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

matching transistors - DIY
MUFF WIGGLER Forum Index -> Music Tech DIY Goto page 1, 2, 3 ... 17, 18, 19  Next [all]
Author matching transistors - DIY
fonik
roglok wrote:
very handsome, matthias!

and small !!! Will have to mount it to a panel .
J3RK
fonik wrote:
okay, while the MFOS circuit is called 'simple', Ian Fritz' circuit really IS simple. and there is no need for a precise voltage supply or matched resistors.
just a set of three resistors, a diode, and a DPDT on-on switch for each NPN and PNP transistors:


Ian Fritz' Transistor Matcher by fonitronik, on Flickr


That is nice! thumbs up I'd much prefer to use that over my breadboarded version. Mr. Green

It would be cool to make a "Builder's Utility Module" that contained this, a tuning calibrator for 1V/Oct and maybe 1.2V/Oct, and maybe some peak level indicators or something along those lines.
frijitz
fonik wrote:
the eagle files, you mean? i would/could ask ian.

Of course, no problem. It's not like some clever invention, after all, basically just a Wheatstone bridge.

Ian
diablojoy
Quote:
That is nice! I'd much prefer to use that over my breadboarded version.

It would be cool to make a "Builder's Utility Module" that contained this, a tuning calibrator for 1V/Oct and maybe 1.2V/Oct, and maybe some peak level indicators or something along those lines.

seconded
an accurate note indicator would be very useful
qfactor
diablojoy wrote:
Quote:
That is nice! I'd much prefer to use that over my breadboarded version.

It would be cool to make a "Builder's Utility Module" that contained this, a tuning calibrator for 1V/Oct and maybe 1.2V/Oct, and maybe some peak level indicators or something along those lines.

seconded
an accurate note indicator would be very useful


An A440 reference tone in this module would be good too! thumbs up
elmegil
This was one of my early projects. Taught me the lesson of "read the specs for the switch, don't just look at it in a picture with no size reference." :-)

It's all wires, resistors and diodes, and a 9V battery on the inside. And one of those switches is a 4PDT :-) It's Ian's circuit....

Edit: having trouble remembering attachments lately, apparently.
qfactor
elmegil wrote:

It's all wires, resistors and diodes, and a 9V battery on the inside.


So this circuit works with a +9V and gnd? No -9V?
elmegil
I misspoke, there are two 9V batteries inside.

I also revised it at one point and put in the trimmer to zero out the difference.

Here are the guts:
marvkaye
I built my transistor matcher from Ian Fritz's schematic, included trimmers to get the resistances matching and then added switches to handle the emitter swaps. After having used the unit for awhile (most recently matching for the expo converters on several Thomas Henry/Fonik VCO555 PCBs) I find that I prefer to install one trannie as a reference and then match the DUTs by recording their variations from the reference part and putting those with the same readings in the same bin. It's actually faster than swapping both trannies and swapping emitters, although I can do it either way. Live and learn. The next update is to replace the DIP sockets with a nice ZIF socket to make putting the transistors in and out a bit easier. With the DIP sockets I have to handle each transistor... there's no way to push the pins in using a pair of tweezers, simply too much force is required.

The circuit is built on stripboard and the DIP sockets have really long pins (wire-wrap type) so they're installed standing tall enough for me to be able to use 3/4" standoffs to secure the board (the 4 black screws surrounding the sockets.)

Anyway, it works really well... I found that in about an hour I was able to match up 4 pairs to +/- 50uV (yep, that's 50 micro volts).... that was before I realized that most people consider their transistors matched as long as they are within 1.5-2mV. Duhhh... Since all my trannies came out of the same batch I probably could have used them without matching and met that spec... I've got leftovers from the process that match to 0.5mV already. Live and learn.



Here's the trans matcher alongside my MFOS V/Oct calibrator... another great project.



Thanks for looking.....

<marv>
Altitude909
perfect timing for this thread, need to match me some 3904s
oldenjon
fonik wrote:
and there is no need for a precise voltage supply or matched resistors.

Wait, Ian's method calls for a matched pair of 100k resistors. What's more is that they should be matched using a 4.5 digit DMM, and a meter with those specs isn't exactly cheap.... Did I miss something?
Altitude909
oldenjon wrote:
fonik wrote:
and there is no need for a precise voltage supply or matched resistors.

Wait, Ian's method calls for a matched pair of 100k resistors. What's more is that they should be matched using a 4.5 digit DMM, and a meter with those specs isn't exactly cheap.... Did I miss something?


You can buy 0.1% resistors and a reasonable price (0.75$ ea last I looked). a few bucks of those is better than sorting through a big pile of 1%.

If you're in the market for a good DMM, I just grabbed one of these: http://www.ebay.com/itm/370769141376?ssPageName=STRK:MEWNX:IT&_trksid= p3984.m1497.l2649

5.5 digit and the one I got was laboratory clean
frijitz
oldenjon wrote:
Wait, Ian's method calls for a matched pair of 100k resistors.

I recommend matched resistors or adding a small cermet trimmer as described in the writeup. If the resistors aren't matched, then you have an offset voltage to subtract and you can end up trying to determine a small difference between two large numbers. The method still works, it's just that you loose resolution, which makes a difference when you get into the uV regime.

Ian
fonik
here are the eagle files.
delayed
Super cool of you to post these. Thank you.
delayed
getting late/early now. wiring tomorrow. kinda think i have this flipped, but since it is a mirror of itself i don't know if that matters. will investigate later.
megaohm
oldenjon wrote:
fonik wrote:
and there is no need for a precise voltage supply or matched resistors.

Wait, Ian's method calls for a matched pair of 100k resistors. What's more is that they should be matched using a 4.5 digit DMM, and a meter with those specs isn't exactly cheap.... Did I miss something?


Check out page four "What Current?".
Make the -V variable.

With my cheap meter measuring 100K resistors I only get:
100.?.
So, one decimal place.
If I measure 10K resistors my meter gives me two decimal places: 10.??.

In this case, I'd replace the matched 100K resistors with 10K resistors (I would replace them with 30K, actually, because my meter still gives two decimal places with that value - 30.??K).

So, by dropping to a lower value resistor you can get better resolution when matching the resistors. Depends on your meter.

Plus, it seems a good idea to have adjustable current because then you can set it according to the circuit you will be using the matched pairs in.
frijitz
megaohm wrote:
So, by dropping to a lower value resistor you can get better resolution when matching the resistors. Depends on your meter.

Percentage-wise 100.x and 10.xx are the same resolution.

Ian
doobedy
jonbstevens wrote:
I breadboarded Ian's when i wanted matched transistors for for CGS DUSG's. I get close to 2 octaves in tune once i've got them trimmed correctly.

I'm not sure what sort of tuning to expect with unmatched transistors, but i'm pretty happy.


Two things:

1) Matching Vbe gets you temperature stability, and has nothing to do with tracking performance as far as I can tell.

2) Does anybody (Ian?) know which transistors have good log conformance, which should get us better tracking.

I just did a couple hours of matching for two DUSG boards, using 2n3904 and 2n3906s, for 8 pairs in total. I haven't wired them up yet but I'm guessing a couple octaves is probably pretty good judging by past experiences.

As far as the rest, I matched the 100k resistors to one decimal point, but I don't think it matters too much as long as you you leave Q1 in it's spot, and compare a succession of Q2s. Like Ian says in his PDF, as long as the difference is the same, they match. No need for supermatched resistors.

I got all my matches down to +/- .1mv according to my cheap meter, but it took a good 5 minutes for temps to really settle down during tests, a small fan might help, but whatever. It is kind of fun. Like the world's worst slot machine.
daverj
Seeing the mention of 9 volt batteries above got me thinking that Ian's circuit could be simplified to run off a single battery instead of two. So I came up with a simple change and PM'd Ian about it. He came back with an even simpler modification to do it. We went back and forth a couple of times and ended up with the following schematics.

I feel guilty even adding my name to the drawings since in the end all I did was prod him into coming up with the changes.

A couple of notes:

-- I did three drawings. The first one uses a pair of DPDT switches and is the most universal. It's the best one to build. The second drawing is for people wanting to build something extremely cheap, without the switches. But it's different for NPN and PNP, plus requires you to manually swap the meter probes during matching. The third one is a variation on the second, combining the two with a single battery but separate sockets and probe points.

-- The sockets are marked EBC but the B and C are connected together so no need to twist the leads around if you have transistors that are in an ECB package instead of EBC.

-- With 1% 100K resistors the trimpot could theoretically be 2K. But since trimpot tolerance is not as good, it should be slightly larger. I picked 5K since it's the next larger standard size. If you can get a 2.5K or 3K trimpot that would be even better. You adjust the trimpot by switching the meter probes back and forth, turning the trimpot until you get the same reading with just the +/- sign changing.

-- All three drawings show the same simplified circuit. Just without the switches in two of them.

-- The same simplified circuit could also be connected to a regular power supply too instead of a battery. It just needs one power rail plus ground for that.





qfactor
daverj wrote:
Seeing the mention of 9 volt batteries above got me thinking that Ian's circuit could be simplified to run off a single battery instead of two. So I came up with a simple change and PM'd Ian about it. He came back with an even simpler modification to do it. We went back and forth a couple of times and ended up with the following schematics.


Thank you all for this!! applause we're not worthy
gaetano
Interesting about the Moog 'flaw', which I didn't notice, as I have only used the Moog version for matching NPN's
Dave Kendall
Many thanks Ian and Dave - a veryuseful circuit thumbs up

Quote:
2) Does anybody (Ian?) know which transistors have good log conformance, which should get us better tracking.


Is there a way we can measure for that?
Would it be something like varying the current (voltage?) to the DUT in set increments and comparing readings with an "ideal" set of values for a log curve?

How important is log conformance for VCOs and VCAs respectively?

Also, how useful/important is the hFE reading in the scheme of things - do the two trannies need to match tightly, and does the hFE value need compensating for in a VCA/VCO circuit in some way?

Sorry for all the questions, but I've been wondering about these things for a long time, and haven't yet found anywhere that discusses them.
A way to avoid expensive matched pairs and get *really* high performance would be truly magical. Well worth the time matching IMO. Ideal if you have time and not money..... wink

cheers,
Dave
ejr27233
As I've only got a 3.5 digit multimeter is there any point in using a x100 gain op amp in the circuit to measure the offset mV?.
frijitz
Dave Kendall wrote:

Quote:
2) Does anybody (Ian?) know which transistors have good log conformance, which should get us better tracking.

Is there a way we can measure for that?

The best log transistors are the special super-transistors that are sold as hi-conformance devices. The best are the MAT0x devices. The LM394 and SSM2010 are also very good.

Log conformance is usually characterized by the spurious series resistance in the emitter path. You will see this given different names. The standard measurement of this resistance is the slope of the collector to emitter voltage vs base current, with the collector open. The conformance is not usually specified for most transistors, except for the super-transistors. I measured a number of other devices a while back, as did someone else on the s-diy list.

Quote:
How important is log conformance for VCOs and VCAs respectively?

Quite important for VCOs. A high-frequency tracking adjustment does a pretty good job of compensating for errors, but it still works best to start with a good device. Most VCAs are linear so log characteristics don't enter in.

Quote:
Also, how useful/important is the hFE reading in the scheme of things - do the two trannies need to match tightly, and does the hFE value need compensating for in a VCA/VCO circuit in some way?

Usually that is not considered. High hFE is required for the ideal equations to be accurate, though. (See Analog Devices Nonlinear Handbook for this.)

Quote:
A way to avoid expensive matched pairs and get *really* high performance would be truly magical. Well worth the time matching IMO.

Once again, matching is for temperature compensation, not for tracking.

Ian
MUFF WIGGLER Forum Index -> Music Tech DIY Goto page 1, 2, 3 ... 17, 18, 19  Next [all]
Page 1 of 19
Powered by phpBB © phpBB Group