Mixing dry/wet signals

[thunderaxe]

hi all, first time posting here.

i want to understand the best way to create a wet/dry blend pot.

my quest started with the CE-2 -- my original thought was that i wanted to be able to control the intensity of the chorus effect by mixing between dry and chorus, but then figured why not go all the way and mix from 100% wet at the far left, to 50/50 mix (standard chorus) in the middle, to 100% vibrato on the far right.

the standard fixed chorus mixing seems pretty straightforward:
(from the electrosmash CE-2 analysis page)



as the description says, the ratio between the two resistors determines the mix, and so i figured intuitively that the way to blend between the two would be something like this:



with the wiper in the middle, it retains more or less the exact same circuit as before, with each signal getting 50K resistance (a slight shift from 47K but i figure that can't hurt) and then moving it to either side shifts it to 0/100K and 100K/0.

but then it turns out i'm not the first person to have this idea, and the walrus audio julia is pretty much exactly this plus a couple of extra options. the relevant part of the circuit with the wet/dry blend, however, looks like this:



can someone explain to me why that is? how does this work? is it better than my simpler idea? if so, why and how?

many thanks in advance!

[m4268588]

Make the VR B100k smaller so that not volume drops too much at 50/50.

[thunderaxe]

Make the VR B100k smaller so that not volume drops too much at 50/50.

can you explain why? sorry, i'm new to this and figuring out the correct pot and resistor values is still a bit of a mystery to me. i imagine that figuring it out has something to do with ohm's law but i don't know how to apply it.

[m4268588]

Even if you reduce the VR value by method I shown, the volume will drop slightly. In case of VR = 2K, it will be 49.5/49.5.

The following method is 89/89.

Detailed explanations in languages other than Japanese are available for a fee.

[antonis]

I think OP needs a further (no fee ) elaboration, so:



@thunderaxe: Consider TP as the reference point of a resistive voltage divider with R1 upper resistor and R2//R3 lower resistor..
(the later occurs 'cause signal from TP is splitted to GND through R2 and to Virtual GND through R3..)

So, signal voltage at TP is Vin X (R2//R3) / [R1 + (R2//R3)] (following standard voltage divider formula)
TP voltage is further amplified by op-amp following standard inverting configuration Gain formula (R4/R3)..

Now, replace R38, R36, R34 & BLEND pot lug 3-wiper in your schematic with R4, R3, R2 & R1 respectively  et voila ..

P.S.
You can (and have to) consider VREF as AC ground for signal frequecies..

[merlinb]

The load on the pot needs to be much larger than the pot resistance to minimise volume change. With these values the change should be imperceptible, but you will need to increase C14 in the CE-2 schematic to about 330n since the cap is now loaded by a 5k pot instead of 47k.

[idy]

can someone explain to me why that is? how does this work? is it better than my simpler idea? if so, why and how?

the second example, at the extremes, sends one of the two signals to (AC) ground, ref. So with that version, you will get no wet or no dry. Other versions will leave some of the other signal in your mix.

[ElectricDruid]

The second example is best explained by R.G.

http://www.geofex.com/Article_Folders/panner.pdf

It's a very useful circuit.

[m4268588]

can someone explain to me why that is? how does this work? is it better than my simpler idea? if so, why and how?

Other versions will leave some of the other signal in your mix.

This will not happen if the signal source impedance is low enough.

[m4268588]

There have been some discoveries about the second method.

It seems that not possible to lower the volume at midpoint of the potentiometer. It will always be 50/50 or more.
The choice of resistance value for VR is limited and must be less than to a certain extent.
The NFB resistor (47k) cannot be changed, and the VR selection is limited, so what can be selected without inconvenience is the remaining series resistor. The equation to select them is very complicated.

Auto calculators may be in demand.

EDIT:
Full implementation is not yet there.
https://mrotqch.web.fc2.com/bin/Bal_Mix_Calc.html

[ElectricDruid]

It seems that not possible to lower the volume at midpoint of the potentiometer. It will always be 50/50 or more.
The choice of resistance value for VR is limited and must be less than to a certain extent.
The NFB resistor (47k) cannot be changed, and the VR selection is limited, so what can be selected without inconvenience is the remaining series resistor. The equation to select them is very complicated.

Sorry, but I don't think much of this is true.

Depending on the resistor and pot values, you can make the volume change in the centre. The *point* of the circuit is that it provides "constant power" panning, so it's usually set up to be nice and flat, but it doesn't have to be.If you change the values to level out the voltage sum, the power sum finishes up with a dip in the middle. And vice versa.
The VR value can be changed. Obviously altering it has consequences for the circuit, so you have to change other values too, if you want things to work the same.
The NFB resistor is the *easiest* thing to change, since it essentially just sets the overall gain of the whole thing and *doesn't* interact with anything else.
Selecting the components is not complicated and doesn't require solving that gigantic mess you posted. The only equation you need is a simple voltage divider with two parallel resistors for the bottom resistance. That's not so bad!

The part where we agree is perhaps on the use of tools. I wrote a PHP script to plot the voltage and power outputs for given resistors/pot values (example shown below). LTspice would be another way. Doing it by hand is certainly possible but somewhat tedious and since we live in an age of technology it makes sense to use it.

[ElectricDruid]

I found my notes about this:

The typical 1/sqrt(2)=0.707 value for the pot gives you a "constant power" pan. This is -3dB in the centre. This gives a slight volume hump if fed with corellated signals (or identical signals) of 3dB.
   
Decreasing the ratio (making the pot small wrt the resistors) flattens the voltage sum curve, which is better for corellated signals. By the time you get to Pot = 0.1 x R, you're very close to the ideal "constant volume" pan curve of -6dB in the centre (we reach -5.26dB). The price you pay is that the signal level drops a lot, so the make-up gain (and consequently noise) increases. We need x12 gain, +21.5dB in this situation. Ouch!
   
Going the other way and increasing the ratio (say, Pot = 2 x R) gives you a curve where even the power sum has a hump in the centre. I'm not sure what this would be good for. It's a pan law of -1.5dB, which I don't think I've ever seen.

This analysis assumes all four resistors are equal (like in RG's paper) but that doesn't have to be the case, and you can tweak it a bit further by using separate values for the resistor pairs before and after the pot.

[thunderaxe]

okay, i think i'm getting it. so i want to apply this to the phase 90, which uses a PNP transistor instead of an op amp for the mixing stage. btw, does anybody know why? was it just to save on ICs, which i assume were more expensive than discrete transistors at the time?

am i safe to assume that it works more or less the same way, and that the feedback resistor between collector and base in the transistor plays the same role as the feedback resistor in the op amp?

the resistors are all 150K here -- how important is this specific value, compared to just keeping the ratios between the resistors? if 30K potentiometers existed i could keep the feedback resistor at 150K, 0.707R = B30K pot, and R = 47K, but since i'm constrained by existing pot values, RF = 113k, 0.707R = B25K pot, R = 36K is the closest i could come up with.

that said, if actual values doesn't matter and it's only about the ratios then R.G.'s suggestion of R=15K, pot=B10k, RF=51K is the obvious choice.

[antonis]

Although Q1 Base couldn't be considered "virtual ground" (compared to op-amp respective inverting input) it should work satisfactorily..

Dunno about DRY/WET DC status so it should be a good idea to place some DC blocking capacitors..

[thunderaxe]

Although Q1 Base couldn't be considered "virtual ground" (compared to op-amp respective inverting input) it should work satisfactorily..

Dunno about DRY/WET DC status so it should be a good idea to place some DC blocking capacitors..

well that schem is from the original phase 90 circuit and it clearly works great as is so i'm gonna assume there's no need for DC blocking caps there.

my idea would be to do a 100% dry to 100% wet mix knob like the following:

[m4268588]

(Vcc-Vbe-Vc)/R7*hFE*R4 = Vc
so...
R4 = 4.5*R7/((9-0.6-4.5)*hFE)

If you don't want to trust me, do it yourself.
I have no obligation to help you.

[StephenGiles]

Couldn't this be very simply done with a dual gang pot? I've seen it somewhere, was it one of Craig Anderton's circuits ?

[Mark Hammer]

The virtues of the different circuits shown depend on what one is trying to achieve.  If the goal is a-little-more-of-that-than-this, the first circuit thunderaxe posted with a 100k pot is suitable. 

If the goal is to be able to have ONLY one of the signal sources, it is insufficient, since some of whichever signal has the higher resistance in series with it will always be included.

The circuit that splits the input resistance in two, and bleeds off signal to ground, is the better one.  It is, in fact, exactly what pretty much any stereo amp with a "Balance" control does.  When the resistance between the junction of the two series resistances and ground is zero, then there is NONE of that signal making it to the mixing node.  The virtue of this circuit version is that it also permits a-little-more-of-this-than-that.

[antonis]

(Vcc-Vbe-Vc)/R7*hFE*R4 = Vc
so...
R4 = 4.5*R7/((9-0.6-4.5)*hFE)

@thunderaxe: What m4268588 is trying to tell you is grounded Emitter feedback biased CE amp basic formula..
In other words, for Collector biased at about Vcc/2, R7 value should be about h_FE X R4..

Unless your Q1's h_FE is 2!! (for R7=113k) or 3!! (for R7=150K), your mixer should be heavily bottomed nearly cut-off..!!


P.S.
If you don't trust neither of us, trust Ray Marston..


Figure 10

[thunderaxe]

(Vcc-Vbe-Vc)/R7*hFE*R4 = Vc
so...
R4 = 4.5*R7/((9-0.6-4.5)*hFE)

@thunderaxe: What m4268588 is trying to tell you is grounded Emitter feedback biased CE amp basic formula..
In other words, for Collector biased at about Vcc/2, R7 value should be about h_FE X R4..

Unless your Q1's h_FE is 2!! (for R7=113k) or 3!! (for R7=150K), your mixer should be heavily bottomed..!!

P.S.
If you don't trust neither of us, trust Ray Marston..


Figure 10

by m4268588's formula, assuming an hFE of 250, in the original circuit where the feedback resistor R7 is 150K, the collector resistor R4 should be ~0.69K or 692ohms. so why is it 56k? again, that's not my circuit, it's the MXR phase 90.