Lockdown turntable and lathe - a DIY project

[pentlandsound]

Hi Trolls! The following set of posts is an attempt to document the development of my latest home-made lathe, a kind of lockdown project that I started in the late autumn of 2020. I'd put together several earlier homebrew efforts up until a few years ago, but had left off for a while until I read of the successful developments by KNOP and other posters on this site using off-the-shelf components (such as all-in-one linear rails), V-Slot or T-Slot framing, and the real game-changer of affordable and accurate 3D printing. I don't have access to sophisticated machining equipment, just a set of very basic hand tools, and I'd have to describe my skills with them as enthusiastic rather than competent. Hence the recent increasing availability of the above modern conveniences was a strong draw to get back in the game.

One of the major problems I encountered right from the beginning in making records at home was that of insufficient torque in the recording turntable. There are some honourable exceptions, but generally speaking most playback turntables cannot cope with the extra drag that cutting or embossing causes - the platter may become wobbly (if it has little mass) or it may slow down and even come to a complete stop. My embossing experiments in the early 2010s used an idler-drive Lenco 75 deck. This worked, more or less, but the speed control had to be set slightly above the nominal to compensate for the drag: this speed was not constant, however, and would increase as the cutter approximated the centre of the record and the resulting moment about the centre pin decreased. The effect was just about acceptable in 7-inch records, but for anything much larger the change in speed became very noticeable.

So I'd set about providing my own solution. Years ago I purchased an inexpensive 200 W BLDC motor and controller on eBay, the idea being to deploy a system of pulleys to reduce the rotational speed from its maximum 3,600 rpm to the normal compass of record speeds. Also on eBay I found quite cheaply a heavyweight platter and bearing from a scrapped Lenco 75, and began the heretical task of belt-driving it. Having constructed a test rig, I found that although the speed reduction worked as calculated, the friction introduced by several sets of belts, pulleys and bearings meant that the speed was far from constant, varying by up to +/- 1 rpm from that desired: the system was clearly totally unusable!

I then considered using the BLDC motor with an idler, in proper Lenco fashion, but mounting the idler on the outside rim of the platter. This was a bit better, but it introduced some noise and I wasn't entirely thrilled with the results. Finally I tried the belt idea again, using just one belt with a tiny 16mm pulley on the motor. Despite this, the motor still had to turn extremely slowly to give sensible speeds at the platter, with definite and observable 'cogging'. At this point I gave up, dismantled my experiments and forgot about them for several years.

In November 2020 I read on Lathe Trolls of one contributor successfully refurbishing a suitcase cutter and flying in the face of traditional wisdom by deploying a stepper motor to turn the platter. The driver circuits available for these motors have advanced dramatically in recent years, with an array of features available to improve torque and reduce vibration. I thought it was worth a try, and found on eBay a NEMA-23 stepper, 23HS41-1804S, and a Trinamic TMC2209 'Step Stick' driver board. The motor draws 1.8 A per phase and offers a maximum torque of 2.4 Nm when half-stepping, but microstepping to quarter-step and below reduces this.

I constructed a test rig consisting of the stepper mounted on a steel plate, set off from the deck by anti-vibration gaskets and pillars. The stepper drove an 80mm pulley and 6mm round belt which was looped around the Lenco platter just under its rim.

00.jpg

This gave a reduction of approximately 3.7:1. The 'deck' was an ad-hoc assemblage of cut-off bits of laminate floorboard mounted in the drawer of an IKEA wardrobe; I promised myself something a bit more professional-looking in due course following satisfactory tests. (As it happens, the laminate flooring offers an unexpected benefit, from a testing viewpoint. The platter and motor are on different boards, which lock together immovably once horizontal, but by lifting and sliding one or other board the tension in the belt can be adjusted.) This motor proved only partially satisfactory, however. At 16 2/3 rpm there was almost no sound from the motor; at 33 1/3 it was just audible but very subdued. However, there came a point, when the platter reached about 40 rpm, at which motor vibration suddenly and very loudly made its presence obvious. I tried adjusting the allowed maximum current, both down, and up a fraction, and used different microstepping rates, but the problem remains: it seems that this particular motor/driver combo does not like to turn faster than about 140 rpm and do so quietly.

01.jpg

I did some more investigation into the NEMA-23 stepper, removing it from my test rig and running it by itself. Without any load, it will run at several hundred rpm, but with any kind of load (even mounted with a pulley unconnected to anything else), vibration manifests at speeds in excess of about 100 rpm, and this becomes unacceptably noisy above 140 rpm. So I modified my test rig to accommodate a NEMA-17 stepper, a much smaller one, that I had spare. This particular one (Wantai 42BYGHW609) draws 1.7 A and has a much lower inductance of 3 mH. Of particular note is the torque curve, which stays more or less constant at 0.3 Nm no matter how fast it's running. A 50mm pulley would mean a speed/torque gearing of approximately 6:1. With the motor turning at 270 rpm a platter speed of 45 rpm is obtained.

Further experiments with this smaller motor indicated that its use in a turntable system would be a practicable proposition. I briefly experimented with using an idler pulley to increase wrap round the motor pulley, but found that it added both mechanical noise and friction. With a 50mm pulley on the motor and a 3mm wide belt of ID 325mm, a centre-to-centre distance of 250mm gives sufficient wrap, without the idler, to avoid slipping of the belt.

The plinth I eventually built comprises three sheets of 18mm MDF interleaved with three sheets of 18mm plywood, all general purpose materials from a builder's suppliers, nothing fancy or 'audiophile'. All six sheets (500 by 350mm) are glued and screwed together, with a partial seventh layer holding the playback arm and rest. To make it look a bit better I applied several coats of Rustin's MDF Sealer, sanding between coats. (I then tried a coat of varnish, but it looked terrible so I stripped it off again.) The plinth is absolutely solid except for three holes bored throughout for motor, platter bearing and playback arm wiring. The platter bearing bolts directly to the top layer. The motor is attached to a small plate of 9mm MDF using an anti-vibration bracket; this plate is bolted to the plinth using rubber grommets to further reduce noise. This assembly positions the belt at the right height to go round the platter.

I'm using an Arduino to control the pulses to the stepper driver, and have organised things so that the speed ramps up and down gradually to and from the desired speed. The controller I built offers seven speeds: 8 1/3, 16 2/3, 22 1/2, 33 1/3, 39, 45 and 78 rpm; to achieve the last of these I had a 70mm pulley 3D-printed to replace the original 50mm one, and adjusted the Arduino timing accordingly. There is also a fine-tune adjustment. The driver is switchable for 1/8, 1/16, 1/32 and 1/64 microstepping, and I have included switches into my controller so that the least noisy setting can be used for a particular speed. (This refinement turned out to be redundant, in that the system works well enough at 1/8 microstepping, at all seven speeds.) Using a stroboscopic disc I found that the speed is much steadier than in my earlier experiments; I used freeware (wfgui.exe) to gauge wow and flutter and it averaged out at about 0.13% r.m.s. Torque at the platter is such that I have to apply some considerable force to it to slow it down at all.

The playback arm is another eBay salvage: its cartridge wires were a little short to reach the base of the plinth, so rather than attempt a complete rewire job I used some screened microphone cable to extend them sufficiently to wrap around the bottom edge of the plinth and terminate with RCA jacks in a little plastic mount. When I connected this to an amplifier I was pleasantly surprised to find that there is no hum! The arm itself is bolted to a strip of 3mm 'Dibond' (an aluminium and plastic sandwich - the aluminium layers are connected to signal earth via wires attached to a bolt passing through the strip).

The original plan for the controller was to use one Arduino to control step pulses to two stepper motors, one in the turntable system and another to move the recording head carriage along a linear rail. I have decided, on reflection, to use two separate Arduinos (Arduini?) and have two control units. There are two reasons for this: the first is that the Arduino has to send micro-stepping pulses every few hundred microseconds, timed by the onboard timers. If the timing were such that the Arduino had to send pulses to both motors simultaneously, one pulse might be delayed or lost; and if this state of things persisted for any length of time, there might be wobbles in the platter speed. The second reason is that a self-contained controller for the turntable makes it convenient to use as an ordinary playback device for general listening.

10.jpg

The 'MDFophone' turntable in 'general listening' mode, with its controller to left.

[pentlandsound]

So now I had a working turntable with good torque, and it was time to design a lathe to fit round it. I put together a basic overhead frame with five pieces of 40x40 V-Slot, and was lucky enough to find, second-hand, a 200mm linear rail and an optical-grade Z-axis slide (comparatively) cheaply on eBay, as they are hideously expensive to buy new. Having read about several successful builds using 3D printed parts, I used a free CAD program (FreeCad) to draw up my own designs, and had them printed in 'alumide' by Materialise.com.

19.JPG

FreeCad screen-grab of complete lathe assembly.

20.jpg

3D-printed parts - Back row: side bracket and dashpot support; cutterhead; main bearing bracket; raise/lower arm; bearing plate. Middle row: rail for limit sensors; cutterhead cover (stylus protector). Front row: small parts in 'model kit' frame; torque tube (in aluminium); two different blocks for connecting the X and Z axes; cutterhead back bracket. The cutterhead cover was added as an afterthought, for transporting the completed cutterhead from workbench to lathe with a possibly very expensive stylus in it. I was also a bit worried about the connection between the X-axis (linear rail) and the Z-axis (fine-tune vertical slide) because, once one part is attached to the connecting block, there might not be enough clearance to get a screwdriver or Allen key in to attach the other. So I had a second block made a bit wider than the first. As it happened the narrower one was fine.

21.jpg

The small parts: brackets and grips for holding swarf tube (for future use when cutting), limit sensor trip flanges (only one actually needed) and three different thicknesses of V-springs (1.2, 1.5 and 2 mm). Scale in cm.

22.jpg

Aluminium torque tube, tapped for M3 setscrew. Scale in cm.

I build a second control box, this one for the linear rail stepper, and equipped it with ammeters for total load and heater circuit (the latter is built, but not yet in use), VU meters, LEDs for linear rail limit switches and a LCD status display. The Arduino in this box connects to my PC and is controlled by software running on it ('Vinyl Burn', now completely rewritten, modelled on CD-R burning software but for making records). This software holds a playlist of files to be recorded on a side, and plays them one after the other through the sound card while sending the appropriate signals to the Arduino to move the linear rail carriage at the appropriate speed.

23.jpg

The two control boxes, for (L) linear rail and (R) turntable.

24.PNG

Screen-grab of PC software.


And this is my 'acid test' playback and tracking test equipment! Probably some teenage kid's pride and joy circa 1963. I rescued this for a very few pounds from eBay; it appears to be electrically safe and works fine. At some point, someone has replaced the original crystal cartridge for a more 'modern' ceramic one. When I received the unit the tracking force was set at a chunky 11 grams. I've reduced it to about 8, but can't go any lighter than that on pressed vinyl without it mistracking. The reason is that the crude arm bearing is merely one plastic cylinder mounted vertically within another - no fancy ball bearings or any such refinement, and it relies on a hefty downforce on a deep groove to overcome friction and make the arm move across. If my home-made records will play on this system, they will probably work on anything!

26.jpg

My first 3D-printed cutterhead was designed around the Visaton BF32 drivers, which several Trolls seem to have used with some success. I used aluminium perfume funnels to connect the speaker drivers to the torque tube, but first had to cut/sand them down to a height of 19.9 mm. The speakers are recessed 3 mm into the cutterhead body. To fix everything together I used B&Q own-brand 'Diall' epoxy adhesive. This is supplied in two separate tubes and has to be mixed before using. Once it's mixed, you have about five minutes to apply it before it starts to set and becomes unworkable. It is quite runny when first mixed, though, and there were a few dribbles here and there - but nothing serious! I held the assembly for several minutes while it set enough for me to be assured that it was absolutely square and symmetrical.

27.jpg

Initial tests were disappointing. There was sound being recorded, but it was at very low levels and with a LOT of the high frequency information lost. For my first test, I used the blank side of an acetate record, and attempted to cut using an unheated tungsten stylus. As soon as the stylus met the surface of the disc there was a very loud rasping or tearing sound; some signal was recorded, and swarf was produced (like navy blue wool fluff) but most of the signal was lost under this noise. I guessed that the age of the blank (~30 years) was probably one reason; another possibly the use of a cold metal stylus.
I tried a Myshank blank next. This was rather better, in that there was no tearing sound. What was strange was that the stylus seemed to work using either its front or back faces. No swarf appeared in either case, so it was definitely embossing, not cutting.
After a number of tests at different angles, weights and stylus orientations - with mediocre results throughout - I unmounted the cutter head and dismantled it. It came apart readily - too readily; the Diall adhesive had not properly 'taken' on the printed torque tube, and was quite soft and rubbery in places. My two perfume funnels were also perhaps a fraction too long, causing the driver cones to be pressed inward slightly; and the narrow ends were not filed exactly level and were bearing against rubbery adhesive rather than the firm sides of the torque tube itself.

I filed a further 0.5mm off the top of the funnels, carefully getting them straight and level at exactly 19.4mm. At this height the funnels would fit neatly into the speaker cones and exactly met the sides of the torque tube, with no inward pressure on either speaker. In other words, the drivers maintained their 'rest' position. I also sanded the surfaces of the torque tube. When it came back from the printers it was coated in a dull gun-metal grey kind of paint; I figured that this was one reason that the adhesive had not stuck properly. Sanding revealed the bright, shiny aluminium surface in its glory. With a better quality adhesive - J-B Weld - I did the gluing again, in two phases: first getting the funnels centred on the speaker cones, and once the adhesive had dried and cured, connecting the torque tube to the funnels and the steel anchoring wire.

The finished assemblage is very rigid - perhaps too rigid? I carried out two 'tap tests', connecting the output of one driver to Audacity, once after attaching the funnel and again after installing it in the head and getting everything glued. In the first case it resonated at about 180 Hz; in the second, 800 Hz. Notwithstanding this rebuild, subsequent tests continued to disappoint: the recorded sound remained tinny and thin, and quite faint, no matter what EQ and other settings I used.

[pentlandsound]

It was time for a rethink. In the mid-2010s I had had some quite good results using somewhat larger drivers, Digikey CMS0401, and as I still had several pairs of these left over from then, drew up FreeCad plans for a new cutter head that would use them. Around this time I discovered Grooveguy's how-to-build-a-cutter-head PDF on this site, and put together a third FreeCad design for a head which would use the same 'coin exciter' speakers as described by Grooveguy in his second build. With an eye to the future, both the new heads (and their mounting brackets) are designed with an 18 degree downward tilt - this would not have any effect on embossing, but might be of value for VTA compensation when I eventually get around to using a cutting stylus. The new torque tubes have a compensating tilt so that, if the tilted mounting bracket is used to mount the head, a cutting stylus ends up vertical; but if the perpendicular bracket is used, the stylus will be held at an 18 degree trailing angle for embossing. As 3D printing in aluminium is expensive, I had the new torque tubes made in the same alumide as everything else this time, deferring the additional expenditure until reasonably decent initial results were obtained. This second order to Materialise.com was in March this year, and I was a little worried in case there might be Brexit-related delays in shipping from Belgium, but the company seems to have worked around any problems and the package arrived, bang on cue, during the second week of April.

With the Digikey drivers, the little plastic saucers covering the voice coils were removed and Ferrofluid added to the inner and outer gap. Push cones for both heads were made from strips of aluminium drink can, and glued into the appropriate conical shapes using cyanoacrylate adhesive, following which they were stuck on to the drivers with J-B Weld, the apexes being set into recesses in their respective torque tubes and also glued into position using J-B Weld. Both structures are stiff - it takes some physical effort to move the drivers - but neither is anything like as immovable as that of the first head.

Figures for moving mass:
Head 1:
Driver: Visaton BF32
Coil: 2 x 0.5g (est.)
Cone: 2 x 0.89g
Torque tube: 0.69g (Alum.)
V Spring: 0.34g
Setscrew: 0.09g
Total: 3.9g

30.jpg

The Digikey head.

Head 2:
Driver: Digikey CMS0104
Coil: 2 x 0.78g
Cone: 2 x 0.4g
Torque tube: 0.47g
V Spring: 0.34g
Setscrew: 0.09g
Total: 3.26g

31.jpg

The Tectonic 'coin exciter' head, in situ on the lathe, with Presto sapphire embossing stylus fitted.

Head 3:
Driver: Tectonic TEAX19C01-8
Coil: 2 x 0.7g
Cone: 2 x 0.19g
Torque tube: 0.42g
V Spring: 0.34g
Setscrew: 0.09g
Total: 2.63g

These moving masses do not include the stylus itself. My earlier experiments from 2016 and before used a modified playback stylus mounted in a section of narrow steel tube. The included angle of the stylus was quite a bit more acute than 90 degrees, but in spite of the incorrect geometry it actually worked fairly well with my old 'hand-made' head, although it required a hefty downforce of about 120g to make an impression on polycarbonate. Tests on the latest two 3D-printed heads using this stylus were rather better than the first, but still a little disappointing in terms of recorded volume and frequency response. The third head was a little better than the second.

And then I found a reference to PIAPTK's Presto sapphire embossing cones on this site, read some very positive testimonials to them and so ordered one from the RecordLatheParts.com website. It arrived the other day, and I have mounted it in the third of the heads, with the 'coin exciter' drivers, 'leaning back' just off-vertical and with a total downforce of 35g as recommended. I also installed a vivarium radiator - a non-illuminating black ceramic whatnot that fits in a lamp socket. I recorded white noise with no EQ, used Audacity to analyse the playback, and created a compensating EQ curve based on what it said.

33.PNG

Audacity analysis of white noise record playback.

I'm very pleased with the results of using the Presto embosser so far - a huge improvement on all that went before. What it is to have the right tools for the job! Here's a playback from my first test session, embossed on a 16 rpm polycarbonate record:

Embossing test_16rpm.mp3

(the decision to use a slow speed was deliberate: the logic here being that if I can get my recordings to sound anything like reasonable at a slow speed then they should sound just as good - with luck, better - at a higher one). There is a slight 'ringing' of the polycarbonate plastic just detectable at the end - I think this may be because the Lenco platter mat is ridged at the 7-, 10- and 12-inch diameters, and the disc is thus only supported by one or more rings of rubber rather than across its entire area. I'm also aware of some rumble coming through from the linear rail stepper, which is bolted directly on to the V-Slot frame. This is not a problem at track-inscribing speeds but becomes apparent during higher-speed carriage movement such as run-out grooves.

40.jpg

The lathe, with the 'coin exciter' head mounted on its transport mechanism. When the head is fully lowered on to the disc, the stylus is at a very slight trailing angle (2 - 4 degrees from vertical).

41.jpg

Close-up of head transport. The counterweight is a collection of 'penny washers' held in place by two 6mm collars.

This post brings the 'project report' up to date (7 May) - I'll document my work henceforth in subsequent posts to this thread. Thanks to KNOP and Grooveguy in particular for their careful documentation of their own projects, and for making their findings public!

David

[markrob]

Hi,

Great job on your project! Really well done and thought out. I've been having good results with steppers and the Trinamic drivers as well. I don't understand your need for two Arduino's to create the step pulses. I am running both platter and overhead with one Arduino Mega 2560 using the on-chip timers to generate the pulse streams with no jitter, missing or delayed steps. How are you configuring and running your timers? If you are interested, I can provide you with my routines.

I directly access the timer registers and use the timer hardware assigned output compare pins to generate my pulses. Once the timer is configured, the CPU is no longer in the pulse generation process and it s all handled by the hardware and controlled by the internal crystal clock. Ramps are also hardware generated with some CPU assist. The timer output compare register is updated at each step via an interrupt process that insures that the CPU is writing the register during a safe time (just after the last step) and well before its value is needed for the next.

Mark

[jjwharris]

This is very similar to how I've approached it, however I've been using a single arduino to control an AD9833 chip which sends a square wave to an integrated servo to control the turntable motor - I imagine this method could also control the carriage motors, however I'm using and accelstepper library.

How are you measuring the cutting head current?

[pentlandsound]

Thanks for your kind words Mark. When I was choosing the stepper driver for my TT motor I decided to go for the Trinamic TMC2209 as it offered higher current handling capability than the 2208. The microstepping options it offers out-of-the-box are also different, being from 1/8 to 1/64 and selected using two inputs on the driver. Timing is controlled by the onboard 16-bit timer of an Arduino Uno by setting the output compare register, with each interrupt generating either a high or low output to the driver's STEP input. The target value of the OCR is set depending on the desired TT speed. I found, during initial testing, that controlling pulses to just the one driver worked fine, unless the 1/64 option was selected for the higher turntable speeds (>=45 rpm), in which case the motor ran slow. By my calculations this would involve pulses being required every 100 - 150 clock cycles (<10 µs). According to this webpage How long does it take to execute an ISR?, an ISR interrupt routine has a call/return overhead of about 2.6 µs, added to the time to execute user-specified code.

As it happens, the motor is fairly noiseless no matter the microstepping rate, but given the ISR overheads and the possibly high frequency of calls I was concerned that burdening the Arduino further with a second set of independent and variable timings for the linear rail motor might cause problems. When I thought about this a bit more I decided to put together separate controllers for the two motors, so that I could use the turntable independently as a general purpose playback deck. That said, I'd be interested in having a look at the routines you've used, if you could PM them to me.

jjwharris - using a multimeter I measured the voltage across one driver (8 Ω) as 1.35 V - this was done while playing a sample of pink noise at what sounded, purely subjectively, like a 'safe' level. Ignoring any reactive elements this gives average power as (1.35^2)/8 or 0.23 W and current as 0.23/1.35 = 0.17 A. A sample of programme material, appropriately EQ'd and played at a level that gives a good-volumed cut, came in at an average of about 1 V. Still on my to-do list is calibrating the pair of VU meters connected parallel to the amplifier's output using a dummy load. It would be more useful to me to have the meters show a maximum (safe) level rather than 'real' VU units. As the drivers are 2 W, 8 Ω I have an 0.5 A slow-blow fuse in line with each driver - experience is a great teacher.

[markrob]

Hi,

I don't use an ISR to trigger i/o events as that can introduce some jitter due to the ISR response times. I let the timer hardware do the work. Once the stepper is at speed, the last value written to the OCR stays in place and the hardware generates the step pulses at the desired rate. During the ramp, an ISR is generated after each pulse. In the ISR, I load a new calculated compare value into the register. Since the counter is now reset, its ok to update the compare register. With your method, you have freedom to use any GPIO as your step signal. For my method to work, you need to use the hardware assigned OCR output pin(s) on the Arduino. Going that way, gives you rock solid pulse steams with jitter measured in ns (due to the stability of the on-board crystal clock.

Mark

[pentlandsound]

I hadn't considered using a PWM stream for stepper control, but it seems obvious now ... I'll investigate this for my next iteration of the Arduino software. Using it for the TT motor should be fine, but I think I'd have to stick with ISRs for the linear rail motor: the PC control software relies on periodic reports from the Arduino of the number of step pulses sent to the motor, so that it can establish when to move to the next music track or start the run-out groove. The ISR increments a step count (and sets/interrogates some flags for possible 'main loop' action). I don't think it's possible to keep an accurate track of the number of individual pulses sent in a PWM stream - the only 'solutions' I've seen rely on sending the pulses to an input pin to be read and counted using an interrupt.

[markrob]

Hi,

That's exactly what I do. My overhead ISR counts each stepper pulse and updates the an absolute position counter. The ISR is also used during the ramping to update the OCR with the time to the next step pulse. This lets me generate a very nice linear ramp when changing LPI (e.g. lead-in, lead-out, track spacing). See the attached picts of the ramps measured using a DC tachmotor on the feedscrew shaft.

The ISR does not take long to execute. During the ramp, there is the need for a square root function and I found a fast 16 bit int function online that works really well. The ramp is based on the formula D= 1/2 AT^2 from high school physics days. Most of the calculation is done outside the ISR to keep CPU time down to a minimum.

On the platter, I don't need tightly controlled ramps and a don't use a position base ramp calculation. Just a crude timed ramp to get the platter up to speed. Once at speed, the last value written to the OCR is the correct value to maintain the speed with no CPU involved.

Overhead Ramp To Full Speed.jpg

Overhead Multi Speed Ramp Test.jpg

[jjwharris]

Mark, your knowledge and skill sets always blows me away.

My thoughts on using a PWM to control the carriage is probably a bit cowboy

By taking the processing load off the arduino, it would control the PWM signal entirely with timing rather than absolute position. However I haven't had an issue with just running the carriage with accelstepper.

My programming knowledge is fairly limited, at the moment, my carriage steps per minute are calculated by a simple distance over time calculations, and positioning. There is no LPI selection, only time per side. (Which I've now realised makes measuring groove width/depth a nightmare)

I'm curious as to how a nema17 would work with a TMC2130 driven by a PWM signal to drive a playback platter. It's something I want to experiment with in the future.

[markrob]

I'm using a NEMA 17 at 18Vdc to run the platter with an ~ 11:1 ratio (380 RPM at 33.33) using LMC2100 in 4X interpolated to 256 steps in spread cycle mode. I shock mount the stepper and the combination of the belt and platter mass seems to do a good job of filtering out the torque ripple. Not in any way equal to a high quality BLDC motor and drive. But good enough for my needs. You need to run spread cycle mode at higher RPM or the driver will lose control of the current.

For the Overhead , I run 16x interpolated to 256 in stealth chop mode. This runs the stepper open loop as a 2 phase synchronous motor. At the much slower speeds, it is quiet. I have the motor direct driving the feedscrew and that is marginal in terms of stepper torque ripple during the lead-outs. I again shack mount the stepper and I couple to the feedscrew with a very springy Ninja flex coupling I 3D printed.

[pentlandsound]

I did a bit more investigation into the rumble that was coming out on my recordings. At normal music groove pitch (8 - 10 lines/mm) it wasn't noticeable, particularly when there was music being recorded, but on track crossovers and run-ins and -outs (0.2 - 2 lines/mm), where the linear rail has to move at high speed, it was much more obvious. The stepper motor (NEMA-14) connects to the linear rail screw by an MXL belt and pulleys, and is microstepped at 1/16.

There were a number of possible contributory factors. The most obvious starting point was to mount the stepper motor on anti-vibration mounts rather than bolting it directly to the frame of the overhead. This new assembly (motor plus bracket plus four rubber mounts) is itself mounted on an aluminium plate and bolted to the frame through rubber grommets.

This removed some of the noise, but it was still a little obtrusive. My next test was to unmount the motor and simply hold it in my hand at the appropriate distance, with the belt still in situ, while embossing some silent grooves.

To my surprise, this did not make any difference. This time I took the motor off entirely, and embossed a few revs of silent grooves while turning the linear rail screw by hand (this is surprisingly hard to do evenly). This time, the rumble disappeared.

It took a while, and several dismantlings/re-mantlings of my overhead, before I realised that it might be the belt itself that was causing the noise. So I replaced the MXL belt with a round belt (a 2.5mm O-ring) and a pair of suitable pulleys. This wasn't quite as quiet as turning the screw by hand (!), but it was much better than the MXL belt. It seems to run most quietly with the belt fairly slack, with just enough tension in it to keep the linear rail moving. I was a little concerned that dispensing with the timing belt might cause irregularity in the spiral, but it seems to work fine, even when cutting unmodulated grooves at 12 lines/mm.

This MP3 demonstrates the run-in rumble before and after the fix. Both records were played back at the same volume setting; the second was embossed at a slightly higher level than the first.

rumble_test.mp3

20210528_1.jpg

[grooveguy]

Good work, David! What's more, you do a great job of documenting and explaining what you've done. A question and a comment, if I may.

First, where did you obtain the drive belt for your recording turntable? It almost looks like surgical tubing; is this something you purchase with a precision splice, or did you DIY that and, if so, how? It seems that the O-ring belt to your leadscrew imparts definite low-pass-filter properties, much to the improvement of the torque ripple noise floor. I'm translating your metric lines-per-millimeter into lines-per-inch by multiplying by 25.4. I figure that your leadscrew must be turning in the 1 r.p.m. range, and as your initial pulleys look pretty much the same diameter, the stepper must be turning at about this rate too. Is that about right? Seems plausible, to give you a nice wide pitch to the lock groove at the end.

Here's something I want to thrown out to you guys, and I hope the others see it as there are some good minds following this thread.

My cousin is something of a mechanical genius with a lot of shipboard experience in keeping commercial freighters plying the waves. I was lamenting my frustrations in building recording gear, and touched on the need to change recording pitch over a nominal 100:1 range. He's had some experience with stepper motors and immediately questioned the situation of torque ripple, as it would immediately translate to lateral groove modulation. He put forth an idea that I've now done a bit of thinking about, the idea of using two stepper motors and a differential gear system to extract the difference in speed between the two. This means that one could be running constantly at, say, 100 r.p.m., and the other at 99 r.p.m. for example. The output from the differential would be 1 r.p.m. All you'd need to do is to is slow down the one motor a bit more to get inter-band spiraling, and stop it for a dash to the lock groove. It would seem to me that you could use rubber belts or some sort of 'soft coupling' to reduce torque ripple, which at the higher rate of speed might make less trouble to begin with.

I looked at precision differential gear assemblies and nixed the whole thing when I saw the price... a thousand bucks isn't unreasonable. However, my Web search kept coming back to, of all things, differential 'rear ends' for electric model race cars. So I got a differential from a local hobby store and gave this idea a try, and although my first 'proof' breadboard was quite crude, it does seem to work. You couple one stepper to one axle, the second stepper to the other, and sure enough, the drive shaft gives you the difference. Well, not actually the drive shaft, it has its own beveled drive gear. But the differential housing actually revolves at the difference in speed between the two motors.

I don't have time to refine this right now, but want to give it a real-world test and see how it plays when I can. Just the idea unloaded here as the proverbial 'food for thought.'

Jim

[markrob]

Hi,

Interesting idea. If you take it further, I'll be very interested in your results. I'm not sure what this will do for torque ripple reduction during lead-out since that is the time when you hear the stepper torque ripple (high RPM). During normal cutting , the RPM is so low that the ripple components are very low in frequency. The steppers seem to work fine in this region. In fact, I wonder if the differential will couple the two stepper motor noise into the differential shaft since they will both be running at high RPM and well into the audible range. Were you able to observe the ripple effects when using this approach?

Mark

[grooveguy]

Hi, Mark; no, my initial test was not a real-world recording situation. And you're right, with both motors running fast, the torque ripple will clearly be in the audible range, and from two sources, too! They may even 'beat' or heterodyne, who knows? Probably best to couple the motors to the differential with low-pass belts, as David found this to substantially reduce the noise.

I was figuring that the advantage would be to increase the torque ripple frequency, supposing it would then be easier to filter out. Thinking about a low ratio, or direct drive, of the leadscrew at a slow speed, I just can't see how you would not have lateral modulation at the stepping rate. Assuming a 1.8° stepper not microstepped, and assuming 1 r.p.m., you'd think there would be a definite clunk-clunk about 3.3 times per second. Microstepping 1/16 would drop the amplitude per virtual step, but drive the frequency up to 53Hz, if my math is right. I did not hear that in David's test, although there was not a lot of blank-groove time without music. I was, however, impressed with his noise floor after the change to a rubber belt.

QUESTION for David: What was the diameter of the "suitable pulleys" when you changed over to the 2.5mm belt? And was the ratio 1:1 as it looks in your photo (with the timing belt) when you changed to an O-ring?

I wish I were more familiar with stepper motors and their drivers. I've never rolled my own driver with an Arduino or PC breakout box, always just trusting off-the-shelf stepper controllers. And there are differences between them for sure.

Embossing results that Trolls have been getting lately gives me great hope, especially as our sources for lacquer blanks and faceted jewel styli are very limited.

[markrob]

Hi,

In the case of the Trinamic driver (I'm using off the shelf SilentStepStick's here) at very low speeds in Stealth Chop mode, the stepper current is very close to a pure sinusoid due to the 256 step interpolation. See page 16 of the attached datasheet

TMC2100_datasheet.pdf

So even at 4x or 16X ,you get the same silent drive performance. That and any low pass filtering by the mechanics keeps the harmonics outside of the audible range (at least on my setup it seems to). As you suggest, you would think that the low pass action would help even more at the higher RPM's, but I think the problem is that the playback pickup in that frequency range is more sensitive to any disturbance as are speakers and ears. On my setup I added a flywheel to my overhead leadscrew to give me some more filtering and its seems to be better, but not totally eliminated. Personally, for a DIY setup, I can live with some hum during track gaps and lead in and out. I suspect you would have to go to high quality servo motor to really attach the issue.

Mark

[grooveguy]

Thanks for that, Mark; I'll look for a standalone stepper driver-thing that uses that chip. Ought to be one out there. I expect this is similar to a driver I bought a couple of years ago, claiming much the same features and performance. Yes, a bit of whine between tracks is no big deal to me. 60(!) years ago, I acquired my first 'pro' lathe, an RCA overhead that mated with RCA transcription turntables. Fixed pitch, about 120 lpi. I removed the gear from the spindle shaft to the leadscrew and had a machinist put a groove in the spiraling crank, about a 3" dia wheel w/ knob. I then coupled this to the shaft of some WW-II brush-type DC motor, which I ran (on AC!) from my old Lionel train transformer. No speed control, but it was pretty steady and gave me microgrooves. Spiraling-out with full voltage from the transformer sure made that motor sing, which was quite audible on playback, but who cares?

[pentlandsound]

Thanks for your comments Jim! In answer to your earlier question, the orange belt I used on my first experiment was a 6mm polyurethane one. I bought it, for an earlier project, from this website https://www.motionco.co.uk/index.php?route=product/round_belt. The sellers will make up the belt to the required length. (I don't have the equipment or steady enough hands to attempt it myself.)

This type of belt might be suitable for most small mechanical projects, but in my tests, the join - although expertly done - was causing a slight wobble in the platter speed every time it went round the pulley. I thought that an 'endless' nitrile O-ring might work better: they are widely available in all sizes and cost literally pennies each. The one I settled on has ID 325mm and cross-section 3mm. https://www.polymax.co.uk/o-rings/o-ring-selector/

To answer your second question: the MXL pulleys on the leadscrew and motor were identical, 12T; and the new round-belt pulleys are also identical, 20mm OD, 6mm ID, with adapters for the 4mm leadscrew and 5mm motor shafts. The leadscrew has a pitch of 1mm: at, say, 9 lines/mm (229 lines/in) at a turntable speed of 45 rpm, the head must move 5mm per minute - so with the 1:1 pulley ratio the motor speed is 5 rpm. At lead-out speed, 0.2 lines/mm (5 lines/in), the motor must turn 9/0.2 = 45 times as fast, or 225 rpm.

speed_of_1mm_leadscrew.JPG

[grooveguy]

Great! Many thanks for all the info and data, all makes sense now. Do continue to refine this and, please, keep us updated on your work. Very promising, indeed.

[markrob]

Hi,

Digikey (amongst others), sells the boards I used on my build (see link). They will drop in to a standard RAMPS board used on 3D printers. In my case, I made a custom shield PCB for an Arduino Mega that had two sockets for the the Trianamic driver boards along with the additional I/O needed for my build. I used the 2100 version which I think was the first device in the series. Since then, Trinamic has released several variants of this driver. All seem to offer the same 256 step Stealth Chop and Spread Cycle modes. Key Differences are interface type and max current and voltages. Silent Step mode works great for very low speeds and low accelerations. Spread Cycle works better for higher RPM's as it adapts the current control on the fly to deal with back EMF and inductance. My setup runs much slower RPM's than yours. I'm direct driving a 8mm leadscrew which results in 3.175 revs/in. At 300 LPI and 33.33 platter RPM, my screw runs at about .35 RPM.

https://www.digikey.com/en/products/filter/evaluation-and-demonstration-boards-and-kits/787?s=N4IgTCBcDaIM4EsA2BTAdgFzhlAHbCAxgNYgC6AvkA