Which is wrong? If either?

[slashandburn]

I found this amongst my saved images, its not my site, I dont ever even remember looking at it. It caught my eye because it basically looks like the front end of a tube screamer to me, which I thought I had a fairly thorough understanding of. Only the voltage bias goes to both the guitar input AND the feedback loop. (as in, the bit of the feedback loop Ive always put to ground goes to 4.5v).  Im curious why. Clearly something fundamental Im missing unless its a mistake. I think Ive seen something similar elsewhere, where instead of GRD or 4.5v it linked back upo with the "top" side (as drawn) of the 1M5. Can anyone shed any light for me?

Hope its okay to post this link.

http://www.mylkstuff.com/USERIMAGES/Schematics/Mylk386-No1-rev1.jpeg

[Bill Mountain]

It can go to either.  In fact if you go to Vref you don't need the 100n cap unless you want it for tone shaping the boost section.

[midwayfair]

It can go to either.  In fact if you go to Vref you don't need the 100n cap unless you want it for tone shaping the boost section.

Continued: but the gain will be lower than if you use the cap and go to ground.

[slashandburn]

Wow okay. Thanks!   Theres no big mystery, then?  Its a case of "it doesnt really matter?".

I thought I had a handle on calculating the gain and high and low pass filters of these setups. (that basic TS style circuit).   Gain will be higher if I goto 4.5V rather than ground?

edit, sorry, the gain will only be lower than if I used the cap to goto ground. I dont understand why, unless the 1M5  resistor then becomes part of the gain resistance? No?  That would be a huge gain cut, wouldnt it?

[Bill Mountain]

It can go to either.  In fact if you go to Vref you don't need the 100n cap unless you want it for tone shaping the boost section.

Continued: but the gain will be lower than if you use the cap and go to ground.

Why's that?

[midwayfair]

It can go to either.  In fact if you go to Vref you don't need the 100n cap unless you want it for tone shaping the boost section.

Continued: but the gain will be lower than if you use the cap and go to ground.

Why's that?

I dunno, but it definitely is on the breadboard ... with a single-supply anyway. I don't know exactly how much less, but noticeably so.

[FiveseveN]

It really shouldn't be any different. I'm suspecting your V_ref impedance is too high.

[Bill Mountain]

It can go to either.  In fact if you go to Vref you don't need the 100n cap unless you want it for tone shaping the boost section.

Continued: but the gain will be lower than if you use the cap and go to ground.

Why's that?

I dunno, but it definitely is on the breadboard ... with a single-supply anyway. I don't know exactly how much less, but noticeably so.

Electronically it shouldn't be any different but I know there's always quirks in the real world.

[anotherjim]

If you need a cap so large it has to be electrolytic, then I think it SHOULD go to ground not Vref. If you take the electro to ground then it will always be correctly polarized, bearing in mind the the amp pins are running about Vref, then the amp side of the cap is always more positive than ground. If you tie an electro cap to Vref, it won't have a correct polarizing DC bias across it.

[R.G.]

That circuit has some issues if I'm looking at it right.

Forget gain for the moment. The feedback gain setting resistor (3.3K through 100nF) can go to either Vref or ground because of the cap. As noted, Vref must have an impedance much less than 3.3K for this to not do gain and frequency funnies if it goes to Vref. As noted, a large electro on Vref, with an impedance much less than 3.3K at the lowest frequency of interest will be needed if it goes to Vref, or especially if other circuit sections also "pour" their gain setting resistor current into the same Vref. If this last happens, then Vref becomes another summing input, probably unintentionally. The moral in this side-drama is to keep the Vref impedance low.

But I digress. The issue I see is that this circuit puts Vref on the guitar through the 1.5M resistor. There is no DC blocking cap at the input for the guitar. If this is connected to an input jack with the bushing grounded, then plugging in the guitar puts the guitar's 4K  to 18K of DC wiring resistance to ground at the opamp input, pulling its effective Vbias down to nil, and probably keeping much, if any of the guitar's signal from being amplified. The circuit as shown only works properly if the input jack bushing is floated up at Vref too.

This has its own problems, as now anything that touches the plug shells (if metallic) will also pull the Vref around, with possibly ugly results.

The input needs a series cap to work well with anything except the floated-guitar special case.

[karbomusic]

There is no DC blocking cap at the input for the guitar.

So what does the 47nF do?

[duck_arse]

are we all looking at this "circuit"? with the caps drawn as green boxes?

[karbomusic]

are we all looking at this "circuit"? with the caps drawn as green boxes?

I am.

[R.G.]

Sorry - I'm going too fast this morning. I saw that as a resistor.

Never mind that "no cap on the input" stuff. 

I can easily fall prey to this as I never use an input without some small resistor in series.

Comments about Vref stand, I think.

[ashcat_lt]

In order for Vref to be low enough Z in this case, it would need atypically small resistors in the Vref divider.  1Ks might almost work as long as they're not also feeding some other low-Z circuit.  680s get you close to the 10:1 rule of thumb.  Smaller is better, but we start to have a very low-Z path between the power rails.  Rails might sag.  Buffered Vref would do it, if you have a spare opamp laying around.

Also, that thing about the electro cap being reverse polarity half the time.

It will actually work about as well to connect that feedback network to the positive rail also.  Probably noisier, but it would "work".

It just seems a heck of a lot easier to go to ground.

[R.G.]

Note that only the AC impedance of the Vref point needs to be much lower than the 3.3K gain setting resistance.

If you wanted full effect of the 3.3K at, say, 40Hz, the minimum cap from Vref to ground (for this stage alone) is C = 1/(2*pi*40Hz*3300) = 1.2uF. The DC impedance conditions are set by the resistors, the AC by the bypass cap. Going to 680/680 would be OK from a DC point of view and AC, but you can make the AC better with just a cap.

[slashandburn]

Ah! Thanks again for all the replies. Ill try to have a look at and digest this later!

[ashcat_lt]

Note that only the AC impedance of the Vref point needs to be much lower than the 3.3K gain setting resistance.

If you wanted full effect of the 3.3K at, say, 40Hz, the minimum cap from Vref to ground (for this stage alone) is C = 1/(2*pi*40Hz*3300) = 1.2uF. The DC impedance conditions are set by the resistors, the AC by the bypass cap. Going to 680/680 would be OK from a DC point of view and AC, but you can make the AC better with just a cap.

Duh!   Right, the AC impedance is pretty much completely dependent on the Vref cap.  I knew that...

Thanks!

[R.G.]

Duh!   Right, the AC impedance is pretty much completely dependent on the Vref cap.  I knew that...
Thanks!

Hey, don't sweat it. I'm the one that didn't notice that there was a cap at the input! 

[slashandburn]

Ah cool I think I follow roughly what that was about!   So for basically, you can put the feedback network to either ground or Vref (or even the positive rail!) but from a novice point of view it's probably simplest to put it to ground?

Another quick question on the same circuit then, could I expect any issues putting a dpdt "bypass" switch on that feedback network? To bypass the whole gain network and turn it into a simple buffer?   Or would it be better to use a dual opamp and use the other half to provide the switchable buffer?   I hope I'm explaining that thought okay.  Basically a switch to toggle from buffer to TS style distortion?