This is a real time build log.. so you're gonna get some LTspice first ... (mods if this isn't acceptable for the competition, let me know)
I have a compressor. It's small, petite, is a Ross-style one which is.. ok.. but it's not great.
My journey started with thinking - that Cali76 stacked compressor sounds good. The Cali76 is based on the 1176 which is a FET based solid state compressor. It's fast and clean, the gain is controlled by a FET acting like voltage divider with a pre amp, a post amp and a 'gain control' which really just decides if the output is too loud and adjusts the FET to reduce the signal (it's like varying the voltage divider). The 1176 is a current based low voltage design and uses an opamp to perform the comparison to adjust the FET.
The Cali seems to be a stack of these allowing a soft stack and a heavier second stack to be setup with different aggression and release.
However the 1176 is really the continuation of a saga.. before that sits the fabled 176 compressor which is all tubes.
The 176 uses a very similar mechanism to the 1176 in terms of its building blocks but the details differ heavily. The 176 doesn't use a FET voltage divider but instead varies the grid bias of the first amplification stage. This essentially changes the volume and creates a "variable gain" (I use quotes because the tube gain itself doesn't change).
The 176, when stripped down to bare bones, is basically a variable volume amplifier stage, a gain stage that creates the output waveform that is output and the gain control that uses a glass diode tube to rectify that signal so we know how big it is, then uses that to vary the volume of the first stage. A louder output wave form = a higher peak, which in turn reduces the volume by decreasing the input amp's grid voltage, making it more negative and reducing the signal out of the tube.
I said negative - but there's no negative power I hear! - correct. The output wave rectification is done to drive the diode's output negative (in the same way as creating a negative bias voltage ref in a tube amp). The result is we get the 'loudest' sound as a negative drop.. which when applied to the bias of the grid causes the entire thing to lower its operating point. Trippy!
Here's an example of the output dropping the bias voltage:
Another example with a number of quiet and loud transitions that are brutally abrupt - this is possibly the worst case scenario:
Time 0 to 4 seconds is a very small 10mVpk signal - that leaves the tube amplifying (the tube is biased at 12V, so 12-12.5 = -0.5Vgrid.. which is like turning up to 10 on the dial)
Time 4-16 seconds is a 1Vpk signal that causes the system to adapt to reduce both immediately (with a little overshoot) and in a longer slope. This is still ok and so it turns the dial down a little ... a bit like turning the volume dial down to 8..
Time 16 seconds gets a 3.16Vpk for a few cycles.. the volume dial gets turned way down.. 8.20-12 = -3.8Vgrid .. so that's like turning it down to 3 on the volume dial.
So the frequency response doesn't seem too bad at any input level and frequency:
So the original version of the 176 has some inter-stage transformers and an output transformer.. which are kind of expensive .. each. So my version of this classic old design has a few mods:
1. I dropped the iron. So there's no interstage transformer or output stage push pull.
2. I'm using 12BH7A-STRs which are linear sounding triodes and can driver up to 20mA, operating point for bet sound is 12mA.
3. I dropped a gain stage.. so I now simply have two stages compared to the 176's three. It still works fine. The important point here is that the last gain stage creates enough voltage swing to drive the gain control and thus control the gain of the first stage.
4. I dropped the tube diodes for humble UF4007s.
5. I added a ecc99 cathode follower line driver. The ecc99 is a brute with 60mA max and a best sounding operating point at 24mA! This cathode follower stage was added to prevent any output line issue later reflecting on the gain control. It means the 12BH7A can drive the gain control and leave the ecc99 driving the line.
6. I use a voltage divider on the output to prevent the system from frying any electronics.
And there you have it.. Stuffed Crust Compressor, it's a very upper-crust and over-engineered solution and a 'pedal' stuffed with tubes and transformers..
A couple of final points.
First - Sound - the tube 176 with push-pull transformer is slower and warmer sounding, the solid state 1176 is faster with more detail due to the opamp+FET control. My version has no push-pull so it's going to sound class A sweet without a transformer cancelling harmonics. There's nothing stopping an opamp replacing the gain control section, or putting a FET+resistor in which would then create a hybrid that has sweet class A sound but then the speed of the FET+opamp creating clarity.
Secondly - danger danger high voltage! - this running a B+ of 330Vdc. I tried doing a 24Vdc B+ but it all becomes very sensitive to minute changes with the real small grid biasing. Tubes like voltage.
Finally, Noise reduction - this has a basic RC filtering, but it's also possible to use a voltage regulator.
I still have some work todo.. before getting the soldering iron out, but this design looks feasible.
I have a compressor. It's small, petite, is a Ross-style one which is.. ok.. but it's not great.
My journey started with thinking - that Cali76 stacked compressor sounds good. The Cali76 is based on the 1176 which is a FET based solid state compressor. It's fast and clean, the gain is controlled by a FET acting like voltage divider with a pre amp, a post amp and a 'gain control' which really just decides if the output is too loud and adjusts the FET to reduce the signal (it's like varying the voltage divider). The 1176 is a current based low voltage design and uses an opamp to perform the comparison to adjust the FET.
The Cali seems to be a stack of these allowing a soft stack and a heavier second stack to be setup with different aggression and release.
However the 1176 is really the continuation of a saga.. before that sits the fabled 176 compressor which is all tubes.
The 176 uses a very similar mechanism to the 1176 in terms of its building blocks but the details differ heavily. The 176 doesn't use a FET voltage divider but instead varies the grid bias of the first amplification stage. This essentially changes the volume and creates a "variable gain" (I use quotes because the tube gain itself doesn't change).
The 176, when stripped down to bare bones, is basically a variable volume amplifier stage, a gain stage that creates the output waveform that is output and the gain control that uses a glass diode tube to rectify that signal so we know how big it is, then uses that to vary the volume of the first stage. A louder output wave form = a higher peak, which in turn reduces the volume by decreasing the input amp's grid voltage, making it more negative and reducing the signal out of the tube.
I said negative - but there's no negative power I hear! - correct. The output wave rectification is done to drive the diode's output negative (in the same way as creating a negative bias voltage ref in a tube amp). The result is we get the 'loudest' sound as a negative drop.. which when applied to the bias of the grid causes the entire thing to lower its operating point. Trippy!
Here's an example of the output dropping the bias voltage:
Another example with a number of quiet and loud transitions that are brutally abrupt - this is possibly the worst case scenario:
Time 0 to 4 seconds is a very small 10mVpk signal - that leaves the tube amplifying (the tube is biased at 12V, so 12-12.5 = -0.5Vgrid.. which is like turning up to 10 on the dial)
Time 4-16 seconds is a 1Vpk signal that causes the system to adapt to reduce both immediately (with a little overshoot) and in a longer slope. This is still ok and so it turns the dial down a little ... a bit like turning the volume dial down to 8..
Time 16 seconds gets a 3.16Vpk for a few cycles.. the volume dial gets turned way down.. 8.20-12 = -3.8Vgrid .. so that's like turning it down to 3 on the volume dial.
So the frequency response doesn't seem too bad at any input level and frequency:
So the original version of the 176 has some inter-stage transformers and an output transformer.. which are kind of expensive .. each. So my version of this classic old design has a few mods:
1. I dropped the iron. So there's no interstage transformer or output stage push pull.
2. I'm using 12BH7A-STRs which are linear sounding triodes and can driver up to 20mA, operating point for bet sound is 12mA.
3. I dropped a gain stage.. so I now simply have two stages compared to the 176's three. It still works fine. The important point here is that the last gain stage creates enough voltage swing to drive the gain control and thus control the gain of the first stage.
4. I dropped the tube diodes for humble UF4007s.
5. I added a ecc99 cathode follower line driver. The ecc99 is a brute with 60mA max and a best sounding operating point at 24mA! This cathode follower stage was added to prevent any output line issue later reflecting on the gain control. It means the 12BH7A can drive the gain control and leave the ecc99 driving the line.
6. I use a voltage divider on the output to prevent the system from frying any electronics.
And there you have it.. Stuffed Crust Compressor, it's a very upper-crust and over-engineered solution and a 'pedal' stuffed with tubes and transformers..
A couple of final points.
First - Sound - the tube 176 with push-pull transformer is slower and warmer sounding, the solid state 1176 is faster with more detail due to the opamp+FET control. My version has no push-pull so it's going to sound class A sweet without a transformer cancelling harmonics. There's nothing stopping an opamp replacing the gain control section, or putting a FET+resistor in which would then create a hybrid that has sweet class A sound but then the speed of the FET+opamp creating clarity.
Secondly - danger danger high voltage! - this running a B+ of 330Vdc. I tried doing a 24Vdc B+ but it all becomes very sensitive to minute changes with the real small grid biasing. Tubes like voltage.
Finally, Noise reduction - this has a basic RC filtering, but it's also possible to use a voltage regulator.
I still have some work todo.. before getting the soldering iron out, but this design looks feasible.
but I should have it all working on the test equipment.
