Calculating Tube Compressor Attack & Release Time

Yes, I've searched here for well over an hour, leaving me to believe that Attack and Release are a function of R*C = seconds.  True??

That said, what if the Attack R = 0?  Seems it can't be such that time = 0, as that cap needs time to charge.  What gives?

Thanks in advance.

do you want to look at current or voltage?

voltage controls the grids, but current charges the cap to give it voltage.

but only 63.2%  of max voltage for the first time constant.

during the second time constant the voltage  rises another 63.2% of the remaining 36.8%, or a total of 86.4%.

during the third time constant, the voltage rises another 63.2% of the remaining 13.6%, or a total of 94.9%

during the fourth time constant, the voltage rises another 63.2% of the remaining 5.1%, or a total of 98.1%.

during the fifth time constant, the voltage rises another 63.2% of the remaining 1.9%, or a total of 99%.

so it takes 5 time constants to get to 99 percent of your max applied voltage.

see that a cap is never fully charged!  ???

now the current and voltage are 90 degrees out of phase with a cap in the circuit, but not in the resistor.

so if you eliminate the R, there will still be a lag in voltage due to the time constant of the cap.

this lag will depend on the frequency of the signal, plot the signal, take 1/4 of the wavelength, and you will have your delay.

square waves change the whole concept, no wonder they have Daven attenuators on the front of the 670,

Thanks CJ - interesting pieces of the puzzle yet again.

Cheers

now when the dead guys were building these compressors, they did some math on a cocktail napkin while eating a cheeseburger at lunch, went back to the lab and plugged in a few values for R and C.

then they probably  measured the response time with a scope and tweaked the thing to get it to sound good,

the fairchild has some combination RC networks which are interesting,

music usually has a complex wave form, so computing the rms voltage coming out of a compressor bridge circuit is difficult,

you can use RC combos that are good for cymbal crashes, or ones that slowly react to bass, somewhere in between would be vocals,

Rein Narma told me that he used to put on orchestra music while tuning up the 660,
since they use those cymbals all the time, you want a quick response time to catch a cymbal crash, but for overall compressing, you do not want an abrupt level change in the music, so you use a quick response for the high end transients, and a slower response to average the level.

so you put a small cap for cymbals, and a bigger cap for average material, side by side, with resistors to balance how much compressing on which band, so you get 2 response times for the same setting, then you use your ear to get the final values for R and C.

CJ said:

Rein Narma told me that he used to put on orchestra music while tuning up the 660,
since they use those cymbals all the time, you want a quick response time to catch a cymbal crash, but for overall compressing, you do not want an abrupt level change in the music, so you use a quick response for the high end transients, and a slower response to average the level.

Click to expand...

The old dual time constant thing, as most noticeably seen on the front panel of the Gates Sta-level.  Appears to have first appeared in RCA circuits at NBC in the late 1930's.  Took another decade and a half in america to become the norm, many broadcast devices stuck with a single time constant much longer. 

> what if the Attack R = 0

There are no zero-resistance sources.

You would think that zero-resistance would make life better. In fact it would lead to paradoxes, or at least irritations. (Compute the terminal impedance of a voice-coil loudspeaker with zero coil resistance.)

A chip opamp can appear to have a very low output resistance... for small signals. Load it with a cap and (if it does not just oscillate) it can only pump the cap at a 20mA rate. Say 1uFd at 20mA, I/C... 20,000 volts per second or 20V per milliSecond. You can have mS attack but not low-uS attack.

While the opamp is slammed to 20mA output it is laging behind the input. The input stage is liable to be saturated, no-gain. When it finally pulls the cap up to the right level it will take some time for the input stage to come back into active zone and NFB to settle. (That's one version of why opamps often act badly seeing a cap-load directly.)

Yeah, big transistor etc. Still you will find it is useful to have some resistance between opamp and time-cap.

The 660 is not kidding when it uses 80+mA tubes working through >1:2 current ratio

fascinatin' stuff          8)