What classic effect circuits react well to dying batteries?

[csj]

I think you're that hippie electron who gave Amdahl the guided tour.

[Eb7+9]

...The two out of phase signals are half wave rectified by pushing current into the bases of the detector transistors; the diodes and pulldown 1M resistors clamp the negative going end of the signal to ground, giving an effective gain of two to the detection, but softening the attack as the clamp diodes ramp the coupling caps to their clamping voltage. (See Halliday and Resnick, "Clipping and Clamping Circuits" in "Pulse and Digital Waveforms" for a more complete explanation). The collectors pull a pulse of charge out of that 10uF cap. The cap charges to +V through that 150K resistors (which, by the by, sets the release time constant), when there is no signal. The signal pulses, being out of phase, effectively full wave rectify in the detector, the 10uF cap smooths that, and the last transistor acts as an emitter follower to drive the buffered 10uF cap voltage into the Iabc pin through the variable resistance sustain pot.

The **only** effect of lowered V+ on that circuit is that the gain of the OTA can't get as high since there is less voltage available to drive its gain up. There isn't any bias effect there.

Ah (I can hear you say) but the ... (FET?? no, bipolar) NPN with its base connected to the OTA, that one is biased! Right??

... the first bipolar stage acting as a follower for the pedal's output  is biased by the resistor network holding the OTA output in its linear range - the signal gain at the collector of that stage drops with supply voltage (same way a FET circuit would) ...

... the OTA bandwidth doesn't vary with bias like an op-amp would because there's no integrator inside of it ...

... indeed if cap charging stage was FET based the channel resistance would play second fiddle to the production of a half rectified signal appearing on the 150k resistor and then being averaged by the cap ... so the drop in detector gain still occurs for the same reason in the same place ... btw, in my otherwise stock pedal the feedforward path is left untied, ie there's only signal coming from the in verted output - so my pedal is in disagreement with your schematic ... otherwise you get clipping bleedthrough from the in-phase path diode to the output ....

... jc

[R.G.]

... the first bipolar stage acting as a follower for the pedal's output is biased by the resistor network holding the OTA output in its linear range - the signal gain at the collector of that stage drops with supply voltage (same way a FET circuit would) ...

... um... the first bipolar stage is a follower as you do note. The signal is taken from its emitter, not from its collector. Whatever takes place at its collector is completely immaterial to the operation of the pedal. In fact, in the stock MXR Dynacomp, the collector is tied to the +9V power supply. There isn't any signal gain at its collector.

And while its base is indeed held to the same bias as the output of the OTA, it is in fact the same bias point as the phase splitter, which is why the Vr is at that odd voltage, as I pointed out in the last post. The follower cannot be affected in gain, phase, or frequency response by the bias point any more than the phase splitter is. Gross changes in the power supply will make the phase splitter distort when the signal gets larger than the bias voltage  so big changes in the V+ will cause distortion(as I pointed out in the last post). But there's no subtle mojo going on here. It's kind of like when there's a glass of milk on a table. If you push it toward the edge, it either falls off or it doesn't, but it doesn't get sour as it approaches the edge.

... the OTA bandwidth doesn't vary with bias like an op-amp would because there's no integrator inside of it ...

First, the bandwidth of *all* bipolar transistors varies with current level. That included particularly the bipolars in the diffamp input of the 3080. Second, the output of the 3080 is a current source, and any capacitance at all on it makes the output into - yep - an integrator, where the available Iabc limits the slew rate.

... indeed if cap charging stage was FET based the channel resistance would play second fiddle to the production of a half rectified signal appearing on the 150k resistor and then being averaged by the cap ... so the drop in detector gain still occurs for the same reason in the same place

? if it were different, it would act different???
I fail to understand how your proposition leads to your conclusion.

so the drop in detector gain still occurs for the same reason in the same place

... so this does not follow.

btw, in my otherwise stock pedal the feedforward path is left untied, ie there's only signal coming from the in verted output - so my pedal is in disagreement with your schematic

Bummer. My otherwise stock pedal agrees. Looks like a different version, huh?

[zenpeace69]

See "Pedalboard Power Supply" at GEO for the first instance I know of of a circuit like this. It's an LM317 voltage regulator adjustable from 7-10Vdc and a resistor in series with the output. Six months *after* I put that up on GEO, someone else applied for and later received a patent on that specific setup. 8-)

I believe I pointed the guy to it then as well.

I think the originator of this thread is wanting to sell a voltage regulating device as well. I hope he doesn't get trouble from the patent holder.  :wink:  :roll:

I do want to sell it.  It isn't the same circuit exactly, though.  I have different values for caps, resistors and pot.  Does that matter?  If someone has a schem design patent does it cover any and all substitutions of component values?  Also, how can you patent the way you wire a voltage regulator?  It is what it is and it only works the way it can be wired up.

[Eb7+9]

... it is in fact the same bias point as the phase splitter, which is why the Vr is at that odd voltage, as I pointed out in the last post. The follower cannot be affected in gain, phase, or frequency response by the bias point any more than the phase splitter is.

... I checked things out a little closer, and you're right there's little gain variation in the follower/splitter stage (Q2 in your schematic) ... but in my pedal the emitter node is used to send out audio at the output only while the collector node sends signal (at unity gain) to the detector circuit ... the feed to the second half-rectifer stage was left disconnected at the factory - it's an early Script Logo MXR ...

... the bandwidth of *all* bipolar transistors varies with current level. That included particularly the bipolars in the diffamp input of the 3080. Second, the output of the 3080 is a current source, and any capacitance at all on it makes the output into - yep - an integrator, where the available Iabc limits the slew rate.

... the OTA is not operating as an integrator as it is primarily loaded by 150k resistor tied to an AC ground (DC referenced, capacitor bypass - AC short to ground) ... also the ft of small-signal is way beyond the audio range of this circuit ... so all this stuff is irrelevant, there is no gain derived bandwidth variatin anywhere in there - the dominant (low-pass) signal suck in that circuit comes from 0.001u cap wired in parallel with the OTA load ...

I was looking at my hand-drawn schematic from way back, with Bipolar devices instead of FETs, and began focusing on the half-wave switch/amplifier stage (in mine only Q4 is active but Q3 is connected as well except for the missing 0.01u cap leading to the base of Q3) ... since there's no emitter degeneration there's no feedback increasing the Rout seen at the collector node ... even though that circuit is operating in a non-linear manner we can extrapolate its "on" characteristics somewhat from linear analysis concepts ... the incremental drive resistance of an emitter follower stage taken at the collector node is given by

rout = ro + (gm * ro) * re ...

where ro, the incremental resistance term tied across its pi-model current source, is approximated by Io/|Va| where Io is the quiescent or DC collector current (not necessarily a stable bias current either) and Va the Early voltage named after James Early if I recall correctly ... these are approximations to circuits biased in their linear region and scuh "constants" largely inapplicable directly in a non-linear context, rather we need to be talking in terms of loads curves and not just zero-crossing quanitities ...

nonetheless, a qualitative analogy is possible based on a per-cycle averaging of non-linear characteristics, and so the absence of emitter degeneration would imply that the per-cycle drive resistance average at that collector be set directly by the device's non-linear drive characteristics directly (ie. not augmented by feedback), and so could conceivably be quite low compared to a device stably biased - lower than customary safe value of 100k for ro in typical small-signal bipolar devices ...

... since the time constant of the averaging circuit is set by the parallel combination of (i) the 150k load resistor, (ii) the non-linear resistance per-cycle average seen looking into the collector of Q4 and (ii) the resistance looking into the base of the voltage-to-current bias converter transistor Q5 (which is much higher than the other two components, ie. Beta times minimum emitter resistance, or 100 x 27k min >> 1Meg min) then we can see the potential for a time constant dominated by the non-linear average of the collector node of Q4 provided it is less than 150k ... otherwise, if we oversimpled things and partly treated Q4 gain stage as an ideal amp/switch then the resulting time constant would be set only by the 150k resistor and 10uF cap alone, yielding a time constant of 1.5 second - which the pedal clearly does not exhibit ...

... this suggests that the turn-on and turn-off characteritics of that transistor does plays at least a strong role in setting the attack/decay times at the averaging node, which directly sets the bias and transconductance of the OTA ... if we take the expression for ro as being principally derived by the total collector current (which it also does in an common-emitter amplifier stage biased in its linear region) and accept that it is higher in value for lower current levels, then we can see that this should be so in such a circuit biased with less supply voltage ... thus a lower supply voltage would lead to increased non-linear ro averages for Q4 and lead to greater attack and release times, all else remaining much the same, but more so on the attack I suspect ... because of the non-linear nature of these manifestation accurate models and a well setup Transient simulation would be required to ascertain these constants accurately - I'm trying out a new simulator these days, I'll see what it gives using 2n2222 models ...

Bummer. My otherwise stock pedal agrees. Looks like a different version, huh?

... it's interesting that your pedal has the second detector feed connected - it does stick a shunting diode straight across the output and I remember it causing some fuzzing when I inserted the missing 0.01u cap in there in mine ... otherwise I've never come across any instance of dynamic distortion in my pedal ...

... jc

[R.G.]

... the OTA is not operating as an integrator as it is primarily loaded by 150k resistor tied to an AC ground (DC referenced, capacitor bypass - AC short to ground) ... also the ft of small-signal is way beyond the audio range of this circuit ... so all this stuff is irrelevant, there is no gain derived bandwidth variatin anywhere in there - the dominant (low-pass) signal suck in that circuit comes from 0.001u cap wired in parallel with the OTA load ...

I believe that's what I said earlier - the 150K/0.001 and the 15K/0.01 cap set the rolloffs. By the way, when the frequency gets above the 1kHz turnover of 150K/0.001, the OTA does, indeed begin to act like an integrator. Plot the response and look sometime.

I was looking at my hand-drawn schematic from way back, with Bipolar devices instead of FETs,

Good, we got that settled.

and began focusing on the half-wave switch/amplifier stage (in mine only Q4 is active but Q3 is connected as well except for the missing 0.01u cap leading to the base of Q3) ... since there's no emitter degeneration there's no feedback increasing the Rout seen at the collector node ... even though that circuit is operating in a non-linear manner we can extrapolate its "on" characteristics somewhat from linear analysis concepts ... the incremental drive resistance of an emitter follower stage taken at the collector node is given by

rout = ro + (gm * ro) * re ...

where ro, the incremental resistance term tied across its pi-model current source, is approximated by Io/|Va| where Io is the quiescent or DC collector current (not necessarily a stable bias current either) and Va the Early voltage named after James Early if I recall correctly ... these are approximations to circuits biased in their linear region and scuh "constants" largely inapplicable directly in a non-linear context, rather we need to be talking in terms of loads curves and not just zero-crossing quanitities ...

nonetheless, a qualitative analogy is possible based on a per-cycle averaging of non-linear characteristics, and so the absence of emitter degeneration would imply that the per-cycle drive resistance average at that collector be set directly by the device's non-linear drive characteristics directly (ie. not augmented by feedback), and so could conceivably be quite low compared to a device stably biased - lower than customary safe value of 100k for ro in typical small-signal bipolar devices ...

... since the time constant of the averaging circuit is set by the parallel combination of (i) the 150k load resistor, (ii) the non-linear resistance per-cycle average seen looking into the collector of Q4 and (ii) the resistance looking into the base of the voltage-to-current bias converter transistor Q5 (which is much higher than the other two components, ie. Beta times minimum emitter resistance, or 100 x 27k min >> 1Meg min) then we can see the potential for a time constant dominated by the non-linear average of the collector node of Q4 provided it is less than 150k ...

Hmmm... so we're down from the entire circuit changing response because of changed power supply voltage to "a qualitative analogy is possible based on a per-cycle averaging of non-linear characteristics" of the change in collector impedance of *one* of the two *detector* transistors providing for a "potential for a time constant dominated by the non-linear average of the collector node of Q4 provided it is less than 150k"

JC, I think that this is becoming one of those cases where if you can't say your piece in a few sentences, a chapter or two won't help.

What's really going on is that the entire circuit gets power supply starved and starts getting marginal clipping. The clipping is kept marginal by the changing gain of the compressor. Apparently country producers like this. There's blather-all little to bias inside the circuit, prompting my comment. No amount of chasing greenies through the detector circuit is going to change that.

By the way, the 150K/10uF sets the decay response time, not the attack.

[Eb7+9]

the OTA does, indeed begin to act like an integrator. Plot the response and look sometime.

yeah, it only acts as an integrator inside the roll-off band - not in the audio mid-band band where it is flat down to DC and level determined by the transconductance-load product ...

JC, I think that this is becoming one of those cases where if you can't say your piece in a few sentences, a chapter or two won't help.

go read it again ... btw, I'm busy with all kinds of shit - go criticize your own style ...

What's really going on is that the entire circuit gets power supply starved and starts getting marginal clipping. The clipping is kept marginal by the changing gain of the compressor. Apparently country producers like this. There's blather-all little to bias inside the circuit, prompting my comment. No amount of chasing greenies through the detector circuit is going to change that.

nope (... burp!) ... the second detector path that came disabled stock in my pedal is causing fuzzing in yours I strongly suspect - I once inserted a cap in mine to check to see what it would do and bammo, fuzz ... check it out !

By the way, the 150K/10uF sets the decay response time, not the attack.

... only if you idealize the circuit - you're obviously ignoring the gist of what I just wrote ... like I mentioned above this 150k/10u combo gives the circuit a 1.5 second "Tau" time constant, five times that for full 5*Tau relaxation would give 7.5 seconds Mr. multiplication man ! - way slower than the pedal actually reacts ... can't argue with the pedal's pudding !!

Like I said, and we have reached several levels of agreements despite first starting off scurying across the floor like two mad mice on meth, I see now that the only place in the circuit where supply variation dependency can have a real effect is on the drive characteristics of the detector stage(s) which must be taken into account to explain proper decay/attack times or else you end up with incongruent numbers as you're suggesting here ...

Again, like I mentioned, I take back what I said about first suspecting gain variations in some of the blocks I first mentioned (that's when we were on meth) ... like I said, I was approaching all this from first experiencing the pedal on dropped supply - not from exhaustive analysis ...

... anyway, this was fun (thnx ! and watch your manners ...) ... until today I'd never really given that circuit a serious look ... (LOL ...)

... jc

[Jay Doyle]

By the way, the 150K/10uF sets the decay response time, not the attack.

... only if you idealize the circuit - you're obviously ignoring the gist of what I just wrote ... like I mentioned above this 150k/10u combo gives the circuit a 1.5 second "Tau" time constant, five times that for full 5*Tau relaxation would give 7.5 seconds Mr. multiplication man ! - way slower than the pedal actually reacts ... can't argue with the pedal's pudding !!

OK, first off you are an order of magnitude ahead of me in your apparent understanding of EE concepts but little ole' me has to disagree with this. The 150k/10uF sets the decay, idealized or not.

If I can try to sum up what you said above about the drive current in the level detector circuit, you were saying that by changing the power supply voltage, you change the current through the BJTs in the peak detector circuit which therefore changes the output impedance at the collectors. OK, I can see that.

But I cannot see where that would make all that much of a difference in decay. On attack an incoming signal clamps the detector transistors hard to ground and the 10uF cap, which was charged to 9V, discharges through the collector-emitter channels of the two transistors, without emitter resistors this is a very low impedance and the transistors are full on and in parallel. Without a signal, the transistors are off and the channel impedance is (as I understand it) really, really high as it is not really a resistance (as R.G. said) like a FET but a controlled current source. As I've always understood it, current sources have huge resistances/impedances.

Sure, this really, really high impedance in parallel with the 150k resistor may drop the total impedance a little bit but not significantly.

The cap still charges back up through the 150k resistor. There is no other place for it to charge. In between extremes the combination of the 150k and the channel impedance is the "R" in the RC time constant for charging the cap; but that isn't decay as there is a signal present and it is changing the voltage on the cap. It isn't decay until the signal dies down to a point where the cap NEEDS/IS FORCED to recharge as the signal is neither charging nor discharging the cap. At that point, the only thing left to charge it is the 150k resistor.

Also, my Ross as well as a clone I just built can easily decay for 7.5 seconds.

Though I imagine I am wrong somehow, but my pedals decay for at least that long.

Respectfully,

Jay Doyle

[Eb7+9]

... ok, with the sensitivity cranked I can hear that span of time going on there ... at lower it seems a little shorter ... my bad ...

... jc