good price on matched transistors and tempco

elmegil · Post by **elmegil** » Mon Oct 13, 2014 1:06 am

Awesome!

InfraXpert · Post by **InfraXpert** » Mon Oct 13, 2014 1:54 am

Great, I would take 20 with thermal epoxy

guest · Post by **guest** » Mon Oct 13, 2014 2:16 am

looks good.

have you considered the pmp4201? last i checked, the pmp4201 had lower noise floor, and higher beta for the same beta and Vbe matching.

EATyourGUITAR · Post by **EATyourGUITAR** » Mon Oct 13, 2014 2:32 am

guest wrote:looks good.

have you considered the pmp4201? last i checked, the pmp4201 had lower noise floor, and higher beta for the same beta and Vbe matching.

is the vbe matching better on paper or real life? have you done any measurements of these? I know there are some discrete vintage synth replacement parts with PMP-4201. at this point I have to think that both DMMT-3904 and PMP-4201 are being used in synths but I do not have any direct comparison data from any synth builders. I can switch to PMP-4201 but I would need a good reason to switch. I do not want to support them both so a decision needs to be made one way or the other. I do not doubt that the PMP-4201 meets the datasheet specs for lower noise floor, however the real world measurements of the DMMT-3904 have already exceed the datasheet specs in some areas.

guest · Post by **guest** » Mon Oct 13, 2014 4:14 am

i did my BA662 clones with the PMP4201's. on a batch of 100, most were around 100uV to 400uV (if i recall correctly, i didnt write any of it down). another group, maybe 25% were 400uV to 1.3mV, and then maybe 10% at 1.3mV to 2mV. but again, this is all from recollection. there probably isnt too strong of a reason to go one way or the other. im going to do another run of the clones, and can write down the spread this time.

sonicwarrior · Post by **sonicwarrior** » Mon Oct 13, 2014 4:23 am

BugBrand wrote:the Linear System matched transistors are very good quality for not too much $$$ (and available in DIP or SO8 etc).

I thought the DIP version was discontiued?

LED-man · Post by **LED-man** » Mon Oct 13, 2014 6:00 am

Good idea.
i'm in for 20 without thermal epoxy.

EATyourGUITAR · Post by **EATyourGUITAR** » Mon Oct 20, 2014 1:33 pm

I will have these assembled in a few days. I have only small quantity right now. these do not have a tempco. I will have tempco pcb's next month in large quantity. if you would like one without tempco they are $2 each assembled + ship. I only have 9 of each one right now.

elmegil · Post by **elmegil** » Mon Oct 20, 2014 1:57 pm

You're missing one for transistors where the base is on one side

Looks awesome, actually, very much would like to get some.

Reality Checkpoint · Post by **Reality Checkpoint** » Mon Oct 20, 2014 2:49 pm

Excellent news.

Thank you for doing this.

Ebolatone · Post by **Ebolatone** » Mon Oct 20, 2014 11:27 pm

I've been looking at those SMD dual 3904; have a quad IC here for experimentation. This is a very good idea!

Graham Hinton · Post by **Graham Hinton** » Tue Oct 21, 2014 5:32 am

EATyourGUITAR wrote:
latigid on wrote: What is the temperature coefficient of the resistor pair?
the resistors are panasonic 3300ppm

Those resistors are +/-5% tolerance and their TC tolerance is +/-10% which is no way tight enough for two VCOs to track without a lot of selection and/or trimming.

If you use wirewound resistors from the same batch the TCs will be the same whatever the tolerance because they are made from the same reel of wire. If you buy them at different times there is no guarantee of matching. SMD components will be all over the tolerance range. Caveat emptor and good luck with the heated oil baths.

EATyourGUITAR · Post by **EATyourGUITAR** » Tue Oct 21, 2014 10:35 am

Graham Hinton wrote:
EATyourGUITAR wrote:
latigid on wrote: What is the temperature coefficient of the resistor pair?
the resistors are panasonic 3300ppm
Those resistors are +/-5% tolerance and their TC tolerance is +/-10% which is no way tight enough for two VCOs to track without a lot of selection and/or trimming.

If you use wirewound resistors from the same batch the TCs will be the same whatever the tolerance because they are made from the same reel of wire. If you buy them at different times there is no guarantee of matching. SMD components will be all over the tolerance range. Caveat emptor and good luck with the heated oil baths.

I reread your post a few times and you did admit that people can adjust the trimmer and it will be fine. this solution of using two 1K SMD tempco's was posted 2 years ago on muffs and no one had a problem with it at that time. we all have different levels of perfection that we are willing to accept with synthesizers. the transistors are matched to Vbe 2mv. not 0.2mv or 0.02mv like a lot of you old cats. this is fine for most people including legendary synth builders like oldcrow. open music labs has a reputation for the best juno voice chips. they really do sound great. you are your self legendary Mr Hinton. You belong in a museum. While I appreciate learning from your posts I do see a trend of being unyieldingly pedantic. I will wait for one of my customers to tell me it was tested in a synthesizer and it sucks before I stop selling them.

the bad producer · Post by **the bad producer** » Tue Oct 21, 2014 10:53 am

These would also be really great if you could have something to fit this footprint:

http://www.cgs.synth.net/modules/cgs108_vca.html

Not sure if that would be possible, as it IS rather tight!

Regarding Graham's comments re the tempco, would it not be possible (can you get these?) to use tempco's with smaller TC tolerance? So saying me who wasn't really aware (bothered?) that tempco's had a TC tolerance before, certainly not seen it mentioned much, or maybe I've forgotten? hhhmmmm!

Is there a relationship in performance between how close you can get to the transistors with an SMD resistor, and how close you can't-really-get with a big fat through hole one?

EATyourGUITAR · Post by **EATyourGUITAR** » Tue Oct 21, 2014 11:29 am

if we are going down that road we need to analyze thermal conductivity of the transistors and PCB substrate as well as the thermally conductive epoxy I use. can someone tell me the typical TC tolerance for TO-92, SOT-363, PDIP8, CDIP8? I think if these problems exist in SMD, they must also exist in through hole. if you are that concerned with the two 1K tempcos having matching TC to each other, you must give the TC mismatch of a through hole tempco and a TO-92 handmatch pair equal scrutiny. after all, this is how you would build them way way back in the days of synthi is that right?

EATyourGUITAR · Post by **EATyourGUITAR** » Tue Oct 21, 2014 12:22 pm

Dave Brown precedes me with testing this exact tempco part for temprature stability. you can read his results here

the bad producer · Post by **the bad producer** » Tue Oct 21, 2014 1:18 pm

Interesting, I hadn't seen that from Dave Brown (it's also reminded me that I need to sort out one of my Mankato's, I think I fried the 2164 being stupid!)

My comment was mainly to say "these are great!" but the thread got my wondering if this is very efficient for heat transfer?! It looks out of place to me!

I've not got around to trying SMD very much yet, but it's always struck me that they seem better suited - being tiny - to sit against another part and share the same 'environment'.

They would be super-useful for those CGS boards too, even though there is no need to have the tempco for that. Just the matched transistor would save time - and TBH my 'decent' DMM needs sorting out before I can match anything to any reasonable level!

oldenjon · Post by **oldenjon** » Tue Oct 21, 2014 2:26 pm

the bad producer wrote:These would also be really great if you could have something to fit this footprint:

http://www.cgs.synth.net/modules/cgs108_vca.html

+1

Graham Hinton · Post by **Graham Hinton** » Wed Oct 22, 2014 5:25 am

EATyourGUITAR wrote: I reread your post a few times and you did admit that people can adjust the trimmer and it will be fine.

I did not, read it again.
By selection I meant testing parts for the correct 3354ppm TC before mounting, which is something the manufacturers cannot do (I have asked them), and by trimming I meant a method like Ian Fritz's.

this solution of using two 1K SMD tempco's was posted 2 years ago on muffs and no one had a problem with it at that time.

The only thing that is a solution for is not being able to buy a 2k part.

we all have different levels of perfection that we are willing to accept with synthesizers.

Perhaps you would like to state what tolerance you find acceptable without descending to immature personal insults?

the bad producer wrote:Regarding Graham's comments re the tempco, would it not be possible (can you get these?) to use tempco's with smaller TC tolerance? So saying me who wasn't really aware (bothered?) that tempco's had a TC tolerance before, certainly not seen it mentioned much, or maybe I've forgotten?

Most people do not understand what the TC resistor is there for or how to test it. It is not, as seems to be believed, to make a VCO stable at one temperature or to make several VCOs stay in tune with each other, but to keep the VCOs at the same tuning when the temperature changes.

When the temperature changes the transistors will cause both the absolute pitch to shift and the V/octave spread to change. The absolute pitch shift will be greater at lower notes because it is multiplied by the Vbe. When Vbe = 0 the VCO will normally be at it's highest frequency, say ~16kHz, so it makes sense to test how much middle C shifts, six octaves down from that with a Vbe of 6 x 17.8mV.

Without temperature compensation you can expect middle C to shift 5.5 semitones with a 25 degC temperature change. With an exact 3300ppm TC resistor this can be reduced to 9 cent pitch shift and a 7 cent error over a 5 octave keyboard, but if the part is +/-10% tolerance it will go from 3000ppm and a -59 cent shift with -49 cent span error to 3600ppm and +40 cent shift with 33 cent span error. Either extreme is a unplayable out of tune keyboard.

Due to the way they are manufactured wirewound resistors from one reel of wire will closely match each other in the same batch. The manufacturer cannot pick a reel of wire that will give the required TC because it changes during winding so they have to test samples after manufacture and grade them. I think I got lucky with the first batch of Q81s I ever bought, the problem only becomes apparent when you go back for more and put two VCOs from different batches together. If you buy a synthesizer with all the VCOs built at the same time they will match each other, but may not remain absolutely in tune over temperature. That's why analogue polyphonics have a tune function and use a self calibration lookup table.

I have had some +/-1% TC resistors made, but I really wanted +/-0.5% for a "drop in and forget" part and the manufacturer couldn't do it.

With film resistors due to the manufacturing process the TC will have a spread and the manufacturers specify a wide tolerance because they know they cannot select finer. You also get the spread within a batch so there is no guarantee that two VCOs built at the same time will track unless the components are tested and matched and that is a pain.

Graham Hinton · Post by **Graham Hinton** » Wed Oct 22, 2014 5:51 am

EATyourGUITAR wrote: can someone tell me the typical TC tolerance for TO-92, SOT-363, PDIP8, CDIP8?

They don't have a temperature coefficient they have a thermal resistance which is quoted by the manufacturer as it varies per device, not per package.

if you are that concerned with the two 1K tempcos having matching TC to each other

You have the wrong end of the stick. The resistors have to have the reverse TC of the transistor they are compensating.

after all, this is how you would build them way way back in the days of synthi is that right?

The VCS3 was originally designed to use dual can transistors. EMS left them out because they were cheapskates and because most of the customers didn't notice, even though it was awful. Those that did put them back in.

Two discrete transistor packages do not achieve temperature balancing, especially with Araldite between them, they need to be on the same substrate to do that. Balancing and compensation are different, both are needed.

valgamaa · Post by **valgamaa** » Wed Oct 22, 2014 7:42 am

Graham Hinton wrote: When the temperature changes the transistors will cause both the absolute pitch to shift and the V/octave spread to change. The absolute pitch shift will be greater at lower notes because it is multiplied by the Vbe. When Vbe = 0 the VCO will normally be at it's highest frequency, say ~16kHz, so it makes sense to test how much middle C shifts, six octaves down from that with a Vbe of 6 x 17.8mV.

Without temperature compensation you can expect middle C to shift 5.5 semitones with a 25 degC temperature change. With an exact 3300ppm TC resistor this can be reduced to 9 cent pitch shift and a 7 cent error over a 5 octave keyboard, but if the part is +/-10% tolerance it will go from 3000ppm and a -59 cent shift with -49 cent span error to 3600ppm and +40 cent shift with 33 cent span error. Either extreme is a unplayable out of tune keyboard.

I agree with all of your maths and the significance of temperature compensation, but your illustrative example is perhaps a little extreme.
Given that it is the delta-Vbe that has the temperature coefficient, and this delta can be either negative or positive, it would be prudent to put the zero-Delta-Vbe point in the centre of the keyboard not at one end. In doing so the centre-frequency will have no temperature dependence due to the expo-pair, and the keyboard extremes are only 2.5 octaves away from this zero-point, rather than 5.
It is also reasonable to assume that a 25 Celcius change form the calibration temperature is larger than most people are likely to experience, and when they do they might even consider retuning their VCO's to remove the absolute error before use.

With a sensibly designed VCO and controlled operating environment (this is a DIY forum) the SMD tempo option is quite reasonable. If I had to choose between saving cost on the tempco or expo transistors, I would use the SMD tempco and a top-quality transistor pair as temperature is only one parameter in the performance of a VCO.

EATyourGUITAR · Post by **EATyourGUITAR** » Wed Oct 22, 2014 8:11 pm

Graham PM me your address I will send you some samples. I would love for you to post your results here. I am confident that they will score somewhere in the middle of the acceptable range for most people. maybe we can test them in some synths in addition to the normal tests that engineers do.

slow_riot · Post by **slow_riot** » Wed Oct 22, 2014 8:27 pm

I've been trying to use a commercial VCO to calibrate a reel to reel, and also an RMS voltmeter that needed repair. I don't think HP or Studer engineers would consider the deviation acceptable.

The musician Blawan mentioned in an interview that he is not using analogue VCOs in his eurorack setup as the tuning was not acceptable. And if you find his industrial assualt in any ways palatable you would also realise that he is not a purist, there is obviously a fundamental problem with his system in the environment he works in.

A personal friend of mine set up his modular system, soundchecked, and then played several hours later, only to find that after the previously empty cold room had filled up with warm bodies his system was out of tune and couldn't get it sorted in time and his performance was a write off.

lysander · Post by **lysander** » Thu Oct 23, 2014 3:41 am

slow_riot wrote: The musician Blawan mentioned in an interview that he is not using analogue VCOs in his eurorack setup as the tuning was not acceptable

That's a rather dubious and overgeneralized statement. There's a variety of VCOs in euroland that have extremely stable tuning, mostly the 2164 based designs.

Graham Hinton · Post by **Graham Hinton** » Thu Oct 23, 2014 7:35 am

valgamaa wrote: I agree with all of your maths and the significance of temperature compensation, but your illustrative example is perhaps a little extreme.

I deliberately chose that example as a reasonable expectation. A 25 deg C temperature change is quite normal in live gigging and at least 15 deg C would easily be encountered in a studio. It takes at least an hour from switching on cold to reach a stable temperature inside a synthesizer and we are talking about inside the transistor, not the ambient temperature.

If I had wanted to overstate the case I would have used the lowest note on an 88 note keyboard as an example.

Given that it is the delta-Vbe that has the temperature coefficient, and this delta can be either negative or positive, it would be prudent to put the zero-Delta-Vbe point in the centre of the keyboard not at one end. In doing so the centre-frequency will have no temperature dependence due to the expo-pair, and the keyboard extremes are only 2.5 octaves away from this zero-point, rather than 5.

I was only considering a 5 octave keyboard, whereas VCOs usually have a good 10 octave range and several may be used for a chord on the same CV. Early analogue synthesizer keyboards had a switch between calibrated V/oct and variable because when the VCOs went out you had to adjust them while playing. (I think there is a video of Joe Zawinul doing that live on a 2600.)

Most VCO designs have the Vbe = 0 point at the highest current/frequency and "attenuate" down. You can offset it, but shifting it six octaves down would make the top end less accurate.

25 cents may be less out of tune than 50 cents, but it is still out of tune.

EATyourGUITAR wrote:Graham PM me your address I will send you some samples. I would love for you to post your results here.

You would not be able to send me enough samples to make a statistically valid test and anyway you should be doing it at your expense. Besides the manufacturer has already done it, that's what +/-10% on their data sheet tells you. I'm sure that you will find some samples that conform close to the ideal, but the process of finding them is very time consuming and what do you do with all the rejects?

The manufacturers don't want to do it and you would expect them to be set up to be able to do so. The way wirewound TC resistors are tested is in a thermostatically controlled oil bath, but they only test a few samples in a batch confident that the others are close. I don't know how they would test film SMD parts because you would need to test every one and make contact with it to measure the resistance.

I would be quite happy to pay more for tested parts just like I am for 0.1% resistors so that I don't have to do it. Once it gets built in it becomes harder to test and reject. If you are making a few VCOs for yourself you can choose to what level you take this, but if you sell somebody a VCO and they may come back for another one years later you need to know how well the two will or won't track.

When Tel Labs were manufacturing Q81s they knew that transistor temperature compensation was an application and aimed for that. These devices on the market now are for something else and I don't know what, but the manufacturers are not interested in our particular application.

MOD WIGGLER

MOD WIGGLER

good price on matched transistors and tempco