honeydripper / diphthonger builders - advise

[duck_arse]

I've just completed a diphthonger build, [largely] based on the honeydripper circuit, which is a conversion from CA3080 to LM13700:
circuit here: http://www.madbeanpedals.com/projects/_folders/FilterMod/schematics/HoneyDripper.gif

my circuit works, makes oohhh and ahhh, much as demoed in Pink Jimi's vid. but - it seems to work fine for plunk plunk plunk note picking, but if I plunk and allow to decay, at times the sweep seems to slam shut, just about at the 'aaa' part , I guess where the filters cross over. this is most noticable when the input is fuzz or distortion, and w/ sweep up around ~80%, but it is not everytime. sometimes it bounces open a few times, as tho envelope ripple riding.

note - the sound doesn't cut out, it's just the swept sound that stops sweeping.

this is really quite annoying, especially as I have the bloody thing all buttoned up in one of my never-to-be-opened builds [sheepskin this time], and I'm forced to a stand-by computer so I can't access my actual real-world as-built circuit dia. to reference my opamp subs.

so a question for the others that have built a honeydripper, using the LM13700 - do your builds exhibit this sometimes slamming shut sweep, in all modes, or not?

and a question for the qualified techs - PRR noted recently the difference between the 3080 and the 13700 was an extra Vbe drop - is there a chance that this is manifesting itself in the sweep shutoff?

thanks for reading, thanks for helping.

[edit :] thanks order

[anotherjim]

LM13700 may be slamming into cutoff because of insufficient voltage behind the drive to the control pins. The control current source reference DC bias voltage sometimes needs to be at least 1 silicon diode drop higher than for a CA3080 circuit. An Lm324 or any other op-amp that can output down to the negative supply, is not helpful in this regard.
A cure might be to raise the reference voltage to the op-amp +in pins by a diode if you can.

[blackieNYC]

How would you raise that voltage by a diode drop?  Change D17 to a 6v zener? And lower D16?

[anotherjim]

Put a diode in series with D17, cathode to ground. Although +10v will also go higher to about 10.7v. I think that will only mean there's a little more headroom in the envelope detector and make up for the higher voltage those OT's need to get the control pin biased.

[anotherjim]

Actually, the more I look at it, it might be better if the LM324 negative power pin is raised with one or even 2 diode drops. This to make sure the OTA control won't have less than the required 1.2v on the control pin, or at least not when you want sound. I haven't built or played with this particular circuit, but I don't think you want or need the OTA's to cut off but just sweep the filters.
Alternatively, try swapping out the LM324 for a more normal dual supply qual amp. TL074 is probably gilding the lilly, but you more likely have some? Anyway, a type that won't output to ground and also give more positive sweep than the LM324 in exchange.

BTW -  what's C16 doing? Probably ripple reduction, but is it really directly driven into ground off the opamp?

[PRR]

> what's C16 doing? ....is it really directly driven into ground off the opamp?

That *may* be an allowable scheme for the '324 family. The output stage is quite crude. I seem to recall they showed 0.1u on the output in an app-note, to stabilize a comparator --- or maybe a loop with added gain inside NFB. But TI has scattered all the old LM docs and I can't find that. I did find a new Note which seems to go into great detail about the '324's awful crossover distortion. http://www.ti.com/lit/an/sloa277/sloa277.pdf

[Rob Strand]

FWIW,  the LM324 only swings close to the negative rail *provided* the output current is very low, IIRC 50uA.  You can see it in the datasheet.

The output stage kind of cheats in that it has a standard "high drop" bipolar output stage but it's bypassed by a second low-current current-source stage which can swing lower.  Once you exceed the current source current the output rises to normal bipolar drop levels.

One way to cheat opamp swing is to add a resistor from the output to 0V or +V in the direction you need more swing. 

If you only need one directional drive you can add "level shifting" diode or zeners with pull-down resistors.

Here the opamp goes to the Vin


The LM723 had the level shifter built in so it could drive power transistors in the negative/0V rail,


Jim Williams used the idea a few times.

In some circuits you don't even need a zener you can just use a resistor in series with the opamp output to provide voltage drop,  and depending on specifics no resistor to ground.

[anotherjim]

Most LM324/LM358 datasheet examples show a 10k resistor to the -ve supply rail, but I only found mention of why in an app' note. Since the outputs can source more current than they can sink, the "helper" resistor pulls down (it could pull up, but this way the resistor current doesn't detract from the amps weaker sink capability). I have found explicit mention of these resistors in a Japanese single rail opamp datasheet with the output arranged differently so it can sink more current than it can source - so the resistor is a pull-up to the +ve supply. Boss used some of these types and you can see an apparently useless pull-up resistor in some schemes like the original DS-1.

[Rob Strand]

Most LM324/LM358 datasheet examples

I think the idea works well on the LM324/LM358 because the output stage doesn't have a strong bias current.  That seems to have been done deliberately so their weak current source bypass scheme works.   However, I think it's for this very reason why the LM324/LM358 has so much crossover distortion.

The pull-up pull idea does work to limited degree on all opamps.

IIRC the NE555 output has something funky in the output stage like the LM324; it's actually a little different, the common thing is two pull-down paths.   The output voltage vs current curves have a characteristic step.   It was that step that caught my eye on the LM324.  (That's the only reason I know about their little trick, before that I thought it could swing to the -ve rail.)

[duck_arse]

aquataur has some to say on that cap hereabouts:
https://www.diystompboxes.com/smfforum/index.php?topic=101188.msg954151#msg954151

well, it seems the joke is on me, at least twice, probably more. I have stock of CA3080's, but built w/ 13700, it's modern, y'no? and all the while I was laying out my boards [split vca and dual oppie/ lm324 and two other oppies], I was cursing all those unused pins on the 13700.

I can't give more specifics at the mo, as I'm still without that hard-drive. however - I used NO sockets. also - much fiddle with the spply section. I made DC in 18V, with lm78l09 + ground leg diode for 10V, resistor divider for the 5V.

and soldered the boards together in place, then soldered the in/out/dc sockets in place. bah! bah, I tells ya.

[anotherjim]

...and soldered the boards together in place, then soldered the in/out/dc sockets in place. bah! bah, I tells ya.

Not the infamous Cordwood? 

Anyone else read that TI app' note Paul linked?
http://www.ti.com/lit/an/sloa277/sloa277.pdf
See chapter 7? Ooops 

[duck_arse]

...and soldered the boards together in place, then soldered the in/out/dc sockets in place. bah! bah, I tells ya.

Not the infamous Cordwood? 

Anyone else read that TI app' note Paul linked?
http://www.ti.com/lit/an/sloa277/sloa277.pdf
See chapter 7? Ooops 

no, not the infamous. just straight up, stock standard, two-boards stacked component sides facing, like everyone does. plenty of cordwood builds laying about waiting for boxes, tho.

[duck_arse]

all right. for those still playing along, and I hope you are .... here is what I built, circuit diagram-wise. you'll have to excuse the odd greyscale-ness of the imgs.

https://imgur.com/a/dethbA0

the LM324 and the LM358 are on one board, the audio and power supply are on another.

[pinkjimiphoton]

stephen,
i built both versions of the circuit. the ca3080 one sounds better and quacks better than the lm13700 (i found the 13600 to work better in the last one)

a couple things... make the diodes germanium. that will help the sensitivity. i will try and find my notes later when i hit the dungeon for a couple other notes on it i may have jotted down.

what you described is NOT normal, and i suspect if you follow the audio path you're gonna find a bad solder joint somewhere, or a micro short. audio probe time!! it SOUNDS
from your description like it's not getting enough signal to the envelope follower part. try putting a fuzzface or similar first before the diphthongizer circuit, and swell in and out and see if the problem goes away. if it does, reflow the board and see if the problem goes away.

hope that helps bro!

[duck_arse]

jimi, I was really, really hoping you would come in and say .... oh, now I remember, that's exactly what my 13700 builds did, and why I didn't like them ..... and, to paraphrase a great man, I don't wanna sound like a dick, but I have more confidence in my build being built and working as diagrammed than I have in the translation of the designed-for-3080 circuit to using-a-13700 circuit.

I'm breadbrding the whole mess [slowly] to see if I can replicate the fault, and then if the same happens w/ 3080's. and then finding a fixxe. and then, probably a rebuild. gah.

[pinkjimiphoton]

hey stephen,

it's not bri's fault... what bean did is right on the money, and WILL work, but it doesn't sound/react the same unless it's  JUST THE RIGHT 13700 or 13600. socket that puppy.

when i told brian i was having issues, he went so far as to burn me a couple boards for the original dripper i built, gratis. one hell of a great dude. i DID finally get the 13700 one going....but only with a 13600, so i'm assuming them chips could be mis-marked or forgeries.

they aren't as sensitive as the CA3080 ones tho... i can't remember the part changes, i did have to mess with some values. i replaced them with trimmers, and DO remember had to go to all germanium for the audio parts of the circuit, i tend to use 1n34a or 1n60 cuz i can get them affordably.

the ca3080 version tho is really nite n day.  i don't have any of them here right now, the last one got traded off about 6 months ago.
but i do recall clearly the switching for the formants wasn't the same, and the overall sound was different. the 3080 quacks much easier and tracks better, too. at least in my experience. didn't have to run a fuzz in front of that one.

[duck_arse]

back to aquataur, if I'd read his conversion dissertation here, I wouldn't be in his mess. be that as it may, I havva nutha hypothetical question fer der gurus ....

is there any reason a person re/doing the audio section in 2019 would not use decent opamps, like a TL072 for the input, and maybe a low noise part like LM833 for the filters/outputs [being as those part numbers are to hand]?

and -  given that a person has [seen on the cro and] heard some awful opamp clipping on the swept peaks, is there any good reason to keep the audio/VCA biased at 5V, what with the whole mess running on +18V? I've tried it at 9V, and it werks, just wanting to know I'm not breaking it somewhere else.

------------
I could not get the breadborded 13700 version to fault long enough to track down the problem it was showing [altho the built buttoned up one does it as it likes ..... bastard!], except to say that the rectifiers section did nothing odd. then I pulled it apart and put the 3080's on the same envelope section, and it is chalk and cheese difference. and I haven't heard the fault since.

but much more messing to come, I fear. hiss problems, supply problems, layout problems.

[pinkjimiphoton]

ummmmm.... decent opamps draw more current, maybe? if i remember, i had distortion on peaks if i used jfet kinda opamps. but it's been a while. crummier parts sound better sometimes, and i think the distortion works better innit with non-jfet based amps.

wish i could comment more, but too far out from my builds to really remember bro

[anotherjim]

You could try improving the range of the envelope amps. LM324 can't drive near to the positive supply, in fact it's probably the worst performer at that end, I don't think it can go closer than 1.5v to the positive rail. That said, it's usually possible to produce a circuit that works with whatever voltage swing that's to hand and presumably that was all figured out in the original.
Might be worth trying CMOS TLC2274, since its better for audio too and won't suck too much power, which a TL074 definitely would.

[duck_arse]

update? progress?

is there any reason a person re/doing the audio section in 2019 would not use decent opamps, like a TL072 for the input, and maybe a low noise part like LM833 for the filters/outputs [being as those part numbers are to hand]?

I've spent the last two days at this hissing machine, trying to track down a de-hiss method. my ears! and to answer my question above: hiss. LM833 is as bad as any other oppie for the filters/out section, and the original specced 4558 is the least hiss. which I only found by accident this evening.

while the lesser hisser is a good thing, I, personally, feel the lack of hiss also takes away some of the vowelly information. without all that top, the sound is a little lifeless. but really, what the hell do I know?^**

as far as sweep is concerned, with the 3080's and the audio biased to 9V, I think the original sweep circuit is entierly adequate^**.

I was getting some bad behaviour from the 3080 circuit today, but I think it has gone again, without being observed in action.