Reengineer Fairchild 670 ratio curves

Hi,
for a DIY compressor project of mine I thought about to mime the legendary ratio curves of the Fairchild 670 compressor and if possible maybe also the attack and release behavior. You can find the ratio curves in the manual and the schematics at: http://audio.kubarth.com/fairchild/index.html -- but use the scanned originals since the redrawn schematic has some mistakes.
The problem is the 670 uses tubes and I don't have any experience with tubes at all, and don't know how to translate the tube circuit into a op amp, or transistor circuit.
Does anybody know how to approach this task as easy as possible?

Happy engineering,
Olli

Hi Ollie,

The signal goes through multiple push pull tubes to increase the gm and reduce the imbalance by averaging out their characteristics.  The 6386 tubes are triodes with a remote cut-off characteristic, I'm not sure if there is a transistor equivalent, probably not.  There is good evidence in this forum that they can be subbed by using 6BA6's triode wired.  These tubes are biased to 6.2V to get onto the straight part of their gm curve.  This signal amp requires a very high performance transformer in order to work to spec.  Part of this output is taken via a pad to an amp with 10~15W output, this level of power is required to charge the timing caps in time to get the level of performance you are looking to duplicate.  The input 12AX7 is biased to cut-off 3.6V so only the peaks get amplified.

I would say that you could make a similar performing control amp solid state but you would have to totally redesign the signal amp.  Remember, a major part of the sound of this compressor is the signal amp which is quite unique.

If you look at PRR's compressor, it might give you a few tips.  He uses 12AU7 tubes which have a slight remote characteristic but you could use real remote cut-off tubes but they have to be well balanced, best to buy 10 and select a matching pair, separate tubes are easier to match than finding good twins.  He then used solid state to amplify the tube signal thereby not needing the fancy transformer.

Don't be scared of tubes, if you love them they will love you back. Just watch out for the HT!

best
DaveP

okmog said:

The problem is the 670 uses tubes and I don't have any experience with tubes at all, and don't know how to translate the tube circuit into a op amp, or transistor circuit.
Does anybody know how to approach this task as easy as possible?

Happy engineering,
Olli

Click to expand...

The tubes at the heart of the 670 are the 6386; they are of the remote cut-off type. Their transconductance depends on the bias voltage on a more or less logarithmic way (gm approx. doubles for every 10 volts bias increase).
This could be emulated with JFET's, but voltages need to be scaled down by a factor 10.
You could also try putting diodes in the NFB of an opamp, and get a log response also.
You may get some interesting results, but you would never get the same response as a 670.
So far, the most successful emulations of a 670 have been digital, and the recurrent comment is they don't sound like a real 670. But then, if you took two 670's, they probably wouldn't sound the same.

Just remembered,

The Fairchild attack/release timings are based on 5 time units not the usual 3.  Check out the time constants of various compressors to see how they match the published spec and you will find this is true.

Also the Control Amp, whatever it is, must have an output resistance close to 100 ohms, this gives you the correct spec cap charging time. 

best
DaveP

Thank you both for your replies -- I've learned a lot!
I will consider this next time I'm working on this project again and will let you know about the simulation results and share the schematics.

Happy engineering,
Olli

okmog said:

Thank you both for your replies -- I've learned a lot!
I will consider this next time I'm working on this project again and will let you know about the simulation results and share the schematics.

Happy engineering,
Olli

Click to expand...

okmog, what result you get with this simulation/experiment?, did you can do a circuit withs data?

Please let us to know about this,
Opacheco.

Remote cutoff tubes need huge CV swings, may as well use the Fairchild control amp...

SS gain control amps would need smaller control voltage, which changes CV slope needed to mimic the 660/670 attack time. With smaller CV swing to some given GR, one would need a higher output impedance control amp to get the same GR attack slope.

If you go with something needing a smaller CV swing, I think a power op amp with output current trim may be a good way to experiment (e.g. OPA548 unity gain stage with output current limiting).

The response of the 6386 is a unique component of the overall 660/670 response and I would expect a lot of experimentation and/or analysis would be needed to mimic the Fairchild response beyond just the time constants and CV slope.

That being said, a 660/670 emulation except with a gain element having a different response might still result in a wonderful compressor, just not exactly a Fairchild.

Somewhere deep in the Fairchild threads some guys found that triode wired 6BA6's had practically the same CV profile as 6386's on a valve tester.
best
DaveP