First of all, the compressor didn't seem to be doing much of anything. No level reduction with the knob turned up to full (even with full input gain and clipping the snot out of the tube), the associated LED stayed green (instead of going red), nothing. A bit of oscilloscope-probing in the signal rectifier area quickly pointed out the issue - there was no real rectification going on. The signal coming out of that stage was still largely sinusoidal, albeit with a small kink at the zero-crossings. Well, that's no good...
Removing the diode in the signal-rectifier for testing didn't reveal anything, it measured as a normal 1N4148 both in and out of circuit. My aforementioned buddy pointed out that the arrangement they had used in this design, a "precision diode" circuit, was somewhat flawed as a concept, as is described in this link over here. I then proceeded to upgrade both channels to the "precision rectifier"circuit (see link above).
A quick browse through some of the parts scattered on the workbench provided a pair of 7.5K resistors and a pair of 15K - might as well give this thing a bit of gain as well (to start the compression earlier, in effect, lowering the threshold and/or increasing the maximum compression). It took me a good few minutes to make triple-sure of how the parts needed to get connected on the board, and which traces did / didn't need to be cut, but i managed, in the end.
Result: the output of said upgraded rectifier now looked as it should. Also, the LED associated with the compressor now was noticeably going red, as well as the level meter dropping out an LED or two, with the compression control going up. That's much more like it...
Speaking of the compression LED (and the fast attack/release indicators) - the series resistor seemed a bit low in value, to my eye. By that i mean, the green side of it (a common-cathode red-green dual LED, the three-legged kind) was quite bright, for what is only an indication of the compression circuitry being engaged. That, in turn, led (no pun intended) to the red part of the LED not being all that visible, only until considerable compression was taking place.
On the bright side (once again, no pun intended; or, well, maybe just a bit), the series resistors were easily accessible, on the two switch / level meter PCB's. The green compressor-on indicator got its series resistor upped to 6.8K (from 4.7K stock), and the fast attack/release indicators got a 4.7K each (up from 2.2K stock). Ok, now we're talkin' - perfectly adequate brightness levels for the purpose, and the red "gain reduction"-indicating LED is much more visible.
Right, on to niggle number two - the tube bias. During the initial post-mod testing, without having measured the test-signal amplitude, scope probing showed that past about half-gain on the "input" control", there was barely anything left from the lower half of the sinewave, when probing the tube's cathode (in effect, the output). Now, i don't know about you, but that's a wee bit too much <ahem> "tube warmth", if you ask me. I'm all for a bit of "grit" or "hair", but a half-wave rectified signal? That's ever so slightly over-the-top...
LTspice to the rescue, once again. Now, i'll openly admit, i'm in no way any sort of "tube guru" whatsoever, and much of the intricacies of their operation are still somewhat of a mystery to me, so a bit of trial-and-error in simulation is what i have to contend with. Messing with the cathode resistor value didn't seem to do much of anything. Since the anode plate's directly connected to the supply, the only other place left to tinker with was the grid.
The simulations showed (or confirmed) that what was in fact happening, was that the output signal of the tube was reaching ground (near 0V). Being supplied between 48V and ground, the output obviously couldn't go lower than ground, hence the severe clipping off of the lower half of the sinewave used for testing.
Employing a trick most often seen in guitar pedals (which are commonly battery-powered, and the opamps used need a "mid-supply" reference), i set up a resistive divider between the +15V and -15V rails, and from their mid-point, connected a 1Meg resistor going to the tube's grid. Since the end goal was to insert a DC offset there, a coupling capacitor would of course be needed, so as not to interfere with the op-amp upstream. And whaddya know - raising the grid by about 5V above where it was otherwise sitting, provided a much improved clipping point (read: at considerably higher signal levels than before, and/or far less severe chopping off of the lower half of the signal).
Aaaaaaaanyway, back to (more) practical matters - scouring through the random resistor section, i found a pair of 2.2Meg resistors, perfect for the grids, and six 220K resistors, for the bias divider. Might as well be low-load for the power rails, to be minimally intrusive, eh? And how convenient - the signals going to the grids pass through a couple wire jumpers, short enough to be spanned by a capacitor each. Hooray for not needing to even cut any traces!
 |
| The two bigger red film caps near the tube daughterboard are the newly-installed DC-blocking grid caps |
First step: "only" use a symmetrical divider, to emulate the "stock" biasing of the tube grid. On the spectrum analyser in my DAW, the 2nd harmonic sinks into the noise floor when the signal at the grid is around 40mVpp or so. That's with the -20dB(ish) pad activated, and with a 0.9Vpp signal at the input, with the gain knob around "2.5 out of 10". Not exactly what i'd call fabulous... With the pad deactivated, and the same input signal, at minimum gain, the 2nd harmonic's quite prominent, some 50dB below the test-sine.
Post-mod (ie. a second 220k in series with the one coming from the -15V rail), with the pad activated, there are no harmonics to be seen, even with the input at full gain. With the pad off, at minimum gain, the 2nd harmonic is clearly poking out of the noise floor, but still some 70dB down from the level of the test signal.Turning the gain up to full (output control untouched) brings the second harmonic to "just" 43dB below the sine, and the third harmonic about 60dB down. That might actually be a teeny bit too far the other way, actually. But still, good to know i'm on the right track.
Given the stock vs. modded readings, i'm going with a Goldilocks-value of 330K there. 220K (emulating the stock biasing) clips way too early, 440K doesn't even facilitate starting to "drive" the tube, so a "happy medium" oughtta do, methinks. Then again, as it turns out, that doesn't even change all that much. At full gain, 2nd harmonic's 40dB down, and 3rd's 55dB down. Dang, still not quite there...
Ok, last ditch effort - a 330K / 440K divider. Regardless of how this turns out, i've had it with messing with this, for now. 220K / 440K (1:2 ratio) was too clean, 220K / 330K (1:1.5 ratio) is still too clean, and 220K / 220K's atrocious. This would be a 1:1.33 ratio, so here goes nothing... Well, i guess it's a bit better - full gain results in 2nd harmonic being 35dB down. But then again, this is all with the initial 0.9Vpp signal into the 1/4" line-level input. Raising the signal to about 3.5Vpp brings the harmonics up a bit more - with the gain around 6.5 out of 10, on the brink of clipping the opamps, 2nd harmonic's "just" 27dB down, while the 3rd is 39dB down, and so on. I think we have a winner here!
Granted, the end-result isn't necessarily the prettiest thing out there, but hey - it works. And that's just as well, i was nearing the end of my patience / persistence with this. All in all, i'm pretty happy with how this turned out. And by the way, despite the anonimity of the tube, it's surprisingly noiseless, at least apparently; imperceptible anyway, as far as i can tell. But then again, it's also not doing any gain of its own, so there's that too.
I'm actually starting to think of implementing a front panel switch for "clean" / "driven", to optionally return the tube to "stock" bias and as such, provide a more pronounced overdrive. Or maybe, if i could fit it , i could even include a dedicated "drive" knob (say, a 100K potentiometer, wired as a variable resistor) in this grid-bias-shift circuitry. But that's perhaps a mod for a later date...
No comments:
Post a Comment