MUFF WIGGLER Forum Index
 FAQ & Terms of UseFAQ & Terms Of Use   Wiggler RadioMW Radio   Muff Wiggler TwitterTwitter   Support the site @ PatreonPatreon 
 SearchSearch   RegisterSign up   Log inLog in 
WIGGLING 'LITE' IN GUEST MODE

My experience with VCO transistor matching and thoughts
MUFF WIGGLER Forum Index -> Music Tech DIY Goto page 1, 2  Next [all]
Author My experience with VCO transistor matching and thoughts
devinw1
During the sorting out of a flaky VCO2 on a MK1 Arp Odyssey, I replaced the expo converter matched pair, which actually did not turn out to be the problem (the problem, TLDR, from my other thread was flaky top-side solder connection to a pin on Z1 via a DIP socket). Anyway, at the time I did not have any equipment to match transistors, so I glued together a TZ81 and a 2N6076 and put them on the board. Well, it worked, but after firing up and calibrating I noticed that if I clicked the transpose switch down, the lowest C was hitting a pretty flat B... So tracking not very well! Over the first 3 octaves with transpose in the middle, tracking was *acceptable*, but not great. So, it would seem I needed to replace the expo converter.

I was going to build the Bob Moog circuit, but I just went ahead and bought this amazing little thing for $75:

https://www.amazon.com/Peak-Electronic-Design-Ltd-dca55/dp/B005NIR8G8

It reads out Vbe to .001 V and is quite repeatable. If you let the tranny settle without touching it, you can get the same reading to 1mV over and over. Sweet!

The first thing I did is remove my replacement transistor pair and measure it. The ebay 2N6076 was .739V and the NTE (blister pack) TZ81 was .776V!.... So 37mV difference! If you have read the Bob Moog matching procedure, he recommends matching to 2mV! Unfortunately I couldn't just replace the original transistor pair, because I had clipped it too short upon removal, but I did measure it, and it was actually 15mV difference...no where near the Moog-spec'd 2mV, but this Osc actually tracked pretty well. And I have no reason to think this wasn't an original Arp pair since it had their telltale white and red paint on it and looked quite old.

Next step was a new replacement... I read on another thread that I can use a 2N3904/2N3906 pair and have a bunch of these and was able to match a set to 1mV via the Atlas device! But, upon installation...nothing. I compaired the pinouts and realized oh s**t you have to swizzle 2 of the pins on these to match the board. I did so, made sure there were no shorts, re-tested the pair and installed it....and nothing. No VCO2. Wtf?

Having given up on the 2N3904/2N3906, I went back to my bag of 15 or so 2N6076 and a bag of 20 2N5172 and the closest I could match was to 7mV and I figured this will probably be good enough since the original pair was only matched to 15mV. I installed this pair, and boom, we have sound, and after calibration, VCO2 tracks very well. Transpose switch down and the lowest C is....as far as my tuner can tell, a C! Yay.

So, my final thoughts from all this:
-The Atlas DCA55 is a killer device for matching transformers and I would recommend it. It also will identify and test a host of other devices (trannys, LEDs, diods, etc..) and I'd recommend it for any hobbyist or tinkerer.
-Yes, you really do need to match your transistors or you may get a oscillator that doesn't track well
-Most *modern* transistors I've purchased all seem to be very tight spec-wise lot-to-lot. meaning, usually Vbe will only vary by 2-3mV over a bag of 20-30 of them. It's when you grab-bag out of 2 different lots that you could get some big Vbe differences.
-Bob Moog recommended matching the expo converter to 2mV, which is probably good practice, but I'm guessing Arp's spec's weren't that tight. From what I've seen, on the Odyssey anyway, to within 5mV would be more than sufficient.
ZatsuOn
Nice! Thanks for the info and experience... Rockin' Banana!
ashleym
Peak do lovely bits of kit. Have a look across their range as it is full of useful devices. A small operation, very nice "engineering" and very approachable.

Dont forget there are plenty of posts here about matching transistors, you can make up your own jig easily and cheaply. Different to the Peak tester.
devinw1
ashleym wrote:
Peak do lovely bits of kit. Have a look across their range as it is full of useful devices. A small operation, very nice "engineering" and very approachable.

Dont forget there are plenty of posts here about matching transistors, you can make up your own jig easily and cheaply. Different to the Peak tester.


For sure, I actually was about to build something, but I don't even have a bench top power supply right now so getting the +/-15V was something to think about. The Peak device was ready to rock at $75 so I figured it was worth it. I can't believe they sell such a nice device at that price and it's make in the UK! Not even what I expected (made in China).
l3v3l6
Just curious, but could one of these be used as a replacement?
devinw1
l3v3l6 wrote:
Just curious, but could one of these be used as a replacement?


Well...that is a matched NPN pair, which is cool, but in this case we need a PNP/NPN pair which are matched to their Vbe's.
l3v3l6
Ah, of course. I'm rather new to electronics and learning as I go along.
devinw1
l3v3l6 wrote:
Ah, of course. I'm rather new to electronics and learning as I go along.


Not a problem... I am sort of in the same boat. While I have an engineering degree, it is in mechanical, so my electronics knowledge was quite limited prior to starting to f**k with old Arps... Ha! I'm learning all the time now!
aabbcc
Not to be a party poopers but According to the specs of the atlas DCA 55 it says:

VBE accuracy: +/-2% +/-4mv
https://www.peakelec.co.uk/acatalog/dca55-peak-atlas-dca-semiconductor -component-analyser.html

Mr moog would turn in his grave Dead Banana
kassu
For matching pairs you don't need absolute accuracy, you need only repeatability (precision), which devinw1 reported to be about 1mV.
fuzzbass
I can see your mileage did vary, but the truly critical matched pair in the expo converter are inside the CA3046. The mirror performance of the NPN/PNP pair is important, but as you guessed, the spec is more forgiving here, and Bob Moog's guidance regarding 2mv was in regard to HIS expo converter, not ARP's.

Most TTSH builders (working on the same circuit) took no special steps to match this pair. I built 21 TTSH VCOs (same circuit) and all of them track at least five octaves well, and another 2 or 3 good enough for 1971. The transistors I used were 2N3904 and 2N3906, just pulled randomly from the bag.

I do have a Peak DCA, and I love it, but I think it is meh for transistor matching.

With the Peak DCA, good and bad match are within the spec'ed margin of error. I'm not saying you must buy a $400/4.5 digit DMM, only that the DCA method is probably about as effective (or haphazard) as pulling adjacent transistors from tape.

Also, the number 1 design problem with these VCOs is that the 1V/Oct and Init Frequency trimmers are 1-turn types with a wide range, and it takes a very steady hand to get them right.
loki
fuzzbass wrote:
I can see your mileage did vary, but the truly critical matched pair in the expo converter are inside the CA3046. The mirror performance of the NPN/PNP pair is important, but as you guessed, the spec is more forgiving here, and Bob Moog's guidance regarding 2mv was in regard to HIS expo converter, not ARP's.


I suggest you spend some time studying the schematic of the 4027-1. The transistors in the CA3046 are used for the comparator that resets the integrating capacitor and for the op-amp that buffers the sawtooth output.

The NPN/PNP pair form a complementary transistor exponential converter. Alan Pearlman obtained US patent 3,329,836 that covers a logarithmic amplifier using that configuration in the feedback loop of an op-amp.

Moog's advice is correct but indirect. To match the change of Vbe in silicon transistors you must match the Is. In transistors with identical geometry the simplest way to accomplish that is to match the Vbe. When selecting from complementary transistors it can be tricky. The 2N3904 and 2N3906 are fabricated as complementary so as a first approximation you should match the Vbe. If you are using transistors with different base-emmiter areas you might need to select them at different Vbe to minimize drift and optimize tracking.

Yes, I know you don't understand what I just told you.
cygmu
[quote="loki"]
fuzzbass wrote:

Moog's advice is correct but indirect. To match the change of Vbe in silicon transistors you must match the Is. In transistors with identical geometry the simplest way to accomplish that is to match the Vbe. When selecting from complementary transistors it can be tricky


That's interesting: I knew that what we were really matching for was Is but I believed that measuring the Vbe at a fixed Ic was good enough to do that -- I didn't realise transistor geometry was a factor. Evidently we need to get into more detail than the Ebers-Moll model to see this, because Ebers-Moll gives a direct relationship between Is and Vbe for a fixed Ic and temperature. Can you give some pointers?
loki
cygmu wrote:


That's interesting: I knew that what we were really matching for was Is but I believed that measuring the Vbe at a fixed Ic was good enough to do that -- I didn't realise transistor geometry was a factor. Evidently we need to get into more detail than the Ebers-Moll model to see this, because Ebers-Moll gives a direct relationship between Is and Vbe for a fixed Ic and temperature. Can you give some pointers?


You are correct and I was not. What I had forgotten is the equation for Is includes terms for the emitter area and the base width:

The equation for the collector current in the bipolar device is Ic = Is exp(VBE/Vt - 1)

The equation for reverse saturation current Is = qADn ni^2/Wb Na

q - electron charge
A - area of the emitter
Dn - diffusion constant for electrons
ni - Intrinsic Carrier concentration
Wb - width of the base
Na - Acceptor Ion Concentration in the base
devinw1
aabbcc wrote:
Not to be a party poopers but According to the specs of the atlas DCA 55 it says:

VBE accuracy: +/-2% +/-4mv
https://www.peakelec.co.uk/acatalog/dca55-peak-atlas-dca-semiconductor -component-analyser.html

Mr moog would turn in his grave Dead Banana


Point taken. I didn't realize the spec was that open. So in theory, I may not be matching as well as I'd like.
devinw1
fuzzbass wrote:
I built 21 TTSH VCOs (same circuit) and all of them track at least five octaves well, and another 2 or 3 good enough for 1971. The transistors I used were 2N3904 and 2N3906, just pulled randomly from the bag.


Another person on a forum recommended using these same trannys on the Oddyssey as well, but as I mentioned the 3906/3904 pair did not work at all on my Odyssey B1 board at all...No sound. I realize the pinouts are different and swizzled accordingly. Is there any reason you can see that this shouldn't work? I agree they are a good candidate because they are supposed to be complimentary. The ones I had all matched Vbe (on the DCA smile) within 2mV straight out of the bags.
loki
devinw1 wrote:
Another person on a forum recommended using these same trannys on the Oddyssey as well, but as I mentioned the 3906/3904 pair did not work at all on my Odyssey B1 board at all...No sound. I realize the pinouts are different and swizzled accordingly. Is there any reason you can see that this shouldn't work?



Should indeed work! I built my first clone of the 4027-1 in 1976. Still have some that I built then. I liked the simplicity of the complementary exponential converter and used it to build voltage controlled trigger generators using 2N3906/2N3904 and 74C14. Used it in my design for the Wasatch Music Systems sequencer.
devinw1
loki wrote:
devinw1 wrote:
Another person on a forum recommended using these same trannys on the Oddyssey as well, but as I mentioned the 3906/3904 pair did not work at all on my Odyssey B1 board at all...No sound. I realize the pinouts are different and swizzled accordingly. Is there any reason you can see that this shouldn't work?



Should indeed work! I built my first clone of the 4027-1 in 1976. Still have some that I built then. I liked the simplicity of the complementary exponential converter and used it to build voltage controlled trigger generators using 2N3906/2N3904 and 74C14. Used it in my design for the Wasatch Music Systems sequencer.


Well, that is very weird. I will have to revisit that again at some point.
fuzzbass
devinw1 wrote:
fuzzbass wrote:
I built 21 TTSH VCOs (same circuit) and all of them track at least five octaves well, and another 2 or 3 good enough for 1971. The transistors I used were 2N3904 and 2N3906, just pulled randomly from the bag.


Another person on a forum recommended using these same trannys on the Oddyssey as well, but as I mentioned the 3906/3904 pair did not work at all on my Odyssey B1 board at all...No sound. I realize the pinouts are different and swizzled accordingly. Is there any reason you can see that this shouldn't work? I agree they are a good candidate because they are supposed to be complimentary. The ones I had all matched Vbe (on the DCA smile) within 2mV straight out of the bags.


Its not guaranteed that the Odyssey you have has the same circuits as the TTSH, it turns out. The TTSH corresponds to 2600 versions 2600P_3.0 thru 2601_1.0. The production times overlap the Odyssey, and some circuits are common. It is the 2810-2823 Odyssey versions that have the same VCO core as the TTSH and aforementioned range of 2600s.
devinw1
fuzzbass wrote:
devinw1 wrote:
fuzzbass wrote:
I built 21 TTSH VCOs (same circuit) and all of them track at least five octaves well, and another 2 or 3 good enough for 1971. The transistors I used were 2N3904 and 2N3906, just pulled randomly from the bag.


Another person on a forum recommended using these same trannys on the Oddyssey as well, but as I mentioned the 3906/3904 pair did not work at all on my Odyssey B1 board at all...No sound. I realize the pinouts are different and swizzled accordingly. Is there any reason you can see that this shouldn't work? I agree they are a good candidate because they are supposed to be complimentary. The ones I had all matched Vbe (on the DCA smile) within 2mV straight out of the bags.


Its not guaranteed that the Odyssey you have has the same circuits as the TTSH, it turns out. The TTSH corresponds to 2600 versions 2600P_3.0 thru 2601_1.0. The production times overlap the Odyssey, and some circuits are common. It is the 2810-2823 Odyssey versions that have the same VCO core as the TTSH and aforementioned range of 2600s.


That would make sense. I was trying this in a 2800 Oddy with the B1 board. A guy on another forum, however, said it would definitely work as replacement for the spec'd TZ81/TZ581 pair. I did not find this to be the case. Interestingly enough, Arp did sub a 2N6076 for the TZ581s on my board... seriously, i just don't get it
devinw1
So, since I am replacing the entire B board in my Odyssey, I decided to desolder the original (I believe) expo pair from VCO1, which tracks VERY well over the 7 octaves the synth has (3 + transpose up and down 2) and was very surprised that the Vbe of these transistors was not even close!

Q4 (2N5172) was Vbe=.737

Q5 (2N6076) was Vbe=.753!

Even with the 2% error on my Vbe measurement on the Atlas device, these 2 are nowhere close!

The HFEs were different by a couple hundred too. Keep in mind, this VCO tracks very well, and I believe these transistors are original due to the ancient red and white arp paint on the top.

There MUST be something different to what Arp was doing to match these things back in the day other than Vbe. I mean, they were going through the trouble to match them and paint them and provide different part numbers to the matched units.... The traditional wisdom that I've seen all over the net about matching Vbe for an expo seems to fall on its face here.

The only other information I can find is Rene Schmitz article about expo converters and they say that Rbulk is important:

https://www.schmitzbits.de/rbe.html
loki
devinw1 wrote:

The only other information I can find is Rene Schmitz article about expo converters and they say that Rbulk is important:

https://www.schmitzbits.de/rbe.html


Transistors with lower Rbe have a larger emitter area. You might want to give the 2N4401 (NPN) and 2N4403 (PNP) a try. These transistors are the "secret" ingredient in many high end audio designs because of their lower noise.
cygmu
I thought the Vbe matching was for temperature stability rather than tracking. Tracking requires good log conformance which is down to Rbe.

Obvious question, then: how stable is the ARP when the temperature changes?
devinw1
cygmu wrote:
I thought the Vbe matching was for temperature stability rather than tracking. Tracking requires good log conformance which is down to Rbe.

Obvious question, then: how stable is the ARP when the temperature changes?


Not great...hahah!

So, my follow up question, is how does one (relatively easily) measure Rbe? eek!
devinw1
loki wrote:
devinw1 wrote:

The only other information I can find is Rene Schmitz article about expo converters and they say that Rbulk is important:

https://www.schmitzbits.de/rbe.html


Transistors with lower Rbe have a larger emitter area. You might want to give the 2N4401 (NPN) and 2N4403 (PNP) a try. These transistors are the "secret" ingredient in many high end audio designs because of their lower noise.


Interesting! Thanks I will look into getting some of those to play with.
MUFF WIGGLER Forum Index -> Music Tech DIY Goto page 1, 2  Next [all]
Page 1 of 2
Powered by phpBB © phpBB Group