run it also at that speed, still it is much more vibrations observable than with cheap Phillips, this time I can better locate them to winding carrying plate, not bearings.
I cant measure torque ripple (yet 😉 ), but I can feel vibrations . If I use it I would like to design something or to reduce it or to dump it. At the moment its tested with pure sinewave from Supaspin, if I manage I will try to utilize its original drive with hall sensors, could be it makes it run less vibrant (for note, these vibrations are not hurricane, just more than best motors so far)
One slightly off subject and idea for Lenco club:
These capstan motors are actually idler drivers; they all directly drive pinch roller in tape machine. The shaft housing is opened to accommodate idler, and shaft itself is precise for the same reason. Might be someone finds use for it as idler driver too.
It's your call...if the Philips motor is working well and there is little or no cogging then investing more money into the Sony motor might be gilding the lily.
Almost any BLDC motor will oscillate in speed when driven as an AC synch motor; it will average out to the center speed, but always be a little fast, then just a little slower. Driving them as a DC motor with closed loop feedback will reduce or eliminate the oscillation, but those controllers can be very complex and require the use of a DSP chip.
This is a polar plot taken from a 2500 PPR optical encoder connected directly to the shaft of a BLWS series BLDC motor from Anaheim Automation. It is being run open loop as an AC synch motor driven by a 3 phase sine wave drive:
The black circle at 3125Hz is the center frequency but you can see significant speed oscillations; the frequency of the oscillations is 40Hz which a compliant belt to the platter will smooth out almost completely.
This is a plot of the same motor run as a DC motor using closed loop feedback and a DSP with Field Oriented Control:
Almost an 18db improvement. The time scale on both of these is 400mS for 360° of chart (each rev of the motor=90° on the chart or 100mS/rev=600 RPM). At 600 RPM (10 Revs/sec) the 2500 PPR encoder produces 25kHz so the output of the encoder is run through a divide by 8 counter to get 3125Hz.
I've found that those 24 pole VCR capstan motors are rather sensitive to the centring between the outer rotor and the coil armature. I couldn't feel cogging until I had taken it apart and removed the coil assembly from its PCB. I conclude that my rough and ready reassembly didn't get the centring right. A job for a dial gauge or perhaps feeler gauges.
Hi,
Did you removed coils from PCB ? I never did that, but keep PCB, coils and bearings housing in one piece.
Might be you can sen us a picture?
Hello,
Thanks for this. I understand plots clearly but missing what motor (and its configuration) was used. Do you have complete paper to share.
Before I was under impression that all BLDC-s will run super smooth with 3phase sinewaves. This was mentioned by so many authors (to admit mostly audio people), but it seems wrong in most cases. I learned by testing several examples. This is one example of such statement praising sinewave, from Digikey:
https://www.digikey.com/en/articles...dally-control-three-phase-brushless-dc-motors
Until this morning I also wasn't aware of DSP use for driving these, all machines I studied had analogue chip feed by hall sensors to create feedback and kind of sinewave. Since there are many motor types here, no wonder manufacturers embed control and drive electronics in the motor, for good or for bad.
When I read your post this morning first thought was Ohhh NO, another bag of worms. But when thinking, it might be bringing one step further to resolution,
Key statements that I remembered are linked here :
https://www.ti.com/lit/an/bpra055/bpra055.pdf
Saying at the end of chapter 4: "Bear in mind that a sinusoidal Back EMF shape motor controlled with a sine wave strategy (three phase ON) produces a constant torque."
And here:
http://korfaudio.com/blog96
"If we spin the platter by hand, and look at the waveform that the direct drive turntable motor generates—we will see a sine wave (or something very much like it). This is the motor's way of telling us what it needs to be driven with. It's easy to oblige, isn't it?"
So what I did is connected 2 phases of several motors to soundcard inputs , switch on oscilloscope and watched back EMF shape, or what is this motor doing as a generator.
- In brackets i put ref No of motor from post #50
- Don't bother wave widths as it is turned by hand, look at the shape as this is what matters
(No1) smooth running and as so far favorite:
(No 3) drumhead motor, also smooth running
(No4) Sony HDVRC capstan, almost smooth running:
(No 5) Modern Papst outrunner, vibrating like hell:
For fun, I also recorded one stepper that I had at hand:
Looking at pictures above, there is very strong correlation between vibrations produced and level of matching generated and driving signal (all were driven with pure sinevavews).
Motors that do not generate sinewave vibrate horribly when driven by sinewave, seems to be clue....
By the way, I have eyes opened for one shaft encoder, if not for motor than for accurately measuring platter rotation.
Even half broken ones are good, as normally it has at least 2 phases... for me 1 phase working is enough...
Yes, I removed the coils from the PCB, thereby destroying the alignment between the coils and aluminium casting that contains the bearing bush, allowing the rotor alignment to be destroyed. Eventual aim is to treat the motor as a synchronous motor - but recent posts here suggest that might not be useful.
From all what I found so far, many Bldc motors can be treated as synchronous, but not all. Stay tuned 😊
Well, I hope it didn't cost you much. It will be difficult to center coils again.
But then, my so far best motor cost me just half an hour to dig it out of VCR, and to return rest of VCR to recycling
Despite coming out of VHS machines, my motors are the same as yours, just with a slimmer capstan and matching bearing assembly. But the coils are identical, as is the rotor. But that's not an enormous surprise as Betacam SP (very successful broadcast format) was developed from domestic Beta (that failed). Most broadcast kit rode on the back of consumer, especially the CCD sensors for cameras (very high reject rate, so selected from consumer).
Shouldn't be difficult to centre the coils. After all, the manufacturer managed it, and I'm used to centring things on lathes using dial gauges. Come to that, I've centred the upper drum on broadcast VCRs (<2um runout required). Usually, it just requires a little thought, delicacy, and a hammer. I kid you not. Light taps with a hammer is how it's done.
I await your motor findings with interest.
Shouldn't be difficult to centre the coils. After all, the manufacturer managed it, and I'm used to centring things on lathes using dial gauges. Come to that, I've centred the upper drum on broadcast VCRs (<2um runout required). Usually, it just requires a little thought, delicacy, and a hammer. I kid you not. Light taps with a hammer is how it's done.
I await your motor findings with interest.
Glad you are familiar with it, many years ago 2 kg copper hammer was my best friend in fine tuning much bigger machines, not that accurate do 😅
The motor in the plots was an Anaheim Automation BLWS231D-24V-2000 24V BLDC motor rated at 2000 RPM. When held in hand, it appears to run very smoothly because the oscillations are small and symmetrical about the center speed, still some vibration can be felt. I also assumed that an AC synch motor would operate at the speed determined by by driving frequency, but if the current is even slightly higher for the torque needed for the load, the motor will accelerate to a higher speed before the magnetic attraction of the rotor to the field pulls it back into synch. This is why AC synch motors oscillate; the average speed is very accurate, but the instantaneous speed is not. My thought was to run a direct drive motor as an AC synch motor which should have perfect speed, but when I did that at 33 RPM, the motor (platter) oscillated terribly for several minutes before settling down. For a belt drive at 600 RPM if the current is matched to the torque requirement, then the oscillations will be small and a belt drive will mitigate the vibration caused by this oscillation. On a direct drive, the only way I found to manage the speed accurately and consistently was to drive it as a DC motor with closed loop feedback. The DSP solution that I used on the SOTA DD uses two nested PI control loops. The inner loop is a constant current (torque) loop that uses FOC to convert the 3 sinewave currents to a single constant (DC) vector that is directly proportional to torque so it is essentially a DC servo for a constant torque PI control loop with very high bandwidth. The reference current term (torque) for the inner control loop is the output from the outer loop which is a PI velocity control loop that monitors the speed of the platter via a 360K counter/rev optical encoder. The filtered error output of this loop becomes the reference for the inner torque control loop so it maintains constant torque at just the level it needs to produce 33.333 RPM. The results are very low W&F (0.004%) as shown on the plot taken directly from the optical encoder:
Compare with the VPI HW40 DD (the signal again taken directly from the magnetic encoder at 355.5 Hz):
This is the spectrum of the SOTA Orioin encoder signal:
And the spectrum of the VPI HW40 encoder:
@Pyramid , thanks again, this post is great! The depth in which you went, and results achieved are impressive!
N.B. for readers just flying through this, Pyramid is solving DD problem, this is about belt drive TT which is much more forgiving, principles nevertheless remain the same and are applicable.
You helped me to understand root cause of the problem that I was solving empirically. If I expressed the issue in own words: motor overshoots on pole to pole level as it is almost free of load, next pole catches it and reduces speed again (on that level).
My solution for this issue (without knowing the actual root cause) was to make main bearing with a drag (I mentioned it few times in this thread) and give to motor a constant resistance to work against.
At the time I was designing this TT I was reading about then actual Bauer audio DSP TT https://www.stereophile.com/turntables/bauer_audio_dps_turntable/index.html . Against philosophy of that time (huge platter, hair thin o-ring, friction-less bearings (sometimes air, sometimes magnetic levitating) and flee motor that sometimes needed manual help to start) author Willi Bauer went to opposite direction: light platter, strong belt coupling and dragging bearing. If you check picture of my TT in post #1 and Bauer's it is clear where inspiration came from. Main thing that I changed is that my platter has 6.5kg, not really light.
My bearing is 20mm shaft and 65mm lenght with such tolerance that very little oil assists rotation. Rotation is smooth, but with drag. It takes 30min for removed platter to sit back in bearing, time needed for air to be released. I also used thickest and tightest belt available as Bauer did.
The drag is what increases the torque to give motor something to work against, reducing jumps between poles, heavy platter and strong coupling are increasing inertia.
In listening, especially when my Benz Ruby was new, this TT confirmed to present more clear sound and better sound-stage than for instance levitating JC Verdier La Platine with some Koetsu, but these comparisons were long time ago. Listening was the only indication there is an problem at that time, as said , did not know where it is so precisely as you explained.
This would close loop to provide constant current to motor instead of constant V as is now, still not constant torque, but with close to constant torque from good bearing should be good enough for belt drive.
The speed loop , I think is not needed, or not practical, for my TT design. Supaspin very accurately measures average speed (over 1, or half or quarter rotation) which is than manually adjustable in 0.01Hz steps. Unless I would also permanently install shaft encoder at platter (and that idea is in my head for decades, but is technically very challenging)), there is nothing to measure it better for feedback.
And 2 observations:
N.B. for readers just flying through this, Pyramid is solving DD problem, this is about belt drive TT which is much more forgiving, principles nevertheless remain the same and are applicable.
You helped me to understand root cause of the problem that I was solving empirically. If I expressed the issue in own words: motor overshoots on pole to pole level as it is almost free of load, next pole catches it and reduces speed again (on that level).
My solution for this issue (without knowing the actual root cause) was to make main bearing with a drag (I mentioned it few times in this thread) and give to motor a constant resistance to work against.
At the time I was designing this TT I was reading about then actual Bauer audio DSP TT https://www.stereophile.com/turntables/bauer_audio_dps_turntable/index.html . Against philosophy of that time (huge platter, hair thin o-ring, friction-less bearings (sometimes air, sometimes magnetic levitating) and flee motor that sometimes needed manual help to start) author Willi Bauer went to opposite direction: light platter, strong belt coupling and dragging bearing. If you check picture of my TT in post #1 and Bauer's it is clear where inspiration came from. Main thing that I changed is that my platter has 6.5kg, not really light.
My bearing is 20mm shaft and 65mm lenght with such tolerance that very little oil assists rotation. Rotation is smooth, but with drag. It takes 30min for removed platter to sit back in bearing, time needed for air to be released. I also used thickest and tightest belt available as Bauer did.
The drag is what increases the torque to give motor something to work against, reducing jumps between poles, heavy platter and strong coupling are increasing inertia.
In listening, especially when my Benz Ruby was new, this TT confirmed to present more clear sound and better sound-stage than for instance levitating JC Verdier La Platine with some Koetsu, but these comparisons were long time ago. Listening was the only indication there is an problem at that time, as said , did not know where it is so precisely as you explained.
Not fully cooked and baked idea: I think this part can be achieved more simplified for belt drive; current sensing resistor > full wave rectifier > VCA between sine source and power amps.
This would close loop to provide constant current to motor instead of constant V as is now, still not constant torque, but with close to constant torque from good bearing should be good enough for belt drive.
The speed loop , I think is not needed, or not practical, for my TT design. Supaspin very accurately measures average speed (over 1, or half or quarter rotation) which is than manually adjustable in 0.01Hz steps. Unless I would also permanently install shaft encoder at platter (and that idea is in my head for decades, but is technically very challenging)), there is nothing to measure it better for feedback.
And 2 observations:
- Anaheim motor you mentioned specified at 2000rpm in post #127: their website is mentioning 4 poles and custom arrangement, doesn't matter. Important is that my motor has many more poles than this one I use. I think this will result in higher frequency and lower amplitude of speed deviation.
- So far I measured platter with test LP at 1kHz (Ortofon test LP doesn't have 3150 Hz track) and consistently got exactly the same as you got on VPI (0.8%). Since encoder as you used is far more accurate than LP, I guess my TT is already better than VPI 🙂
My experience is that they can be made to rotate very smoothly. They're actually two motors on a common shaft. From trigonometry, cos^2 + sin^2 = 1, suggesting that they ought not to cog. But the equation requires that the amplitudes must be equal and the angle must be exactly 90 degrees (neither achieved with capacitor phasing). Inside the motor is a single rotor and its vertical position within the pole pieces is critical. Even if you have exactly the same current in each coil, unless the rotor is in the correct position, you get cogging. One of the early Connoisseur motors had a screw underneath the bottom of the shaft for adjusting rotor height to minimise cogging. I've driven such a motor from a supply that provides equal amplitudes and has a phase tweak. It's amazing how the vibration disappears when the rotor is at the right height (equal forces from both coils) and the phase tweak is adjusted (to correct mechanical misalignment of the two motor halves). I tried the same supply on a Connoisseur BD1 motor that had rectangular rather than triangular pole pieces and couldn't eliminate the cogging, so that suggests that the shape of the pole pieces is important.
I'm afraid I don't follow the argument for why a BLDC motor should not track the rotating field. What's the difference between it and a true synchronous motor?
All AC synch motors with PM rotors will have this issue including a BLDC motor operated as AC synch. It is caused by cogging which is the change in magnetic attraction between the rotor and the iron core stator poles/gaps and by the mismatch in applied torque vs what is needed to maintain constant speed. The motors are still operating synchronously as the (average) speed is determined by the frequency, but the speed oscillates up and down about the center speed (see the plot above for the BLWS motor driven by a Condor 3 phase sinewave supply). I observed this even with a coreless motor being run as an AC synch motor so it isn't caused solely by cogging. If the applied current produces torque in excess of what is needed to overcome losses, the platter will accelerate and move ahead of the rotating electric field. The farther ahead it moves, the more the field pulls back against the rotor until it swings back the other direction and falls behind the field and the cycle repeats. The speed will eventually settle very close to the synchronous speed but on a DD with a heavy platter, this took several minutes and any changes in load would send it back to oscillating again. On belt drive tables, the inertia of the rotor is small and the speed is higher so the oscillations are small and high enough in frequency that the belt will absorb most of the vibration when connected to a heavy platter.
Running the motor as a DC type eliminates the back and forth oscillations but to maintain constant speed, the torque (current) has to be exactly the correct amount or the platter will either slow or speed up. The load on the platter will be a moving target due to changes in stylus drag or changes in bearing oil viscosity and drag so even a tightly controlled current source will not hold exact speed if the load changes. Without a velocity feedback loop, the speed would drift over time. I'm not aware of any DD tables that don't use feedback to control the speed for this reason.