How much drift is a good thing?

From circuitbending to homebrew stompboxes & synths, keep the DIY spirit alive!

Moderators: Kent, luketeaford, lisa, Joe.

Post Reply
skee
Common Wiggler
Posts: 58
Joined: Sun Sep 22, 2013 11:39 am
Location: Dublin

How much drift is a good thing?

Post by skee » Wed Sep 09, 2020 6:25 pm

Just completed my build of the Soundtronics Midi Ultimate and all of the modules are up and running. However, when calibrating the MFOS based VCOs, I found that all three drifted around 6 to 7 Hz at the calibration frequency of 800 Hz over three or four minutes, making calibration quite difficult. This was after a 20 minute warm up. The power supply seems stable enough, and changing settings on the other modules has no effect on the rate of drift, which seems quite random within its range. This percentage of movement looks the same at all audio frequencies.

Poking around with my finger showed that the only sensitive components were the exponential converter transistor and it’s associated tempco components. So I put a blob of Plasticine around them, and this reduced the drift to two or three Hz at one kHz. The room temperature is fairly steady.

So as a this is the first time that I have got this far with a complete build, my question is whether this is an acceptable level of instability, and beneficial to the character of the instrument, or should I continue to search for other causes, and what might they be?

Don T
Super Deluxe Wiggler
Posts: 1282
Joined: Sun Feb 02, 2014 3:11 pm

Re: How much drift is a good thing?

Post by Don T » Wed Sep 09, 2020 6:45 pm

Place it inside an enclosure and test with a tuner over a period of 20 minutes. The lack of air movement around it should make it more stable. If testing this way does not yield satisfactory results, then explore other options, such as thermal paste or thermal epoxy

Borogove
Learning to Wiggle
Posts: 44
Joined: Sun Feb 25, 2018 10:56 pm

Re: How much drift is a good thing?

Post by Borogove » Wed Sep 09, 2020 7:24 pm

The difference between two pitches a semitone apart is about 6%; we use "cents" as a ratio unit for 1/100 of a semitone. Tones and cents are exponential units but over short intervals we can pretend they're linear; 3Hz at 1000Hz is 0.3% -- 1/20 semitone, which is 5 cents. 5 cents is generally accepted as the "just noticeable distance" for average listeners, and hearing two tones 5 cents apart at the same time yields perceptible beating but not unpleasant dissonance.

So if 3Hz at 1kHz is reliably your worst case drift, and your drift is proportional to your tuning (i.e. 0.3% of whatever frequency you initially tune to), you should be good enough for rock-n-roll?

Try getting it in tune with a guitar and/or a digital synth, play them together for a while, and see if it sounds good. Ultimately, that's the important thing.

KSS
Super Deluxe Wiggler
Posts: 3411
Joined: Mon Jan 25, 2016 7:28 am

Re: How much drift is a good thing?

Post by KSS » Thu Sep 10, 2020 5:12 am

Do you have a photo of your expo transistors and their tempco? How did you arrange them before adding the plasticine?
Treat utility modules as stars instead of backup singers.

Treat power supplies like Rockstars instead of roadies!

User avatar
devinw1
Super Deluxe Wiggler
Posts: 1575
Joined: Tue Aug 07, 2018 11:20 am
Location: Portland, OR
Contact:

Re: How much drift is a good thing?

Post by devinw1 » Thu Sep 10, 2020 11:18 am

That sounds kind of high after a warm up period. Is your room temp or airflow fluctuating a whole lot while you're doing this?

skee
Common Wiggler
Posts: 58
Joined: Sun Sep 22, 2013 11:39 am
Location: Dublin

Re: How much drift is a good thing?

Post by skee » Thu Sep 10, 2020 4:07 pm

VCO2 & VCO1.JPG
Photo attached. The unblobbed transistor pair and tempco are on the left. I gave it an overnight soak test, and it dropped about 8Hz, but the temperature had fallen by about 10 C, so maybe not so bad?

In contrast a 555 similarly set up moved down less than .5 Hz. The air in the room is pretty static and the blob lends a fair bit of thermal inertia. It's in an attic but well insulated. I have no other musical instruments or tuners to compare it to, I'm using the frequency meter in the scope (GwINSTEK GDS-1052-U) and a 20 quid DMM to measure.

Any recommendations for a suitable thermal conductive potting epoxy in a small volume? Some of the prices I've seen are higher than the price of the VCOs.
You do not have the required permissions to view the files attached to this post.

User avatar
guest
Super Deluxe Wiggler
Posts: 4511
Joined: Mon Aug 19, 2013 11:49 am

Re: How much drift is a good thing?

Post by guest » Thu Sep 10, 2020 4:53 pm

the MFOS vcos also use the voltage rails for the reference currents and offset CVs, so if the rails move at all, youll get frequency drift. 8Hz @ 1kHz(?) is within reason for a thermistor compensated VCO with transistors in independent packages.
openmusiclabs.com

KSS
Super Deluxe Wiggler
Posts: 3411
Joined: Mon Jan 25, 2016 7:28 am

Re: How much drift is a good thing?

Post by KSS » Fri Sep 11, 2020 2:03 am

skee wrote:
Thu Sep 10, 2020 4:07 pm
Photo attached. The unblobbed transistor pair and tempco are on the left.
Thanks. That's kinda what i expected to see and it clearly shows the problem. The two transistors need to touch. It's the gradient between them that makes the difference. The tempco then sits over the top in this -less than ideal- case.

What you have instead is the electronic equivalent of social distancing! Good for a pandemic, not so much for expo pairs. One transistor mated to one side of a tempco, and the the tempco mated to the other transistor. There is a huge disconnect between the two transistors. And that's you'r problem source. Their mating faces is far more important than the tempco. Ma and Pa Q can't kiss with the swaddled baby R between them!

Here's what to do.
If you lean the Ma Q on the left backwards -pretend she's a limbo dancer and bend her legs 90 degrees. Then bend the 2nd Pa Q over as if bowing to Ma Q's limbo expertise! This will mate their faces in a loving embrace. Now ideally the baby R tempco is at the same place their faces meet. But that's not possible so -at the very least- put it over the top -Pa Q's back- and use some white heatsink grease bwtween their tranny faces-THIN!- And then use a SLIGHTLY bigger glop for the tempco to transistor connection. That's a likely 'good enough' solution. But it's still not ideal.

Ideally this tempco would be parallel to the face of the trim pot and between it and the Q pair. With its height sitting tangent to the plane where the two tranny faces meet. This will give a "line" contact at the meeting plane and will require less heatsink grease. It will also put the tempco in a semi-'shielded' area between the trim and the tranny's to reduce the effect of ambient airflow. And allows for a tiewrap to hold it all together.

To do the ideal tempco position -and it's worth it!- you'll need to bend the baby R's legs quite a bit and maybe even sleeve them with some insulation to keep them from touching the tranny legs. Or sleeve the tranny legs instead. Don't worry about them being in close physical proximity. But they can't touch electrically. Probably need to de-solder it.

The designer of that PCB and silk screen -seemingly- did not understand the whole practical aspects of expos and tempcos. Fortunately you can fix this. And you should.
I gave it an overnight soak test, and it dropped about 8Hz, but the temperature had fallen by about 10 C, so maybe not so bad?
No, it's bad. But the fixes detailed above should fix it. While we're on the subect, I'm not sure why so many "test" expo pair-tempcos with steady temperatures? Throw some heat and cold at it. If it can't handle those reliably, you can't gig with it.
Any recommendations for a suitable thermal conductive potting epoxy in a small volume? Some of the prices I've seen are higher than the price of the VCOs.
Just use a small amount of thermal grease. You don't need -and won't benefit from- epoxy assuming you use the method detailed above. Which will also allow you to put a small tiewrap arounf both transistors AND the tempco, holding them firmly in relative position. If you don't have an electronics store near you -and don't want to order online- ask for silicone or dielectric grease at an auto parts store.

Note that you also don't benefit from large blobs of heatsink grease or epoxy. You *could* benefit from building a little box over the expo pair-tempco and one wall and support is already in place using the trim pot. This would really make for a more ideal situation, but it's possibly only gilding the lily. Good to keep in mind though.
---------------------

Moog also used the back-bend, bow-down tranny pair layout. For the moog modular VCA. It's a sometimes useful arrangement when the more common 'kissy-face' arrangement is not practical. But never put the tempco between the tranny faces! That's just going to make things worse.
Treat utility modules as stars instead of backup singers.

Treat power supplies like Rockstars instead of roadies!

User avatar
mskala
Super Deluxe Wiggler
Posts: 2654
Joined: Thu Mar 27, 2014 7:33 am
Location: Toronto
Contact:

Re: How much drift is a good thing?

Post by mskala » Fri Sep 11, 2020 8:26 am

KSS wrote:
Fri Sep 11, 2020 2:03 am
Just use a small amount of thermal grease. You don't need -and won't benefit from- epoxy assuming you use the method detailed above. Which will also allow you to put a small tiewrap arounf both transistors AND the tempco, holding them firmly in relative position. If you don't have an electronics store near you -and don't want to order online- ask for silicone or dielectric grease at an auto parts store.
Here's one of mine. On some recent builds I've also been adding a blob of polyurethane glue on top - not to improve the thermal connection (the contact handles that) but just to prevent the wire tie from slipping off.
DSCN1086.jpg
You do not have the required permissions to view the files attached to this post.
North Coast Synthesis Ltd.
Latest Web log posting: Genome Jazz

skee
Common Wiggler
Posts: 58
Joined: Sun Sep 22, 2013 11:39 am
Location: Dublin

Re: How much drift is a good thing?

Post by skee » Mon Sep 14, 2020 6:00 pm

OK. Many thanks for all the replies. So yesterday the synth has been disassembled back to the pcb (not a trivial task on the Ultimate. 50 knobs, 40 hex nuts and a dozen or so screws and other fixtures all have to be removed to get access) The two transistors and the thermistor on VCO 1 were desoldered (tricky with all of the other components in place, one melted pot spindle will have to be replaced) after a lot of fiddling, all the parts are back, as per KSS’s scheme, with a smidge of thermal grease and a tie wrap to hold it all in place.

This evening saw switch on and after 20mins it was set to 1kHz. Drift is now down to about 2 Hz over a 5 minute period, but combined with a plasticine igloo over the transistor/tempco bundle it is less than 1 Hz over 10 minutes. One observation is that the heat imparted to the plasticine from shaping it with my hands resulted in a 15 Hz rise in frequency, which took about half an hour to re-stabilise.

So it’s on for an overnight soak and see what changes occur overnight.

Borogove
Learning to Wiggle
Posts: 44
Joined: Sun Feb 25, 2018 10:56 pm

Re: How much drift is a good thing?

Post by Borogove » Mon Sep 14, 2020 8:45 pm

You could also consider using a single matched-pair package like the AS194 or AS394.

User avatar
dot matrix madness
Wiggling with Experience
Posts: 458
Joined: Fri Aug 23, 2013 2:48 am
Location: Germany, SB

Re: How much drift is a good thing?

Post by dot matrix madness » Tue Sep 15, 2020 3:50 am

mskala wrote:
Fri Sep 11, 2020 8:26 am
Here's one of mine. On some recent builds I've also been adding a blob of polyurethane glue on top - not to improve the thermal connection (the contact handles that) but just to prevent the wire tie from slipping off.
DSCN1086.jpg
Typically I apply some thermal conductivity paste (very few because it's messy stuff and gets everywhere) and then put
a heat shrink around the transistor pair and tempco .
Licenced to solder since 1993

KSS
Super Deluxe Wiggler
Posts: 3411
Joined: Mon Jan 25, 2016 7:28 am

Re: How much drift is a good thing?

Post by KSS » Tue Sep 15, 2020 5:05 am

Does plasticine have any electrical properties which couldmatter in this use? Resistance? Hydrophilic? Containing salts? <--this is the one most concerning long term
Same for Sugru, or any other thing one might glop on to attempt thermal stabilization.

I don't know this answer. But we should try to find out.

Personally, I'd make a little plastic or styro box using the Trimpot as I wrote earlier. 3 sides and a top should do it, with the PCB as bottom and trim as 6th face of the enclosed space. Use silicone to assemble it and to hold it to pcb. Means it can removed for repair easily.
Still air inside will heat and stabilize.

Pill bottles have been used in the past, but the PCB here seems better suited to rectangle box.

@OP. Seems you're getting there!
Treat utility modules as stars instead of backup singers.

Treat power supplies like Rockstars instead of roadies!

skee
Common Wiggler
Posts: 58
Joined: Sun Sep 22, 2013 11:39 am
Location: Dublin

Re: How much drift is a good thing?

Post by skee » Thu Sep 17, 2020 5:10 pm

Well on checking the scope after the overnight run, the frequency had only shifted by a few Hz, but then, as I watched, it started drifting around again. Very disappointed.

So last night the oscillator section of the pcb got a good scrub with an old toothbrush and some surgical alcohol. Back on test but no noticeable improvement. Then I discovered that two frequencies could be monitored at the same time, albeit at lower resolution, so two VCOs were hooked up. After a couple of hours staring and trying to discern any patterns or correlated events, it seems that whatever’s the cause, it affects both VCOs most of the time. Sometimes they will drift apart, but I believe that this is from the temperature differential effect discussed previously, as only vco 1 has had the ‘bend and bow’ treatment so far.

Putting this alongside what Guest commented, it may be supply rail fluctuations. Even though my 20 quid meter shows nothing, the scope does show an oddly formed sine wave coming out of the wall plug unit. So a replacement linear transformer is on order, along with a mica cap (getting desperate here, they are quite expensive) to replace the polystyrene in the integrator. I’m also looking at the pcb to see if it’s possible to cut in and insert a regulated supply to each VCO. Is this going too far, perhaps?

Don T
Super Deluxe Wiggler
Posts: 1282
Joined: Sun Feb 02, 2014 3:11 pm

Re: How much drift is a good thing?

Post by Don T » Thu Sep 17, 2020 6:18 pm

I wouldn't replace a styrene cap, I doubt there'd be much difference, if any at all, with a silver/mica cap.

Just curious, what is your power supply like?

KSS
Super Deluxe Wiggler
Posts: 3411
Joined: Mon Jan 25, 2016 7:28 am

Re: How much drift is a good thing?

Post by KSS » Thu Sep 17, 2020 6:21 pm

Re: Is this going too far?
It may be. You've got the overall drift down to a few Hz overnite, and had a decent 10 minutes/1Hz. At this point, you probably feel like a cook in the kitchen with guests waiting for the meal?

I'd take a step back, finish the food preparation and sit down to a meal of knobbin and wigglin. Put the thing in its case or cabinet or whatever you're using for housing and give it a go as a sound source. Play it some and see how it feels. Take off yourtech hat and put on your musician hat for awhile.

After that, I trust and believe you'll know the best next step. For you. But do give yourself a break from perfecting. For now.
-------------
Do the Bend-Bow update of the other pair. The Mica won't necessarily be better than the Polystyrene, but it will handle soldering heat better. Meaning unless you overheated the PS cap, the mica's probably not going to do much for you. But it *is* easy to overheat PS caps. And before adding any moreplasticine, check the label for ingredients like salts, water. It 'may' be adding to your problems.

Post Reply

Return to “Music Tech DIY”