Old Dynair line amps: ideas for improvement

[PRR]

Large-signal Hfe is not the same as small-signal Hfe... but for an approximation, 40-50 sounds perfectly right.

I wuz gonna say you forgot to acount for feedback... but there isn't much, is there?

Q3 or Q4 give voltage gain near 150-300 over 13, about 10 or 20.

Input impedance of (say) Q3 is about 40*13 or 520 ohms. In shunt with 3K3, 450 ohms.

Impedance in Q1 emitter is about 82+8= 90 ohms.

450/90= 5.

Forward gain Q1e to Q3c is 50 or 100.

Reverse path is 3K6/82= 45.

So it has nearly no overall feedback. (Unless the load goes open.... that's probably all that R13 R12 are for.)

And the nominal NFB through R13 R12, which we presume will boost the effective input impedance, aint happenin.

Therefore the input Z is about 8K as you say.

> you don't always need ~30dB of gain

That's what pads are for. A pad also keeps the input iron working easy (or is that a "fault"?). If you know you need 6dB gain, build a 24dB 10K pad. If you don't know, stick a 10K or 25K pot in front (keep wires short) and diddle.

 

[PRR]

me> Input impedance of (say) Q3 is about 40*13 or 520 ohms. In shunt with 3K3, 450 ohms.
Impedance in Q1 emitter is about 82+8= 90 ohms.
450/90= 5.

Ah.... that's why they needed a trimmer. Since Q3 Q4 load Q1 Q2 so heavily, forward gain is proportional to Q3 Q4 Beta. No reasonably economic selection is going to set gain consistent to 1dB, which was probably the goal (swap cards and hardly notice).

Also Beta varies with temperature.

So they said "Dave can spend his life trimming the %$#@! modules!" and left it at that.

Drop Darlingtons at Q3 Q4, get 6X the forward gain. Now the gain is limited by the 3K6 and 82.5 resistors, +/-1dB card to card and summer to winter. THD and IM may go down 2 or 4 times lower.

 

[NewYorkDave]

I just realized, my IC figure for Q1 and Q2 was high. I was calculating it via the drop across the collector resistors and not taking Q3,Q4 base current into account. With .23V dropped across ~81 ohms, IC=2.83mA. Therefore, HFE(DC, at room temperature)=2.83/0.0675=42 and Zin=42*89=3.7K, Diff. Zin 7.4K. This is slightly lower than previously figured but these are all ballpark figures anyway. We can call it "8K" and leave it at that, I'd say.

Drop Darlingtons at Q3 Q4

Sounds like a good idea. When choosing a Darlington for that position, anything I should look for besides suitable max ratings for power, current and voltage?

As I mentioned earlier, a higher-beta transistor (even a common 2N3904) seems to me a good replacement for Q1, Q2 as well.

I realize this is getting into the realm of redesign, but I'm not adverse to that.

 

[NewYorkDave]

Just for fun (and to alleviate boredom during those last unproductive minutes before lunch hour) I plugged 2N3904 and 2N4401 [NOT a Darlington, I know] into the sockets, put on goggles and carefully powered it up--watching out for overcurrent and smoke, of course.

It works, if you tweak the base bias voltage divider for Q1,Q2 to bring their collector voltage down to 7.7V as with the original transistors. With that done, the rest of the voltages fall in line with the original circuit. But the THD is worse--twice as bad, actually, even after careful trimming of the base bias and the "balance" trim.

That's all I have for today. It was just a fun way to waste 15 minutes, and I have plenty of 2N3904 and 2N4401 around.

 

[PRR]

> the THD is worse--

Ah. Dumb young snots like us assume voltage-transfer.

Transistor is a current-transfer device.

Low impedance at Q3 base makes Q3 current almost exactly Beta times Q1 current. If Beta is fairly linear at these currents, we have Vin/82 times Beta and a fairly linear output.

Why it gets worse with high Beta is not clear.

Love it for what it is, or re-design. Shame the input winding isn't split; that would allow input base bias to be set by output emitter resistor without audio degeneration. That would tend to set output emitters at 0.6V-1.0V. Short the 100 ohms and pick the "12 ohms" to set a good current, maybe 25 ohms. Use Darlington outputs so it really does run voltage-transfer and to get some headroom on input collectors. Now the 3K-82 NFB loop really should dominate gain.

Or say "%$#@! it!" and drop an ADSL chip on there. Good low input noise, ample output drive, and vanishingly low THD/IMD.