OK, where to start.....this is gonna be a long one.
I've spent the last 3 days with my head buried in this thing despite the 50 million other things I should have been doing. I just couldn't help myself lol, but I've gotten pretty intimate with it now and made some major progress. Bottom line, this thing was not intended for mixing purposes. It's designed to smash the hell out of mic signal so no matter how quiet or loud the guy in the booth is talking, the control room can hear him. In its default state it is very impractical for studio use IMHO. It
could be something really awesome (and reality is something new), but it needs some TLC to get it there.
Things that need addressing:
1) Metering - mandatory for practical studio use. (See below)
2) Audio output in feedback loop - this makes things funky. Because auto makeup is occurring with increased compression, but it's inproportionate to the amount of gain reduction taking place at lower G/R. Ratio is lower around the knee, which seems fairly soft, then increases to basically infinity the more you compress. When you reach more heavy gain reduction, then it's more or less unity proportion. So basically the more you compress, the hotter your output signal gets, which means you end up having to attenuate your output signal to get you back to unity gain. Branching the audio output before the feedback gain (IC T4 in original schematic) is pretty much mandatory.
3) Threshold point - default schematic starts compression point extremely low. I can't remember exactly what it was, but it was something stupid, like -30dbu or something. Feeding it a +4dbu signal means you're seeing like 25db gain reduction or something silly like that. Great for talkback use, not so great for the mix. Attentuating the signal down below threshold is stupid, because you'll have to bring it back up the same amount as well as any makeup gain for the amount of gain reduction you're doing, so assume you're compressing 5db that would be like +30db you'd have to boost the output (and +30db noise along with it). Only solution - move the threshold.
Keith - you were onto something with feeding the sidechain a fixed CV. This moved the threshold some 20db up. Funny, how once I did this it's starting to look familiar - (SSL ratio section and
) only with half wave rectification. Here's the thing though, so far I have been unable to manipulate it whatsoever. Right now feeding it the 10V CV through a 10K resistor puts the threshold point at about -8dbu. Any change in this resistor value, whether up or down results in no change to the threshold point at all. It stays fixed at -8dbu......wierd. I need to do some investigating as I admit I'm not clear on what's going on here. Any ideas?
Ideally though, this might be a good threshold point anyways if my input stage is 1/2 gain. As then feeding it a nominal +4dbu (1.23V) signal would result in -2db after the input stage (610mV) which would put you around about 6db of G/R, which is probably a good guesstimate as to
nominal gain reduction use, but will result in best noise results, as you'll not have to apply much if any additional input attenuation or input gain (or input noise
) to get a decent amount of gain reduction.
Not sure though if keeping input stage at unity gain and moving the threshold up 6db (ie -2db threshold point) would result in better S/N. Actually, nah......If feeding it a +4db you'd have to attenuate the input signal down 6db still via your input knob to be at that same 6db of gain reduction point so no difference. It
would cut into headroom though.
4) Adjustable "threshold knob" verses adjustable "input gain knob" and metering issues:
Adjustable threshold knob - it isn't possible, well...... not if you want metering that is. And here's why. The amount of gain reduction is affected both by input volume and feedback gain (threshold knob) to the fet, and non-linearly at that. With 2 variables, you'll get different feedback voltage to the fet for the same amount of gain reduction depending on the input signal level, making metering impossible IMHO. In order for metering to work, you'll have to duplicate the fet section, feed it a fixed amount of DC, and meter the source/drain difference output (Thanks Keith for pointing me in the right direction and to the 1176 metering, that was the key) Basically if you know "X" gate voltage to the fet equals "X" compression, you can scale the fixed DC to the fet accordingly, compare the difference and feed it to the meter. Because the fet will attenuate the fixed DC proportionately (provided it's scaled appropriately and the fets are matched) to how it's attentuating the audio since they're both seeing the same gate voltage, and same amount of bias voltage, it will scale the fixed DC the same. The trick here is biasing them both the same. You can't use the same actual bias voltage on the metering fet as you're using on the audio fet, so you need to duplicate it and scale it exactly the same. Turn on time of the fets need to be fairly matched, otherwise the metering fet may think it's compressing sooner or later than it actually is.
Anyways, back to the threshold knob verses gain input knob. To explain a little clearer, and lets assume we have adjustable threshold AND input gain (which we will always have input gain variable whether we have an input gain knob or not because whatever gear is feeding this thing will be outputing "whatever level" signal):
Scenereo A - With threshold knob at min feedback gain (gain of 2), feeding it approxamately +17dbu input signal results in about 14db G/R and approx 500mV rectified voltage to the fet's gate.
Metering fet sees 500mV. Indicates 14db gain reduction.
NOW..... Scenereo B - Put the feedback (threshold knob) to max gain (Gain of 11):
Feed it only a +6dbu signal results in the same 14db of gain reduction, however, measure the voltage at the fet's gate and what do you get? - 690mV DC.
Metering fet sees 690mV, indicates say 20db+ of gain reduction.
As soon as you change the amount of feedback gain the relationships between signal level and gate voltage all change. You can't have 2 variables. And since you have no control of the input signal level, you have to eliminate the feedback variable.
Now you can rely on that when I have a -10dbu signal I'll get "X" voltage at the gate. When I have a +4dbu signal I'll get "Y" voltage at the gate. When I have a +10dbu signal I'll have "Z" voltage at the gate. Once you know these things you can scale the metering circuit to get the proper readings.
Anyways, using all my findings and the 1176 as some inspiration, I've designed what so far seems to be a reliable metering circuit, at the moment it is tracking within 2dB with
unmatched fets. Before I go any further I need to build a fixture for matching, but I believe using matched fets should result in very accurate tracking. Only issue with it so far is it drops negative eventually (around 25 - 30db gain reduction), which exceeds the throw of a normal VU meter so it's a don't care, but problem is if you do happen to hit that point needle will begin to move back up. So need to figure out a way to keep it from swinging to it's negative rail. Little green here, but pretty sure can put in preventative measures via some diodes, need to experiment a little.
Now for the issues I'm having (if anyone has any input, I'm all ears)
Only real issue I'm having which is critical to fine tuning this thing - FET biasing. This is my first FETAnd before anyone says "just do it like you would an 1176" - hold that thought. With how loud an input signal? Because it matters. Feed it a 1V signal, adjust bias as you would an 1176, you get X bias voltage. Feed it a 3V signal, adjust like an 1176, you get a different bias voltage. I guess what I could do is find what gives me the best distortion figures, but I don't have a distortion analyzer here:
dustbro - or anyone else that may have any access to an original. Is there anyway you can check what your bias voltage is? I know it's FET dependant, but at least it'll give me some real rough idea if I'm in the same ballpark or way off. (Hopefully you're somewhere in the 4V-5V area.)
Other various design related items:
10V reference voltage - changing this to a fixed LM78L10 regulator. Since multiple items will now be hanging off of this it needs to be a reliable voltage point, which the zener can't really provide at this point.
Pretty large negative DC offset (50mV) has been measured feeding the feedback loop (and audio branch now) - (IC T4 positive input). Will need to change proceeding cap to bipolar or back to back lytics. Also, value here on original is way too low, will loose most of the low end, so this has been adjusted.
Variable Attack - yes it's possible, more info on this later today
Variable Release - yup, same as attack
Stereo link - I think it's possible. I haven't even started to work out the logistics, but think I can do it. Attack/release settings will need to be set the same on both channels, not sure what adverse affects could happen if they aren't. I don't own an 1176, isn't it the same deal? What happens if you set the settings on each channel different with it stereo linked? Or possible for both to go to a common R/C time constant when linked, like SSL Buss comp.
Fitting this in a 500 form factor is no problem. My proto board is about half the size. No mix circuit though or metering circuit onboard my proto, so add those for the final version plus some iron should fill up the other half with some wiggle room.
Speaking of iron, this thing's begging me for it. Over and over, it won't shut up, and I can't get any sleep. I have various cinemag's on the way, maybe I should try some jensens as well, not sure what the lead time is on those. Or if anyone else has any recommendations or even wish to lend anything for auditioning that would be awesome, would save the hassle of buying tons of iron I have end up having no use for.
Basically, for input need either a 1:1 xfmr going into a 1/2 gain dif amp to drop 6db, or 2:1 step down xfmr. Not sure which way to go, still a little new when it comes to the magic iron mojo and all the design logistics (mainly impedances) involved. From my understanding a 2:1 step down would result in low non-modern primary impedances (ie 600ohm, etc). Don't know if this is really an issue as some people seem to make it, 90% of the best stuff you guys are always after has 600 ohm input. Using a 1:1 into a 1/2 gain diff amp would still get me my 6db drop on the input and desired headroom gain, using something like a cinemag CMLI15/15 would get me modern 10K+ input impedance, don't know about the mojo on those, anyone care to comment?
For output - going to try a THAT1646 and iron as well. If getting the mojo from the input, then there doesn't seem the need for output iron as well to me if it doesn't get us anything, a THAT1646 would be much more economical.
OK, that's more than enough for now. I imagine that will need a few read throughs as it is. More later.