Loaded VRef?

[karbomusic]

I saw something similar in another thread which matched something I'd been considering. In the image below (cropped to keep attention on the important piece), I'm feeding 3 dual opamps + change for a total of eight connections to VRef (see black arrows). Would it be wise from either a decoupling sense and/or other to split this VRef about half way down creating a second VRef on the right half using second divider?

Everything works but I'm sometimes on the edge of oscillations when tweaking if not careful; I have that mostly worked out but wondering if I'm tugging on VRef too much and/or opening up myself for signals running around in ways that might make it and me unhappy.  



Hope it's not too juvenile of a thought. I can work with it as-is but it feels like this is one of those differences between technically working and well thought out.

[R.G.]

It's not a juvenile thought, but it's unlikely that you're having the problem you think you are.

First, read http://geofex.com/circuits/Biasnet.htm for some background.

Resistor/resistor/capacitors rely on the resistors and how much current they let through to hold the DC conditions in an acceptable range for the loads on the network and the variation of DC loading or very low frequency changes in the bias network drain. They rely on capacitors to keep the bias network stable with AC loads.

More capacitance, and better high frequency caps paralleled onto the bias bus usually work pretty well.

If you return an opamp output to the bias voltage, perhaps like a volume pot going from the output to Vbias and not to ground to save a capacitor, this can let significantly more current into the bias node. That can wiggle the bias node up and down and cause feedback. It all depends on how much current the capacitors on the bias node are trying to get to ground without allowing the bias supply to change.

So the real advice is to estimate your bias network currents.

Failing that, you can use separated bias generators as you note, or use a  buffered bias, the output of a spare opamp with only DC into its input. The output of this becomes a DC source with quite low impedance, so you can drive more and more difficult bias points with it. Or if you have only one "difficult" bias section, either make a second bias for it.

In really difficult cases, you can split off resistors from the bias point to a separated bias bypass cap for one or more of the circuits. The bias voltage is dropped a bit by the DC currents pulled by the "difficult" section,  but the cap supplies the changing AC currents locally and the resistor isolates the AC currents from the main bias point so it doesn't hurt the other sections.

But just try more capacitance, or a capacitor at each end, plus maybe a 0.1uF ceramic. That bias line is too short for reflection effects and too small for much loop pickup.

[karbomusic]

Thanks R.G,

Adding more capacitance was one of the things I had done with this latest round of circuit changes and results were positive. I already have 3 other working units so this is really, more than anything, iterative refinement. All of the above makes sense and I'm now more comfortable sticking with what I have. It would also be easy to split the VRef after the fact if needed. Thanks again.

Edit: Thanks for the link, it will come in handy and I'm sponging it away immediately.

[karbomusic]

OK, I just hit this again and... Excuse me whilst I ramble and think out loud...

It's only when bypassed due to the circuit input being effectively free hanging when bypassed. I'm OK with that other than.... when the unit is in bypass there is a faint whine bleeding into the bypassed signal. For the life of me I ...

1. Can't find out exactly where it is bleeding over, I don't think I'm doing anything "too stupid" layout wise but more importantly...
2. Can't understand how I might stop it to begin with (need some more academic reading on opamp feedback when input isn't grounded).

I do know it is due in part to increased gain.

I could obviously convert the switching to ground the input when bypassed but that feels dirty. For one, I'd need to change my JFET bypass (similar to millennium) to a 3PDT just so I can ground the input and so on. The frequency of the whine does appear to follow any resistance changes anywhere, meaning that any knob or switch changed slightly alters the pitch  (it's center point is around 2-3kz as a guess) . I'm currently using 220uf on main and VRef rails. I did "clamp in" a 330 in parallel as a test but doesn't appear to make any difference.

I suppose I could take 5 steps back and use a buffered bypass but would prefer to learn more about exactly what is occurring before redesigning things. I'll try splitting out the VRefs tonight since it is a 10 minute test to cut one trace and solder in 2 resistors and a cap. I can dig out the schematic if needed  (need to remote in and retrieve) but it's basically a TS sandwiched between two OPA2134s with a TLC2272 in the middle. The 2134s buffer/split/overdrive/clean blend. At the output OPA2134 I apply gains of 2.2 (overdrive output), 8.0 (clean output). I'm going to guess that the final 2.2 I add to the OD output isn't helping things.

I think I should just step back and do better at analyzing the original circuit. Maybe my idea of adding some gain at the end is silly but part of the point was the ability to get the blend POT "just so" and drive the amp harder as needed. Thusly, I started configuring to be able to handle up to 1.2 volts peak input without clipping @ 18V and achieve a total output of 1.5-2.0 volts with the volume at max, again the goal was overdrive + clean boost + blend + ability to push the amp as needed. I'm really, really close here, only this whine when bypassed has reared it's head, everything else is purrrfect.

/ramble

[PRR]

> ground the input when bypassed but that feels dirty.

In my religion, grounding un-used inputs is bliss, not sin.

3KHz whine is probably not Vref, since Vref has many uFd on it.

Audio signals go through air. Just like radio signals, Not so good, but a high-gain circuit in a small box is begging to pick-up its own output on its input. In theory there is a certain phase needed to oscillate. In practice Murphy ensures there will always be a frequency where phase is good to oscillate.

[karbomusic]

In my religion, grounding un-used inputs is bliss, not sin.

I can certainly accept that idea but I hate adding a 3PDT just to ground the unused input. If that makes the angels sing I suppose I'll learn to live with it. Still feels like killing a fly with a sledgehammer though but likely a price I'm paying by choosing true bypass. Thanks for the wisdom Paul.

[PRR]

If you are flogging JFETs to do bypass already, surely one more JFET can grab the input and mute it when bypassed?

Since I can't see what you are doing from here, hard to give specifics.

[karbomusic]

If you are flogging JFETs to do bypass already, surely one more JFET can grab the input and mute it when bypassed?

Since I can't see what you are doing from here, hard to give specifics.

I'm doing this below. I'm embarrassed a little because I can see things that are redundant/amateur and know there are likely plenty I don't see but it is a work in progress so how else can I do better without getting spanked? Basically it's a TS circuit with the in/out buffers replaced. And... the input buffer splits the signal OD/Clean. The output buffer does a gain bump and blends the clean/OD at the end. IC2 is really a TLC2272. I assume I have a redundant buffer on the clean but it's an endless circle of things to do and haven't addressed that yet.. Ducks...

[duck_arse]

... yes yes?? you can't just leave a hanging "Ducks" like that!

if you can stretch as far a millenium bypass, you can do a shunt to ground (4 shunts to ground ?), no 3pdt needed .....

http://www.geofex.com/article_folders/mill2extn/mil2plus.htm

[karbomusic]

... yes yes?? you can't just leave a hanging "Ducks" like that!

if you can stretch as far a millenium bypass, you can do a shunt to ground (4 shunts to ground ?), no 3pdt needed .....

http://www.geofex.com/article_folders/mill2extn/mil2plus.htm

Holy crap, that articles nails that piece if I go that route. I may rework the current bypass to something similar in the article and make it one of my standard builds. What I have now isn't exactly the millennium but darn close if not exact so this should be workable. I might just tweak that into something and etch a few for the future. To be clear though, I had reserved myself to thinking free hanging input oscillation a design failure so I wanted to cover that angle first. Just trying to do the right thing ya' know.

Thanks.

[karbomusic]

Okie Dokie... Let's see if I have this right after searching and noticing this is nothing new...

- I'm obviously exploiting opamp's outside of their designed audio intentions (saturating the crap out of it) and adding even more gain on top of that.
- Due to the above statement, it's plausible if not guaranteed that I could see oscillation when the circuit input is floating.
- I need to get creative in grounding the circuit input when bypassed to prevent this or use a lower gain circuit.

I can do that, I feel better now assuming my assumptions above can be assumed correct.   I'm not wild about the extra circuitry to make this happen in a DPDT so I either grin and bear it or go back to 3PDT and forgo my beloved alpha DPDTs for circuits that don't have this issue.

Unless I missed the boat above, I now have closure, thanks for the advice from everyone.

[PRR]

> things that are redundant/amateur ...it is a work in progress so how else can I do better without getting spanked?

Gentle spank....

Buffer buffer Hi-Z input-- don't need all that.

Vref resistors to places that already sit at Vref (via another amp).

IC3B would seem to be biased to output=38VDC (IC3A to 9.9VDC), but can only go to 8VDC, so is surely stuck?



Now you have just *one* thing connected to Vref, and it's a 10Meg.

Also I think C2 68pFd could be much bigger, 100 200 even 500pFd. You want gain on guitar tones, not bat-tones. And this huge gain above the guitar band (along with a compact layout) is probably your high audio whine.

But what I was asking about is the bypass system.

If GeoFex has your answer, cool.

[karbomusic]

> things that are redundant/amateur ...it is a work in progress so how else can I do better without getting spanked?

Gentle spank....

Needed.

Buffer buffer Hi-Z input-- don't need all that.

Agreed, will fix. Once I have something working, I'm slower to modify (it's a carry over from my programming background). However, I now have a semi-permanent bread boarded version of this circuit where I can now test such changes instead of mucking with one of the ones already built.

Vref resistors to places that already sit at Vref (via another amp).

IC3B would seem to be biased to output=38VDC (IC3A to 9.9VDC), but can only go to 8VDC, so is surely stuck?

Once I figure out how to calc and understand what you just said, I'll be all over it.   All I actually know is I can run it from 9-18VDC and at 18 I can get 1.7 ish avg output without clipping the parts that aren't supposed to be clipping assuming a 1.2 or so max input from the guitar. At least that's what the scope tells me. I also have a requirement to be able to spank an amp input hard as needed/depending hence the gain at the very end.

Now you have just *one* thing connected to Vref, and it's a 10Meg.

I don't understand unless you are speaking in reference to the input only, there are 7 connections to the VRef rail so I'm sure I'm missing what you mean here. Don't waste too much time on me though, I do not want to be a leech but I do like to learn to fish on my own.

Also I think C2 68pFd could be much bigger, 100 200 even 500pFd. You want gain on guitar tones, not bat-tones. And this huge gain above the guitar band (along with a compact layout) is probably your high audio whine.

I've gone to 120, I can go to 500 and test just to see.

But what I was asking about is the bypass system.

If GeoFex has your answer, cool.

My sensibilities still tell me I should try a little harder to fix it at the source. So that article isn't ideal for me, it's an ideal and elegant solution for the question asked that generated the article but if I had my way, it would never whine to begin with. Time will tell, this project is the one that is a living one so it'll always be in progress and teaching me new things. I already have 4 of these built in various incarnations with a couple in the wild already (those don't whine since the have slightly lower gain) but who's counting. :  

Thank you guys for the patience and advice, I'm a sponge, really so everything anyone types, I soak up so it isn't wasted, again thank you.

[PRR]

> there are 7 connections to the VRef rail so I'm sure I'm missing

Did you study the drawing I posted? Most of your Vrefs were pointless, others easily designed-out.

[karbomusic]

> there are 7 connections to the VRef rail so I'm sure I'm missing

Did you study the drawing I posted? Most of your Vrefs were pointless, others easily designed-out.

No I didn't realize you had changed it, now I see it and will digest and test, thanks.

And if you don't have IC3's outputs slammed high, then your build does not match your drawing.

I probably do and don't realize it? Here is a quick scope output I just took. C3 (pink) = input, C4 (blue) = output. Clean output is from IC3B. It is entirely possible I confused myself. I'm trying to dig out my old notes but getting late. I may not have mentioned I run it at 9V but sometimes 18V. Ignore the 8.2k label its from previous testing and forgot to delete:

Input:200mV @9VDC



Input:600mV @18VDC

[karbomusic]

OK, I looked at your changes, nice they are. The only thing I see is that the IC2 is supposed to be a TLC2272 not an OPA2134 (buried in 4th post - reply #3). I wanted a FET input, didn't think the TLC2272 has one and was basically following the TS in buffer / overdrive / out buffer mantra.

[PRR]

> And if you don't have IC3's outputs slammed high

Sorry, I was in an alternate reality when I wrote that; now deleted.

You returned the NFB legs to Vref. That works, but clamping the end of low-resistance NFB legs can be a heavy burden on Vref. For an extra cap per stage you can go to ground, where you really want to be, and use a much lighter cap on Vref.

> I wanted a FET input, didn't think the TLC2272 has one

Datasheets. '2272 is CMOS. This means potentially even higher input impedance than JFET (either one more than we ever need in audio). CMOS has a reputation for higher hiss, but the process has got a lot better over the years, and '2272 seems to be on-par with '072 which is generally quiet enough for guitar circuits.

[karbomusic]

All excellent info thank you very much. Will sponge all this over the weekend.

[alfafalfa]

Audio signals go through air. Just like radio signals, Not so good, but a high-gain circuit in a small box is begging to pick-up its own output on its input.

What does this mean for the size of stompboxes getting smaller all the time and input and output getting closer to eachother. So it must be asking for trouble I reckon ?

[R.G.]

Audio signals go through air. Just like radio signals, Not so good, but a high-gain circuit in a small box is begging to pick-up its own output on its input.

What does this mean for the size of stompboxes getting smaller all the time and input and output getting closer to eachother. So it must be asking for trouble I reckon ?

First, stompboxes getting smaller all the time is a little silly. The limit to the size of a stompbox is the size of the stomp switch and controls, if any. The human foot is not getting any smaller all the time. I have done an orange squeezer on a bit of PCB stock less than 1" square for a special application. That didn't make my foot any smaller. Think about it.

Second, inputs getting closer to outputs isn't necessarily any worse. Most stompboxes have a random bundle of wires from inputs and outputs to the stomp switch, and on the stomp switch the contacts are separated by the same distance they always were. So the issues are the same: high gain, high impedance, small separation, all require careful attention to wiring and layout, and shielding if necessary.