Dual pedal, Analog and Digital in same enclosure

[CodeMonk]

Anyone ever build a dual pedal with one being analog and the other digital in the same enclosure?
I'm guessing you might have to use a 1:1 transformer for at least powering each one.

Just a though I had late last night.

Maybe more "gotchas" but I just woke up

[PRR]

Why? Lots of stuff is analog and digital in one box and one power source. PCs (with sound cards or VGA outputs). Car radios. Cell-Phones!!

You are going to have to get specific before you can begin to think why you might have problems.

[R.G.]

If you want to be really tricky, put digital and analog stuff on the same PCB.

Oh, wait. All the digital effects do this.   

What it takes is good attention to power decoupling and grounding practices.

Make that *GOOD* attention to grounding practices.

[CodeMonk]

Its just that I had an early morning thought (before caffeine and with maybe 3 hours of sleep) of building dual delay that ran in parallel in the same enclosure.
And using the ROG Splitter/Blender with it.
Maybe I just need a drink, or 12.

Its probably something I'll never build, or if I do it would be at least a year or more down the road.

And yeah RG, the grounding part of it was something I was wondering about.
Got any suggested good reading on that?

[mth5044]

And yeah RG, the grounding part of it was something I was wondering about.
Got any suggested good reading on that?

I have the same question. RG, I have your PCB Layouts book from many moons ago, and I recently went through it again, but didnt see anything on grounding planes/ analog/ digital. I suppose that is even more niche than musical analog circuits, but layouts for things like BBD modulation and whatnot probably benefit from correct grounding as much as decoupling.

There is some good info in a thread by DrAlx on his flanger project on power manament, but I can't find much on ground planes.

[electrosonic]

I like the "Circuit Designer's Companion"

http://www.amazon.com/Circuit-Designers-Companion-Third-Edition/dp/0080971385

I have an older addition, but it has lots of good info without being too academic. Pricey though, the edition I bought cost much less.

Andrew.

[R.G.]

The Circuit Designer's Companion is very good. I highly recommend it.

First: you're dealing with VHF and UHF - if not higher - radio frequencies. The edges on logic signals are designed to be as sharp and fast as possible, and as Fourier taught us, vertical edges and sharp corners mean high frequencies. The designers of logic chips have usually made their signals as sharp and fast as they can. That means that the PCB traces are inductors, capacitors, transmission lines, and antennas, simply by their shape and spacing. This means at least:
- any conductor of approximately 1/4 wavelength at any of the frequencies up to the highest harmonic generated is an antenna, and will radiate noise
- any conductor near a ground is a tuned circuit, and will cause ringing squarks of RF when excited unless damped
- the inductance of a conductor is far, far more important than its resistance in terms of causing voltage drops due to current flow.

That last is why you have to do ground planes. The wider and flatter a conductor is, the lower its inductance, and planes are the logical conclusion. Actually, you'd like to do matching power and ground planes, but with a good decoupling strategy you can get around this for teeny-tiny digital stuff like goes into pedals.

Decoupling the power supply is a huge deal. You want reservoir caps of 1uF to 4uF tantalum (for low ESR and ESL) no further than 1-2 inches of power+ground trace from any digital chip. That keeps the power 'bucket' local. But it's not local enough for fast edges, so you want 0.1uF monolithic ceramic, 0.01uF mono ceramic, and 0.001uF mono ceramic right at the power pins of the digital chip, from V+ to ground pins. Ideally, you want zero lead length and trace lengths on these caps - seriously - and so SMD is a real step forward in keeping things quiet. You'd like to have one of each kind of cap on every power pin, but there's not enough room in only three dimensions to do that, so you settle for interspersing 0.1, 0.01, and 0.001 to vary the ESR and ESL on the traces, and usually pick 10uF to 22uF aluminum electros instead of tantalum because of tantalum's nastinesses.

The point of decoupling is that there is inductance and resistance in the traces back to where the power comes from, so you want local 'buckets' of charge to satisfy the high frequency current needs of the chip, and only pass along a highly smoothed version down the traces to the power supply. Since it's impossible with most chips and caps to get the cap directly from the power pin to the ground pin, you use a ground plane under the whole chip, for low inductance, and place the caps with one pin as near the power pin as you can get it, and the other pin to the ground plane. The electros have hopelessly big ESR and ESL, but they store a lot of charge, so you might as well place them further away - an inch or so. The mantra of "short and direct" for leads and trace routing exists because every fraction of a millimeter of trace adds inductance.

OK, so now I have you good and scared of digital edges. What about the analog ground?

There is only one way to make a (non-super-)conductor not have zero voltage across it: make the current through it zero. We can't do that, because both the digital and analog circuits have to flow their own operational currents through ground or they don't work. What we can do is force the analog currents to flow in an analog ground, and the digital circuits to flow in a digital ground, and that is what all mixed-signal stuff does. The analog and digital grounds are separate all the way back to the power supply, in a demonstration of star(ish) grounding. That would be perfect if the digital stuff and analog stuff in the same box didn't trade signals, but for A-D and D-A work, they have to share a signal ground. The way that works is to make the analog power ground and signal ground have modest inductance back to the power source, usually by ignoring special considerations for this, and by connecting the analog and digital grounds only right at the A-D and/or D-A reference ground pin.

And good A-D/D-A chips all have an analog ground pin marked on them. Usually.  That is the one connection of the digital ground to the analog signal ground. Anything else is pot luck. May be quiet, may be noisy, may interfere with aircraft navigation radio and/or cell phones. Or pick up one of those funny "numbers" radio stations.

In "PCB Layout for Musical Effects" I deliberately avoided digital and ground plane issues. It would have hopelessly muddied the waters for someone who was wrestling with how to make schemos into layouts to start with. PLME is a "101" course. Ground planes are effectively a whole course in themselves, and the two are separated by another course on PCB consequences of RF signals. I didn't think I could handle a three course set.

As some advice if you're off into this:
- conceptually separate your entire circuit into digital and analog; the analog includes incidental switching that's not clocked, like footswitching and such
- there will be inconclusive places in the circuit where it's hard to tell analog for digital; those are your boundaries
- physically separate digital from analog sections on the layout, and put a digital ground plane under the digital stuff
- to the extent possible, make the digital plane unbroken, not sliced by non-ground traces; on single sided board this is fond hope, usually not possible, but try
- it is possible to make one plane if you have carefully separated the analog and digital sections into different areas of the board, but if you do this, put the plane's power feed so the ground currents from the digital side do not flow through the analog section of the plane; that amounts to putting the main input power filter cap right on the imaginary line between the analog and digital sections of the combined plane
- if you have done good chip-level decoupling for the digital stuff, it's OK to have the digital plane be a big island with traces back to the main power feed to the board - **IF** you have done a good job of local decoupling

[cloudscapes]

I frequently combine digital circuits with analog circuits, usually to achieve the same "final" effect. You don't need transformers to isolate the two.

1. Digital components are (or should be) powered by regulators, like the 7805. That's one layer of isolation.
2. You usually use star-grounding. Where various components that use ground are connected independently to the common ground point, like a star, rather than in a chain. That's another layer of isolation. Sort of.
3. Filtering/decoupling capacitors everywhere. Around the voltage regulator, and next to every IC right on the rails. That's another layer of noise cleaning.
4. Sometimes have a 100ohm resistor on ground and/or power and the digital component, to just to take a bit of an edge off the power rails. Another layer of isolation.
5. Ground planes. They're used for better inductance, but also used for shielding. Another layer of noise filtering.
6. Separate the digital parts of a circuit from the analog as much as you can. Don't run sensitive analog traces (like audio) under digital parts, and vice versa.

Combine these tricks and it's perfectly doable to combine the two.

[mth5044]

Awesome information, thank you guys.

[CodeMonk]

Thanks RG.
Looks like a ton of great info.

I only glanced over for now it though.
(I just woke up and my brain needs caffeine)

And....

OK, so now I have you good and scared of digital edges. What about the analog ground?

I don't scare easy
I thrive on pushing my brain and knowledge.

BTW RG, I sent you a brief PM
(I'm not really ready to go public with stuff yet, but rest assured, when I get it all working, I will be posting everything here.