Matching transistors & measuring the results

OK folks!

I built up version II of the Beta matcher. It works very nicely!

Here are the Gerber and drill files.
http://www.mv-makoto.com/adcom/Matcher.zip

Here is a photo of my unit in operation,

BetaMatcher.JPG


...here are some PDFs of the schematic and board,

http://www.mv-makoto.com/adcom/MATCHER-SCH.pdf

http://www.mv-makoto.com/adcom/MATCHER-BRD.pdf

...and here are the raw Eagle files.

http://www.mv-makoto.com/adcom/Matcher-Eagle.zip


Have fun!

Scott

Edit: split from this thread:Yet Another Adcom GFA-565 Thread
 
Last edited:
OK, I've just received a pile of MPSA42/92 variants. Here's the Hfe from a cheap multimeter, about 1ma. I really wish I had a curve tracer, but the Hfe differences between these devices is pretty clear, and these trends should hold at higher currents.

I sampled 10 of each brand, and got Hfe's in these ranges:

Central Semi MPSA42 - 96-120
MCC MPSA42 - 135-170
Fairchild KPS42 - 150-190

Central Semi MPSA92 - 116-126
MCC MPSA92 - 150-174
Fairchild KPS92 - 225-266 (Woo!)

The Fairchilds seem to be winners! They seem to fall well within range of the "L" grade 2SC3478/1376" (135-270 Hfe [email protected]). I see mostly L's used in the 565's, sometimes K's. (200-400)

Interesting how the Central Semi's have lower gains, but fall within a tighter range. I wonder if that's good in some way.
 
Hi Phloodpants,
Tighter parameter control is a very positive thing. One thing that will spread your groups is temperature. Even the warmth of your fingers will throw the results out. So experimental error is huge here.

-Chris

I use small plastic cover over the matcher board when I am testing devices. The air currents tend to change the temp and the devices often react differently tot he change. Ideally we should measure the temp of an operating amp,a nd test them all there..
 

anatech

Administrator
Paid Member
2004-06-06 8:31 pm
Georgetown, On
Hi Scott,
Yes, but you are testing in diff pairs, and once your shielded devices settle, they tract thermally.
Ideally we should measure the temp of an operating amp,a nd test them all there..
I have measured various equipment. The temperatures vary widely from 25°C to 50°C. The hotter temp is ridiculous while the lower temperature was from a Marantz 2225. Most Marantz amplifiers ran the diff pairs in clean air with vertical air currents from bottom to top. The worst offenders were newer amplifiers designed in the 90's and later were air came in the bottom and came out the top - eventually. No clear shot from bottom to top. A Nikko Alpha 450 was up there too, as well as other amplifiers that ran the voltage amp board upside down.

I'm pretty sure our members can look at their equipment and visualize the air flow path.

No matter what the temperature is, matching transistors the way you are doing it is the best way. Pretty much the only sane way unless you have multiple jigs set up in environmental chambers to control the ambient temperatures. Even then, chances are good that pairs matched at ambient room temperatures (20 ~ 22 °C) will fare as well as the one matched at various elevated temperatures. Remember that they are most likely from the same production lots from the same manufacturer and the doping levels will match (by definition). Therefore the two parts will track each other as long as you can keep them at the same temperature. A little thermal compound and some heat shrink tubing will help there, and you could also stick foam around that if you want to go crazy. (don't do that, it's stupid for audio). You could even stick your bonded pair in a chamber with a heater, leaky thermal insulation and temperature servo. That way you could set the pair to run at 40°C for example. In most homes and lab environments, they could be relied upon to stay at the set temperature. Overkill if all you want to do is listen to music.

Why leaky thermal insulation? So you can drop the temperature to take into account the heat generated by the matched pair within a reasonable time frame. That means your oven is always generating some heat (excepting "bang-bang" controllers). You really want a proportional temperature controller!

-Chris
 
I must say that matching these suckers is a very time consuming task. My observation is that some devices settle right away, and some start drifting as soon as you insert them, and can take about a minute to settle down. When they drift past 20 mV I just note the 20+ offset and move on. Some settle for a while than take off again for a few more mV and then settle back to some stable value. It is tempting to try to rush things, but they are going to settle on their own time....
 
I match them one way, and then for those matches closer than 10 mV I put them in the opposite sockets. This allows me to verify the close matches, and also to remove any residual offset in the jig.

I probably posted this before, but here is the logic in that:

The measurement in position A (transistor 1 in socket 1, and transistor 2 in socket 2) is:

Collector voltage difference=DV12=VC1-VC2+Vo

Where Vo is the inherent offset in the jig.

The measurement in position B (transistor 1 in socket 2, and transistor 2 in socket 1) is:

Collector voltage difference=DV21=VC2-VC1+Vo

So, to get rid of the Vo term we simply subtract these two results:

DV12-DV21=VC1-VC2+Vo-(VC2-VC1+Vo)=2VC1-2VC2+(Vo-Vo)=2(VC1-VC2)

So, the true offset is:

VC1-VC2=(DV12-DV21)/2

I only do this second test for pairs that are already close, since if they are not better than 10 mV I do not consider them a pair.

I do this in spreadsheet that automatically color codes the cells that pass the first (10mV) test, and then it performs the computation above and color codes good pairs (Which I defined as better than 2mV true match).

For a batch of 32 devices, this requires about 400-500 measurements..VERY tedious!!!
Scott
 
OK, after finally matching some Darlingtons I fired up the new board. It fired all right.. Great sparks and smoke! After some sleuthing, I found to my dismay that in the intervening 2 months of sitting on my bench while I was doing other projects (namely my multi rate digital audio board project) a stray bit of solder had splattered onto the circuit traces on the bottom of the board. Not at all sure how that occurred, but there it was, a splat right in the vicinity of Q114, which shorted a variety of elements, wiping out two transistors in the current source and Q114. It took a few days of dinking around to figure out what was wrong, but now the amp works. And so far there is no sign of the oscillations that the prior board had.

I am still concerned that something is not quite right with this amp. On turn-on the rail voltages are slightly low (about 83.5 volts vs about 84.5 to 85 on my other amps). I have the entire output stage disconnected, so the only power consumers are the control board and the soft start board. If I let it run for a few hours, I notice that the rail voltage continues to fall slowly.. down to about 81 volts(!!). This has obvious implications for the ultimate output power from the amp, since it limits the clipping output to about 55 volts peak (185 W rms). On my other amps this is about 75 volts (344 W rms).

This amp had some serious output stage issues when I got it, and I have since replaced all of the output devices. I wonder, however, if the heavy short circuits it must have experienced back in its earlier days may have damaged the bridge rectifier, so that it has higher internal resistance...

It is either that, or the control board is drawing too much current. I do not see any hot parts on the board, though, and given that the supply should be capable of sourcing 6-7 amps at 85 volts, it seems something is amiss somewhere..

In terms of board current: I am seeing a drop across the input rail resistors (R153/154) on the current sources of about 10-15 volts, which implies a current of about 10-12 mA. This seems high, but surely not high enough to draw down the main rails. I have been using my variable resistors in the current source outputs (R144/145) to adjust and balance the currents in the driver legs. From analysis, the current through R125 and R126, which is the combined current of the diode stack and the output driver devices, should be about 16 mA, and thus the drop across those resistors should be 0.8 volts. with a little back and forth (they are interdependent because the ground point between R131 and R132 is virtual), I can set those currents nearly perfectly*.

So, the entire board seems to be drawing about 25-30 mA...Surely not so much as to pull down the supply rails!!

*Note when these currents are balanced, then the output DC offset is also exactly zero, so it seems smart to set up the amp so the bias is balanced, and then rely on the servo to take care of thermal and aging issues. The only other way to deal with this would-be to try to match all of the parts on the positive and negative sides...What adjusting the current sources does is to change the current through the diff pair loads (R106 and 107), and, at zero audio input this fixes the bias points at Q109 and Q110, which then controls the current through the driver stages. When these are matched, the currents cancel in the output stage and produce zero DC at the speaker terminal. I think I am going to refine the board to support trimmers for these resistors...
 

anatech

Administrator
Paid Member
2004-06-06 8:31 pm
Georgetown, On
Hi Cogeniac,
I have never before measured the current draw of these boards, but your readings don't seem to be out of whack. No, these wouldn't affect the main rail voltages at all.

It is possible to have damaged bridge rectifiers that could draw enough current to reduce the supply voltages. They could also have a larger series voltage drop. Also, look at the mains input circuit for resistive connections or components. The relay isn't above suspicion either, although I've never had to replace one. Somewhere you will find a simple problem that is difficult to find. Life can be like that.

-Chris
 
I found two issues with the rail voltage/output signal level issue. First, I had jury rigged the power feeds to the board so I could move it out into a panavise on my bench. The connections were just twisted and taped. I soldered those and the rail droop went away.

I also changed the 1SS178 diodes, and that solved the overall output level issue. I had changed these out for a higher Vf diode when troubleshooting the board, and forgot to swap them back. The higher Vf meant that Q110 and Q109 were saturating at a lower input signal level. As soon as I changed them the output swing went to nearly 80 volts!

This is much higher than usual, and it is because I have D101 and D102 out of the circuit... So there is no active soft clipping in place. I'll try putting them back now that the board is rock solid.

I am also anticipating measuring the trim pots in R144/145. Once the amp was stable, I set the currents on Q111/112 using these pots. They are interdependent because the diode stack doesn't have a hard ground between the positive and negative ends (i.e. between R131/132). So you have to go back and forth numerous times. Adjust the drop across R126 and then R126 (to 0.8 volts for 16 mA quiescent current.. which seems to be the proper bias point based on my analysis). If you set one to 0.8 volts then the other will be at, say 0.7, so you adjust that one to 0.8 and then the first one will be at 0.85 or so.. So you go back and forth Increasing one and decreasing the other until both are at 0.8 volts. I'll measure the pots today and see how far they are from each other, and from the original resistor specs (499 ohms)...Probably just put the 499 ohm devices back in...The Vishay trimmers are nice for troubleshooting though!


The good news is that the replacement board layout seems to work fine (at least when it doesn't have solder splatters on the traces!)

My next plan is to replace the boards in all of the 565s I have (using the new CMX SMT layout), pick the best three and probably sell the others. I'll provide the original board with each amp in case the new owners want to spend their time chasing electrolyte issues... :-D
 

anatech

Administrator
Paid Member
2004-06-06 8:31 pm
Georgetown, On
Hi Cogeniac,
Excellent news. You did a fair amount of troubleshooting and accepted your own issues, we call TIM (Technician Induced Malfunction). :)

I sometimes use pots in circuit to troubleshoot. Mostly when developing a new circuit.

Good plan with those new boards. You should compare the performance between the two designs.

-Chris