Funking up an FX-25

[Mark Hammer]

I recently re-established contact with a friend I used to play in a band with when we were both 4 decades younger.  He's been a working musician in the interim and is an outstanding bass player.  He's coming to town in a month to play a local club and I wanted to give him an envelope controlled filter that was optimized for bass, since he had expressed a desire for one.

Not having a lot of time to spare, because of other commitments, I thought I'd take something I already had in duplicate, and make it more suitable for bass, and especially for someone who looks forward to being able to pop like crazy with his thumb and channel his inner Larry Graham.

The DOD FX-25 is a decent sounding flter, but is more guitar-oriented, so I decided to do something about that.  Using this drawing as a reference, note the following:
The normal output, taken from pin 8 of the LM13600, is the bandpass output of a kind of state-variable filter.  By taking the output from pin 9 instead, we actually end up with a lowpass filter, which is far more suitable for bass since you never lose your bottom.  Note that this will be louder than the usual bandpass output, making the effect/bypass balance all wrong.  To address this, R22 is replaced with a 50k/47k pot and 47k-51k fixed resistor between ground and the ground lug of the pot, to turn down the output a bit.  You don't want it quite, so the fixed resistor addresses that.  If you want to make effect louder than bypass, or use the volume to restore level for bandpass, you can do that.

Resonance is set by the combined series resistance of P2 and R15.  The higher their combined resistance, the higher the resonance.  The filter has the interesting behaviour that increasing resonance also shifts where the filter centre frequency is.  I replaced R15 (22k) with a 30k resistor, and may well replace that with a 33k unit.  The shift in range was sufficient to not require any change in the two filter caps, C9 and C10.  At max resonance, the unit now has a more vocal sound, and is more suited to bass.

D2 and D3 (but D3 in particular) do the rectification to obtain an envelope signal.  The stock diodes have a forward voltage of around 500-600mv, which means that unless the rectified signal is greater than that, nothing is fed to the filter.  Since slapping and thumb-popping does not necessarily get you the same amplitude as a full string pluck, simply turning up the sensitivity of IC1a is not necessarily the answer.  By replacing D3 with a germanium diode having a forward voltage of 250-300mv, I made it possible for subtler "string actuation" to have an effect, and make the unit appear more responsive.

Lastly, slow decay is great for playing a chord, but lousy for bass popping.  I replaced C7 with a 10uf cap and a 330k fixed resistor in parallel.  That provides a slightly faster attack than stock, and drains off the stored charge in the cap quickly.  I imagine one could use a 100k fixed resistor and 500k pot in series, or pehaps a 3-position toggle to select resistance values, if they wanted to have variable decay.  In this instance, I wanted it to just play nice out of the box, and not require any learning whatsoever.

Total parts involved:  1N34a (or similar) diode, 10uf cap, 330k resistor (I used 1/8w soldered to the copped side), SPDT toggle (for bandpass/lowpass), 50k pot, 47k resistor.

I'm no bass player, and these mods do not move the FX-25 to the front of the filter pack, but trying to fake some things along my E and A strings provided pleasing results without the involvement of making a Mu-tron.  The board fits nicely into a 1590B.  I hope it pleases my buddy just as much as I think it will.  Thanks to Alex Petrini for the very helpful drawing and layout.

[StephenGiles]

I can't seem to see the drawing Mark.

[chicago_mike]

drawing doesnt show up for me either. 

[aron]

I really like the FX-25. I have had a number of auto filters, but this one works for me.

[Mark Hammer]

It should show up if you go here:  http://topopiccione.atspace.com/PJ11DODfx25.html

[PRR]

Useful analysis, techical and musical.

Especially the point about tenors using bandpass and bass liking response all the way down.

> replaced C7 with a 10uf cap and a 330k

The 22u-10u directly gives the faster decay. The 330K may do little, because the decay seems to be dominated by R12 R13 10K into their diodes. And I -think- if "330K" is reduced much below 100K (say 60K) then the OTAs just turn off? No sound comes out? R9 1Meg flows a minimum current which a heavy decay resistor would steal? The simple way to change decay time is changing C7.

The diodes in series with those 10K resistors will rise in impedance as you return to idle. So a small decay resistor "might" smooth/stretch the tail of the decay. Said resistor might ideally return to a small positive bias (0.5V) rather than ground.... the difference may be inaudible.

Both D2 and D3 deduct from rectified voltage. Changing D2 to Ge or Schottky might get a wee bit "appear more responsive" effect. And given free time, a dual-rate time network could decay fast after pluck and slow after chord. AYK, there's a million "better" rectifier and filter designs, but a bazilion other commitments.

The C4 R6 hi-pass is 1KHz, right? And C5 R7 P1 will often be near. I'm surprised that works for bass. I guess you do want a hi-pass, you don't want the filter to open-up if there's no highs to bring out.
> drawing doesnt show up

The Topopiccione webserver apparently know when an image request comes directly from a "strange" site, and declines to deliver the image. Open the link in Mark's second post and scroll down to the link for "DOD FX-25 schematic".

[Mark Hammer]

Useful analysis, techical and musical.

Especially the point about tenors using bandpass and bass liking response all the way down.

Thanks!

> replaced C7 with a 10uf cap and a 330k

The 22u-10u directly gives the faster decay. The 330K may do little, because the decay seems to be dominated by R12 R13 10K into their diodes. And I -think- if "330K" is reduced much below 100K (say 60K) then the OTAs just turn off? No sound comes out? R9 1Meg flows a minimum current which a heavy decay resistor would steal? The simple way to change decay time is changing C7.

The diodes in series with those 10K resistors will rise in impedance as you return to idle. So a small decay resistor "might" smooth/stretch the tail of the decay. Said resistor might ideally return to a small positive bias (0.5V) rather than ground.... the difference may be inaudible.

I was under the impression that R12/R13 would limit the current going to their respective pins, such that the 13600 wouldn't be burnt out.

The R9/C7 pair, as I understand it, are somewhat like the 150k/1uf pair in the Dynacomp, in that R9 charges up C7 and the envelope detracts from it.  Thinking it over, maybe my preferred route is to do what the various current crop of compressors do and speed up the rate at which the cap recharges by reducing R9 a bit, instead of using a resistor in parallel with C7.  I'll carry out that experiment and report back this evening, barring distractions.

Both D2 and D3 deduct from rectified voltage. Changing D2 to Ge or Schottky might get a wee bit "appear more responsive" effect. And given free time, a dual-rate time network could decay fast after pluck and slow after chord. AYK, there's a million "better" rectifier and filter designs, but a bazilion other commitments.

You got it.  Q & D is the order of the day here.  I also wanted to be able to use a board I already had made and wired up.  Desoldering a component and replacing makes it a simpler exercise with faster results.

The C4 R6 hi-pass is 1KHz, right? And C5 R7 P1 will often be near. I'm surprised that works for bass. I guess you do want a hi-pass, you don't want the filter to open-up if there's no highs to bring out.

You know, that would have been my first thought too, but then when I did a similar calculation on another noninverting stage several years ago, and expressed surprised the circuit could even work with such a high bass rolloff, several more educated members here drew my attention to the fact that calculations for caps such as C4 are predicasted on certain ideal assumptions about the impedance of the source, and do not turn out to be what a simple calculation involving only C4 and R6 would predict.  So, AFAIC, all bets are off.  I won't even pretend to know what the actual rolloff is.  I will note, though, that at maximum sensitivity, the calculated bass rolloff of IC1a, as set by C5, is 338.6hz, and at min sensitivity its 15hz.  Were C4/R6 to result in the rolloff you predict, that would kind of make the value of C5 rather moot.

Thanks for a useful post.

[Mark Hammer]

Okay, confirmed.  Reducing the 1M resistor produces a faster decay.  I dropped it to 390k.

[PRR]

> calculations for caps such as C4 are predicasted on certain ideal assumptions

Yes, you must account the source and in many cases it is significant.

But here we get the direct output of IC1b, ~~1-ohm, insignificant compared to 150K.

I say the rectifier just gets "squeaks". You say it plays well. Can't argue with results.

> value of C5 rather moot.

My fingers were tired so I only guesstimated that C5 was "don't care" with C4 R7 as they are now. If you re-visit this (4 decades from now?) and decide that "more thud" in the sidechain is worth a try, first C4 and then C5 would be upped.

I don't doubt that a low-cut is useful here. Probably I underestimate how much highs come off a well-slapped/snapped bass string.

> R12/R13 would limit the current going to their respective pins, such that the 13600 wouldn't be burnt out.

Sure, that's part of the job.

They also equal-split the current from R9; and convert rectifier voltage to two equal currents.

Take the rectifier away. 9V flows through 1Meg to two 10K to two LM13600 bias-diodes to ground. The current in the bias-diode sets that OTA's current, which sets gain. "Gain" here is infinite at DC and teeny at high frequency; then the R20 NFB loop sets overall gain flat (nominally R20/R23 = 2.2) from DC to some audio frequency, dropping above that, double-dropping so that the corner peaks-up damped by P2. (Or if you take the bandpass output, a peak sloping down at high and low frequency.)

So 8.4V across 1Meg is 8uA total.

This splits two ways R12 R13. If we simply tied pins 1 and 16 together, small mismatch between diodes would un-equal split; 10K each swamps the diode impedances well enough. So 8uA total is 4uA per OTA.

Now put back the rectifier.

With no signal, D3 blocks, still the same OTA bias current.

Say BIG signal. Charge C7 up past 5V, say 5.7V. R9 current drops from 8uA to 4uA. However the R12 R13 currents rise to 5V/10K= 0.5mA or 500uA each OTA. The OTA gain, and integrator gain-bandwidth, rise by factor 500uA/4uA. That's enough to shift the peak two decades. I dunno what the design peak was/is, but 80Hz to 8,000Hz would change sound from pure thud to pure tizz (or as low-pass, from pure thud to everything with extra tizz).

So I see R9's job as setting the minimum bandwidth at no-signal.

R12 R13 set the OTA current when rectifier raises C7; are also the primary discharge for C7. A sloppy analysis says 10uFd+10K is a 100mS decay, which is in a reasonable range for per-note music massaging. The "true" decay is not simple; if you 'scope C7 you see it drop fast to ~~0.6V and then slower, and exponential whereas the OTA+10K is a linear function.

> faster decay.  I dropped it to 390k.

The time-change is hardly different. I think what 390K does is sets the idle bias, and thus the idle peak-freq, a little higher, and the ear hears it "get there, bottom-out" sooner. But my analysis does not totally convince me, and the ear often hears complex change very different than a technical study suggests.

> Q & D is the order of the day here.  ...a board I already had made

Understood. I was just taking a break from salting the ice and hanging the curtains, and your post gave me something to chew on.

[petemoore]

  FX-25 in Pin 9 Bass Attack Mode, cool stuff, what you want is what you need, anything else just gets in the way.

[Mark Hammer]

> calculations for caps such as C4 are predicasted on certain ideal assumptions

Yes, you must account the source and in many cases it is significant.

But here we get the direct output of IC1b, ~~1-ohm, insignificant compared to 150K.

I say the rectifier just gets "squeaks". You say it plays well. Can't argue with results.

> value of C5 rather moot.

My fingers were tired so I only guesstimated that C5 was "don't care" with C4 R7 as they are now. If you re-visit this (4 decades from now?) and decide that "more thud" in the sidechain is worth a try, first C4 and then C5 would be upped.

I don't doubt that a low-cut is useful here. Probably I underestimate how much highs come off a well-slapped/snapped bass string.

> R12/R13 would limit the current going to their respective pins, such that the 13600 wouldn't be burnt out.

Sure, that's part of the job.

They also equal-split the current from R9; and convert rectifier voltage to two equal currents.

Take the rectifier away. 9V flows through 1Meg to two 10K to two LM13600 bias-diodes to ground. The current in the bias-diode sets that OTA's current, which sets gain. "Gain" here is infinite at DC and teeny at high frequency; then the R20 NFB loop sets overall gain flat (nominally R20/R23 = 2.2) from DC to some audio frequency, dropping above that, double-dropping so that the corner peaks-up damped by P2. (Or if you take the bandpass output, a peak sloping down at high and low frequency.)

So 8.4V across 1Meg is 8uA total.

This splits two ways R12 R13. If we simply tied pins 1 and 16 together, small mismatch between diodes would un-equal split; 10K each swamps the diode impedances well enough. So 8uA total is 4uA per OTA.

Now put back the rectifier.

With no signal, D3 blocks, still the same OTA bias current.

Say BIG signal. Charge C7 up past 5V, say 5.7V. R9 current drops from 8uA to 4uA. However the R12 R13 currents rise to 5V/10K= 0.5mA or 500uA each OTA. The OTA gain, and integrator gain-bandwidth, rise by factor 500uA/4uA. That's enough to shift the peak two decades. I dunno what the design peak was/is, but 80Hz to 8,000Hz would change sound from pure thud to pure tizz (or as low-pass, from pure thud to everything with extra tizz).

So I see R9's job as setting the minimum bandwidth at no-signal.

R12 R13 set the OTA current when rectifier raises C7; are also the primary discharge for C7. A sloppy analysis says 10uFd+10K is a 100mS decay, which is in a reasonable range for per-note music massaging. The "true" decay is not simple; if you 'scope C7 you see it drop fast to ~~0.6V and then slower, and exponential whereas the OTA+10K is a linear function.

> faster decay.  I dropped it to 390k.

The time-change is hardly different. I think what 390K does is sets the idle bias, and thus the idle peak-freq, a little higher, and the ear hears it "get there, bottom-out" sooner. But my analysis does not totally convince me, and the ear often hears complex change very different than a technical study suggests.

> Q & D is the order of the day here.  ...a board I already had made

Understood. I was just taking a break from salting the ice and hanging the curtains, and your post gave me something to chew on.

I appreciate the thought put into your response.  Much for me to chew on.  I have to confess a great deal of naiveté with respect to the workings of OTAs.

[PRR]

> naiveté with respect to the workings of OTAs.

They are a very odd part. And what I once understood is slipping. And although after a while I could bend my head around an OTA, explaining is tougher.

Here's an essay to chew: http://www.uni-bonn.de/~uzs159/ota3080.html

How are you on Basic Grounded Emitter BJT stage? And concept of transconductance (Gm)?

Although I prefer to think of 1/Gm as "apparent emitter resistance". Because I see voltage gain as ratio of collector to emitter resistance.

The raw BJT has high Gm, which means a small emitter resistance. And it is proportional to current. About 30 ohms at 1mA. About 30,000 ohms at 1uA.

Note one big problem. If emitter resistance is 30 ohms, and we cause a +/-1mA signal swing, to 2mA and zero mA, the Gm doubles on one side and falls to zero on the other side. Distortion is gross. It only takes 30mV to cause a 1mA swing across 30 ohms. So our maximum clean input is somewhat less than 30mV. We see this in fuzzboxes. Grounded emitter with no signal attenuation fed a 20mV to 500mV guitar signal fuzzes all the time. We'll usually want some added emitter resistance or other loss so we have a range from clean to fuzz.

Ignore the problems of DC bias and swing for a moment. If the load in the collector circuit were 30K, then at 1mA the gain would be 1,000 and at 1uA it would be unity. Or if the load were a pure capacitor, the gain would drop with frequency, but at any given frequency the gain would be 1,000 times higher at 1mA as at 1uA.

If we just have one BJT and try to use it as a variable-gain amplifier, 1uA to 1mA, the collector voltage goes all over the place. Also most methods of influencing the DC bias current require mixing into the Base which is also our audio port.

The RCA 3080 (and the 13600 which adds a buffer) uses two BJTs as the Gm devices, a current mirror to set the idle current without messing the input, and three current mirrors to combine the two BJT outputs and get to an output point.

http://www.idea2ic.com/LM13600/LM13700.pdf page 3
(this is LM13700 which adds yet another frill.)

If you set a fixed idle current, it is just an op-amp, except it has an "infinite" output impedance. If you simply closed a NFB loop around it, the output impedance becomes "zero" and voltage gain is about what the NFB says. But it's not a great opamp. Very load sensitive, and not a lot of gain.

What is interesting is that by loading it, no NFB, and changing the idle current, we can adjust gain over a very wide range. And the gain is Zl/Re, where Re is Q4 Q5 1/Gm and Zl is whatever load you hang on it.

Figure 11 shows something like you have. The OTA itself is loaded with 150p cap. Gain falls with frequency. The R Ra NFB loop tries to force a flat gain, but when OTA-cap gain falls too low, overall gain falls. Low-pass. Gain also falls as you reduce current. Now the R Ra NFB loop loses control at a lower frequency. Variable low-pass.

You set the gain by a current flowed into pins 1 16 "Amp Bias". The voltage-drop from Amp Bias to the V- rail is semi-fixed, diode drop. And  my memory is failed: it is 2 diode drops, ~~1.4V (not 0.6V as I was misremembering).

For fixed bias you just run a resistor up from the Amp Bias port to a positive voltage. Fig 11 shows 30K to the far side of +/-15V supplies, nearly 1mA, roughly the maximum safe current for the part. For a simple manual bias-change, we could rig a pot across the rails, connect the 30K to the wiper. Or to idle at low current and push to high current, we could run a 1Meg from a positive rail and maybe 10K from a 0 to 10V source referenced to V-. All these voltage-based bias schemes fade as the voltage approaches the 1.4V of the Amp Bias port diodes. You can rig current-source to force precision. Or in this case it seems the designer used the 1.4V threshold so that small <1.3V inputs caused no shift, then a larger >1.5V input swings significantly.

Back to the ~~30mV input distortion. FX-25 has those R23 R10 resistors which are an 11:1 pad. So things are keen up to about 330mV peak input, which covers many pickup signals. And because of symmetry, the 13600 doesn't get to a gross 10% THD until 100mV peak input. Of course taking an 11:1 cut then gaining-up again puts you close to noise. Selecting this input pad is a key detail in OTA use, tradeoff between hiss and distortion. The 13700 version adds pre-distort diodes which can keep you under 1% THD to fairly high level, but then THD rises abruptly; no free lunch.

And speaking of lunch, I need to get up before noon, so I better quit here.

[rankot]

If anyone comes to this discussion looking for schematic, here it is:

[Fancy Lime]

Ha! I was worried for a second that Marks mind was going because I remembered that he posted the same mod years ago. Only to realize that this IS the post from years ago. Seems I'm the one loosing it. Well, at least I got a laugh out of my dysfunctional brain.

BTW Mark, what did your Buddy think?

Andy

[Rob Strand]

Ha! I was worried for a second that Marks mind was going because I remembered that he posted the same mod years ago. Only to realize that this IS the post from years ago. Seems I'm the one loosing it. Well, at least I got a laugh out of my dysfunctional brain.

Don't worry we all do it with these old threads.   It's funny when you read down the thread then hit a post you recognize, or perhaps even see one of your own posts 

[PRR]

....If you re-visit this (4 decades from now?)....

{January 13, 2010} Almost three decades to go........

[rankot]

{January 13, 2010} Almost three decades to go........

Right, but I'm building one right now

And I think it's much better to refresh those old threads then to scatter information in zillion different ones.

[Mark Hammer]

Ha! I was worried for a second that Marks mind was going because I remembered that he posted the same mod years ago. Only to realize that this IS the post from years ago. Seems I'm the one loosing it. Well, at least I got a laugh out of my dysfunctional brain.

BTW Mark, what did your Buddy think?

Andy

I think he liked it, but it's ten years later, and he's retired now.  But we keep in touch.  I'll have to ask him again.