Mark Hammer's Flatline/Punchline (compressor/expander) mods issues

[alparent]

I think I found mt problem!

I built the Flatline with the Punchline mods as suggested by Mark on is site.

On the compressor side all was OK but when I switched to the expander...nothing?
I checked my build time and time again .... but as soon as I switched to the expander ...... the LED (of my LED/LDR) stopped flashing?

After starring at the layout, for I don't know how long, to figure our what I was doing wrong, it hit me!

If you look at this layout http://hammer.ampage.org/files/Punchline.PDF

If you look at the blue cut line, you're not just cutting the LDR leg from R5 but you are also cutting the Neg. leg of C2. And I think this is the problem? Please correct me if I'm wrong.
I think R5 and C2 should always be connected. Only the LDR leg should be toggled between R5 and R4.

Mark also explained the gain range issue between the compression and expansion.

Keep in mind that the amount of combined parallel resistance change required to achieve compression and expansion will be different.  I may have made it sound too easy in the "Punchline" document.

The stock circuit has a 220k feedback resistance and a 10k-110k variable ground leg.  At maximum ground leg resistance, we have a gain of 3x, and at minimum we have a potential gain of 23x.  If the LDR in parallel with the 220k has a dark value of, say, 2megohms, then changes in the LDR resistance will have an impact in their total parallel resistance, and the gain that comes out of that.  With gain pot set to max (R=0 ohms):

When LDR = 2M, parallel resistance = 198k; gain = 21x
When LDR = 1M, parallel resistance = 180k, gain = 19x
When LDR = 500k, parallel resistance = 153k, gain = 16x
When LDR = 250k, parallel resistance = 117k, gain = 13x
When LDR = 100k, parallel resistance = 69k, gain = 8x

And so on.

Place the same LDR in parallel with the 10k ground leg resistor (so 220k feedback resistor works on its own) and:

When LDR = 2M, parallel resistance = 10k; gain = 23x
When LDR = 1M, parallel resistance = 9.9k, gain = 23x
When LDR = 500k, parallel resistance = 9.8k, gain = 23.2x
When LDR = 250k, parallel resistance = 9.6k, gain = 24x
When LDR = 100k, parallel resistance = 9.1k, gain = 25.2x

So, with max gain and the same LDR, you get less change in gain in expansion mode than you do in compression mode.

I think I understand the what's what......But what can I do about it?  

[Mark Hammer]

Gah!! You are absoutely correct.  Talk about slow on the uptake.  I had an error for over 10 years and never noticed! 

[Mark Hammer]

Thinking about it some more, perhaps one could steal an idea from the EA tremolo, and use a single transistor variable-gain stage for the expansion part.

So, instead of the C2 output feeding the volume pot directly, it feeds a bipolar or JFET/MosFet gain stage that has a bypass cap from emitter/drain to ground (like this: http://www.generalguitargadgets.com/diagrams/eatrem_sc.gif ).  The gain of that stage is partly dictated by resistance of the path the electrolytic/bypass cap takes to ground.  Make the resistance smaller than the emitter-to-ground resistance, and you increase gain.  Harness that  resistance to an envelope follower and you can use it to accentuate the dynamics.

[YouAre]

Blast from the past. Hey Mark! It's Murad. (I was the younger Middle Eastern looking guy from the Small Bear Brooklyn meetup in the fall)

I'm Building MadBean's Afterlife Compressor (Based on Flatline). Link below:
http://madbeanpedals.com/projects/Afterlife/Afterlife.pdf

I want to perform the Punchline mods to this, and observed the documented lack of dynamic range if we were to stick with the stock resistor values. What if we up'ed the series and shunt resistances to account for the parallel resistance issue present with the stock values? I understand at this point, it would no longer function as a compressor with the LDR in parallel with the series resistor. This is assuming we'd only want "Expansion," and compression isn't required. Switching resistor values can be another exercise for another day.

Using the formatting from the post above, if we change the resistances for R3 and R4 (Referencing Afterlife schematic) from 220k and 10K (Respectively) to 4Meg and 220K, we would get the following.

When LDR = 2M, parallel resistance = 198k; gain = 21
When LDR = 1M, parallel resistance = 180k, gain = 23x
When LDR = 500k, parallel resistance = 153k, gain = 27x
When LDR = 250k, parallel resistance = 117k, gain = 35x
When LDR = 100k, parallel resistance = 69k, gain = 59x

That might be TOO much range, so I'm sure we can alter the gain of the envelope detector to play in a nicer range. That math is easy enough. Do you foresee any problems with this operation?

Thanks for the help!

[Mark Hammer]

Hi Murad!

I guess the starting point is to ask what LED/LDR combination you're using and whether it is appropriate for the task at hand.  Where an LDR with a bit of sleepiness to it can be useful for compression, one generally wants a somewhat faster speed for expansion.

Actually, let me say that differently.  Faster OR slower LDRs can be useful for compression, but sluggish LDRs suck for expeansion.  You may also want to reduce the value of C3, or at least experiment with smaller values for a speedier response.

As you probably know, the gain of IC1a depends on the joint value of the feedback resistance (R3 and LDR) and the ground leg resistance (R4 + pot fraction).  For a sudden increase in gain, you can either increase the feedback, or decrease the ground leg.  Since the circuit, as shown/designed, is built around the assumptionthat harder picking will result in a reduction in LDR resistance, I guess that sort of makes the decision for us.

But it doesn't make all the decisions.

It is, after all, the ratio of those two general resistances that sets the gain.  So, for me, the challenge is where to stick any parallel or variable resistances, such that a useful quick change in gain can be produced.

Let's say your LDR ranges from 50k to 5k, at full picking strength and circuit sensitivity.  With R3=220k, and the LDR inserted where the 100k Sustain pot currently is, your gain will range from (220k+50k+10k)/(50k+10k) = 4.7x to (220k+10k+10k)/(10k+10k) = 12x.  Is that too much, not enough, just right?  I honestly have no real idea, but I suspect it is a bit much, especially if the response isn't super fast.  If we were to replace R4 with either a pot, or maybe a 3-position toggle to pick one of 3 ranges, where R4 was, say, 47k, 27k and 15k (47k default, with 62k or 22k added in parallel), we would have potential gain changes/contrasts of  4.4-9.8, 3.9-6.9, and 3.3-4.9. 

Let's ignore the specific absolute gains (because you DO have a volume pot, after all), and translate this sudden gain change into ratios.  These represent ratios of 2.2:1, 1.8:1, and 1.5:1.  That is, the output signal , in each of the 3 toggle positions, would either be 1.5 times, 1.8 times, or 2.2 times louder than what you fed in.  The absolute level, of course, would vary with switch position, but you can compensate for that with volume.  Note that compander chips typically apply a 2:1 ratio, so this gets us a little more expansion, and two shades of less expansion, than a compander chip provides, which seems about right, now that I'm thinking about it in those terms.

Of course, all these calculation are contingent on the described hypothetical properties of the LDR.  Depending on what LDR you are using, the same calculations could be done, and adjustments made to R3, R4, values in paralle with R4, and possibly even something in parallel with the LDR.  Again, the goal is to use the known properties of the LDR at fulldark and full bright, to calculate how much gain change it will yield.

Hope that's not too convoluted.

[tubegeek]

(I was the younger Middle Eastern looking guy from the Small Bear Brooklyn meetup in the fall)

Ahem. Younger than whom, exactly?

[YouAre]

Ahem. Younger than whom, exactly?

Younger than most of the gear you work on 

Hi Murad!

I guess the starting point is to ask what LED/LDR combination you're using and whether it is appropriate for the task at hand.  Where an LDR with a bit of sleepiness to it can be useful for compression, one generally wants a somewhat faster speed for expansion.

Actually, let me say that differently.  Faster OR slower LDRs can be useful for compression, but sluggish LDRs suck for expeansion.  You may also want to reduce the value of C3, or at least experiment with smaller values for a speedier response.

Noted! I had not even thought of that. I understand that a slow acting compressor will have somewhat of a swagger when clamping a gained up signal (which should have more sustain). But for slow acting expansion seems less useful because it's beefing up a signal after it's already begun to decay. Won't sound natural. I will use this tidbit to select an appropriate LDR or compensate with the envelope detector's cap as necessary.

As you probably know, the gain of IC1a depends on the joint value of the feedback resistance (R3 and LDR) and the ground leg resistance (R4 + pot fraction).  For a sudden increase in gain, you can either increase the feedback, or decrease the ground leg.  Since the circuit, as shown/designed, is built around the assumptionthat harder picking will result in a reduction in LDR resistance, I guess that sort of makes the decision for us.

But it doesn't make all the decisions.

It is, after all, the ratio of those two general resistances that sets the gain.  So, for me, the challenge is where to stick any parallel or variable resistances, such that a useful quick change in gain can be produced.

Well, assuming we're using the LDR described above, we're kind of stuck using it in parallel with the shunt resistor. I can't think of any arrangement with the series resistor where a reduction in brightness with increased attack would be useful for this exercise.

Let's say your LDR ranges from 50k to 5k, at full picking strength and circuit sensitivity.  With R3=220k, and the LDR inserted where the 100k Sustain pot currently is, your gain will range from (220k+50k+10k)/(50k+10k) = 4.7x to (220k+10k+10k)/(10k+10k) = 12x.  Is that too much, not enough, just right?  I honestly have no real idea, but I suspect it is a bit much, especially if the response isn't super fast.  If we were to replace R4 with either a pot, or maybe a 3-position toggle to pick one of 3 ranges, where R4 was, say, 47k, 27k and 15k (47k default, with 62k or 22k added in parallel), we would have potential gain changes/contrasts of  4.4-9.8, 3.9-6.9, and 3.3-4.9. 

Let's ignore the specific absolute gains (because you DO have a volume pot, after all), and translate this sudden gain change into ratios.  These represent ratios of 2.2:1, 1.8:1, and 1.5:1.  That is, the output signal , in each of the 3 toggle positions, would either be 1.5 times, 1.8 times, or 2.2 times louder than what you fed in.  The absolute level, of course, would vary with switch position, but you can compensate for that with volume.  Note that compander chips typically apply a 2:1 ratio, so this gets us a little more expansion, and two shades of less expansion, than a compander chip provides, which seems about right, now that I'm thinking about it in those terms.

I don't think we can ignore the absolute gains, because really high gains will likely smack into the rails. Not pretty. Since we don't have clipping diodes to clamp that signal we want our gains to be more "reasonable" than that of a non inverting clipping stage of a standard overdrive pedal

Also, if we have TOO much of a gain range, that volume pot would make the pedal unusable. By attenuating the super high gains (50? Let's just use that as a reference), we're also making the "standard gain" (20?) inaudible. I had not thought about this before until you shared your insight. Thank you.

Of course, all these calculation are contingent on the described hypothetical properties of the LDR.  Depending on what LDR you are using, the same calculations could be done, and adjustments made to R3, R4, values in paralle with R4, and possibly even something in parallel with the LDR.  Again, the goal is to use the known properties of the LDR at fulldark and full bright, to calculate how much gain change it will yield.

Hope that's not too convoluted.

I will be certain to recalculate these resistances utilizing a faster LDR. Thank you for this insight. You're always perfect to bounces ideas with.



Does anyone see any issues with using rather large resistance values in a non inverting op amp stage? I'm not too familiar with the concept of thermal noise from large resistors, but is there going to be an issues with using R3 as 4MEG and R4 as 220K? The gain isn't that high, but the resistance values are rather large.

[Mark Hammer]

Ahem. Younger than whom, exactly?

Younger than most of the gear you work on 

BAZINGA! 

[Mark Hammer]

Does anyone see any issues with using rather large resistance values in a non inverting op amp stage? I'm not too familiar with the concept of thermal noise from large resistors, but is there going to be an issues with using R3 as 4MEG and R4 as 220K? The gain isn't that high, but the resistance values are rather large.

See page 8 here: http://hammer.ampage.org/files/Device1-8.PDF

[YouAre]

Does anyone see any issues with using rather large resistance values in a non inverting op amp stage? I'm not too familiar with the concept of thermal noise from large resistors, but is there going to be an issues with using R3 as 4MEG and R4 as 220K? The gain isn't that high, but the resistance values are rather large.

See page 8 here: http://hammer.ampage.org/files/Device1-8.PDF

OK, so the indicated page in the referenced article details how to attenuate high end using the feedback capacitor to limit high end. Being that there is no capacitor in the original design, it may not affect us? I understand that with a 4MEG series resistor, that there's no common capacitor that we can use to get a usable high end roll off. (a 10pF cap would give us a 4KHz roll off). It's a little too low.

What about "thermal noise?"  I understand we will have more at these higher resistance levels, but I'm still not familiar with what it "looks" like. Is it something that can be filtered out with a simple Lowpass tuned to like...10KHz?

[Mark Hammer]

Actually, it's the article starting on page 8.  The content on page 10 shows how noise varies, depending on the combination of op-amp type and associated input resistances,

[YouAre]

Actually, it's the article starting on page 8.  The content on page 10 shows how noise varies, depending on the combination of op-amp type and associated input resistances,

Interesting, so the main effect of thermal noise isn't really related to the feedback but to the biasing resistor...So we have a nice balance between high input impedance and thermal noise.

[Mark Hammer]

Which is why you have to pick your chip based on the surrounding resistances, or base your surrounding resistances based on the chip you decided on (for whatever reasons those might be).

[YouAre]

Which is why you have to pick your chip based on the surrounding resistances, or base your surrounding resistances based on the chip you decided on (for whatever reasons those might be).

Mark Hammer, infinitely insightful.

Thank you for never giving me the fish, but teaching me how to fish...by fish hooking me and dragging me to the river. 

[tubegeek]

Ahem. Younger than whom, exactly?

Younger than most of the gear you work on 

BAZINGA! 

Sure, Mark, encourage some young turks to turn on your, shall we say, "experienced," fellows. What could possibly go wrong with THAT I wonder? For you? In overalls?
(sob) ... I thought you'd have my back, man ... (sob)...

[Mark Hammer]

Jeez louise, Jeremy!  Can't a guy enjoy a cage match every once in a while? 

[tubegeek]

Jeez louise, Jeremy!  Can't a guy enjoy a cage match every once in a while? 

A reasonable question, I must admit.