Let's Design and Build a (mostly) Digital Theremin!

"It's like #2, the entire thing from an inch or so up shifted ~2mm and stayed shifted to the end of the print.  I gripped the bed and the X-carriage underneath but there doesn't seem to be any easy play there between them via the spring knobs (though I do see what you're talking about).  The Y-axis belt was riding slightly to the right side on the front idler, and I played around its bolt so it's running more or less in the middle now, but I kinda doubt that was it.  It's got me a little skittish / gun shy." - Dew

With all of the things that can go wrong during a long print, a layer shift is particularly aggravating.  I would look at a few specific things if this is truly limited to a Y-axis shift - nozzle play, bed play, and Y-axis motor torque.   I am assuming that you are printing from an SD card here, but if not try that first.  Printing from your computer directly without a dedicated server can introduce all kinds of opportunities for printing glitches.

1) Have the unit turned on with a cold extruder, and use the controls to move any of the motors to some random position (so that all the motors are active) where you can grab the nozzle and try to wiggle it in all directions.  This is a bottom-line test that should show any sources of gantry/extruder backlash which can be caused by the hotend mount, the wheels, or the belt tension.

2) Next run the hotend up to about 260C and use a pair of pliers or an adjustable wrench to see if the heater block or nozzle is at all loose or free to rotate.  If it can rotate with moderate force then try to tighten the nozzle while holding the heat block.  If you find that the nozzle is already tight within the heat block then try to rotate the heat block to tighten the stainless steel heat barrier into the heat sink (I'm assuming this is a threaded joint, but check first).  If it tightens, snug it up without overstressing the wiring. Then loosen the nozzle inside the heat block, rotate the heat block back to the normal angle, and retighten the nozzle.  That's kind of a wordy procedure; a video would be more useful if you haven't had to do this before.

3) Now check the Y-axis bed backlash by again making a fresh stepper move (to make sure that the steppers are still active and holding) and try to push the bed in the Y-axis (and the X-axis too, since it has some slop).  Look for any movement in the table due to rollers, belt, or the z-height adjustments. 

4) If the mechanics look solid, then really bear down on the table or try to turn the Y-stepper pulley to check how much force it takes to break the motor's holding torque.  You can also try to resist the y-axis while it is in motion to check the moving torque.  If it seems weak, there is a hardware motor current adjustment, but this seems unlikely at this point.

Thanks Roger!

I'm always printing via SD card (wearing that poor thing out).

1 thru 3: It's all tight and IMO OK.

4: The Y-axis stepper motor static torque is really strong!  It's holding so well that I'm afraid I'll stress out the Y-axis belt by getting it to move.  But if I apply a steady -Y force to the Y-axis with my hand (attempting to push the table into the machine) and issue a movement via "Prepare | Move Y | Move 1mm" the torque clearly isn't as strong, as I can get the stepper to slip without significant fear for the life of the belt.  I would guess that this is normal - i.e. that the motor has to use up some of its static torque to accomplish a rotational step?  It takes a fair amount of effort to get it to slip, more than I would imagine happens during printing, though acceleration in some scenarios could perhaps be surprising in terms of total dynamic force?

If the G-codes are relative rather than absolute I suppose it could be a data error.

I tried downloading the craftbot craftware slicer but their new user web thing seems broken so I haven't gotten my hands on it yet.

[EDIT] I got IdeaMaker downloaded and running, and it has a nice toolpath preview.  PrusaSlicer doesn't?

Heh. A Gfx card that's better than some old on-board Mobo gfx chip shouldn't cost too much. Can find something that's <= 5 years old on craigslist or whatever these things you have there in the US?

Although somebody just showed me a screenshot of his OpenGL 4.0 laptop gfx and that vertical slider in Cura present (on windows 10, though).
Are you sure about this 4.1 thing, or is that the case specifically on linux or what?
Though I faintly remember, sometimes, especially integrated cards support a certain OGL version only partially...
Have you seen specific stated requirements of certain OGL features?
It's not the software I remember using in the past for this, but it looks similar, to check what an installed "card" supports:
https://sourceforge.net/projects/openglchecker/

Home Depot... can't you order stuff like that, if not at specialized online resellers, well, Amazon? I have bought drills and fan sanders (of types my closest "Home Depot" equivalent(?) did not have) there, OK, Euro Amazon, no idea if it's any different. It's certainly not less economical than driving to some place and back, and probably wasting time browsing things physically without neato search filters.

Interesting: PLA in the sun for a year. Maybe that's not what anyone would do with a Theremin prototype (not sure how much UV gets through regular windows, not much IIRC?), but good to know.
https://www.youtube.com/watch?v=qqNfa_zExRU

"A Gfx card that's better than some old on-board Mobo gfx chip shouldn't cost too much. Can find something that's <= 5 years old on craigslist or whatever these things you have there in the US?"  - tinkeringdude

Yah, probably just easier to buy something new from Amazon or newegg.  The thing with a PC as old as mine is: where do you stop?  Even a $40USD or so investment for just this issue could easily snowball into more.  I was really hoping to make it to the next gen AMD APU and upgrade the guts (uP, mobo, & mem).  Last upgrade was a 0.5TB SSD (totally worth it).

"Have you seen specific stated requirements of certain OGL features?"

Yes, in forum posts where folks were having this issue.  There's a compatibility mode checkbox in Cura that forces the issue (downgrade and remove slider) and I've tried that both ways with no difference.

"...to check what an installed "card" supports"

There's a command in Linux: "glxinfo | grep OpenGL" which gives this on my system:

OpenGL renderer string: AMD RS880 (DRM 2.50.0 / 5.3.0-62-generic, LLVM 9.0.0)
OpenGL core profile version string: 3.3 (Core Profile) Mesa 19.2.8
OpenGL core profile shading language version string: 3.30
OpenGL version string: 3.0 Mesa 19.2.8
OpenGL shading language version string: 1.30
OpenGL ES profile version string: OpenGL ES 3.0 Mesa 19.2.8
OpenGL ES profile shading language version string: OpenGL ES GLSL ES 3.00

Linux uses something called MESA to emulate OpenGL in SW.  Not sure why it can't do v4, but I've apparently got the latest version on my PC.

Another command: "inxi -Fxzd" gives this:

Graphics:
  Device-1: AMD RS880 [Radeon HD 4200] vendor: Gigabyte driver: radeon
  v: kernel bus ID: 01:05.0
  Display: x11 server: X.Org 1.19.6 driver: ati,radeon
  unloaded: fbdev,modesetting,vesa resolution: 1920x1080~60Hz
  OpenGL: renderer: AMD RS880 (DRM 2.50.0 / 5.3.0-62-generic LLVM 9.0.0)
  v: 3.3 Mesa 19.2.8 direct render: Yes

So my mobo RS880 chipset is equivalent to Radeon HD 4200.  IIRC it uses some of main DRAM for video memory.

"Home Depot... can't you order stuff like that, if not at specialized online resellers, well, Amazon?"

Yeah, I guess I like to see what I'm getting in the way of paint before I actually buy it, and I feel a little weird having people ship that sort of thing to me.  And shopping for anything from Amazon (or newegg for that matter) lately is really time consuming.  There's no curation going on with their offerings, and the search / sort doesn't do a good job when the categories aren't correctly filled out in their database.

"PLA in the sun for a year. Maybe that's not what anyone would do with a Theremin prototype (not sure how much UV gets through regular windows, not much IIRC?), but good to know."

Probably not a good idea to stick just about any raw plastic in the sun.  Some of those constructs seem like single wall "vases" (one long extrusion, no Z-axis steps?) which are pretty flimsy in the first place.  And I'll definitely be painting my stuff.

Waffle House

Intrigued by a video extolling the virtues of a certain slicer which allows one to finely control things like exposing the infill, I whipped up a quick 100mm square, 7.5mm tall, radus at corners and a 45 degree chamfer on front and back, with no top or bottom layers and 10% infill:

Gyroid infill is on the left, and is something I've never experimented with.  Cubic infill is on the right.  Gyroid is definitely not as rigid as cubic, though it is funner to watch print, and might make a good fan filter!

The reason for this experiment is to come up with some sort of way to print the plate antennas but have them lightweight and insulated.  Some slicers let you stop in the middle of a print to manually insert things like nuts, magnets, or PWBs, and then continue printing to seal them inside.  Maybe stick a sheet of aluminum in there?  The antennas should probably be designed before the antenna connectors on the D-Lev body are designed, just to know what I'm dealing with (cart before the horse).

Yah, probably just easier to buy something new from Amazon or newegg.  The thing with a PC as old as mine is: where do you stop?  Even a $40USD or so investment for just this issue could easily snowball into more.  I was really hoping to make it to the next gen AMD APU and upgrade the guts (uP, mobo, & mem).  Last upgrade was a 0.5TB SSD (totally worth it).

Ah, you were planning a complete upgrade anyway. Probably makes sense, that should last for a while.
SSD is really nice not only for booting, but also compiling C/C++ projects with a lot of small files...
I wonder how much the OS paging stuff all the time reduces their lifetime on its own, even if you yourself don't do much writing. Haven't looked at any real life longevity studies of those things.

I guess you don't want to wait until one of them things here is commercially available, if ever
Might be an interesting intro, though, as it#s geared towards EEs. I'm not one, but I knew nothing of how that tech is supposed to work and found it interesting also.
https://www.youtube.com/watch?v=jDW9bWSepB0

"Ah, you were planning a complete upgrade anyway. Probably makes sense, that should last for a while."  - tinkeringdude

My "current" PC is an amalgam of core parts bought in 2009 (AMD Athlon II X2 250 Regor 3.0GHz; GIGABYTE GA-MA785GMT-UD2H AM3 Micro ATX Motherboard; G.SKILL 2 x 2GB DDR3 1600) and aside from this issue I've been 100% fine with it (though I don't game anymore).  It's usually up and running 24/7 and has really hung in there.  I try to keep my eye on what's going on in the PC world of HW, and only lately have I been tempted to upgrade to a Ryzen 3 APU.  It's always a bad time to upgrade as there is generally some new desirable technology that isn't quite ready / cheap enough to make it onto low end motherboards, but right now seems fairly opportune.  Actual in-stock offerings seem rather slim though, probably due to C19.  I'm a "bang for the buck" kind of person when it comes to PC HW, and I'm probably more interested in electrical efficiency than computing power, as most PCs are powerful enough to make it 10 years or more (for me anyway).  Moore's Law seems to be plateauing out at a very useful point (or maybe my expectations aren't super high).

"SSD is really nice not only for booting, but also compiling C/C++ projects with a lot of small files...  I wonder how much the OS paging stuff all the time reduces their lifetime on its own, even if you yourself don't do much writing. Haven't looked at any real life longevity studies of those things."

SSDs can evidently take a butt ton of writes before failing outright.  But I wonder about long-term retention, as FLASH data retention time specs are based on the number of writes.  I mean, if I (and/or my OS) beat the thing up over 10 years or so and I pull it out and set it aside, will I be able to read it after another 10 years go by?  I haven't encountered any of that sort of testing.

"Might be an interesting intro, though, as it#s geared towards EEs. I'm not one, but I knew nothing of how that tech is supposed to work and found it interesting also."

Thanks!  I made it to about 30 minutes in, but what I heard was quite interesting and grounded (as these things go - I guess I'll be forever semi-allergic to college professors discussing their research in a painstakingly detailed manner).

Breaking The Sound Barrier

Try as I might (~10's of hours at least), I just haven't been able to get a realistic "ooo" vowel sound from the D-Lev.  Spent some time reading papers yesterday and today, a couple had nice tables of vowels, including formant frequencies and amplitudes and everything, with charts plotting the first formant against the second, etc.  I dial up the "ooo" sound and get bupkis.  Some of the other vowels sound OK-ish but not as real as what I've been able to just mess around and find on my own.  And the amplitudes in the tables make no sense: the second down 20dB?  Kinda crazy.  -50dB for the third?  You'd never even hear it.  And of course there are no damping / Q figures, no throat radiation, etc. so you're still pretty much on your own.  These same old tables get reproduced over and over in other papers and on the web.

A Sound-on-sound article said to crank the Q up to max (!) and also had wildly suppressed formants (and I find the author's cutesy writing style grating, we're not children).

A paper on breathy vowels seemed to be a mother lode of sorts, but it seems I'm past where they went.  They took a pulse train and gated noise with it, then applied atan to the noise to level off peaks.  Then they high pass filtered the gated atan noise and low pass filtered the pulse train with the same cutoff frequency to produce an overall "white" (non-tilted, spectrally flat) result, then applied a -6dB de-emphasis, and ran it through a formant filter bank.  I'm low pass filtering the noise with the treble control, which is a rough sort of de-emphasis, and my glottal source has dynamically variable de-emphasis.  And they had the same old formant tables, so nothing all that new unfortunately.

The "ooo" vowel doesn't seem to have significant nasal formants, as I can pinch my nose and it doesn't sound much different.  The harmonic content of the glottal source has a lot to do with the setting of the formant levels, and for those papers where they're really cranking down on the formants, I imagine it might be because their source is overly bright.  There's unfortunately no real standard here.

Tomita's "Venus the Bringer of Peace" (Holst) has a nice synth female "ooo" at the start.  And I've heard Theremins and synths doing passable female "ooo" so it's obviously doable and I must be missing something key.  Human vocals are touchy, get it a little wrong and the realism falls apart.

As a sanity check I ran the noise source through the formant bank and watched the spectra in friture - it checks out fine.

[EDIT] Wanted to add, also read about LPC (linear predictive coding) which is used to highly compress speech (hence high-priced horrible cell phone audio) - it yields coefficients that don't correspond to physical parameters and aren't very tweakable.  Compression based on actual formants could work and provide tweakable parameters, but accurately extracting the formants from female and children's voices (where the glottal stimulus can be above the lower formants) is evidently too problematic.

Vowels Gone Wild

Taking the "ooo" bull by the horns, I recorded my own voice doing "aah - ooo  aah", pulled it into Praat, and examined the formants:

You can see the bottom two formants dipping during the "ooo" phase.  Praat gives formant frequencies that don't sound right to me, so I adjusted them by back-and-forth listening between my own voice and the D-Lev.

I think the problem with "ooo" is it's not nearly as distinct as the neutral "aah" vowel sound, and so it's rather relative to what you are expecting to hear in relation to the other vowels.  Here's my recording: [MP3].  The first is obviously my voice, the second is the D-Lev trying to do the same thing.  An octave of "aah" sounds real enough to me, but an octave of "ooo" gets a little lost.  After that are some goofy moving vowel presets I've been playing around with.

The real vowels sound a bit breathy, though. Could make the picture somewhat less clear for Praat, I guess?

Also, after an octave or especially more above your voice's "resting pitch", if it was roughly there (though you started a bit lower and then shifted up 2 half steps or so), I'd expect to the vowels becoming less clear.
Try to actually vocalize it at that pitch.
It's a known effect among singers, although I'm not sure whether it's because of actual changes made in vocal setup to reach the high notes that cause it, or just that a smaller number of (humanly audible) harmonics is "scanning" the voice's response curve, as it were, making things less discernable.
At some point the first formant might also be skipped by the fundamental frequency for some vowels?