The GPa... At last another compressor!

[Kipper4]

Always a fan of new compressos so I might just take he plunge and breadboard it.
Thanks Andy

[Kipper4]

quick question will j113 do instead of j112?

[Fancy Lime]

J113 should be fine. However, the useful range of the Compression control will be mostly in the lower half of the pot travel. To get a better feeling range, try lowering the pot to 50k and add a 47k resistor between the pot and the holding cap. Or leave it at 100k, add a 100k resistor and halve the holding cap values. Or leave the pot and caps as they are but add the 100k resistor between them and also add a 220k in parallel to the pot+R series arrangement. Whatever your parts drawer says is most convenient.

Andy

[Kipper4]

cheers Andy.

[amz-fx]

I have tested a number of compressors that use the two-transistor-rectifier, including the Dynacomp, Boss CS-3, Ibanez CP10, etc and they all respond much the same. There is a limitation due to the two grounded-emitter transistors in the fwr, which only have about 0.6v range between off and hard on. This results in a compression curve that looks more like a limiter than a compressor. The threshold control works much as expected but the compression ratio stays quite high.

The DOD 280 reacts similarly since it uses a single transistor to drive the LED of its photo-resistor, but the compression curve is uneven due to the response of the photocoupler (though still with a high ratio).

I suspect that players adjust the average signal level so that it falls on the small knee of the rectifier, and they then have compressed peaks and normal sustained tones. This works well enough for most.

I have not tested the GPa but I may make a prototype and see how it reacts since it is fet controlled with some good mods to the rectifier section.

Best regards, Jack

[Vivek]

Thanks Jack

I see that when Vbe is exceeded, the transistors turn hard on, and will drain the integration cap rather quickly.

Should a resistance be introduced in the discharge path ?

[Vivek]

I want to run this analysis on my SPICE model:

Take input sine waves, 1Khz, amplitude 0mvp to 1500mvp, incrementing in steps of 10mv

For each amplitude, inject signal into compressor
Wait 820ms so that attack phase is surely over, then record next 10ms
Plot the peak voltage at output (green), peak voltage on integration cap (blue), peak voltage at gate of FET (red)

Would that be meaningful ?

Any other tests I should do ?

It would be so nice to correlate the SPICE outputs with actual oscilloscope readings

[Fancy Lime]

I have tested a number of compressors that use the two-transistor-rectifier, including the Dynacomp, Boss CS-3, Ibanez CP10, etc and they all respond much the same. There is a limitation due to the two grounded-emitter transistors in the fwr, which only have about 0.6v range between off and hard on. This results in a compression curve that looks more like a limiter than a compressor. The threshold control works much as expected but the compression ratio stays quite high.

The DOD 280 reacts similarly since it uses a single transistor to drive the LED of its photo-resistor, but the compression curve is uneven due to the response of the photocoupler (though still with a high ratio).

I suspect that players adjust the average signal level so that it falls on the small knee of the rectifier, and they then have compressed peaks and normal sustained tones. This works well enough for most.

I have not tested the GPa but I may make a prototype and see how it reacts since it is fet controlled with some good mods to the rectifier section.

Best regards, Jack

Hi Jack,

despite the very similar side chains, I would put the OTA-based Dyna Comp family on the one hand and the DOD 280 and GPa on the other in slightly different categories because. The former should theoretically asymptotically approach limiting, if the side chain was infinitely fast. In reality they can indeed do full on limiting. The latter category has the gain control element in a non-inverting opamp stage and therefore can only reduce gain to 1 at minimum. Controlling the maximum gain of the opamp and the action of the gain control element (vactrol or JFET) separately, we get access to both the ratio and threshold, albeit with a soft knee and some parameter interactiveness. The threshold setting is influenced by the ratio setting but not vice versa. This is the reason why I opted for a switch instead of a pot for the ratio; makes it a bit less fiddly to use. Of course, if you choose a very high ratio, as is possible on the DOD 280, the result sounds pretty much like limiting again. The charm to me lies in the possibility to also choose rather low ratios instead.

Cheers,
Andy

[Vivek]

I want to run this analysis on my SPICE model:

Take input sine waves, 1Khz, amplitude 0mvp to 1500mvp, incrementing in steps of 10mv

For each amplitude, inject signal into compressor
Wait 820ms so that attack phase is surely over, then record next 10ms
Plot the peak voltage at output (green), peak voltage on integration cap (blue), peak voltage at gate of FET (red)

Would that be meaningful ?

Any other tests I should do ?

It would be so nice to correlate the SPICE outputs with actual oscilloscope readings

and this is what I got when I ran above analysis

R11 and R14 in circuit
C15 and C16 in circuit
Sensitivity at 50%
Volume at Max





The compressor curve (Green) seems to have 2 knees.

A) Output begins with gain of 6.3x. When Output peaks reach 0.49Vp, there is first knee and output looks limited to around 0.49vp to 0.528Vp

B) From input = 519mvp to 600mvp onwards, integration Cap voltage starts to decrease to zero. After 600mvp of input, gain is 1


Does this match expectations ? Does this match Oscilloscope readings ?

I hope I did not introduce some errors when doing the analysis

[Fancy Lime]

Hi Vivek,

I'm not quite sure what I am seeing. Can you plot the input signal along with the other curves? Unfortunately, I don't have a scope, so I cannot compare to that.

Also, what exactly are we trying to learn from these curves again? If you want to understand the reaction of the compressor, I think it may be useful to compare different Compression settings for the same input.

Cheers,
Andy

[Vivek]

Hi Andy,

my primary reason to plot these curves is to study the properties of the GPa so that I can learn something. Secondary could be validate your design, check if there are any glitches, find out if it can be improved or new controls added if necessary.

Let's see what happens if we input a sine wave of 550mvp. 1Khz as input :

Input starts at 0.1sec and stops at 1.1 sec

R11 and R14 in circuit
C15 and C16 in circuit
Sensitivity at 50%
Volume at Max

Green is output
Blue is voltage on the integration caps
Red is voltage on Gate of FET


X axis is time


We see that the attack is quite fast.

When we plot other input amplitudes, we see that for a range, GPa acts like a limiter rather than a compressor

We see that voltage on integration caps first drops very fast at start of note, but then has a knee and continues to drop slowly. I feel this knee is due to C10 and may be rectification action due to nonlinearities and differences in the swing of the FET. If we remove C10, the blue curve is closer to what we see in other compressors.

Now if we zoom into the part between 820ms and 830 ms, that's the range I analysed.

X axis is time

We can read that for input = 550mvp, at time 820ms from start of input signal
Vp of output is 550mvp
Voltage on integration cap is 2.695V
Vp at gate of FET is 5.82Vp

I had earlier posted the graph of all data above, for inputs from 0 mvp to 1500mvp, in steps of 10mv


X Axis is Vp of input signal

If we zoom into 550mv, we see

X axis is Vp of input signal

We can read that for input = 550mv, at time 820ms from start of input signal
Vp of output is 550mv
Voltage on integration cap is 2.695V
Vp at gate of FET is 5.82V

From all above, we have generated the Compressor Curve of the GPa (Green curve with two knees on graph that has input Vp on X axis) when sensitivity knob is at 50%
We see it has two knees. It acts like a severe limiter in some range of input signals. It has gain of 1 for large input signals.

Yes, next set of analysis runs will be compressor curves at different sensitivity pot settings. It can take LTSPICE maybe 1-2 hour to calculate all that on my PC.

[Vivek]

Plot of Compressor Curves for different settings of the sensitivity pot


X axis is Vp of input. Lines show Vp of output after 820ms of signal, for different amplitudes of inputs, for different settings of sensitivity pot.


Green = Sensitivity pot at 0 (wiper touching power supply). Gate never gets biased in this situation. Curve has knee around 2.5Vp output due to clipping rather than compression

Blue = Sensitivity pot at 25%

Other lines at 50%, 75% and 100% are all clumped together

We can see that there is lots of action between 0 and 50% of pot travel, but after that, its same-same. When the FET has pinched off due to enough gate control signal, further increase of control signal has no further effect.

Besides controlling compression curve, the sensitivity pot allows adjustment for different FET with different parameters.

[Fancy Lime]

Hmm, interesting. Something is clearly wrong here, but I am not suer if it's a problem with the design or the simulation. It should not be possible to get a gain of less than 1, no matter what the JFET does.

The second knee is most curious. The gain seems to rise at the same time that the gate voltage drops sharply for the second time. This should not be possible unless there is something wrong with the JFET. The only thing I can think of at the moment that might cause this behavior is gate-source breakdown but that is specified at 35V for the J112. Any chance your J112 model specifies 2V where it shoul say 35V? Seems unlikely...

I'll have to think about this some more.

Andy

[amz-fx]

At T1 the fwr transistors start to turn on. The knee is very small here and the output quickly goes into limiting (as I mentioned in a previous post). There is limiting up to T2 where the fet has pinched off and can no longer adjust gain. When the fet is fully off, the circuit becomes a unity gain buffer, as would be expected from a non-inverting op amp.

For signals with a Vp of 0.5 or greater, it is only acting as a buffer.

This is typical of non-inverting based compressors when the control element reaches its limit to restrict gain.

I wonder how it would respond if you increased the gain of the op amp from 6x max to a much high amount? like 60x for example

Best regards, Jack

[Vivek]

I agree with Jack's interpretation.

PS some other compressors have similar issue when control element is saturated.

However they adjust range of system such that control element gives up gain control for signals more that let's say 2Vp input ie for very high input.

And they have large range of proportional control where control element behaves linearly with input signal.

The idea of transistor discharges cap is more binary than proportional. It's more of a threshold detector than envelope follower.

[amz-fx]

When the control resistance is in the feedback loop, a log taper will give the best response. When the control resistance is on the inverting input, an anti-log taper is best.

This may have some bearing on the response.

Best regards, Jack

[Vivek]

I would tend to believe

Due to very strong action of transistor discharging cap

Maybe it is not necessary to have 2 transistors there.


I will run a simulation without inverter and 2nd transi

To check if there is any change as compared to one transistor to discharge the integration caps




Or even just have only a peak detector with one diode going to holding cap instead of transistor that discharges cap

But then it will become the Rockman compressor!

[PRR]

....for input = 550mvp,... Vp of output is 550mvp ...

So at 550mV, output is EQUAL to input.

And same for higher inputs, all the way to clipping.

As Jack and others have pointed out, this thing goes to UNITY gain. No lower. It may as well be (is) a buffer.

And if you want level control, do not put more than 550mV in. (So a limiter ahead of this limiter?

For a 550mV max player this may work fine.

But an all-purpose/player limiter may want more range than 140mV-550mV.

[Vivek]

....for input = 550mvp,... Vp of output is 550mvp ...

So at 550mV, output is EQUAL to input.

And same for higher inputs, all the way to clipping.

To be very clear :

If we input 550mvp input
We get about 2.6Vp output for few cycles. These are clipped at the peak
then the output falls very rapidly to 550mvp

My compression curve graphs were made at 820ms after start of signal, to be very sure of being out of the attack phase.

[Vivek]

So SPICE says

If we remove inverter Opamp and second transistor from the "envelope" section of the schematic

There is very little difference in the compressor curves.

This means that one transistor working on one half of the waveshape is good enough to quickly drain the integration caps,

and adding inverting Opamp and second transistor to use other half of the waveshape to drain the integration caps makes extremely small difference.