[Project] Turing Machine Mk2

[KSS]

@emmaker
Too many styles of 'leaf' sockets to reliably say. Heights vary. Sometimes widely.

general tip re: sockets. Regardless of height, always use double-wipe types.

[HeWhoWantsJeans]

I just replaced the trim pot and all of the IC's and I still cannot get the Turing Machine MK2 to do anything but be fully random or fully locked. The window for calibration is utterly tiny to the point of being worthless. I guess I can attempt to try a lower value trim pot (since the window with 1M is so damn small) or... I guess I put this and the expanders in the rubbish bin.

[patched]

I just replaced the trim pot and all of the IC's and I still cannot get the Turing Machine MK2 to do anything but be fully random or fully locked. The window for calibration is utterly tiny to the point of being worthless. I guess I can attempt to try a lower value trim pot (since the window with 1M is so damn small) or... I guess I put this and the expanders in the rubbish bin.

By trying multiple times with very much caution, I managed to get it to almost work correctly... It is not fully locked at CW, but
close enough for some uses (like a slowly evolving apreggiator) and works as well as a random sequencer at noon. Just be very very careful during calibration, it's a tedious process but with some luck you'll stop at the right position.

Dibs on it if you scrap it tho

[Faye]

I just replaced the trim pot and all of the IC's and I still cannot get the Turing Machine MK2 to do anything but be fully random or fully locked. The window for calibration is utterly tiny to the point of being worthless. I guess I can attempt to try a lower value trim pot (since the window with 1M is so damn small) or... I guess I put this and the expanders in the rubbish bin.

Have you tried another 2N3904? (I recommend reading through page 9 in this thread...)

[HeWhoWantsJeans]

I just replaced the trim pot and all of the IC's and I still cannot get the Turing Machine MK2 to do anything but be fully random or fully locked. The window for calibration is utterly tiny to the point of being worthless. I guess I can attempt to try a lower value trim pot (since the window with 1M is so damn small) or... I guess I put this and the expanders in the rubbish bin.

Have you tried another 2N3904? (I recommend reading through page 9 in this thread...)

I hadn't swapped the transistor, yet. I can give that a shot - thanks.

[HeWhoWantsJeans]

Just an evening bump: swapped the 2N3904 and the calibration window is still tiny and unreliable. Sometimes with the trim pot fully CCW I can't even get a steady tone.

However, I'm leaving it right now on a 'steady square wave' to settle but I'm doubtful that this resolved it, unfortunately.

[HeWhoWantsJeans]

Just a weekly bump: still the same behavior, unfortunately.

[bostonguy]

The trimpot is definitely too large. In my builds, I substitute a 50k trimpot and the calibration window is still very small. I’d recommend getting a 50k or 25k one and swapping it out.

[HeWhoWantsJeans]

Thanks. I've just ordered up 25k and 50k replacements to try out.

[cptnal]

Another one with The Calibration Issue here - completely random or completely static with no transition in between. Has anyone opened a GitHub ticket as suggested in the manual? I'm open to the trimpot idea, but it seems odd that Thonk would supply a kit with a part that's not up to the job.

And I've read through page nine as recommended a few times now. Is replacing the 2N3904 a proven solution? (Edit: HeWhoWantsJeans suggests not.) Is there a way to test whether it's bad?



Update: Going through my work again I measured 88K for R7 on the jack board, where it should be 470K. I've raised this on the GitHub page and will report back in case this is of any use to anyone.

[emmaker]

Update: Going through my work again I measured 88K for R7 on the jack board, where it should be 470K. I've raised this on the GitHub page and will report back in case this is of any use to anyone.

If you measuring in circuit without disconnecting one end of the resistor you are probably getting a bad reading. When measuring things in circuit like this you are measuring everything else connected to the resistor. I'd verify the resistor color code and if that is wrong replace it.

Adding the 1M trimmer is a poor solution to the problem of the noise transistors highly variable output levels. Proper solution would of been to test a bunch of transistors and get a ball park figure for the noise peak to peak level and then calculate the gain stage requirements. A combination of a fixed resistor to get the gain close with a lower value trimmer to tweak it in.

You will probably need the schematics to follow this.

If you have an oscilloscope or know someone that can do some measurements for you this may help. But note this is doing the math on the circuit and not actually measuring things. So it should get you closer.

• Measure the peak to peak output of U1C (pin 8) to get an idea of how much you want to set U1D's gain via VR1.

• Doing the math for the comparator voltage reference (R23, R22, R35, CHANGE, R6, IC7A) the voltage applied to the comparator (IC7B) is about +/-4V or about 8V peak to peak. NOTE: You probably won't be able to measure this at IC7A-1 or IC7B-6 since it's at virtual ground.

• We want to have a solid '0 or 1' coming out of the comparator when the CHANGE pot is at the ends. So that means we want the comparator reference voltage to be more or less than the noise voltage at the ends. So lets say we want the noise to be +/-3.7V or 7.4V (Vopp) peak to peak. For reference U1C-8 is Vnpp (peak to peak) an U1D-14 is Vopp. With some math: Vopp = (VR1 / R32) * Vnpp, VR1 = (Vopp * R32) / Vnpp, VR1 = 7400 / Vnpp. So that should give you a rough value for VR1 to get the proper gain.

Some other things to try with an oscilloscope.

Just measure the peak to peak noise voltage at U1D-14 and tweak it into around 7.0Vpp to 7.5Vpp output and see what happens. Digital scopes with peak to peak measurement makes this pretty simple. Analog scopes you have to work for it but you can get the info.

Put the noise and comparator circuits on a solderless breadboard or build it on protoboard with a socket for the transistor and the ability to measure VR1 (remember you can't measure in circuit, you have to disconnect one end of VR1). You can try different noise transistors with this and tweak VR1 in to get a rough value.

If someone sees any flaws in the math or logic please comment.

Jay S.

[cptnal]

Update: Going through my work again I measured 88K for R7 on the jack board, where it should be 470K. I've raised this on the GitHub page and will report back in case this is of any use to anyone.

If you measuring in circuit without disconnecting one end of the resistor you are probably getting a bad reading. When measuring things in circuit like this you are measuring everything else connected to the resistor. I'd verify the resistor color code and if that is wrong replace it.

Adding the 1M trimmer is a poor solution to the problem of the noise transistors highly variable output levels. Proper solution would of been to test a bunch of transistors and get a ball park figure for the noise peak to peak level and then calculate the gain stage requirements. A combination of a fixed resistor to get the gain close with a lower value trimmer to tweak it in.

You will probably need the schematics to follow this.

If you have an oscilloscope or know someone that can do some measurements for you this may help. But note this is doing the math on the circuit and not actually measuring things. So it should get you closer.

• Measure the peak to peak output of U1C (pin 8) to get an idea of how much you want to set U1D's gain via VR1.

• Doing the math for the comparator voltage reference (R23, R22, R35, CHANGE, R6, IC7A) the voltage applied to the comparator (IC7B) is about +/-4V or about 8V peak to peak. NOTE: You probably won't be able to measure this at IC7A-1 or IC7B-6 since it's at virtual ground.

• We want to have a solid '0 or 1' coming out of the comparator when the CHANGE pot is at the ends. So that means we want the comparator reference voltage to be more or less than the noise voltage at the ends. So lets say we want the noise to be +/-3.7V or 7.4V (Vopp) peak to peak. For reference U1C-8 is Vnpp (peak to peak) an U1D-14 is Vopp. With some math: Vopp = (VR1 / R32) * Vnpp, VR1 = (Vopp * R32) / Vnpp, VR1 = 7400 / Vnpp. So that should give you a rough value for VR1 to get the proper gain.

Some other things to try with an oscilloscope.

Just measure the peak to peak noise voltage at U1D-14 and tweak it into around 7.0Vpp to 7.5Vpp output and see what happens. Digital scopes with peak to peak measurement makes this pretty simple. Analog scopes you have to work for it but you can get the info.

Put the noise and comparator circuits on a solderless breadboard or build it on protoboard with a socket for the transistor and the ability to measure VR1 (remember you can't measure in circuit, you have to disconnect one end of VR1). You can try different noise transistors with this and tweak VR1 in to get a rough value.

If someone sees any flaws in the math or logic please comment.

Jay S.

Many thanks for that.

When I read it first I thought it was way above my pay scale since this is pitched as a beginner's project, but the more I read it through the more sense it makes. I raised a ticket on GitHub and Steve made the same point about measuring resistors in circuit. Haven't heard back since posting pictures though. He's probably fainted after seeing my soldering.

I'm going to get Googling on how to test transitors....

[whoop_john]

What a real Turing machine looks like…

Array Cell Address Check_2.jpg

[cptnal]

What a real Turing machine looks like…

Array Cell Address Check_2.jpg

Mine doesn't look anything like that. Probably did something wrong somewhere...

[Thonk Support]

When I read it first I thought it was way above my pay scale since this is pitched as a beginner's project

we don't pitch it as a beginners project at Thonk

Will get to your email, sorry we have a stupid big queue at the moment

[cptnal]

When I read it first I thought it was way above my pay scale since this is pitched as a beginner's project

we don't pitch it as a beginners project at Thonk

For sure, but there is a picture in the manual of what a good solder joint should look like in. Surely not something that concerns your seasoned DIYer.

Maybe "pitched" is putting too fine a point on it.

Will get to your email, sorry we have a stupid big queue at the moment

Great, thanks. I'm not being impatient. Well I kinda am, but I understand you're busy. So a bit of troubleshooting experience in the meantime is all good.

[Thonk Support]

When I read it first I thought it was way above my pay scale since this is pitched as a beginner's project

we don't pitch it as a beginners project at Thonk

For sure, but there is a picture in the manual of what a good solder joint should look like in. Surely not something that concerns your seasoned DIYer.

Nope that's something in every build document we produce and something a seasoned DIYer should remind themselves of every time they pick up the iron

[Quinie]

Just finished mine and can't get it to calibrate either. The trim pot has a very small window.. just to small.. I installed a 1m trimmer (bom/pcb).
When turning the trimmer it's nearly fully counter clockwise and then it stops making sound/ random. When going back to clockwise it suddenly goes to completely randomness and sound. And still a lot of the trimmer left to turn to.

I'm using a erica synth diybassline (output for the square), and also tried with my module tester. Can't get it to calibrate. Does the 50k or less trimmer work for others here ?

[Thonk Support]

Hey everyone, it does seem like we have people having calibration problems at a potentially increased rate in the past couple of months, although we do know lockdown has had people burning through their longstanding build backlogs too so we can't tell yet if that is a factor in it happening right now. We are looking into it. It's interesting to know who is building from a kit and who is building from just a Panel/PCB set.

[cptnal]

Hey everyone, it does seem like we have people having calibration problems at a potentially increased rate in the past couple of months, although we do know lockdown has had people burning through their longstanding build backlogs too so we can't tell yet if that is a factor in it happening right now. We are looking into it. It's interesting to know who is building from a kit and who is building from just a Panel/PCB set.

Confirming mine was a kit from your good selves.

[Quinie]

Panel and PCB set
Can provide pictures of build (or even de machine it self by mail if you wanna have it for debugging).

[patched]

Panel/PCB here

[bostonguy]

It is VERY IMPORTANT to use genuine quality semiconductors on the Turing. I would emphasize to anyone building from panel/pcb to source your semis from a reputable source. If you're using a bunch of cheap ebay/aliexpress TL074's that work fine for you in the audio path of other builds, you could find it doesn't work in the Turing. Same with the other Turing semis. I'm speaking from a painful experience I had a few years ago (with the Turing!). Since then, I never, never ever would build a Turing without genuine semis. Yes, they seem more expensive.... until you factor in the hours of time and the payments to Hair Club for Men.

[Triglav]

Hey everyone, it does seem like we have people having calibration problems at a potentially increased rate in the past couple of months, although we do know lockdown has had people burning through their longstanding build backlogs too so we can't tell yet if that is a factor in it happening right now. We are looking into it. It's interesting to know who is building from a kit and who is building from just a Panel/PCB set.

My unit, built from a kit in 2016 is also pretty hard to calibrate, but it’s not a big issue.

[J_Behan]

Hey everyone, it does seem like we have people having calibration problems at a potentially increased rate in the past couple of months, although we do know lockdown has had people burning through their longstanding build backlogs too so we can't tell yet if that is a factor in it happening right now. We are looking into it. It's interesting to know who is building from a kit and who is building from just a Panel/PCB set.

Mine was from a pcb/panel. Can't remember where I sourced my TL074s, probably from tayda.
I swapped out the 1M trimmer for a 50k but still can't calibrate it.
When I feel up to it, I might try removing the noise transistor and adding a socket, so I can try a few more types.
It might be an idea to suggest socketing that transistor in the build guide maybe?