Need help on a difficult order switching puzzle :(

[bean]

I'm designing a reusable template for a client that employs a very cool switching idea but I've run into a roadblock. Could use some bigger brains to help confirm my suspicion or find the right solution.

Setup:
- Dual circuit, single input and output.
- Order switching for A and B circuits via a 3pdt toggle.
- Effect insert between the two circuits (so an additional pedal or pedals can be placed between the two effects regardless of order and the bypass state for those inserted pedals is independent of this scheme).
- The client is using relay switching for each bypass but that is not relevant to the puzzle.

My suspicion: it cannot be done with the "typical" 3pdt order switching scheme. The send and return insert points would end up having to switch places depending on if the order is A to B or B to A because of the way the 3pdt toggle is wired for order switching. IOW, there is no common pole in the 3pdt for the returns of effect 1 and 2 for an "send" point. If that makes sense 

Here's the schematic. Any insight would be most helpful as I have already burned a few brain cells on this.

[Ripthorn]

Are you asking if there is a common point that can be tapped for the effect loop send using just the 3PDT switch for the order switching? If that is the case, then I agree with you that I don't think there is one.

[bean]

Are you asking if there is a common point that can be tapped for the effect loop send using just the 3PDT switch for the order switching? If that is the case, then I agree with you that I don't think there is one.

That's the gist of it, yeah.

[R.G.]

1. Is metal-contact switching mandatory? There are single chip switch matrixes that will do this and much more.

2. How sensitive is this to cost? There are solutions that appear more complicated, but are actually simple after logical reduction.

3. Will this be placed in a commercial product, or is it for one (or two...) custom builds?

You've described what I think is most readily solved with a crossbar matrix. I sketched out the basics of a non-blocking crossbar in the article at geofex on any-to-any switching. 

[bean]

1. Is metal-contact switching mandatory? There are single chip switch matrixes that will do this and much more.

2. How sensitive is this to cost? There are solutions that appear more complicated, but are actually simple after logical reduction.

3. Will this be placed in a commercial product, or is it for one (or two...) custom builds?

You've described what I think is most readily solved with a crossbar matrix. I sketched out the basics of a non-blocking crossbar in the article at geofex on any-to-any switching.

So far mechanical switching is the only option discussed. But, they might consider other options. Space is actually a critical consideration, though, since two effects with board mounted pots and switches AND two relays with switching circuitry is being crammed into a 1590BB. And, it's all through-hole! It's for a line of potential commercial pedals ~ I'm just doing the setup and layout work for them to get it started.

[R.G.]

It's definitely possible with one of the video matrix routing switches. A single chip gives you signal routing from any 8 inputs to any 8 outputs simultaneously (that's the non-blocking thing). A single "baby" uC does the control, and in my designs watches the footswitches, debounces them and runs the LEDs. The single matrix chip and uC are probably smaller and may be cheaper than the number of relays and switches.

[ElectricDruid]

I agree - I think it's impossible with a 3PDT.

The reason is there are four elements which all change. If the two effects are "A" and "B" and the FX loop is Send/Return, then:

• A_in comes from the Input jack or FX_return


• B_in comes from FX_return or the Input jack


• A_out goes to the FX_send or the Output jack


• B_out goes to the Output jack or the FX_send

I don't see how that could ever be done with only three switching contacts. Unless there's some redundancy in there that I'm not seeing, and I'm pretty sure there isn't.

[Ripthorn]

Using a crossbar, or crosspoint, like RG suggests is super flexible, but from what I saw when I ordered some last year, they are all QFN SMT packages, which is typically outside the capabilities of most DIY'ers, especially if everything else is through hole. I have not seen any through hole cross point switches, personally, but they may be out there.

[R.G.]

They may be obsolete by now but there used to be 8x16 crosspoints in 40 pin dips, for telecom applications. They passed pedal audio just fine.
But the current crop may be all SMD. I think maybe we all have to get used to through hole going obsolete, in fits and starts. Most companies that introduce new ICs will make them only in SMD.

A useful dodge to a situation like this is maybe to use the SMD spreader boards that adapt a QFN or similar to through hole.

In the long run, anyone who even thinks about selling pedals really ought to bite down hard and go learn to use SMD. It's not going away. You need good magnifiers (or I do, old fogey that I am), probably smaller soldering irons and tips, liquid flux, and a steady hand. I've watched fine pitch QFPS and QFNs soldered by hand. It's a skill that can be learned.

[PRR]

> I have not seen any through hole cross point switches, personally....

As R.G. says: they were around, readily available, working good, back in the last century. They came up as 3rd-choice in several of my wild projects. Some back warehouse of CBS or the military video-production platoons has crate-fulls.

Hey: Futurlec??

MT8816A 8 x 16 Analog Switch Array IC    Datasheet    Stock:Yes    PDIP40    $4.50
https://www.futurlec.com/cgi-bin/search/search.cgi  "MT8816"
https://www.futurlec.com/Zarlink/MT8816AE.shtml

[Ripthorn]

I have moved towards SMD for a lot of things; my favorite MCU is SMD only and my projects are getting more complicated. I will admit that grabbing one of those through hole cross points is tempting for one project I have in mind, as space is a little less of an issue on that one. Thanks, PRR!

[Rob Strand]

Back in the days of Motorola and through-hole the Communications data book had chips for phones and for telecommunications industry infrastructure.    There were through-hole cross point switches in there for sure.    Not the things for new projects.

The last cross-point switch I used was from Analog Devices (smd).   Lets face it the switch technology nowadays is much better than the bad old days.

[R.G.]

MT8816! That was the one! Worked fine. It was a little bit of a pain to program IIRC. But it did the job.

[Rob Strand]

You could probably use a 3PDT switch where one of the switches drives a one DPDT relay (or solid-state equivalent).

[R.G.]

Hmmm. One of the ways to make a crossbar is to use a number of one to N switches. A CD4053 is a dual one-to-four switch. A CD4051 is a single one to eight switch. The problem here is (if I haven't miscounted) a four by four. I think, without actually doing the drawings, that it might be done with two CD4053s, or maybe four or five 4051s. That stuffs the problems back into through hole dips, although at the price of some more complexity on driving the address lines of the MUX chips.

Edit: belay that. (Note to self, do thinking when NOT consuming evening beverages.) The 4053 is a quad one to two; the 4052 is a dual one to four, the 4051 is a single one to eight. The address lines are common. So it would need to be at least four 4052s, each half used.

[PRR]

The CD405X are "one-of-X" switches.

Crossbars generally, and MT8816 specifically, seem to be "any/all-of-X". At an extreme you can connect everything to everything. Which may not be useful, but covers some situations better than one-of switches.

The flip side is of course a long string (128) of set/unset instructions. Not gonna do that by hand even sober. It is a nano-CPU and some pre-thought to get most useful configs at a fingertip (toe-tip?).

[R.G.]

There are various kinds of crossbars. One of the sub-varieties can be viewed as a set of 1 to N switches. In this case, the input jack can be connected to: the output jack, FX1in, FX2in, Send jack, or the output jack (assuming that it will be done with bruted force in "bypass all". It never connects to either effect out or the Return jack.
In addition to the input jack connections, FX1 output connects to either the Send jack, FX2 input, or the output jack.
In addition to the previous, FX2 output connects to one of the Send Jack, FX1 input (for order reversal), or the output jack.
In addition to the previous, the Receive jack connects to one of FX1 input, FX2 input, or the output jack.

I'm doing this in my head, so I may have missed something that I'd see in drawing it out. If you connect the "one to" input of a one-to-N to the input jack, FX1 output, FX2 output, Receive jack, and output jack, that one "switch wafer" can do all the possible switching for the input jack. So a one-to-five will do the job.
A one-to-four will work for FX1 output and FX2 output. A one-to-three would work for the Return jack.

So I think (with my mental drawing space overloaded) that four one-to-five (or more) would do the necessary switching. It's a complex-ish control setup to make it work. It needs a uC to filter the switch requests for bypassing and reversal, then setting the one-to-X to the right addresses, but ... maybe...

The manual version of this is shown in the "programmable FX switcher" here:http://www.geofex.com/article_folders/fxswitchr/fxswitchr.htm

I mention crossbar, because this is a sub variety, limited and blocking compared to a full nonblocking crossbar.