Pedal switching matrix with the AD75019

[R.G.]

"Really need" is hard to call. Isolation is always welcome when there are lots of things which have separate power supplies and then have their signal grounds connected.

In a moderately benign setting you might get away with a servo-ed ground setup, using one opamp per output to remove the ground offsets.

[G. Hoffman]

Oh, and a lot of thanks to R.G. for the starting point on a lot of the blocks, particularly the transformer isolated outputs!

Wait, WHAT?

I totally missed that one.

It certainly does not hurt, electronically speaking, to have all those transformers isolating the sends.

But it DEFINETLY will hurt your wallet. Do we really need all those?

Well, for this I kind of threw everything at it, with the idea of making it kind of bullet-proof.  But the transformers in question are actually pretty cheap - less than $4 a pop.  Not too bad when you are already using a $30 switching chip!  Once I've actually breadboarded it and (hopefully) got it working, I'll probably try to figure out what I actually need.


Gabriel

[Hatredman]

Even though, there is a bunch of them. I have a problem with transformers. They pick up hum, distort, have limited BW, are heavy and claim a lot of PCB real estate.

So my question remains: are all of them really needed?

[Thewoodguy]

Yeah I would have to think that is a bit over kill

[amptramp]

Keep in mind there are always ways around audio transformers for isolation.

You can go linear optoisolated or use RF modulation and demodulation, both of which can maintain bass performance down to or near DC.  PLL chips like the NE565 are good building blocks for an FM modulator / demodulator system.  They can be coupled by RF transformers or capacitors.  $4 transformers are going to sound like audio strangled through $4 transformers unless you happened to get a great deal on decent audio quality transformers, but the usual 600 ohm: 600 ohm audio coupling transformer cuts off above the low E string guitar frequencies and rolls off in the low harmonics of the high E string frequencies.  Decent transformers are also heavy - if you get good ones, you will have to wheel this thing around on a pallet truck.

[R.G.]

$4 transformers are going to sound like audio strangled through $4 transformers unless you happened to get a great deal on decent audio quality transformers, but the usual 600 ohm: 600 ohm audio coupling transformer cuts off above the low E string guitar frequencies and rolls off in the low harmonics of the high E string frequencies.

Insert a big asterisk here. $4 audio transformers are going to sound like $4 transformers when used all by themselves. The limits on audio passband are almost always primary inductance and leakage inductance, with a nod to inter-wire and inter-winding capacitance. In a small, cheap audio transformer, inductance is usually the big issue, and it limits low frequency response.

To come up with better frequency response from the cheapest isolation trannies I could find, I used Xicon's $2.50 audio types. These are rated at 300Hz to 3kHz. A little measurement on them showed that the ones I got went from about 240Hz to 28kHz at -6db. There was a hump at 22kHz. The low frequency problem is classical: the primary inductance siphons off signal current from the signal source that doesn't get into the M-field in the core. For the rated source impedance, this happens at the point where the inductive reactance is equal to the source impedance. If you're living with a passive primary drive, you're stuck there.

But if you have active primary drive, you can just lower the source impedance by driving the primary with an active buffer. An opamp will do this nicely, and I got to a low frequency rolloff of about 60Hz doing this. Another scheme is to use an amplifier with a matched low frequency hump that counters the tranformer's rolloff, just boosting the low frequency content to make up for what is lost. Either of these only work until the core starts hitting saturation, but saturation is usually not an issue so much with the 1" cube stuff and guitar or line level signals. It's worth noting that this same trick, drive the lows harder to make up for low frequency rolloff, has been used to equalize woofers to much lower than their normal low frequency rolloff.  Lower source impedance has the advantage that you don't have to know the transformer's characteristics to compensate for them.

It's a trick, but it does work within its limits. Opamps are far cheaper than fancier transformers.

[Thewoodguy]

I'm going to order a one this week and start programing it first. Then mess with that stuff. Ahhh programming see you guys in 2 years haha

[PRR]

> optoisolated or use RF modulation

Less-radical techniques include differential amplifiers.

Input diff-amps are will known. Output ground-sensing amps are also possible, as R.G. mentions. .

These are problematic here because they rely on "matched" resistances, including source resistances. In guitar-land impedances are high and we we never know their real values, so we have to buffer.

> Do we really need all those?

"Sometimes"!

Consider three different cases.

1) You take/get signal from a remote location, possibly on another power system. Ground potential differences can be 100V or more. This is (should be!!) super-rare in stage-work, even up to stadium gigs, but stuff happens. It is a constant problem in long-line work, radio broadcast audio feeds and telephony in general.

2) Everything is on the "same supply" but ground and neutral are crossed. Any power load causes ~~1V AC ground difference between box "ground". This is fairly harmless to humans but clobbers audio with hum.

3) Power supply is properly connected. Ideally "ground" has ZERO current, thus zero voltage difference. However "zero" is only constrained to "safe leakage". Many-many boxes throw milli-Amps of leakage to ground. Safe to us, but 10mA times 0.1 Ohm grounding conductor resistance is 1 milliVolt of hum. As low-level guitar can be 20mV, 1mV of hum is significant, very annoying.

Working solutions backward:

3) Good diff-amp implementation can put the ground-difference 100:1 down, which may get you from "annoying" to "hardly audible". (However poor diff-amp may give less than 10:1 reduction.)

2) For Line Level signals peaking over 1 Volt, and not-large power current in grounding conductor, a good diff-amp can get hum down to "tolerable" for lower-fidelity work. In practices no diff-amp is really satisfactory, and transformers (or opto- or RF) are needed.

3) When ground differences exceed a few Volts, there is also a Safety issue for the operators on both ends. Transformers (or equiv) are required.

Note a common thread: ground differences due to ground here being different from ground over there. The First Defense for most such trouble is to put ALL your stuff on ONE power-strip. (A good one, where the ground holes grip tight and are really connected to each other.) Often you need no further gimmickry. Sometimes there's low hum and 50-cent diff-amps mitigate this.

1:1 transformers output transformers at opamp impedances can have *extreme* frequency response. Like 5Hz-400KHz. If you don't expose them to >100V ground differences. Transformers *tightly* bound to an opamp can have response much better than the $3 specs say, as R.G. (and others) have proved: make the opamp work to fight transformer flaws.

Input transformers from high source impedances are a much tougher problem.

> Do we really need all those?

Do we even want ALL those? Early audio was simple: if the butcher understood your order, all was fine. But when talking movies got fancy, and an audio path might extend through a dozen or more transformers, innocent engineers had to admit there was a problem on transients (tap-dancing was a craze). Improved transformers were much better and many-iron paths became common in fine audio. However the improvements are not cheap.

So as a rough scale-out rule-of-thumb, figure that twice as many transformers should cost at least FOUR times as much, because with twice as many in the chain each one has to be twice as good to maintain tolerable degradation. That's ideal; in today's market there is a long jump from Cheap to Good. The next step up from $3 may be $20.

That says, roughly: if a $3 part is good enough for an A/B/Y path, a 15x15 matrix with potentially 15+15 transformers in one over-wrought path needs $2,700 of transformers. This alone suggests reserving transformers for the interconnects where they are really needed.

[Thewoodguy]

> optoisolated or use RF modulation

This alone suggests reserving transformers for the interconnects where they are really needed.

Interconnects?

[G. Hoffman]

$4 transformers are going to sound like audio strangled through $4 transformers unless you happened to get a great deal on decent audio quality transformers, but the usual 600 ohm: 600 ohm audio coupling transformer cuts off above the low E string guitar frequencies and rolls off in the low harmonics of the high E string frequencies.

Insert a big asterisk here. $4 audio transformers are going to sound like $4 transformers when used all by themselves. The limits on audio passband are almost always primary inductance and leakage inductance, with a nod to inter-wire and inter-winding capacitance. In a small, cheap audio transformer, inductance is usually the big issue, and it limits low frequency response.

To come up with better frequency response from the cheapest isolation trannies I could find, I used Xicon's $2.50 audio types. These are rated at 300Hz to 3kHz. A little measurement on them showed that the ones I got went from about 240Hz to 28kHz at -6db. There was a hump at 22kHz. The low frequency problem is classical: the primary inductance siphons off signal current from the signal source that doesn't get into the M-field in the core. For the rated source impedance, this happens at the point where the inductive reactance is equal to the source impedance. If you're living with a passive primary drive, you're stuck there.

But if you have active primary drive, you can just lower the source impedance by driving the primary with an active buffer. An opamp will do this nicely, and I got to a low frequency rolloff of about 60Hz doing this. Another scheme is to use an amplifier with a matched low frequency hump that counters the tranformer's rolloff, just boosting the low frequency content to make up for what is lost. Either of these only work until the core starts hitting saturation, but saturation is usually not an issue so much with the 1" cube stuff and guitar or line level signals. It's worth noting that this same trick, drive the lows harder to make up for low frequency rolloff, has been used to equalize woofers to much lower than their normal low frequency rolloff.  Lower source impedance has the advantage that you don't have to know the transformer's characteristics to compensate for them.

It's a trick, but it does work within its limits. Opamps are far cheaper than fancier transformers.

The basic block for these, as I'm sure RG noticed, came from the transformer isolated ABY box on GeoFX.  I've used them as simple transformer isolated spliters before on guitar with very good results, and I just dropped the same block in here, but with only one output, instead of two.  They really do sound surprisingly good, given the cheap transformers (which have gone up a bit in cost since RG wrote the article!), and I really like the certainty of transformer isolation when you have as many opportunities for ground loops as exist in a switcher like this. 

Gabriel

[G. Hoffman]

1) You take/get signal from a remote location, possibly on another power system. Ground potential differences can be 100V or more. This is (should be!!) super-rare in stage-work, even up to stadium gigs, but stuff happens. It is a constant problem in long-line work, radio broadcast audio feeds and telephony in general.

2) Everything is on the "same supply" but ground and neutral are crossed. Any power load causes ~~1V AC ground difference between box "ground". This is fairly harmless to humans but clobbers audio with hum.

3) Power supply is properly connected. Ideally "ground" has ZERO current, thus zero voltage difference. However "zero" is only constrained to "safe leakage". Many-many boxes throw milli-Amps of leakage to ground. Safe to us, but 10mA times 0.1 Ohm grounding conductor resistance is 1 milliVolt of hum. As low-level guitar can be 20mV, 1mV of hum is significant, very annoying.

Power is never an issue at stadium gigs - because you bring your own power supply.  You take three phase from the venue (or, if you're having a good day, you have your own generator!), and use your own power distribution.  Separate power distros for lights, sound, and video.  It makes life really easy, unless the lighting guy wants to run his DMX line to front of house through the audio snake!!!!!

But that's the world I come from - running lights and sound in high end corporate and music venues.  Transformers are EVERYWHERE.  Granted, usually more expensive transformers, but not always.  I've killed a lot of ground loops with a $30 in-line Shure isolation transformer.  Saved my butt on a regular basis.  For a mono version of this, you're looking at 9 transformers on the sends and the output.  If you design your ground returns right on the circuit board, everything else SHOULD be fairly easy to keep quiet, but when you start messing with the outside world....

Still, if you really don't want them, just jumper across the footprints.  The op-amp driver will be plenty capable of driving any pedal you'll ever have.  Personally, I like a bit of belt and suspenders.


Gabriel

[Thewoodguy]

Has anybody looked at how the boss es8 is built? It would seem that it uses one of these matrix switches. I would like to open one of those up to see how they use it.
Gabriel, do you know what your price for parts on this project is up to?

[G. Hoffman]

Has anybody looked at how the boss es8 is built? It would seem that it uses one of these matrix switches. I would like to open one of those up to see how they use it.
Gabriel, do you know what your price for parts on this project is up to?

Not exactly - most of it was stuff I have on hand already (I like to buy in bulk!).  I think the main cost is always going to be the AD75019 - at almost $30 a chip, well...

I figure, even making it with mostly SMD parts (I love SMD boards, myself!), it's going to be a pretty big board, though, so I'm thinking of making it pretty modular, and MAYBE even using a board edge connector to separate input/output boards.  In my head (which is the only place the mechanical design lives yet), it is a lot like a fully modular console - along the lines of an SSL, Neve, or API in the studio world, or the really high end Yamaha (PM4K, etc.) or Midas (XL4, etc.) analog consoles in the live world.  That way, all the repeated stuff can be made in higher quantities, and you only need to make the ones you really need.


Gabriel

[G. Hoffman]

Has anybody looked at how the boss es8 is built? It would seem that it uses one of these matrix switches. I would like to open one of those up to see how they use it.
Gabriel, do you know what your price for parts on this project is up to?

Oh, and I'm not sure about the Boss, but I'm pretty convinced the FAMC Liquid Router is using this chip.  For $1,400 and a 6-26 week wait.  (Probably worth it, to be fair, as he does really nice work - I'm a big fan of my Liquid Foot controller - but damn pricey, and not real easy to control.)


Gabriel

[Thewoodguy]

Just looked in to that liquid switch. Really curious if he did that with just one 16x16. The info isn't very good on what you can do with it. It would seem he is doing summing inside the chip, seeing it has 16 loops. Unless he is using multiple/bigger chip. On the "editing" video he has, he is doing all kinds of parellel loops and crazy stuff. I would never spend 1400 on that but would love to have it!

[G. Hoffman]

Just looked in to that liquid switch. Really curious if he did that with just one 16x16. The info isn't very good on what you can do with it. It would seem he is doing summing inside the chip, seeing it has 16 loops. Unless he is using multiple/bigger chip. On the "editing" video he has, he is doing all kinds of parellel loops and crazy stuff. I would never spend 1400 on that but would love to have it!

Well, I've never tried the Liquid Router, but the build quality of my Liquid Foot is top notch, which tends to give me a pretty good feeling about him (and I pulled it apart before I ever started using it - as one does).  His documentation isn't always great, but his forum is good, he tries REALLY hard to be responsive to any problems people have, and when people have good ideas for software updates, he incorporates them as possible. 


Gabriel

[Hatredman]

This alone suggests reserving transformers for the interconnects where they are really needed.

My point exactly.

I can see it being needed at the amp outputs of the matrix, because a faulty ground on the backline can kill you.

But on the loop sends and returns, where you will not connect to the venue's systems but only to your own pedals, maybe it's overkill. On those spots maybe a ground lift switch would be cheaper and less cumbersome than a transformer.

[Thewoodguy]

This alone suggests reserving transformers for the interconnects where they are really needed.

My point exactly.

I can see it being needed at the amp outputs of the matrix, because a faulty ground on the backline can kill you.

But on the loop sends and returns, where you will not connect to the venue's systems but only to your own pedals, maybe it's overkill. On those spots maybe a ground lift switch would be cheaper and less cumbersome than a transformer.

Ok thanks for the clarification. But if we are going to go that far, wouldn't we be telling everyone they need them on the out put of any pedal?

[G. Hoffman]

Bench clean, lets get started!

(I would have started a week ago, but my amp started acting up, so I had to fix that first.)



Gabriel

[Thewoodguy]

Keep us updated! Wish my work bench was clean like that