PROJECT: Crossbar

[Peter Snowberg]

What's the Crossbar?

I'm probably going to offer this design commercially so I'm going to make it available for DIY use and if you want to build them commercially we need to talk licensing.

The Crossbar is an effects signal switcher which uses a special circuit called a crossbar switch to route signals from a guitar, through up to 7 out of 8 possible effects loops, IN ANY ORDER, and finally out to the amp. The switch here is actually an array of 128 independent CMOS switches arranged in a matrix and driven by a microcontroller. Switch resistance is less than 75 ohms and signal swing can be up to +/- 6 volts.

There are lots of ways the user interface for this device can be configured and naturally it's nice to be able to have the stompswitches remote from the gear. This is the real motive behind introducing MILAN when I did.

The Crossbar is composed of two parts, the controller and the switching matrix / signal routing circuits.

The basic controller model is a floor based stompbox housing three or four switches, each with an accompanying LED, and optionally a dual 7 segment display. These switches and LEDs are not linked inherently in any way. The switches transmit MILAN information and the LEDs respond to MILAN information. The box has two external jacks; one 1/4" TRS for MILAN, and a 2.1/5.5mm coaxial power connection. DIY versions of this box may contain more switches and LEDs.

The switching matrix is a larger format box because it supplies up to 20 1/4" jacks for routing signals through up to 8 loops. The matrix also includes 1 or 2 low noise, high impedance buffer circuits for the inputs as well as a couple of NE5532 buffered amplifier sends. The matrix uses MILAN information to either select and deselect individual effects loops, or to select preset configurations. A dual color LED per loop indicates status and a dual 7 segment LED shows the preset number in use. A couple of switches provide preset programming.

The use of MILAN allows the controller information to be used by other devices. At the same time, the matrix is able to respond to control messages from other equipment.

OK, now for the big question......

Does anyone want to build this?

[DavidS]

Sounds pretty sweet, but how does this fit in with your WaveFront DSP signal router/blender/mixer/eq project? THAT's the one I'm REALLY chomping at the bit for. Is the DSP stuff still  a ways off?

I think I'd probably be pretty interested in the Corssbar, though...

[Peter Snowberg]

The DSP stuff will enter into the picture after this part has been successfully completed.

MILAN is a big part of the picture so I want that to be here first.

The Wavefront chips have lots going for them, but they're not the end all and I don't see the commercial possibilities that existed a couple of years ago when I designed the stuff originally. They're still good DIY learning tools but this jump into digital around here is a big one so no DSP for a while.

We need a solid foundation with microcontrollers and digital communications first and then look out. 

[DavidS]

OK, Crossbar it is then!

[David]

Yes.
Yes!
Yes!!
Yes!!!

[R.G.]

Unless you have already done the design and layout on the crossbar, I can save you some time.

I was into crossbars heavily back when I did the prelim work for ASMOP. I actually set up an 8x8 crossbar using reed relays for crosspoints. Talk about a glutton for punishment.

In the process of thinking about how a crossbar would work, I simplified it down to eight one-of-eight selector switches, and possibly a few other housekeeping switches. It does the same job as a full crossbar, less the ...er... depraved modes available from a crossbar that you'll want not to use anyway.

The one-of-eight is a CD 4051. Eight and possibly nine of these gets you effectively an any-order eight way loop selector.
http://geofex.com/Article_Folders/ASMOP/asmop1d.gif is the controller interface, and
http://geofex.com/Article_Folders/ASMOP/asmop1e.gif is the switch matrix

The control is simple. You need three bits for each cd4051 to set it to one-of-eight. I used CD4094 shift registers to shift in all of the switch setting data; a 74HCT595 might be better, but both work. Four pins on the uC do all the work for the matrix. You need a data pin, a shift clock, and an "transfer" clock to move the shifted data into the output pins on the shift register. The fourth pin is a selective mute on the end of the audio line to edit out any switching transients at a microsecond level.

I believe this does all of the useful work of a crossbar - at least for an 8x8 mono crossbar. It won't do stereo; you'd need another eight CMOS chips for that 

[R.G.]

Oops, forgot to add - there are also some crossbar switch chips. I have some samples from Maxim that seem to be OK. They're mostly stereo, but you can ignore the other side.

Believe me, wiring up a crossbar from individual swithes is not something I'd gladly do again.

[Peter Snowberg]

Believe me, wiring up a crossbar from individual swithes is not something I'd gladly do again.

Nor would I.

[Peter Snowberg]

Only two people interested in a switching matrix??? Hmmm. 

[R.G.]

It seems like a shame, doesn't it?

I got to the point that I'd toss out the info and see if there was any interest before I pursued things much. I got ..two.. responses to my CMOS crossbar.

That's one reason I never pushed the uC harder, other than noting where it was a better answer for a few years.

[Khas Evets]

For me, it's not lack of interest but lack of understanding. I suspect there are many people (like me) here who are just cutting their teeth on analog circuits. Microcontrollers are still on the distant horizon.

I think the Digital & DSP section is great, but I'm nowhere near understanding this yet. Maybe a beginner project would get us interested and un-intimidated.

[dano12]

Only two people interested in a switching matrix??? Hmmm. 

I would be very interested. It seems that many of the skills are still beyond my ability. I read R.G's excellent article, but the amount of work seems daunting. If we could get a basic PCB layout working and get ggg to sell them, that would be a huge boost.

I am just finishing this:

and I quickly gave up on a switching bus because of the complexity. Instead, it has in/out jacks in the back. Not ideal but very effective. If I could have put a simple CMOS or relay based switchboard in, I would have considered it.



I also very intrigued by the idea of starting to build circuits and mount them in standardized bus packages. Something like the great VectorPak stuff that Craig Anderton used to use. Unfortunately, the cost of VectorPak pieces are unbelievably high. (A great alternative is offered by PAIA with their very cost effect FracPac system.)

I would love to enjoy the same flexibility and cool designs that analog synth folks do--a big modular panel that allows me to quickly build and mount effects circuits using a common bus. And to use a cool switching circuit to control fx order.

It just doesn't seem to be enough critical mass until someone (or small group of folks) puts together a prototype system and unleashses it on the internet.

I, for one, am a bit bored with the whole "fit it in a smallish Hammond enclosure" mentatlity. We could learn a lot from those awesome big panel analog synthesizers with all the great twiddly bits.

[Dave_B]

Well, the good news is that this thread has been read nearly 200 times.  About five of those hits are mine, but that's beside the point.   

It's obvious to say, but this is a whole different thing compared to the analog side of the forum.  The cost of admission is higher.  The learning curve is  steeper.  A person can carefully build a lot of analog boxes without spending a lot and without learning too much about electronics.  Over here, we're buying $80 programmers, diving into assembler and C, and trying to make an LED blink without setting off the watchdog timer.   

When we get a couple of killer-apps (with layouts) under our belt, I think things will pick up.  Even then, we're going to be offering pre-programmed chips for a while, I'd bet.

It's still going to be fun!

[Peter Snowberg]

Thanks for the input.

I understand how this might be intimidating but if the microcontroller is already programmed it's just another perfable design. This thing is actually surprisingly simple thanks to LSI (Large Scale Integration). All the switching is handled by a single chip. A pair of opamps provide the buffering and the microcontroller runs the show. There's a little more to it than that but not too much. It doesn't even use SMD parts. 

Perhaps we need a little implosion to achieve critical mass on this stuff.

R.G., I guess that's what you get for being too many years ahead of the curve.

[Dave_B]

Well, having said all that, I do think the crossbar is a great place to start.  Just last night I was wishing I could turn on my overdrive and turn off my chorus with one push.  With any luck, I'll have that momentary/toggle switch incorporated.  Momentoggle for short? 

[R.G.]

For those of you who don't want to tackle something that big, you may find some commercial product in the near future that will help  you out... 

For those of you who want to do it yourself;
To just flip the states of several effects at one time, read:
http://geofex.com/Article_Folders/rmtswtch/rmtsw.htm

Doing any-order is a bit more complicated
First  you have to understand the any-order-of-eight selector switching. Read
http://geofex.com/Article_Folders/fxswitchr/fxswitchr.htm
until you understand how a rotary switch can do any-order by selecting different banks of rotary switches. In this all-manual implementation you don't need a uC to do the programming. You set up several banks of rotary switches, and then pick a bank of rotary switches with the single-contact reed relays. Notice that this is a viable option even today; it requires no electronics, only some careful wiring. ]

The key idea is that a rotary switch switches one input wire to the input of any of eight effects. The input jack (which paradoxically must be viewed as the guitar's output)  gets the first rotary switch pole, and it can be switched to the input of either seven effects or the output jack, which counts as one of the places for the input to go. The output of each effect also gets its own rotary switch, so it can go to the input of any of the other effects. You leave out the wires that would let you connect the effect to its own input, of course. By setting the rotary switches, you can systematically connect the output of any of the effects to the input of any of the other effects. There must be at least one chain of effects from the input jack to the output jack or no signal will go through. That's where the user interface gets sticky.

From there, all you have to do is substitute the CD4051 one-of-eight selector switch for the rotaries and use a uC for controls; then you have an any-order matrix. The uC reprogramming the CMOS "rotary switches" each time substitutes for switching in and out banks of rotary switches.

Then look at
http://geofex.com/Article_Folders/ASMOP/asmop1d.pdf
where you'll find a pre-existing PCB layout for the any-order of eight switching matrix. There is a $2.50 PIC on there, so all's you gots to do is program it. In fact, you could program an AVR and adapt the layout. Once you understand how the switching works, it's  A Simple Matter Of Programming. With the layout as shown, you could just enter that layout into your favorite board editor and have a single sided PCB to do the work within a couple of hours. I believe that the layout as shown would work in press-n-peel. You might want to slice it into two halves for easy of etching. Or you could have a commercial board made. Or, frankly, you could easily wire-wrap it or hand perfboard it.

Now that I think about it, the implementation as shown practically begs to be put on stripboard. The PCB is almost entirely lateral strips with horizontal jumpers. It may be the most natural coversion to stripboard that I've seen.

[aron]

I'm interested as well, but I look at some of those projects and WOW, they look complicated!

[Alex C]

I would like to state my interest as well, and that I'm in the same boat as many others: no experience with digital.  I am really excited by all of the possibilities, though (as cheesy as that sounds), and this looks/sounds very promising.  Thanks Peter and R.G.

[R.G.]

I'm interested as well, but I look at some of those projects and WOW, they look complicated!

I would like to state my interest as well, and that I'm in the same boat as many others: no experience with digital.  I am really excited by all of the possibilities, though (as cheesy as that sounds), and this looks/sounds very promising.

This is an interesting point. We're confronting some of the basic responses to logic systems here.

The "wow they look complicated" is valid. They DO look complicated. However, there is a repetition there that hides an underlying simplicity. What matrix, crossbar, barrel and other highly interconnected systems are composed of is one basic concept repeated over and over. My implementation of any-order is based on a one-of-N selector switch. Once you understand that a selector switch can take one signal and select any one of a number of places to send it to, you have the concept nailed.

When one makes eight copies of a selector switch, things start looking messy. It's when you add on top of that a slight difference per copy that normal human eyes start glazing over. But it's a lot like you learned to build effects. This is a resistor. It has two ends. Either end is identical. This is a capacitor. It has two ends; some capacitors you have to remember that one end is positive. This is a transistor, three leads. Only one more - but they are very picky about which is which.  Now we take four resistors, two capacitors, and hook them together in special ways. And solder...

It's all building blocks. Aron, as you know, if you skip a building block, it WILL come back and bite you. It's not that any one block is crucial or incredibly complicated. It's that you have to assemble blocks. If you carry the attitude that you have to learn a few basic blocks, and that each new project may have one or two new blocks involved, you'll do fine.

It's like in Hack - how does one attack a long worm? Recursively.

Not knowing digital is a bit more basic. It's the analog to not knowing anything about electricity. To do effects on more than a "hook wire 23 to resistor 15 and solder; wrap resistor 15 around pot lug 3 and solder:" basis, you simply have to know a little about electricity. It's not tough, as the blocks are things like:
- electricity must flow in a complete loop from the electrical source through the circuit and back to the source.
- things can be more positive or more negative; what changes that is where you decide "ground" is
- "ground" is any place you decide is going to be 0.000V for a reference.
- grounded metal shields help keep hum and RF out
The list goes on, but you see where this is leading, right? There is a certain irreducable minimum of blocks you have to learn.  There is an infinite number of blocks, and as you collect them, the next blocks get easier to pick up.

For those not having an inkling about digital, you have to learn a little. For instance:
- Digital circuits depend on interpreting voltage levels as a 1 or a 0, on or off.
- Usually we decide that a higher voltage is a "1" in the positive voltage system
- Rarely and seldom we decide that a lower voltage is a "1", in the negative logic system. If you ever want to totally confuse experienced logic people, make them work in negative logic. 8-)
- Logic signals can be combined to get specific logical results. For instance, an AND gate puts out a 1 if and only if all of its inputs are 1. If any of them are 0, then the output is a 0. An OR gate puts out a 1 if ANY of its inputs are 1. A NOT gate puts out the opposite of its input.
It's the same process. You learn one block, then go to the next.

But - I can hear you thinking - when do I have enough blocks?
I'm still collecting blocks after almost four decades of this. But the first ones were the toughest, and you only need a few basic ones before you start to say... oh, I see!

I think that we ought to advise people who have any inkling that they want to mess with uCs that are concerned at all that they don't know anything about digital to go buy a copy of the CMOS Cookbook by Don Lancaster. It goes over a lot of this in very readable form. I have worn out two copies. This is available from abebooks.com for as low as $1.60 plus shipping. It's kind of like having a multimeter.

Good points, and very apt for the folks here.

[Peter Snowberg]

I think that we ought to advise people who have any inkling that they want to mess with uCs that are concerned at all that they don't know anything about digital to go buy a copy of the CMOS Cookbook by Don Lancaster. It goes over a lot of this in very readable form. I have worn out two copies. This is available from abebooks.com for as low as $1.60 plus shipping. It's kind of like having a multimeter.

While everyone is at it, it's not a bad idea to pick up the Active Filter Cookbook too, also by Don Lancaster. Two books on the must have list. 

Digital may look complex, but you don't have to worry about biasing transistors. 

I'm getting close to posting the schematic and maybe that will quell some digital apprehension.