help me understand the tube screamer / SD-1 / DS-1 front buffer? opinions?

[darron]

i always assumed that the buffer at the font of boutique time screamers was leftovers that got included from the original active switching circuit.

so i made an overdrive with just a TL072 and no input/output buffer, like on GEOFEX: http://www.geofex.com/article_folders/TStech/tsxfram.htm

later i breadboarded this with the front buffer and used a switch to blind test bypassing it. to my ears i could hear a small, nice difference with the buffer in.  nothing HUGE but worthwhile to keep in. I used massive coupling caps that would pass all frequencies.

so am i wrong? did i trick myself?

1) the TL072 has a JFET buffer at the front of each input already.

2) i thought that using the non-inverting input of an op amp works nicely as a buffer, as everybody does this in boutique buff+boost pedals. maybe at unity gain though, not at gains in the hundreds?


so i'm thinking of building an overdrive using a quad op amp to replace the transistors at the front and back. but it kind of feels wrong/redundant to me running one opamp directly into the next stage.





also, why is there a decoupling cap and new bias resistor between stages? can't the first transistor just feed the op amp directly? IE the same way as the clipping op amp feeds into the tone control op amp. does this help with noise decoupling between? any expect opinions / thoughts? i never bothered much with overdrive, in 12 years, because there were so many nice designs already and didn't want to reinvent the wheel.


thanks

[PRR]

> why is there a decoupling cap and new bias resistor between stages? can't the first transistor just feed the op amp directly?

http://www.geofex.com/article_folders/TStech/tsxtech.gif

Perhaps it could.

But is the JFET output exactly/close-enough 4.5V? No, there is clearly a 0.6V drop from 4.5V, and more from the base bias resistor. How much Rb drop? Hmmm.... can't say. It depends on Q1 hFE. If I score one with hFE near 1,000, about 0.2V. If my cheapskate Purchasing Department buys a crate of hFE<50 semi-rejects, it could be over 2V drop from Vref.

An op-amp WILL follow exactly (with 500K bias you want a FET opamp). However this scheme probably dates from days when transistors were cheaper than chips (at least if your Purchasing beat the scrap bins for non-dead material). Two TL072 is now as cheap as jellybean NPNs (with resistors caps and joints) and may be a better plan. (TL074 was always too many pins and too much gain in one small lump, unless you just wanted a buffer-array.)

> kind of feels wrong/redundant to me running one opamp directly into the next stage.

The guitar could drive the clipper directly; except when the clipper is working HARD even its input will be glitchy. That means you can't get an "unmolested" no-effect signal without a buffer somewhere in front, or more switching.

The Level pot is 100K so has <25K impedance at the wiper. Many pedals have more. However it still seems good practice to buffer the output. Especially with the JFET switching.

The 0.1u and one of the 510K at Q2 could be omitted, either transistor or opamp. This may be relic of design changes not totally re-factored for minimum cost.

[darron]

thanks PRR for the in-depth answer. i thought that maybe everybody had disappeared over the silly season.

The 0.1u and one of the 510K at Q2 could be omitted, either transistor or opamp. This may be relic of design changes not totally re-factored for minimum cost.

great, i was wondering about that one and was going to try it. the SD-1 works this way, with the 10K volume put tied to V/2 rather than ground and doubling as a bias resistor.

The guitar could drive the clipper directly; except when the clipper is working HARD even its input will be glitchy. That means you can't get an "unmolested" no-effect signal without a buffer somewhere in front.

i'm glad you said this, so it's along what i was thinking already. but then when i look at the TL072 data sheet and it shows the internal schematic break-down, it looks like the IC's FET at the front is exactly like what we're doing with the external transistor/fet. well, if it sounds better i guess it's hard to argue not to do it.

An op-amp WILL follow exactly (with 500K bias you want a FET opamp). However this scheme probably dates from days when transistors were cheaper than chips (at least if your Purchasing beat the scrap bins for non-dead material).

i found this schematic for an old (original?) 14 pin OD-1 where they still decoupled between the buffer and the drive op amp. i'll try feeding it directly (like you said, op amp should carry the same DC voltage at the output):

Two TL072 is now as cheap as jellybean NPNs (with resistors caps and joints) and may be a better plan. (TL074 was always too many pins and too much gain in one small lump, unless you just wanted a buffer-array.)

ahh...   i actually was planning on using a TL074, thought it would be a cool idea. so you think it might be a recipe for problems with glitching and cross-talk? especially with the heavy gain demand. i figured that two of the stages were buffers and one wasn't doing a lot...






Thanks again. you really addressed my questions and gave me some reinforcement to ponder.

[Keppy]

ahh...   i actually was planning on using a TL074, thought it would be a cool idea. so you think it might be a recipe for problems with glitching and cross-talk? especially with the heavy gain demand.

The original TS uses a dual opamp without problems, so the packaging shouldn't be a factor. I think Paul was talking about another problem where, in the original circuit, the buffer had to feed the gain stage even when the effect was bypassed, and so the characteristics of the gain stage input would wreak havoc on an unbuffered signal and wreck the bypass tone. This could affect the unbypassed tone as well, but not as noticeably.

Bear in mind, though, that the diodes in the feedback loop are usually what clip, rather than the opamp itself, so this doesn't necessarily explain the difference in tone that you heard.

i found this schematic for an old (original?) 14 pin OD-1 where they still decoupled between the buffer and the drive op amp. i'll try feeding it directly (like you said, op amp should carry the same DC voltage at the output):

Opamps generally have some small DC offset at the output. It's negligible, except when you're amplifying it hundreds of times. You know, like we ALWAYS do. I have one design that with ideal opamps wouldn't need a coupling cap, but with real-world opamps doesn't work without one. I've only run into this problem with cascaded gain stages, though, and I don't think a TS has quite enough gain to cause problems without the caps if you're using opamp buffers. But, at max gain, with a relatively high-offset opamp, I could be wrong. Try it and see.


Incidentally, did you set up the unbuffered test to have the usual TS input impedance of ~500k? If you kept the 10k biasing resistor of the gain stage, that drastic change in input impedance could easily explain the difference you heard.

[darron]

Oh right, I get you, that Paul probably meant effecting the clean bypass. Even the SD-1 has complaints of some bypass bleeding, especially when you mod it for more gain. And yeah, people are wanting for 1,000x gain sometimes! But like you said, the diodes stop it asking from 'asking' for that much gain at a certain point. and i agree, a little whisper of dirt traveling inside a dirt box probably isn't a problem. it could even help.

You make a good point that since there's no problem with the dual op amp too, so I don't see why to expect one with a quad. I misunderstood what Paul meant.

Incidentally, did you set up the unbuffered test to have the usual TS input impedance of ~500k? If you kept the 10k biasing resistor of the gain stage, that drastic change in input impedance could easily explain the difference you heard.

let's see.. the first one i did was a breadboard of an OD-1, which used 100K impedance into the op-amp. maybe you're right, i should have used the higher 470K which the first buffer transistor sees. i'd like to hope the 100k wouldn't make much difference though. i can imagine 10k from a TS circuit could be more prominent/drastic. i needed more though and testing there... thanks.