Matched transistors in Dynahi

You can go ahead and power-up without matching the transistors. Just don't connect your precious headphones until you validate that the DC offset at the output is low enough (less than 10mV would be good). If the transistors are not matched closely enough, the DC servo will fail to correct the offset. In that case, you will need to do some surgery.

Another point about well matched transistors is that the amp would be most linear (least distortion) even before overall negative feedback loop is applied. This is considered a good thing.

As per amb's suggestion, matching the JFETs can be done using the schematic in the link you mentioned:

Then,
So, how would one go about this? I luckily have a DMM that has PNP/NPN sockets, and a hFE setting, but haven't looked into how to test it yet, but I'm sure the meter can do it. How would one set up a test circuit to test this "at the intended collector current"?

To measure Vbe, would this test circuit be useful?

If so, what values of V1, R1, and R2 would be optimal test values for the dynahi design?

Close, but not quite right because your circuit won't provide forward bias for the transistor. Rather than drawing another diagram, I refer you to this article (courtesy of Elliot Sound Products):

http://sound.westhost.com/project31.htm

Look in figure 1, which is the most basic hfe tester for an NPN transistor. For PNP, reverse the polarity of the power and the meter. You can use a pot for Rb. When you adjust Rb, the collector current will vary. The hfe is the ratio of the collector current divided by the base current. The meter shown in the diagram will display the collector current, and you can compute the base current by measuring the voltage drop across Rb and use Ohm's Law: current = voltage / resistance. To measure the resistance of Rb you will need to disconnect the power. Also, make sure you start off with Rb turned to maximum resistance and work from there, so that you don't fry the transistor (and the ammeter). If you use 9V DC power, a 1M ohm pot for Rb will probably be a good value to test with.

As for the designed currents flowing through the various stages of the Dynahi amp, these are my calculations: The input differential stage CCS BJTs (2SA1145/2SC2705) should flow about 2mA. The cascode BJTs (2SC3381/2SC1349) should flow 1mA each. The VAS stage (2SA1145/2SC2705) should flow about 15mA each, and the output stage power BJTs (2SC3421/2SA1358) should flow about 85mA each.

Note that a small 9V battery isn't going to supply 85mA very well, so a bench power supply would be appropriate for testing those transistors.

Given this, you can adjust Rb in the Hfe test circuit until you achieve the desired collector current, then do your measurements. You can also measure the Vbe at this time.

The output transistors should be mounted on a heatsink for hfe measurements at 85mA. Since BJTs have positive temperature coefficient, they're going to warm up under test, and the warmer it gets, the hfe will go up, possibly leading to thermal runaway.

You could, of course, build the full-boat tester as described in the above web link and enjoy the convenience of switchable ranges and other niceties. Pay special attention to the WARNING paragraph, as it applies to the basic tester as well as the fancy one. If used improperly you could fry the transistor under test.

Well thank you very much!!!

So basically, using a pot for Rb, turn going from the highest resistance till you read the meter in this figure indicating the value that you quoted for each respective transistor:



Once that value is set, record the voltage across the Base and Emitter pins to get Vbe.

Then calculate as
hFE = (mA METER) / (Irb)
where Irb = VRb / Rb while Rb is measured with the power off.

With PNP devices measured using the same diagram but with the Emitter and Collector reversed.

Is that right?

Could you guys post the range of variation you are getting on the values measured from your pre-matched set of transistors, and then from your matched sets?

Which 2SA1145/2SC2705 are connected to the input differential stage? Is that the set in pairs, or the single ones connected to A/B on the schemeatic?

Yup.

No, just reverse the power supply polarity and the polarity of the meters.

The single ones, biased by the LEDs. They act as constant current sources for the JFET input differential pairs.

[edit]
Oh by the way, as PRR pointed out on headwize, NPN devices are typically going to have 40mV less Vbe than their PNP counterparts. So you won't get perfect Vbe matches. But that's ok, it's much more important to get the hfe matched as closely as possible.

I was just about to test them this afternoon, but I wanted to double check the pin-out configurations on all the devices before I put them in the test circuit.

Unfortunately, the datasheet I looked up from B&D are in Japanese!

Anyone know the Collector, Base, and Emitter pin sequences off-hand? Or else I will have to cross-reference the schematic and the board traces for a couple hours to check...

Dynahi transistor datasheets

2sa1145

2sa1349

2sa1358

2sc2705

2sc3381

2sa3421

2sj109

2sk389

See also:
Gilmore amplifier datasheets page

I must say, HAPPY HAPPY

JOY JOY



I finally have a plan of action and everything I need. It looks like all the trans' are the standard pin layouts for their packages. I've been sidetracked today by some other chores, I should have measurements of components for 4 boards by next week maybe...

Ok.. I've just had time to check my J-FET devices for Idss, the results are:

2SJ109: Part, side 1 (mA), side 2, (using a 9V battery):
#1 - 8.3, 8.6
#2 - 10.0, 10.0
#3 - 10.1, 10.2
#4 - 10.0, 10.5

2SK389:
#1 - 6.9, 7.1
#2 - 7.4, 7.7
#3 - 7.9, 7.9
#4 - 8.9, 9.0

Hmm... it's difficult to closely match the 2SK's to the 2SJ's... if I make the best matches as above, (1-1, 2-2, etc.) should it be alright?

chillysalsa,
Yeah, it's not surprising that the N-channel devices don't match the P-channels all that well. But that's alright, because each JFET is being "regulated" down to passing only 1mA in the circuit by the BJT CCS. So you should pick the JFETs that have the best match between the two devices within the package, and use the best match you could find for the two CCS BJTs.

Ok.. so I think it's safe to just stick with the JFETs I havce then, and be more discerning with the 2SA1145/2SC2705 matches.

I've also measured the 2SC3421s, my hFE values were:
142.2
157.0
160.2
161.3
161.3
161.5
161.6
167.2
167.4
167.5
167.6
167.8
168.2
169.2
170.4
174.5
177.5