While waiting on the ebay gods...

[dmills]

So, I am sat here bored, waiting for the post to deliver the bearings and some other minor mechanical doings to allow me to do the mechanical stuff to finish my build.

Being bored, and not being in a position to do much in the workshop (The other lathe, the Colchester is also waiting for parts!), I decided to sort out that most annoying of projects, control panels... I hate designing them, so, so fiddly to get looking right.

In particular I decided to tackle the controls for the motors, vacuum, heat and feed rate, feeds and speeds mattering as much when cutting a record as when cutting metal.

Here is what I came up with:

Capture.PNG

The three big windows in the middle will have three sets of BCD thumb wheel switches to set the feed rate, with the buttons below used to switch between them as desired. Pressing 'Variable' will transfer the control of this drive to the computer for variable groove spacing (Incidentally I have a way to do that with an all analogue recording path WITHOUT needing a special tape transport).

The cutter heat controls will be a pair of 10 turn pots, where the 'ramp' parameter controls the rate at which the temperature falls with time since you started the cut, seems that lower surface velocity (closer to the centre of the disk) should intuitively need less heat. Plan is to PWM that circuit and measure the resistance during the idle time so as to close that loop.

Lock and lift stops the crossfeed, waits an appropriate amount of time to cut the lock groove then lifts the head.

Switches are EAO 18 series.

There will, I think probably be a small micro involved in this thing, I could probably do it the old way, but it would be a board full of 74 series and probably less reliable.

And yes, there will be volt free contacts for things like half speed to signal the audio electronics that curves need to change!

The equivalent panel for the audio doings is to follow once I have finalised that part of the system.

Now, what vital control have I forgotten?

Regards, Dan.

[2bitcomputer]

(Incidentally I have a way to do that with an all analogue recording path WITHOUT needing a special tape transport).

I'll bite - how are you planning to do a no preview head preview??

[dmills]

Well, seems to me that nobody cuts from tape without listening to it first....

So the idea goes like this, splice a marker into the leader at least one turn ahead of the start of the audio (more doesn't much matter, but it needs to be at least one turn at whatever speed you are cutting), this (on my deck) will likely be a change from opaque to clear leader or similar which can be read by a photocell.
This is to give a rough timing reference.

Stick the lathe computer in 'rehearsal' mode and play the tape thru, the computer digitises the audio (post IRIAA filter) starting when that photocell sees an edge (This is NOT the audio that will actually be cut!).

Stick the computer in 'Cut mode', rewind the tape, and go for it, now the computer sees the optical trigger and starts playing the recorded file 2 seconds ahead, to keep everything locked it can do some correlations between the live cutting audio and what is in memory to avoid any drift (Time code would be nice, but not everything has it).

From there you have the analog audio from the tape driving the cutter chain, the digital audio from the PC (1 Turn in the future in comparison) for doing the feed rate calculations, and it is all good.
In practise I would probably do the feed rate purely in the computer and output a simple tone proportional to the required feed rate from the sound card (Because that fits how the tacho on my cross feed motor works), but whatever works.

Feels like a few thousand lines of C sort of project, no big deal, and some tape machines already have a photocell setup (Normal use is to apply the brake when rewinding).

A 50:1 harmonic drive has just turned up in the post, lovely bit of engineering, so back to the main project....

[Radardoug]

There would be too much drift if you just let the tape recorder run. You need to lock it to a digital reference, like timecode.

[opcode66]

Incidentally I have a way to do that with an all analogue recording path WITHOUT needing a special tape transport

LOL ha ha ha ha ha ha. Magic? How exactly are you going to delay a signal via analog circuitry without addiing noise or distortion? Unless you have created something entirely new that will blow the lids off off electrical engineers worldwide I simply can't believe that. Sorry. What you can do is process the wave file and create a pitch map and use that to control the lathe. But, that isn't all analog. More details or I'm calling B.S.

[dmills]

Nothing I can see wrong with digitising the tape during rehearsal and then running that a turn ahead to control the groove spacing for an AAA cut.

Timecode is nice if you have it, but not everything does, so I was thinking in terms of a black and white striped disk on the back of the capstan motor and a photocell if needs be, or as I say run a correlator in the software and track the tape position that way, you know exactly where the correlation peak should be, so it's just lots of sums and a good SRC or VCXO (There are machines that are really good at that shit).

Hell, the capstan motors on decent machines are generally synchronous drives, so just feed a sample of one phase of the capstan drive to the computer, instant tape speed tracking....

The audio for cutting path stays entirely analogue (Because folks seem to care about that), but the cross feed drive is done digitally (it generally is anyway because that is usually how you skin a self tuning PID loop these days).

Seems like a few kloc of C, and a sound card (Which does not even have to be very good!), audio goes in, tone to control the feed rate motor comes out, a GPIO contact would be needed to get the rough alignment marker in.

Back of an envelope calcs for cpu requirements:

Correlation window of say 0.1 seconds, 48 kHz sample rate (more then enough for this), so ~5 ks correlation length, Order is N^2 so ~25M multiplies per correlation, even at 100Hz measurement rate a modern PC will laugh at that workload.

One of the nice things about the correlation approach is that you can implicitly measure what the operator is doing on the transfer console to allow the software to correct the spacing for the effect of things like gain changes.

Nope, for the avoidance of doubt I am NOT advocating a couple of seconds worth of allpass networks, that would be nuts.

Regards, Dan.

[markrob]

Hi Dan,

I really like the idea. If you go to that trouble, why not pre-calculate the pitch control signal based on the digital capture of the source and output the control signal to the pitch servo directly (e.g. via MIDI, other serial digital method)? It would make the real-time calculations and hardware much simpler.

Getting the two processes in sync is the trick. I'm thinking that you would want to use some sort of inaudible slate tone or maybe and optical marker on the tape to trigger lead-in and lead-out for each track. The slate info would also let you sync the pitch process on a track by track basis rather than once for the entire side and reduce the need to deal with timebase drift.

Mark

[dmills]

Sure, you could pre calculate it, but it ain't hard to do realtime, and realtime means you are at least somewhat aware of what the transfer console is doing.
Doing a precalc run to check that the material is cuttable and is going to play back on a realistic consumer deck is however reasonable (Overall record length, S channel excursion ,excessive HF for your cutting heads and suchlike).

My traverse motor servo is frequency controlled hence why outputting a tone to control the speed looks sane to me, but MIDI could work (just annoyingly laggy as you only get 3kB/s so the motor controller dynamics will play a bigger part).

If going for digital input then just writing a full up lathe controller makes sense, import everything from file, set the gains and gaps, lead in, lead out, do you want to cut a lock groove at the end of a certain track?, all that stuff, then do the sums, possibly a few times per track with different amounts of phase rotation to find the one that cuts best, but that ain't an all analogue cutter signal chain anymore (Useful, and I may write it eventually, but not quite the same thing).

I don't see what the sync problem is if given a correlator to work with and they ain't hard to write, pilot tone or timecode is a downer when you are being delivered content on tape without it, and there should be no real need, just give me a pulse within a few hundred ms of the same place a few seconds before the audio and I figure I am good to go (And that I can do with some clear and opaque leader and a razor blade).

Regards, Dan.