Horizontal Lathe CNC into phonographer

[Sasasamb]

Hi! My name is Alessandro. Thanks @steve for letting me in.

I am an artist and I am trying to use a horizontal CNC lathe to build a phonographer. I would like to engrave bars of steel into phonographic cylinder. To be afterword reproduced by an other machine: an electric motor able to run at the same speed of the lathe being engraved. The reading stylus? Still no idea. It is like a new world for me. I am a musician and a designer and I listen to vinyl records since when I was -1.

I will start my experiment at the end of August. As I am waiting the technicians to come back from holiday in order to ask the possibilities we have with that machine. One of my questions for example is: can the engraving tool move fast enough?

So I am here to ask if anyone of you thinks that this could be a good idea or not. Do you think that conceptually theoretically it may work?

Of course the stylus is not going to engrave the steel because a sound is allowing it to vibrate. In fact it is going to be a musical wave previously translated into a drawing using with some code (G-code maybe) and then engraved millimetre after millimetre (at a crazy speed tho) into the steel.

Thanks for your interest. And sorry if I have not been really professional in the description of the technology. For me it is still a discovery! Have a good day!
A.

[Sasasamb]

I start to doubt I may be really naif somehow by thinking that drawing a line, engraving it, could seriously lead me to an engraved playable groove.

This leads me to the question: if I draw a random line from scratch that seems like an audio waveform will it possibly sound like a song? Is it the chaos theory that says that there is still at least one possibility that that may happen?

In fact I am starting to think that in a conventional classical way (vibrating stylus as consequence of a sound being emitted) there is a series of unpredictable engraving details that you may not be able to really translate bu following my experiment and a simple translation into CNC language of a waveform may not result into a playable groove.

Thanks. Best

[markrob]

you can get a playable groove using CNC G codes, but keep in mind, the quality will be very poor. The detail in a groove cut by conventional means has features down in the hundreds of nanometers range. Hard to get CNC resolution down anywhere near that far.

[farmersplow]

I think in addition to the exact resolution markrob described, the start/stop frequency of the stepper motors is another problem.
The resolution of the cut groove could "theoretically" be improved by a lever transmission - which, however, deteriorates the quality to the same extent. Regarding frequencies, the best stepper motors have 4.5 kHz start/stop frequency. Under load, about half (i.e. 3 kHz maximum). This does not achieve a reasonable quality. And on top of that, the angular tolerances per step for stepper motors are +/- 3'.
This will not work with the current technology.

Greetings from Austria
Thomas

[dmills]

For shits and giggles I cut a record on a Hardinge HLV, turned up a mandrill to take the blank with a stub to fit in a collet, and mounted the cutting head in a tool holder, then use the power facing mode to cut the groove.

Speed was a bit 'iffy' and the headstock belt on that machine is showing its age a bit, also you could hear the apron gearing, overall not exactly satisfactory but it did sort of work. A proper 'suspension box' would have helped and the variac was difficult to manipulate to get a good groove pitch, also limited to just under a 10 inch disk by the centre height.

[Sasasamb]

you can get a playable groove using CNC G codes, but keep in mind, the quality will be very poor. The detail in a groove cut by conventional means has features down in the hundreds of nanometers range. Hard to get CNC resolution down anywhere near that far.

Thanks. In fact I just found some programs able to convert an audio file into G codes. I understand it may become poor in quality, and I expect that, my attempt is to create an archive of documents, of voices of people of a small town. And I know a audiorecorder would be much better but I am pushing people into develop a thinking around the opposition between the high quality high volatility of a digital medium opposed to a sturdy less clear almost infinite piece of analogue stuff.

[Sasasamb]

I think in addition to the exact resolution markrob described, the start/stop frequency of the stepper motors is another problem.
This will not work with the current technology.

Thanks Thomas, your explanation was very detailed and I tried to grasp as much as possible. This week I will talk with the engineers and programmers of this company that uses CNC lathes and see after all this what they think. When you say “this will not work” you mean I wont be able to have a great quality? Or that no intelligible sound will ever be heard back?

[Sasasamb]

Interesting! That inspires me hope. But makes me think of one question: did you record a cylinder? Or a plate? And how did you manage to have the right instruction to cut the groove? Was it a CNC lathe? Thanks!

[Sasasamb]

As an update: I spoke with different engineers who actually said it is theoretically possible. I am happy to talk to you here as in fact you work on the other end. You are actually on that engraving stylus and I appreciate your answers.

Just few words concerning the project. I am an artist and it is a sort of residency inside this company. The aim of the residency is to put in connection the company and the town where the company is. They do gearboxes. So a lot of noise. They have lathes but they also have robots and maybe mills. So the idea was to record some voices of the town and save them on metal. To create an archive but also a commitment by means of which the company will always serve the town with recordings of new voices along the years.

Thanks again!

[dmills]

It was NOT a CNC Lathe, it was an old school manual toolroom machine (And frankly rather knackered). Look up Hardinge HLV for a picture of the kind of thing.
I cut a small disk (hence the limitation due to centre height), could have done a cylinder I suppose, but I have nothing to plat those back.

The cutting head was an actual record cutting head hooked up to an amplifier and eq in the usual manner, I was just messing about in the workshop.

[Sasasamb]

I see. And I saw the lathe. Nice experiment indeed.
May I ask you what kind of material did you use for the disc?

[farmersplow]

I think in addition to the exact resolution markrob described, the start/stop frequency of the stepper motors is another problem.
This will not work with the current technology.

Thanks Thomas, your explanation was very detailed and I tried to grasp as much as possible. This week I will talk with the engineers and programmers of this company that uses CNC lathes and see after all this what they think. When you say “this will not work” you mean I wont be able to have a great quality? Or that no intelligible sound will ever be heard back?

I really love the idea of your project! So I don't want to say that it can't be done. It might be possible to find a solution with strong quality compromises (and by that I mean worse than the worst Gramophone). I now have a week's holiday at the beach and will think about how it could be implemented.

A first approach could be: Instead of "half speed mastering" one could apply 1/100 speed mastering. You would have to convert the frequencies of the signal accordingly, but you would have more time to cut the sounds and could thus create the necessary frequencies of 2kHz.

I have a few questions about this:
1. at what speed (min-max) can the lathe rotate?
2. how high can the accuracy of the engraving knife be set (µm)?
3. should the final turning be done on a round disc (face) or on a roller? (I have made a small sketch).
4. how sharp are the sharpest engraving knives? (Radius at the tip?)

Horizontaldrehbank.png

Playing could be done via a steel needle (like a gramophone). With a round disc, a gramophone could be used directly? With a roller, you would have to modify a gramophone a little - but it can also work.

I'll think about it. I have no idea how that could work.

Greetings from Austria
Thomas

[Sasasamb]

Thanks Thomas

I really appreciate your help. Your drawings are very beautiful. I am still waiting to get back from the company, that is just back from holiday so I think they maybe overwhelmed.

You know, I think I will just go to them and ask straight away your questions. So it will be easier to grasp the situation for you too.

I agree on the worst-than-grammophone quality. For me it is important to just understand what this voices are saying. And this voices at least will be properly recorded.

I will let you know. And in case some one is reading about this experiment thanks for following this story.

Best!
Alessandro

[Sasasamb]

Hi! Today I have spoken with the guys in the company (6 engineers. I told them your questions and they are investigating on the tools. The sharpest possible and smaller possible. In terms of “resolution” they say they think they can not meet the nanometer but they can stay maybe somewhere around the micrometer.

These are the two machines they have: the dmg 2000 and the Multus 400.

Here the links to the two models.
https://en.dmgmori.com/products/machines/turning/horizontal-production-turning/nzx/nzx-2000

This is indeed older: https://www.gindumac.com/product/okuma-multus-b400-w_DK-TUR-OKU-2007-00001

Maybe you can see something about the machine capabilities (or non-capabilities) here that ring in your head. In fact the answers to your question (in particular for the DWG that seems newer and promising maybe…) are easier to find here. I hope!

On one thing they all agreed tho: we need a software that translates the audio file into a readable code for the machine. Into a readable series of positions for the engraver. And I think I may have found something that I need to check out. Some older threads pointed at some software on Github apparently able to do that.

Last but not least: let’s say that we will not reach the nanometers we need. What happens in the micrometer zone? Everything becomes “louder”? Less precise? I just wonder wether a gigantic groove would be read by a gigantic stylus.

Thanks!

[markrob]

Hi,

I decided to try and write some code in c# to do a proof of concept. I have a small Taig mill here running via Mach3. Not what you would call high precision. The stepper resolution is .000125". So, with that in mind, I tried to do a simple embossing CNC program that would cut 3" diameter locked groove loops. At 33 RPM, that gives you 1.8 seconds of sound. I used a conical carbide embossing tool that I ground. I wrote some simple code that takes an 8 bit raw format audio file as input and generates G code output. I used Adobe Audition to process the audio into the needed format.

Based on some calculations, I settled on a 5Khz sample rate and a 2.5Khz audio bandwidth with a 3" diameter loop. The mill probably reduces the bit depth to around 4 bits of resolution. I scaled the audio to produce a +/- .006" peak excursion and a depth of about .005". To test, I embossed a CDR (polycarbonate), but I'm sure it would work with an aluminum or brass blank as well.

Here is a link to a video of the embossing in action:
https://drive.google.com/file/d/1tNESUdSJdI2FMq2NfqJTCL5PyQeDd-IX/view?usp=sharing

And a capture of the result:

Test.mp3

The CNC Gcode that did the work:

test.txt

And a link to the code on GitHub:
https://github.com/markrob1066/Mill-Record-Cutter

Not what you would call Hi-Fi, but you can understand the voice.

Mark

[Sasasamb]

Hi Mark, your audio test is very moving. I really wish to get to something like that. If you agree I may try using your code. You said I need to use a 8 bit audiowave. May I?

Two questions: you talk about embossing not engraving. How do you read embossing?
Do you think this machine can be used to do something?

I must be honest: it has been hard for the engineers to understand my proposal. I tried my best but although they will investigate on tools and sharpness they still struggle with the comprehension of the method. That is why I come back here to have an exchange. I fell understood. If only they could produce something like your test they would see what I mean.

Thanks!

[Sasasamb]

I forgot the link. Sorry https://en.dmgmori.com/products/machines/turning/horizontal-production-turning/nzx/nzx-2000

[markrob]

Hi,

Feel free to use my code and modify as needed. If you do, let us know what you did and how it worked out for you. If you have specific technical questions, feel free to ask and I'll do my best to respond.

You can also try to run my G Code file on your machines to see if they do a better job.

I think embossing makes the most sense for this application. You are not removing material, but just displacing it. But, go with whatever works for you.

I wrote the code to be as simple as possible to test the concept. To that end, I decided to use raw audio format as input (no headers, just data). Wav, AIFF, etc. files could be made to work, you would just need to write the code to parse it and feed the data into the G Code conversion calculation. You could also add code to do the DSP filtering (IRIAA, conversion from velocity to excursion, band limiting, dynamic range compression and limiting, and decimation) needed end up with low sample rate, 8 bit audio data. To avoid writing that code, I just used Audition, which has great processing tools to do all of that. It also allows for saving in raw format. Many other DAW and editing software can do this as well. Also note that this was the first pass result of running the program. I'm sure many parameters could be changed that might improve the end result. I was pretty happy with the results as they are.

Regards,

Mark

[markrob]

Hi,

I made some minor changes to allow the user to setup CNC parameters safe height and feedrate, and embossing depth. Also did some cleanup to the code to get rid of some unneeded dialog boxes. Also added a readme to the git as well.

Mark

[Sasasamb]

Dear Mark, thanks for the update I have sent the new software version to the company engineers so that they can try out something today. I will update you about the results to possibly tomorrow. Let’s see what happens. Thanks again.