Mixer / buffer : power supply, vref questions, Ohm's law for ignorants

[Nicotep]

I'm building a matrix mixer based on 4 "ggg mini mixers". I plan to use 2xTL074 to save a little space.

I'm adding buffers to all 4 inputs, based on the Klon Buffer, but with an inverted phase output. Again, 2 other TL074 are needed.
Another difference with these original designs is that I power them with 18v rather than 9, for increased headroom.

The original schematics are here : http://www.generalguitargadgets.com/pdf/ggg_mixer_sc.pdf and here :


Questions :
1.
- instead of making different Vref's all over my circuit, I'd better use only one, right ?
- I just read the nice article here : http://geofex.com/circuits/Biasnet.htm and then it becomes more blurry to me... I understand I'd better use lower values for R1/R2 to have a stable Vref for all my TL074. I saw their datasheet says each opamp draws 10mA max., though I don't know what to do with that value. Do I only need to consider the whole circuit will draw 4x 10mA x4 = 160mA ?
If I follow the examples on Geofex page, it means my R1/R2 values would be 18/150 = 112.5ohm/2, so it means using 56ohm resistors. Never saw such small values in any schematic.
2. In the Klon Buffer, R3 is meant to give the input impedance ? Or is it wrong ? And why aren't there any of these resistors in ggg mixer, but the Vref directly ?

3.
Then I don't know how to size my capacitors.
The caps are used to keep noise away from power supply right ? Would it be a good idea to oversize them ?

Thanks a lot. Learning slowly and I'd like to know more.

[antonis]

1.
- Right..
- Wrong.. 10mA refer on working current, NOT bias..!! For FET input amps, just ignore it - consider bias current zero..
(but better make R1 & R2 for previous "right" answer 10k each to make a more stiff Vref for many "zero" bias currents..  I know that despite the number of zeros added the result is always zero but let it be for the moment..) 

2.
Right..
Bias resistor(s) are only needed for non-inverting configuration.. (the one that signal goes to the same input as bias Vref does..)
In ggg mixer signal goes to inverting inputs so it doesn't "leak" to Vref (AC GND), hence no need for "blocking" resistors..

3.
Wrong..
C2 & C3 are part of  Input & Output High-Pass filters..
Their size comes out of 0.159/(R3 X f) & 0.159/(R4 X f) respectively, where f = cut-off frequency..
C1 plays a dual role.. It smooths Vref and also forms a Low-Pass filter with R1 to attenuate any +9V supply 100/120Hz ripple due to mains rectification..

[antonis]

Let's make it a bit more plain..

In GGG mixer, signals go to pin 2 (inverting input) and then into pin 6 (also inverting input) resulting into non-inverted output (two inversions, each of 180⁰..)
In posted buffer, signal goes into pin 2 (non-inverting input) so In-Out signal are in-phase..

In GGG mixer, signal only deals with pin 2 & 6 impedances which are considered of very high value with no other way to ground..
In buffer posted, signal deals with pin 3 AND Vbias point, which point is considered ground for signal due to C1 (short for AC)
R3 is needed for "isolation" purpose.. It's high value blocks most of signal amount from leaking into ground, hence sets input impedance..

P.S.
In case of +9V pure DC, absolutely zero input current  and no noise issues, C1 & R3 could be omited and R1 & R2 value could be raised up to Megaohm range for upholding high input impedance..

[PRR]

> a stable Vref for all my TL074. I saw their datasheet says each opamp draws 10mA max.,

It is <2.5mA per opamp, which is 10mA per TL074....... but this is NOT the Vref load! Which for TL07x chips is about zero. A pair of 100k resistors can supply Vref for a hundred TL074 chips.

[Nicotep]

It is <2.5mA per opamp, which is 10mA per TL074....... but this is NOT the Vref load! Which for TL07x chips is about zero. A pair of 100k resistors can supply Vref for a hundred TL074 chips.

Thanks ! That's good news for future monster matrix mixers.

P.S.
In case of +9V pure DC, absolutely zero input current  and no noise issues, C1 & R3 could be omited and R1 & R2 value could be raised up to Megaohm range for upholding high input impedance..

- Wrong.. 10mA refer on working current, NOT bias..!! For FET input amps, just ignore it - consider bias current zero..
(but better make R1 & R2 for previous "right" answer 10k each to make a more stiff Vref for many "zero" bias currents..  I know that despite the number of zeros added the result is always zero but let it be for the moment..) 

Well I'm actually looking to prevent any noise issue in most conditions and make a stable supply, that's why I try to decrypt these circuits despite my vast lack of knowledge   

I'm taking the opportunity to take another example, which comes from one of David Rolo's pedals (ignore the left part with the 3.3V regulator) : why C19 & C20 ? Another low pass filter with R23 ? Could he then not use a bigger resistor to make C19 smaller ?
Or is it something else ?
So many questions

[niektb]

I'm taking the opportunity to take another example, which comes from one of David Rolo's pedals (ignore the left part with the 3.3V regulator) : why C19 & C20 ? Another low pass filter with R23 ? Could he then not use a bigger resistor to make C19 smaller ?
Or is it something else ?
So many questions

Yes it's a low-pass filter indeed! To have the same cutoff frequency R23 could be increased and C19 could be lowered. !However! Remember that a pedal draws current so all the current used by components (that are connected to V1) passed through R23. This causes a voltage drop across R23 (remember Ohm's Law) (U = I * R) (which means less analog headroom) and power dissipation in R23 (P = I^2 * R). The later means that the resistor could blow if the power dissipated is higher than the power rating of the resistor

[Rob Strand]

I'm taking the opportunity to take another example, which comes from one of David Rolo's pedals (ignore the left part with the 3.3V regulator) : why C19 & C20 ? Another low pass filter with R23 ? Could he then not use a bigger resistor to make C19 smaller ?
Or is it something else ?
So many questions

There's many reason to use an RC filter.

As far as the 3V3 path goes it's hard to know the motivation for adding R22 and C18/C17.
- R22 will help reduce power dissipation  in the 3V3 regulator.   It will also limit fault currents.
- R22/C18/C17 will stop low frequency current changes on the 3.3V circuit from feeding onto the 9V rail.
- The regulator will help stop high frequency noise on the 3V3 rail getting onto the 9V rail, but
  R22/C18/C17 will not help that much more because it is already low.

R29/C19/C20 is a filter for the 9V it stops noise/ripple on the 9V inlet getting through to the 9V rail.
R29 is judiciously chosen.   if you make R29 larger, which will allow a  smaller C19,  you get more voltage
drop across it.   That makes the battery look flatter than it is.

[antonis]

What niektb & Rob said..!! 

If you've noticed, exactly the same RC filter is used for 3.3V regulator Input but voltage drop is rather desirable here, 'cause it lowers regulator's power dissipation..
[P = (V_IN -V_OUT)/I]

Of course, total power dissipated remains the same but it's splitted in R22 & Regulator, resulting into lower power rating item..

[Nicotep]

thanks to all of you

[Nicotep]

OK so just to be sure :

R29/C19/C20 is a filter for the 9V it stops noise/ripple on the 9V inlet getting through to the 9V rail.
R29 is judiciously chosen.   if you make R29 larger, which will allow a  smaller C19,  you get more voltage
drop across it.   That makes the battery look flatter than it is.

R29 -> You mean R23 right ?

I still don't understand the use of C20... 2 capacitors in parallel get their value added, no ? So the difference in the filter obtained is so small it doesn't justify the use of this extra component ...? Or C20 has a different role to play ?

I'm missing something... again .... 

[antonis]

Yes, R23..!!
(Rob is notorious for his typos.. )

C20 is usually a ceramic 10nF - 100nF decoupling cap for High frequencies shunt (bypass)..
https://electronics.stackexchange.com/questions/172447/where-did-the-value-of-0-1uf-for-bypass-capacitors-come-from

[Rob Strand]

R29 -> You mean R23 right ?

I still don't understand the use of C20... 2 capacitors in parallel get their value added, no ? So the difference in the filter obtained is so small it doesn't justify the use of this extra component ...? Or C20 has a different role to play ?

I'm missing something... again .... 

Yes, I do mean R23 (no idea how R29 got in there.).

You will often see capacitors in parallel on schematic.   There's many reasons but I'm only going to elaborate on the case for power supply caps.

1)  If the caps are of equal value then that usually means you are just making a larger out of two small ones.
      You can see this from the parallel capacitance formula  Ctotal = C1 + C2. 

2)  The harder to understand case is when small caps are put in parallel with larger caps.   For power supplies, that pretty much has nothing to do with adding the capacitances.  In reality capacitors are non ideal parts.   As frequency increases capacitors start to work less like capacitors.  Also larger values tend to poop-out at lower frequencies.  So by putting a smaller capacitor in parallel with a larger capacitor it extends upper frequency that the combination looks like a capacitor.   More precisely is it about making the impedance of the cap combination low at higher frequencies. Sometimes you might even see three caps in parallel.

3) Another case is when capacitors are placed at different physical positions on the pcb.   More often than not you will see the larger caps near the power inlet and smaller ones spread around the board.   The idea here is very similar to 2 except with the added issue that the pcb tracks add to the problem.   The key idea though is a smaller cap near an IC will  ensure the power stays at a low impedance *at* the location of that IC.

[Rob Strand]

(Rob is notorious for his typos.. )

My fingers don't keep up with the voice in my head     (not voices in my head  ).
Gets worse when I haven't slept properly.