Need ideas for a looper relay

[R.G.]

[...] regarding using CMOS logic chip for switching analog inputs and outputs is there any concern with NOISE?
[...]Boss_SD2_
Uses a TC4011 quad 2 in NAND gate (CD4011) and JFETs for channel switching. Seems to me that could be adding noise or raising the noise floor when increasing the gain. Is that possible?

Noise is always an issue. The word "noise" is much like the word "weed" - a weed is any plant you don't want growing there, and noise is any signal you didn't intend to be there.

Noise in this context comes in three flavors, hum, hiss and switching noise. Hum is introduced by you not doing the proper job grounding and shielding the whole mess of connectors and wires in this setup. Hiss is a consequence of either bad contacts (usually poor soldering) or resistance thermal noise. You're usually stuck with whatever thermal noise the resistances you had to choose give you. The CMOS switches contribute minimal thermal noise because they're either on, and a thermal noise resistance of a couple of hundred ohms, or off and the thermal noise is there, but doesn't reach your audio chain.

The real noise to worry about is switching noise. All switches contribute noise. The beloved hard-contact metal switches in "true bypass" setups can cause a clicking noise in the signal because they switch so quickly. So can CMOS switches. This is fixed by switching only at zero crossings of the signal, which CMOS switches can do, and some of them actually do implement. Metal footswitches can't be timed so accurately. All electronic switches suffer from feedthrough of the control signal. This is minimized in CMOS switches by biasing the switches to nearly the center of their voltage range, and by careful layout. JFET switches can be simply slowed down so they gradually change resistance from on to off and vice versa. This is the reason for those diodes, resistors and caps on the gates of switching JFETs. This does not eliminate control noise feedthrough, it merely moves it down to below the audio range where you can't hear it.

So which noise were you meaning?

[PRR]

> assumptions about the chip and its power supplies

Hard to squint, but I *think* you are feeding Vcc (digital power) with a resistor divider?

The Icc current is rated at 70mA. That's at 5MHz, which we don't need to hit, and may not do very often (idle current is <1mA). While a hi-Z source may work, it might cause digital chaos "for no reason". Suspect that budgeting a '05 regulator might be wise. The analog Vbias of course could be two 10K resistors with decoupling.

Considering ALL the analog Vbias/de-Vbiasing for 30 ins and outs, 90 parts, it *may* be worth considering a power voltage inverter for +/-9V analog supplies. Then (with luck) the ins and outs are no-parts.

> there any concern with NOISE?

No. (Aside from ALL the general "noise" issues of any audio project as R.G. explains.)

The switches are just like jumper-cords (near-enough, 300 Ohm added resistance which is low in guitar-cord interfacing). Putting digital next to analog is asking for odd buzzes, but with this chip (and for most analog switching) we only have digital transitions at the "change", not constantly. The buzz-burst to load the AD75019's state is shorter than the shortest sound we can hear (amounts to one cycle of 19KHz).

Hard-cutting from one signal to another (even same signal different gain) always adds a "tik". The solutions for this are far more complex than we are likely to want to pay for 256 of. We can use a "digital switch" which is really semi-analog so we can slow-up the transition with external R-C network. Which means 256 control pins and 256 R-C networks. We can "zero-cross" for simple audio systems (car radio), but sometimes a mutual-zero never comes (my car radio "volume" chip can have a long delay before it gives up waiting). 15 sources is a lot to monitor (whole lot of ADCs and data), and all different (effected different) a good zero-cross is unlikely to come "soon".

[R.G.]

I did a quick and dirty PCB evaluation. It is much smaller if you can delete the bias pulldown resistors and in/out caps, but I got the chip, and 32 I/O networks on a 4" x 1.6" area, using 0805s and 1uF NPs for caps. Might as well use SMD as the chip only comes in a 44 pin PLCC.

I felt a little silly doing that because the 32 jacks have to be spread out. Even a vertical two-stack of jacks can only be squeezed up to a box about 14" wide. An all-lateral jacks panel would need to be over two feet wide just to hold the jacks all in a row.

Of course, it is more fun than doing crosswords. 

32 jacks, 16 footswitches and [whatever] for indicators is still TBD. I thought about using one single digit LED display per effects loop to show which order it was in the loop, but really the display of what order the effects are in is a problem. It probably makes sense to grab a technologically-bypassed laptop and convert the whole thing into a display and controller for the footswitch looper. That gets you enough display horsepower to actually show you what order the effects are in. These things are often free (as in you already have one in the closet) or nearly so.

[PRR]

me> 256 control pins and 256 R-C networks

Duh. For this purpose, we could just have 16 soft-switches, one at each input. Fade-down all inputs in a couple mS, hit the matrix switching, fade all inputs back up. That's essentially one more bit, and could (with thought) be done on the data overflow pin of the matrix.

R.G.> enough display horsepower to actually show you what order the effects are in.

Agree. I've done more with less display (8 lamps for a whole mini-computer, because paper output was billable and CRTs were not in sight), but not in performance.  An off-stage drag-drop interface to pull pedal-pictures off your camera or the interwebs, then on-stage display showing the pedal-faces in actual order. (Would look like you are playing video poker with your feet: 3 to 8 rectangles flipping and moving.)

[Harold]

I was interested to see what could be done with the AD75019. Here's one thing: an 11 in/out, any order effects switcher:

Before you go off ordering the chips, note that (1) the power supply setup is a very special case, and won't work for just any power supply (2) there is a huge amount of work to get those deceptively simple "Sck", "Sin" and "Pck" pins to work right from a microcontroller, and (3) implemented properly it goes down to 2x2 and up to 15x15, not counting one in and one out for the input/output jacks.

It's funny - sometimes feel like I'm living in an amazing paradise. All the stuff that was conceptually simple but out of reach for circuit complexity reasons keeps appearing as single chips. Tee hee! 

Well, I ordered these chips. And got them to work pretty fast and easy with my Arduino and the SPI library. Timing was a bit off in the beginning, but I got that tackled too. The AD75019 still sits on my breadboard, in a PLCC44 > DIP44 adapter with +12VDC from a wall-wart, and -12VDC through an ICL7660S power converter.

The AD75019 likes an input signal impedance < 600R, so it seems to be wise to buffer the return signals.



The buffer is biassed at GND, so C2 might be redundant, as well as C3. R1 might be obsolete too, as there's no DC in that area that needs to be bled to ground.


What I'm still not convinced about is if the switches need pull-down resistors. It's pretty easy to switch all unused loops to ground in bypass, both the return buffer output, and the effect send, but I'm not sure R2 and R4 are needed as pull-downs.


Edit: ESD protection might be an issue to. All pins with external connections should have two diodes to protect them from excessive ESD shocks.

[PRR]

U1 wants a bias resistor on the + in pin.

I don't see what C2 R3 do.

[Harold]

U1 wants a bias resistor on the + in pin.

Excellent point!

I don't see what C2 R3 do.

R3 = Current limiter for the switch, especially when S1 is grounded.
C2 = filter any DC offset from the buffer U1, but it's probably not needed.

[Harold]

U1 wants a bias resistor on the + in pin.

Excellent point!

Ehm, or maybe not?

The + input of U1 is biased to ground already, because of the bi-polar power supply.

[PRR]

> The + input of U1 is biased to ground already, because of the bi-polar power supply.

How does it know where "ground" is?

How does the chip's input current flow?

[Harold]

You are suggesting a resistor to ground, between C1 and the + input?

[R.G.]

Ehm, or maybe not?

The + input of U1 is biased to ground already, because of the bi-polar power supply.

No, it's not. I think you may have this backwards in your head.

Opamps *can be* biased to any voltage within their power supply range. They are carefully designed to have no internal preferences about bias at all, and no internal "bias point" that they automatically go to. Opamps *must be* fixed at some bias point between their most-positive and most-negative supply voltage. They really don't care where in that range they're biased, given only that it's within the "input common mode voltage" range, which is smaller than the power supplies for most opamps.

Having a bipolar supply as opposed to a single supply means that the opamp *can be* biased by a resistor to ground. There is a class of opamps which are especially designed to *allow* their inputs to be biased at ground with a single supply - they're sold as "single supply" opamps. But they also work with a bipolar supply and the + input tied to any voltage between the most-positive and most-negative.

So Paul is correct - if you want this thing biased in the middle of the power supply range, you have to "tell" it where to sit by using a resistor to the reference voltage, which is likely to be ground in this case.

You are suggesting a resistor to ground, between C1 and the + input?

Yes, he is.

[Harold]

Thanks, RG! Clear explanation, as usual

I have all required components here, so I'll start with a layout for this all. I really love where this is going; the AD75019 is quite a nice chip to build this around.

I'll post some pictures when I get some work done.