Germanium Diodes in a Dyna/Ross Comp?

[markm]

I recently saw a "mod" for using germanium diodes in the Dynacomp/Ross Compressor circuit
in place of the regular silicon 1n914,
What exactly will this do?
What's the benefit?
     MarkM

[alderbody]

I tried it in my clone and it works.
Since i liked the tone i kept it like that.

but the funny thing is that i did it on a "trial and error" basis and not by following somebody's mod.

I used a Ge diode type i found in banzai. (a NOS fast diode, if i remember well)

[markm]

It changed the tone

[alderbody]

Just because i had also replaced some caps and transistors in there, i can't be quite sure if this was the reason it sounds significantly better now.

It feels like it responds much faster to the changes in the input and it generally has a softer sweeter sound.

But i repeat that this can't be happening 100% because of the Ge diodes. (maybe even 0%...)

[R.G.]

Germaniums with a lower forward drop make the rectification in the envelope detection portion of the compressor more responsive at lower signal levels. It may also make it start compression at a lower level - I'd have to do more looking to figure it out exactly.

[markm]

So I suppose it would make the circuit a bit more sensitive and softer perhaps.
I think I read that it makes it respond more like an optical compressor.
I could be wrong but, I would think it would also cause a bit more ripple as well.
  MarkM

[Mark Hammer]

I doubt that such a diode change would produce any of the qualities of an optical compressor.  As RG notes, however, since it would alter the point at which the signal reaching the base of the transistor after it becomes constant in amplitude (i.e., "clipped"), you can expect it to change how much compression is applied at which input level.  This will change the feel somewhat, although I gather that the original feel would be restored if one simply changed the input signal a bit.

Whether Ge or Si, the two diodes need to be reasonably matched to keep envelope ripple to a minimum.  So, if you have a choice between a pair of Si diodes measuring 515mv and 518mv voltage drop, and a pair of Ge diodes measuring 190mv and 218mv drop, you're smarter to go with the Si's for sonic quality.

[A.S.P.]

you folks make it difficult to try and help;

one could post a a link
to the circuit in quest, so people who are not acquainted to acronyms like "Dyna/Ross Comp", and would have to do long search & click adventures (that T*ne-P*d site is really cryptic...)
to find what`s the talk about,
could have a quick look at the schemo,
and maybe quickly could analyze,
that the diodes don`t "squeeze, squash or bend the signal or the control voltage,
and thus don`t lower the threshold for the onset of compression
(like would be the term to use in this case of suspections...),
nor bend the "knee",
but IMHO just cut off the negative going halves of the oppositely phased signals emerging from phase-splitter Q2 (similar to: "clamping"),
in order to get those left over positive halves amplified by Q3 & 4,
which are being added at the respective collectors.
These transistors *might* give an "exponential" character to the thus
fullwave rectified/amplified and smoothed by 10µF (don`t you just hate undesignated components?) signal amplitude derived CV.

do I misinterprete something here?
plz correct me, then.

maybe the DC at which the bases of Q3 & 4 "sit" will be on a different level
with other diodes,
and hence shifts the Qs` operating points to get a different "slope" for the CV,
(since the emitters don`t have loval feedback resistors for linearizing),
and maybe the Diodes cause such a shift during the negatve ("cut off") halfwaves,
but I doubt very much, that the signal/CV undergoes a "tubescreamerlike" experience
like it seems to have been commonly assumed.

[R.G.]

do I misinterprete something here?
plz correct me, then.

Yes, you do. They don't just cut off the lower halves, they actually boost the signal by something less than 2x by clamping the bottom peak up to ground.

The diodes clamp the negative going half cycle to ground, forcing the entire waveform up from ground. It makes for essentially doubling the signal level there. It's a variation of the old "DC restorer" circuit from early video stuff.

The more ideal the diode, the more closely the bottom peak of the AC signal is clamped to ground. Germanium has earlier turn on at a lower voltage than silicon, so smaller signals initiate the clamping, and hence smaller signals get the signal level boost from having its negative peak clamped (upwards) to ground. So smaller signals will turn on the transistors, and could make the circuit more responsive at lower signal levels.

The two diode/transistor circuits are a variation of full wave rectifier, acting on the filter cap at the joined collectors.

Good DC restorer clamps actually pump the incoming capcitor up to almost ground at steady state, not losing a diode drop like you'd expect. This relies on the diode's conduction at very low voltages and currents. But because of the transient nature here, I doubt that it ever gets to anything like steady state.

[A.S.P.]

but IMHO just cut off the negative going halves of the oppositely phased signals emerging from phase-splitter Q2 (similar to: "clamping"),

crap - not "similar" to clamping, but: "clamping" in deed...
(not cutting off negative halves,
but shifting the complete signal to positive values,
so there is nothing negative - or at least nothing substantial negative left over).

yes: minus about that part of the signal,
that lies about a "diode-drop" beneath ground reference.

(Thank You, R.G., for the rescue!).

did a little "real life"-simulation,
and found that the Ge-version does have a bigger portion of the clamped original signal above ground, than the Si-version.
(i.e: the effective average positive energy content in the Ge-version is higher than that of the Si-version (is that: RMS?).

With both kinds of diodes, both signals at the diodes
(which are not used in the "tubescreamerlike" forward  "method", here),
don`t get distorted, or bent like would be be assumed in the common
"Germanium-discussion",
but rather provide an average DC-level to the transistors,
which is higher in the Ge-version.

With both diode-variants,
the effect to obtaining a CV,
is providing a DC-level (depending on the signal-amplitude to the transistors),
who in turn amplify any positive voltage in the range of +~0.5V to ~+0.8V
at the bases, into a voltage of ~+8.3V to +0.6V at the joint collectors.

My assumption (not yet posted, luckily...) of a differential expontially amplifying thing happening here,proved to be wrong...
but lead to an interesting experiment:

instead of having emiiters Q3 & 4 tied to ground directly,
insert a ~4k7 pot (outside lugs to the emitters, wiper to ground),
and balance out the "ripple" at the joint collectors.
(this will lead to a different "operating point" at the Q`s curve,
as well as a lower amplification of the wanted DC-level,
so it`s not a drop-in replacement for this compressor...),
The ripple will be able to be balanced from a sawtoothlike wave
to a sinelike wave of half (alternating) amplitude, while remaining substantially the same DC-level fluctuation depending on signal-strength.

Decrease the smoothing cap from collectors to ground
from 10µ to 22n, for fast response.
Yes - this will increase the ripple on top of the DC to whatnot...
Now decouple the ripple AC-wise from the CV via another cap,
invert the ripple, and add the inverted ripple AC-wise
to the "DC+ripple"...

Gives a very fast-responding CV without any ripple down to below 25Hz of signal-frequencies!

Tested with signal-generator sine/triwave from 17Hz - 17kHz,
wobbling the amplitude from 0V~ to 2V~ quickly by hand.

Sorry for the inconvenience,
but may be useful for other envelope derived purposes...

[markm]

Well,
This is kinda over MY head a bit.
So.....I take it that it makes a difference huh?

[A.S.P.]

with GE-diodes, the gain-reduction will start at lower signal-strength,
and max compression will be reached at lower signal-strength.

sorry about my link-comment,
which was written off-line,
and didn`t re-insert the link after copying...
CompaRous-link

[A.S.P.]

another interesting experiment:
pot (~22k), outer lugs across the rails,
wiper via 470k to base of Q4
"prebiases" the envelope,
and thus lowers the "threshold".
(might require an overall DC-level realignment, though).