Germanium Amplifier

[Mike H]

Yes yes fetch the ducking apparatus, spawn of Beelzebub etcetera ~ where's Dave?

This has been a 'back burner' for some while ~ 3 years? Could be longer ~ memory is such a fleeting thing...

Only when I started out in this electroniks lark is woz all trannies back in them days don't you know, the new-fangled in thing, and germainium at that. I was still at school. In all that time I probs had about maybe 15% successes and the rest dismal failures (sometimes spectacular). Biggest problem was you couldn't trust circuits in books. Most were 'theoretical for illustrative purposes only' and plain didn't work in the flesh. Or I was inadequate to the task... insufficient information an' all that

Some diagrams I copied into exercise books. Well the ack-chewl printed books were from the library, OK?

So, I kept a couple (nostalgia? Or neuralgia) and one of them's got these two smallish diagrams in it. I never tried building 'em. Seemed fishy. Nah much too simple, it'll never work...

And who or what the blinking flip is Transis-Tronics when it's at home?

Prepare to be educated ~ also known as TEC, they were the first company in the whorld to produce an all solid-state hi-fi amp, is what, wot they did in 1961.

The two diagrams I've got, one is for a phono EQ pre-amp 'front-end', 'tother is the tone control section. Nary hide nor hair of a comparable TEC type power amp to go with 'em however. Which kind of put the mockers on the project for a bit, 'specially if I wanted to keep it more period authentic, like. Know warra mean. 1970's trannie amps I can run off in me sleep, not sure about 1960's ones though.

Google to the rescue. I found a US patent, filed on behalf of Transis-Tronics in 1962. See it here:

http://www.freepatentsonline.com/3142807.pdf


At least we now have a better idea of how they used to do it in them days. Presumably, while low power npn's could be got, remotely high power ones definitely not (or not until late '60's / early '70's?) therefore phase-splitting must be done by a transformer.

(I tried simulating the patent but again it's not an actual working circuit, just theory, so didn't work out too well. But then I wasn't using germanium devices either.)

Speaking of which, I've also abandoned all ideas of involving actual germaniums ~ they're fragile and generally nasty. And expensive!

No, why not, do it with modern silicon. Better behaved and less aggro in the long run. In theory...

To LTspice...


Cont.d...


Reason for edit: title changed (15/12/2010)

.

[Mike H]

Where was I? Was interrupted.

Got hung up on the phono stage, I was trying a modification by splitting the EQ into two parts, treble part passive, but in the front end, then realized there'd probs be an issue with noise like that.

Anyroad, here is the phono pre-amp pretty much back to original.

Yep just two trannies. Ingenious biasing as well. I think this is what threw me originally, just looked so odd.

Modifications: Reversed the positive earth supply as for silicon npn's. Also I can't think that way round any more.

R2, 47k > 100k, as R1 is already 100k.

The original seems to come off all the same supply rail, I split it so the sections are separately filtered. Hence C2, R10 added ~ final supply about 16 - 17-ish.

C4 & C5 seemed quite big, so values halved and R6, R7 doubled.

C6 100uF > 330uF. After some time I realised this one's reactance largely controls LF gain. So eventually I figured that after establishing what the extreme treble output level is, tweak C6 until the bottom end is the same. Everything else falls into place.

Note extremely low resistor values and hence high-ish supply current. Another aspect that 'didn't look right'. Certainly not compared with what was fashionable in the mid-1970's. I believe this is deliberate for minimum interference pick-up, as it's all low impedances. Relatively speaking. Again an ingenious idea.

.

[Mike H]

Tone controls section.

Modifications: VR1 was 100k and VR2, 3, 50k. This seemed a bit out of proportion but I expect the 100k was to provide a reasonably high impedance input. I added the jFET buffer stage (hadn't been invented yet in them days) so VR1 can be 10k.

VR2, 3 doubled to 100k and associated resistors, and halved C4, 5. Also gives Q1 an easier time. Otherwise values as orignal.

Note only 1 off each C4 & C5, again this is unusual compared to what became 'the norm' a decade later. I think this way works better.

.

[Mike H]

The piece de la resistance. This bit held me up for quite a while as I wasn't sure what it should look like.

However a quick flurry of research in the last week or so and voila. We have a plan.

Pieced together from scraps of crudely copied drawings in my old exercise books and 'tInternet.

Two interesting observations:

#1. LTspice sez it will do 10 Watts r.m.s. into 8 Ohms, and that with just 3 transistors and a single 30V supply. Though I cheated a bit the outputs are Darlingtons, as they're much easier to drive needing only a very few milliamps of base current.

A much more fancy, and more recent kind of configuration using all CCS's and wotnot could only manage 5 Watts.

#2. the two series DC blocking caps, C7 & C8, was a bit of a fashion once again back in the old days, occurring in magazine projects a bit. Simply thought it might be easier to source and find space for 2 x 1000uF rather than one 2200uF. However as the junction = half the supply Volts, no 'switch on thump', which was also a feature cited in the mag articles of yore.

Also something else ~ the supply current is only one third to one half the output load current, because the output devices mostly discharge one cap and charge the other through the load. Tha's clever... didn't know that before

I've got a decent wooden sleeve case that I can caniballize from something else to put it in, but need to sort out how to construct a chassis to fit, so don't expect too much of a flurry of activity over the coming months.

.

[Mike H]

Found the thing with the wooden case on it and took it off, so that's earmarked, also got a toroidal mains for it as well.

Also think I've figured how to fabricated the chassis frame to go inside. Internal dims are about 12 x 8 x 3.5" high.

.

[Mike H]

A Study In Old Technology Push-Pull Configuration

Where a transformer is used to do the phase-splitting

The following is probably typical of the early to mid 1960's

However it makes various assumptions

1. That the forward Voltage knee of both each transistor and its biasing diode are identical.
2. Deriving the DC bias in this way (as it stands) is a bit hit-and-miss to say the least, in sim it's 58mA, but what about real life?
3. That both transistors have identical characteristics, otherwise there's no guarantee that the output point will be at half of the supply Voltage. Preferably.

It should be mentioned though that there is Voltage gain going on here, not just current gain, as, due to the way in which the transformer secondaries are connected to use the transistor emitter as the reference point, so it is a true power amplifier. In this case, has a Voltage gain of about 13 times.

[Nick]

Interestingly (well I think so anyway) doesn't that circuit suffer from the same problem as simple pre futterman OTL stages, in that the output is unbalanced, the top transistor is a emitter follower and the bottom a common emitter, so they will have different gain and the impedance the output sees looking up and down is different?

[Mike H]

It occurred to me (in a flash of inspiration) that each transistor half is like a constant current generator, but made variable (in opposing phases obviously) by the transformer winding current being added or subtracted.

If that's the case then why not make the bottom half the "CCS", and have the upper half just be Voltage biased? I.e. the output junction point will make supply/2 when the top transistor has pulled up the bottom transistor collector until the current is equalized.

Now C1 just provides an AC path to Q1 emitter for the "earthy end" of L2.

There is a slight problem though, for some reason I haven't figured out yet the top half is amplifying slightly more than the bottom half. Or the bottom half slightly less. Maybe it's AC signal current leakage through the diode bias?

To be continued....

[Mike H]

Interestingly (well I think so anyway) doesn't that circuit suffer from the same problem as simple pre futterman OTL stages, in that the output is unbalanced, the top transistor is a emitter follower and the bottom a common emitter, so they will have different gain and the impedance the output sees looking up and down is different?

Funny you should say that cos I thought it was OK, bu-ut, it started getting like that if the top transistor was a pure emitter follower, see second diagram!

[Mike H]

Oh wait I know what you mean, well I think that was always the case where each o/p device is the same polarity type, NPN or whichever, even in a direct coupled circuit. In that case an opposing types transistor splitter is used to actually control each o/p transistor. Even then the crossover distortion region can be slewed in one particular direction, because of it (saw a graph about it once!)

However in this case, because of the transformer drive and the way the secondaries are connected across base - emitter in each case, they are both common emitter amplifiers?

 

[Mike H]

More later....

 

[Nick]

Interestingly (well I think so anyway) doesn't that circuit suffer from the same problem as simple pre futterman OTL stages, in that the output is unbalanced, the top transistor is a emitter follower and the bottom a common emitter, so they will have different gain and the impedance the output sees looking up and down is different?

Funny you should say that cos I thought it was OK, bu-ut, it started getting like that if the top transistor was a pure emitter follower, see second diagram!

Do you mean the one with the feedback to the driver stage?

If you look at a Futterman OTL you see much the same feedback.

[Mike H]

Oh right, erm, no feedback in mine! Unless you mean C1? Which is like a bootstrap (to use the vernacular)

OK just looked at the currents in each half ~

1. the original, version 1

2. the modified with top transistor as emitter follower only:

Green = top [R1], pink = bottom [R4]

Definitely a discrepancy and it comes from the transformer secondaries, there's about 24mV out of the top one compared to 37mV from the bottom one, so the power is definitely draining off somewhere for the top one

[Mike H]

Yep OK it's the transformer secondaries being overloaded as it's only a little transformer with high winding resistance, so even a modest base current is too much


Back shortly....

[Mike H]

A variation on a theme

The 2SK1058 MOSFET version

R6 biases up Q2, which then acts like a CCS. Q1 acts as just a source follower so the o/p point DC is as per whatever it is at junction of R3/R5.

For the same signal level at V2, get 6 times the output from this one As the loading on the secondaries is almost negligible, just capacitance mostly (600pF)

Wouldn't they have liked to have these in the 1960's