Solving the UA726 problem . Let's get it done for good

[auxren]

I made a simple breakout board for the SSM2212. Works great: https://github.com/auxren/UA726_Replacement

[Haralds:Werk]

For my Next Generation Formant project i needed a replacement for the obsolete uA726, because i wanted to stay as close as possible to the original Elektor Formant VCO with my NGF VCO Core 2. This means a heated exponential matched transistor pair. There are a variety of possibilities to achieve this goal. Most of those solutions uses now obsolete parts or are hard to build. But there is still on well known solution with with the 3046 transistor array. And best, it is still available. At least the SMD Version is still in production. The DIL version can be found as well, but is no longer in production.

The circuitry I found on the net did only use one transistor for the heating. This leaves one transistor unused. My thought is using two transistors makes for a faster heating up and more stability against environmental change. To my surprise I can not find a solution which uses two transistors instead of one. Here is my take on the circuitry.



This circuitry is part of my NGF VCO Core 2. It make use of two transistors to heat the 3046. It works with the SMD and DIL Version as well. The given resistor values keep the current values and the power dissipation below the maximum ratings. T3 is used to measure the actual temperature. The voltage drop over T3 is direct proportional to the chip temperature. It is compared to the voltage at pin 2 of the 3046 which is derived from TR6. The temperature is easily adjusted with TR6. Between heating off and maximum temperature. The testing bridge is driven with a stabilized 10V voltage source. T4 and T5 are used as heaters. R41 and R43 limit the maximum current. The different values are selected with purpose to keep pin 13 of the 3046 (the substrate) the lowest negative point at the chip.

The below picture shows the graph temperature in deg. Celsius vs. Voltage in mV at pin2. It is quite linear. You can easily derive the needed voltage for your preferred temperature from the graph. The data were taken from a LM3046 SMD mounted on a PCB. The figures for the DIL Version are slightly different. They will show up on my website as well as some more details ASAP. Temperature was measured with Fluke 63 IR thermometer.

[guest]

good idea to use both transistors. it would be nice to have an image of the die inside the chip, so the transistors could be picked to be on opposite sides of the matched pair. if they are both on the same side, you get a thermal gradient across the matched pair which causes an offset (which you get in the single transistor case anyways).

[Haralds:Werk]

good idea to use both transistors. it would be nice to have an image of the die inside the chip, so the transistors could be picked to be on opposite sides of the matched pair. if they are both on the same side, you get a thermal gradient across the matched pair which causes an offset (which you get in the single transistor case anyways).

The thermal conductivity of Si is quite high. About 150W/(m*K). The die of the chip is very small so i don't think the thermal gradient will cause any problem here. I vaguely remeber a picture of the die and a discussion ot this topic, but i can't nail it down.

[guest]

i remember reading something about linear tech (?) interdigitating the collectors of their diffpair input stages to minimize thermal gradients. so i assumed it had some small effect. a 0.1C difference in transistor temperatures is a 2cent shift in frequency up at 5V CV.

[Haralds:Werk]

If there is a temperature difference within the die it will stay constant. This will not cause any problems for pitch shifting.

[cygmu]

But if the two transistors in the expo pair are at different temperatures you'll still have a dependence on their I_S won't you? (I only have a loose grasp of this).

Would that be irrelevant in practice once the oscillator is trimmed for V/Oct response?

[Haralds:Werk]

Would that be irrelevant in practice once the oscillator is trimmed for V/Oct response?

Matching transitors and heating is for temperature compensation. It has nothing to do with tracking. Tracking is determined by the exponential conformance of the logging transistor and by the speed of the core reset.

[cygmu]

Matching transitors and heating is for temperature compensation.

Indeed, but I thought that the point of matching was to eliminate the dependency on I_S, which is itself temperature dependent.

Using matched transistors that are at different temperatures will not eliminate this dependency.

I suppose my question is whether this matters at all, if you have accurate enough temperature control. Presumably the use of matched transistors at very close temperatures (assuming the temperature difference we're talking about here is relatively small) helps to reduce, if not eliminate, the temperature dependence, such that the heating can do a good enough job of keeping things steady.

[guest]

its really a small amount, and not worth worrying about, but still there. there will be I_s differences, and the gradient increases as external temperature decreases (q = k(T1-T2)), so it wont be constant. but again, a few cents at most, and putting the chip in something insulating makes it mostly go away. id still like to have a look at the die, though. maybe ill decap one at some point.

[jhulk]

heres my version thats in prototype stage being tested

[jhulk]

ok between don t myself and others looks like this is a success

we are going to create build docks next and state the perfect resistor to replace the trimmer.

on these board the foot prints for the through hole parts are also the correct size for 1206 smd types.

these will go into production next week should have them back in a few weeks

there will be built and diy versions available

if you want them prebuilt dont t will be building them and synth cube will be usa stockist and uk built versions will be by amsynths aswell as uk and eu stockist for the diy version

if there is any one in australia who wants to stock let me know.

[aladan]

ok between don t myself and others looks like this is a success

we are going to create build docks next and state the perfect resistor to replace the trimmer.

on these board the foot prints for the through hole parts are also the correct size for 1206 smd types.

these will go into production next week should have them back in a few weeks

there will be built and diy versions available

if you want them prebuilt dont t will be building them and synth cube will be usa stockist and uk built versions will be by amsynths aswell as uk and eu stockist for the diy version

if there is any one in australia who wants to stock let me know.

As Australia's biggest uA726 lover*, I am happy to step up to the plate here

* happy to argue the point if anyone disagrees on this

[cygmu]

we are going to create build docks next and state the perfect resistor to replace the trimmer.

This is very good!

Isn't the trimmer supposed to set the temperature? If so, doesn't the perfect temperature depend on the environment in which the unit is being used -- you need to raise the transistors above the ambient temperature to have any control don't you?

[Don T]

we are going to create build docks next and state the perfect resistor to replace the trimmer.

This is very good!

Isn't the trimmer supposed to set the temperature? If so, doesn't the perfect temperature depend on the environment in which the unit is being used -- you need to raise the transistors above the ambient temperature to have any control don't you?

Every implementation of the uA726 I've ever seen had a fixed resistor setting temperature. I mean, that was the entire purpose of the thing: Turn it on and not have to worry about temperature adjustments!

The idea here is to find a resistor value that sets the chip temperature somewhere around 100ºF/40ºC. That's far below the chips max temperature rating, and not many of us are going to put up with a studio or stage that hot! Besides, even in a performance/recording venue that is that hot (been there, done that), the pitch control and/or adjustment of any other instruments has far been exceeded, especially acoustic instruments.

Here are a couple photos of the prototype I built:



No, neither the dime or the circuit board is photoshopped! If anyone ever needs proof that I am totally insane, here ya go: That is not only hand-built, but the circuit board is also hand-made. Yep, etched, drilled, vias filled in, etc., all by hand. I think it took an entire hour afterwards for my eyes to un-cross...




Prototype in place on Card 7 of a Buchla 208. I'm running my 208 as I type this, with a Card 7 that has a uA726 installed, and I'm about to take measurements of the 726 with a Ryobi infrared thermometer. I'll install the Card 7 above later and try to set it to a similar temperature.

I have another one of these prototypes to build, and I'll try 1/8 Watt resistors to save space.

Hopefully I'll have more results soon!

[synthdude]

No, neither the dime or the circuit board is photoshopped! If anyone ever needs proof that I am totally insane, here ya go: That is not only hand-built, but the circuit board is also hand-made. Yep, etched, drilled, vias filled in, etc., all by hand. I think it took an entire hour afterwards for my eyes to un-cross...

Bravo very nice!

[aladan]

Isn't the trimmer supposed to set the temperature? If so, doesn't the perfect temperature depend on the environment in which the unit is being used -- you need to raise the transistors above the ambient temperature to have any control don't you?

Every implementation of the uA726 I've ever seen had a fixed resistor setting temperature. I mean, that was the entire purpose of the thing: Turn it on and not have to worry about temperature adjustments!

The idea here is to find a resistor value that sets the chip temperature somewhere around 100ºF/40ºC. That's far below the chips max temperature rating, and not many of us are going to put up with a studio or stage that hot! Besides, even in a performance/recording venue that is that hot (been there, done that), the pitch control and/or adjustment of any other instruments has far been exceeded, especially acoustic instruments.

While what you say is true, the temperature of the inside of electronic devices can and often does exceed the ambient outside temperatures (for an extreme example, consider the temperature of PC CPUs!) I'm not saying 40 degC won't be just fine (I doubt many VCOs inherently get that hot, but badly ventilated synths with linear power supplies sweating under stage lighting just might), but I thought I should make the observation in case this needs to be thought about further.

[Don T]

Every implementation of the uA726 I've ever seen had a fixed resistor setting temperature. I mean, that was the entire purpose of the thing: Turn it on and not have to worry about temperature adjustments!

The idea here is to find a resistor value that sets the chip temperature somewhere around 100ºF/40ºC. That's far below the chips max temperature rating, and not many of us are going to put up with a studio or stage that hot! Besides, even in a performance/recording venue that is that hot (been there, done that), the pitch control and/or adjustment of any other instruments has far been exceeded, especially acoustic instruments.

While what you say is true, the temperature of the inside of electronic devices can and often does exceed the ambient outside temperatures (for an extreme example, consider the temperature of PC CPUs!) I'm not saying 40 degC won't be just fine (I doubt many VCOs inherently get that hot, but badly ventilated synths with linear power supplies sweating under stage lighting just might), but I thought I should make the observation in case this needs to be thought about further.

If it winds up needing to be tweaked, that's fine with me. One thing I was also trying to keep in mind is minimizing the amount of expansion/contraction due to heating/cooling cycles, especially given this is a lot of surface-mount stuff.

Some preliminary temperature drift measurements. Buchla 208r V2, all modulation shut down, pulser, EG, sequencer, and random shut down, main oscillator at fixed initial pitch (whatever voltage step 1 on the sequencer happened to be, didn't measure). Pitch measured at speaker with Tonal Energy Tuner (iPhone app). Ryobi infrared thermometer. 208r not mounted in boat

Left to right: Time, Temperature in degrees F measured at panel, Frequency-

9:42PM - 65.9 - 353.5Hz
9:46PM - 68.6 - 353.4Hz (furnace had come on)
10:45PM - 67.0 - 354.2Hz
10:50PM - 67.2 - 353.6Hz
10:57PM - 66.2 - 353.2Hz

[Ebolatone]

I would be very interested to know what a uA726 does in that same long test, Don T, for direct reference...outstanding stuff. Glad your eyes finally uncrossed

[Don T]

I would be very interested to know what a uA726 does in that same long test, Don T, for direct reference...outstanding stuff. Glad your eyes finally uncrossed

Oh, I'm planning on it. Going to see how the tempco version stands up also, even though it probably won't do as well as either heated version.

[momo]

Very nice work Don T!

[Don T]

I would be very interested to know what a uA726 does in that same long test, Don T, for direct reference...outstanding stuff. Glad your eyes finally uncrossed

Here we go, the results for the uA726:

Left to right: Time, Temperature in degrees F measured at panel, Frequency-

12:31AM, 68.7, 313.9Hz
12:51AM, 67.4, 313.5Hz
1:06AM, 68.6, 313.2Hz
1:30AM, 68.4, 313.2Hz
2:28Am, 67.2, 313.3Hz

Notes: Temperature outside today was noticeably warmer, about 10 degrees F, so there was slightly less variation of temperature inside the house. The uA726 runs about 10 degrees hotter than the heated THAT 300 prototype, so plans are to tweak the prototype to run a little warmer and test again.

[webboy]

Just wondering if there was any news on this? Thanks!

[Don T]

Just wondering if there was any news on this? Thanks!

There will be in a week or two. A second prototype is assembled and ready to test, but I'm out of town and away from the workbench for a few days.

[Ebolatone]

Interesting; the uA726 is a touch looser...have a good time out of town!