VCO temperature stability testing chamber

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

Moderators: Joe., lisa, luketeaford, Kent

User avatar
devinw1
Super Deluxe Wiggler
Posts: 1376
Joined: Tue Aug 07, 2018 11:20 am
Location: Portland, OR
Contact:

Re: VCO temperature stability testing chamber

Post by devinw1 » Tue Mar 31, 2020 10:09 am

KSS wrote:
Tue Mar 31, 2020 12:38 am
The question is whether it matters as much as all that.
Vintage analog synths have been using the simple method with 3% tempco parts -which are often much worse in 3350 PTC adherance than the 3% implies.
And yet we have generations of 'wonderful' music from them.

You can find on this site and others a few poste like yours seeking ways to improve the accuracy over temperature variations. It's not a bad goal. But it might be a how many angels can dance on the head of a pin kind of thing.

Then of course there is the discussion from some of the "happy accidents" that happen from precisely the technology not being perfected due to limitations of the time/conditions, whatever. Sometimes these things make sounds that end up being iconic.

On the other hand, it's natural as an engineer to try and dial something (a machine, a circuit, etc) in as best as possible to "perfect" (perfect in quotes because there's ALWAYS that next thing you can do better :D). It's just how it is. There is often a lot to be learned about other machines/circuits just by doing this.

So, I don't have a strong opinion on this, just saying I get what you are saying. I, being and engineer, spend a lot of time and enjoy dialing stuff in and making it really sweet, but on the other hand I really enjoy my old, not so perfect machines from the 70s and can make plenty of great music with them. :tu:

KSS
Super Deluxe Wiggler
Posts: 2116
Joined: Mon Jan 25, 2016 7:28 am

Re: VCO temperature stability testing chamber

Post by KSS » Tue Mar 31, 2020 11:46 am

I think you both make good points.
Like I said, it's a good goal.

User avatar
Nantonos
Super Deluxe Wiggler
Posts: 1275
Joined: Sun Mar 03, 2013 1:00 pm
Location: Boston, MA

Re: VCO temperature stability testing chamber

Post by Nantonos » Tue Mar 31, 2020 11:15 pm

Regarding whether it matters, that depends on the type of music and the musician and circumstances. Someone who makes music that sounds like cats being tortured while lightning strikes a radio may have different goals to someone making subtle, melodic music with nine-octave Ondes Martenot-style glissandi or playing together with classical instruments in an ensemble. Or someone playing a gig where the equipment arrived late, 15 minutes before the concert starts having sat for 12 hours in an unheated, well sub-zero cargo hold.

I recall reading that the Synthi-100 at the Radiophonic workshop was so sensitive to ambient temperature that a day's work could be ruined by someone opening a door and letting in a draft. Even after the room temperature has re-stabilized, it didn't sound the same as it had before.

For temperature, there are several temperatures which need to be logged. The ambient temperature around the oscillator, of course; the temperature inside the case, as that will be different and many cases have poor or no ventilation. The actual surface temperature of the circuit, especially the expo converter, which heats up more on higher pitches because there is more current. And then the ambient temperature around the meter; good meters will state on their calibration certificate what temperature the calibration was performed at, like 23C. The temperature inside the meter will be higher. My meter (Keysight 34465A) shows the current temperature inside the meter, and also the time and the difference between that and the time and temperature of the last auto-calibration.

> An engineer/designer might decide that it is fine if, for the VCO they are designing, the exact initial tuning of every note depends on the notes just before it

I have wondered about running two automated sweeps of the frequency range on an oscillator. On one, the voltage and thus the frequency is held low for a couple of minutes, rises to the voltage to be tested, is measured, and goes back to the low voltage again for a minute. The second sweep the same, but resting at a high voltage between measurements. It would be fascinating to see whether the expo converter self-heating produced a measurable difference in the two sweeps.
ondes | current rack
I am afraid a firmware change will not be able to turn a rather expensive 16-bit DAC into a 16-bit ADC, and flip all those op-amps :)

ricko
Common Wiggler
Posts: 230
Joined: Sat Dec 03, 2016 5:36 am
Location: Sydney, Australia

Re: VCO temperature stability testing chamber

Post by ricko » Wed Apr 01, 2020 2:52 am

> It would be fascinating to see whether the expo converter self-heating produced a measurable difference in the two sweeps.

The sweeps would have to be long enough at the top to generate heat and going down fast enough to still gave heat at the bottom. That is why I think alternative notes are better than sweeps (and short chirps would be useless.)

The problem that engineers have been solving is long term stability (i.e at equilibrium), but not short term stability (ie worst case of disequilabrium), AFAICS. It doesnt get measured (in fact it is often systematically excluded), when it is measured it is measured tangentially, and when it is measured right there is no real way to compare designs and units; that is my impression after looking for the last few years.

Why did Wendy Carlos move on from analog: in a large part it was driven by a desire to explore microtunings, is my impression. Her Moog had a tuning box to allow each keyboard note to be microtuned, but still it was not good good enough.

The long-term stability is certainly an issue, don't get me wrong. And I understand the argument that it is better to fix problems people are aware of, rather ones that supposedly no-one is aware of. But if you make VCO with excellent long term drift, have you measured it for initial stability? (Is there slight motorboating from the heater circuit, for example?) And if people are not aware that short-term phrase-dependent stabilization is a different beast to long-term drift, and have been assured correctly told that their VCO is low drift, how could they report it?

KSS
Super Deluxe Wiggler
Posts: 2116
Joined: Mon Jan 25, 2016 7:28 am

Re: VCO temperature stability testing chamber

Post by KSS » Wed Apr 01, 2020 3:37 am

Because they hear it. In 60 years some people with perfect pitch -and well trained ears- have used synths.

The things you're talking about are not new and have been considered and discussed before. Analog Heaven mailing list would bea likely resource. Which is not to say you shouldn't continue to seek out whatever level of perfection suits you or that you may achieve. at the end of the quest you may find that the result is not as satisfying as it seems now. Microtonal tunings may benefit more than standard tuning.

User avatar
guest
Super Deluxe Wiggler
Posts: 4186
Joined: Mon Aug 19, 2013 11:49 am

Re: VCO temperature stability testing chamber

Post by guest » Wed Apr 01, 2020 9:20 am

ive measured both initial stability and short term thermal drift. the main thing that makes both of these better, is to reduce the thermal mass of the transistor pair and temperature sensor. use the smallest SMT parts you can find. next, put your temperature sensor in direct, thermal contact with the SMT. finally, insulate the both the temperature sensor and SMT pair. a matched quad in SMT is sort of the ideal version of this, as the sensor is the other transistor. but it still benefits from insulation.

thermal ovens are great for long term drift, and horrible for everything else. the limiting factor for thermistor circuits is the thermal mass, and the limiting factor for Vt multipliers is the frequency response of the feedback loop, which can be quite fast. ive found Vt multipliers to be the best of all the options.

finally, short term drift can be almost elminated with non-compensation techniques: basically dont draw as much current, and Vce low. the mechanism for short term drifts is the self-heating of the thermistor or exponentiating transistor. its not too difficult to keep the thermistor current under the level that would result in noticeable self-heating, and if you keep Vce to 0.6V (which most circuits do, but you could also push this down to 200mV) then 1uA of current is only 600nW of power, which is pretty darn small. so keeping Ic low not only improves HF linearity, it also can elminate the thermal drift of playing high notes. there are some practical limitations to how low you can go.
openmusiclabs.com

ricko
Common Wiggler
Posts: 230
Joined: Sat Dec 03, 2016 5:36 am
Location: Sydney, Australia

Re: VCO temperature stability testing chamber

Post by ricko » Wed Apr 01, 2020 10:04 am

Guest: interesting. Do you recall, roughly how strong was the frequency deviation, at worst, please, and how many seconds until near-equilibrium?

User avatar
guest
Super Deluxe Wiggler
Posts: 4186
Joined: Mon Aug 19, 2013 11:49 am

Re: VCO temperature stability testing chamber

Post by guest » Wed Apr 01, 2020 10:26 am

i dont have access to any of my notes right now, but i recall it varying greatly depending upon transistor package. for the metal can types, it was absolutely horrible (large drift, long stability times), for the tiny SMT stuff, it was maybe 200ms for drift. the drift percentage can be huge depending upon the coupling with the thermistor, like a few percent.

insulation the package has its pros and cons. by insulating the package, you help ensure that the thermistor is the same temperature as the transistor, but you also increase the settling time. if the thermistor is doing its job, then settling time no longer matters, so i usually err on the side of more insulation.

you can get a rough estimate of settling time by doing a "thermal circuit" calculation, treating the mass of the package as the C, and its thermal conductivity to the ambient air as R. this gives an RC time constant for temperature change. the thermal conductivities are usually given in the datasheets. you can estimate the thermal mass of the package by taking its volume, and multiplying that by its specific heat and density.
openmusiclabs.com

ricko
Common Wiggler
Posts: 230
Joined: Sat Dec 03, 2016 5:36 am
Location: Sydney, Australia

Re: VCO temperature stability testing chamber

Post by ricko » Wed Apr 01, 2020 10:41 am

@Guest: Thanks. A few percent not a few cents! That is more than I would have expected, actually, though you are saying that is worst case.

I guess someone could make a compensation circuit for this kind of drift too, if it was a problem in their design.

(I read that plastic DIP chips it takes 5 seconds for heat to make it to the surface.)

Do you know, has anyone measure the 3340 VCO chips in this regard?

Cheers
Rivk

User avatar
guest
Super Deluxe Wiggler
Posts: 4186
Joined: Mon Aug 19, 2013 11:49 am

Re: VCO temperature stability testing chamber

Post by guest » Wed Apr 01, 2020 11:09 am

some of the VCO topologies use a buffer after a capacitor, rather than a charge integrator. this can give an average Vce of 5V, and if they drive the Ic up to 1mA (common on older synths, not done as much now), then youve got 5mW of heat production. a TO-92 package is around 200C/W and an SMT 400C/W for the junction to ambient air thermal resistance, which gives 1-2C rise, or 1-2% drift (depending upon how long it takes the thermistor to catch up).

ive tested the 3340, but not specifically for transients. it was a while ago, and im trying to remember how it faired (you notice the transients while testing for long term drift). i seem to recall it being pretty good at initial warmup, but not so good at HF induced drifts. the best thing i tested was a Vt multiplier made from an AD633 and a MAT14. that thing was just rock solid no matter what i did to it. there is low frequency thermal noise as well, and it had way less (meaning less note jitter).

i have some of this written up here:
http://www.openmusiclabs.com/files/expotemp.pdf
openmusiclabs.com

User avatar
devinw1
Super Deluxe Wiggler
Posts: 1376
Joined: Tue Aug 07, 2018 11:20 am
Location: Portland, OR
Contact:

Re: VCO temperature stability testing chamber

Post by devinw1 » Wed Apr 01, 2020 11:13 am

I would highly recommend anyone interested in this topic read the link to guest's paper in the post above, if you haven't. It's extremely in depth and useful.

ricko
Common Wiggler
Posts: 230
Joined: Sat Dec 03, 2016 5:36 am
Location: Sydney, Australia

Re: VCO temperature stability testing chamber

Post by ricko » Wed Apr 01, 2020 6:41 pm

Yes, it is a very good paper, and I am sure everyone interested in this subject is well aware of it, and appreciates Guest's actual measurements and comparisons in his analysis. Heroic!

P.S. Minor thing. My editor's pedantry kicks in: I wonder if it might improve the article if the use of "drift" was tightened up. The NIST defines (in their A to Z publication) "frequency drift" specifically as "An undesired progressive change in frequency with time." So the error found at different static temperatures is not really "drift", just the equilibrium point or boundary or average or cutpoint of any drift (even though the latter gives part of the formula for the former, of course.) I think the same applies to resistance drift, or current drift, etc: no time axis, no drift, perhaps. I found it a little confusing, last year, because I thought 'woohoo, finally a paper about drift', but my confusion does not invalidate any part of the paper, of course!

User avatar
guest
Super Deluxe Wiggler
Posts: 4186
Joined: Mon Aug 19, 2013 11:49 am

Re: VCO temperature stability testing chamber

Post by guest » Wed Apr 01, 2020 6:57 pm

perhaps if i revisit that paper at some point, i can change the wording a bit. unfortunately, i wrote that on a computer that was 3 computers a go, and im not sure where the source files are. im horrible at keeping track of things.

if youre intersted in doing some tests of your own, the way i looked at drift was to pulse the CV input for a period of time longer than the circuits electrical time constant, but much, much shorter than the thermal time constant. i used a scope for this, which usually doesnt have the resolution for these sorts of measurements, but i would figure out where the value was going to land, then zoom down to 20mV/div or so, and add whatever offset was required. i would then calibrate with a few known voltages referenced to my 6 digit multimeter, and then perform the tests. this gives the baseline, then you can do a regular step response, and wait for the temperature to settle out, putting the scope on 1s/div, or whatever is relevant to that transistor.
openmusiclabs.com

Post Reply

Return to “Music Tech DIY”