Yoshimi on the Raspberry Pi

[Eddy Deegan]

Progress Report:

Next, the performance. I started out cautiously at first with a few lightweight sounds. Everything was fine, so I went for some of the bigger ones. Still no issues. Took a deep breath and loaded up the most complex and processor hungry patch of the lot. Breezed through it, to the extent I threw caution to the wind and and started playing arpeggios. It's only on a mini keyboard at the moment, so I was limited to a degree. However, I'm frankly astonished at the performance. I couldn't detect a single Xrun, dropped, or stuck note.

I had it on a 5V bench PSU (fed via the GPIO pins)) and the current consumption remained well below 1A, mostly about 700mA.

I'll order the case and other bits now, and get to work on a full build. I'll also make a dedicated P4 image with all the goodies.

Sounds like great progress Will!

I'm seriously looking forward to getting my hands on this as much as I'm looking forward to getting 'em on the Novation Summit and I will create a demo video of the Yoshimi appliance too.

[The Elf]

I'm starting to take an interest, especially since I have other potential uses for a Pi, partly as a virtual Commodore 64 and partly as a network aid. I quite fancy one of the cases that doubles as a mini-screen.

I'm a total novice to Pi and Linux, so any advice would be gratefully received.

[Folderol]

Parts have arrived, and I can fix the internal layout right here.

The view is from above, with the dotted items upside down above the others. As you can see there is not a lot of space. This isn't (yet) fixed in stone, so if potential users think something is definitely wrong, now's the time to shout.

The box is pretty much an exact fit under the monitor, so the DC and HDMI leads are only 1/2M long.

The PSU I'm using has class 2 medical certification (surprisingly cheaper than the alternatives). The mains side will be earthed, but, apart from a 2.2M bleed resistor shunted by a 470pF cap, the chassis - and therefore all the connections - will float, so no ground loops.

[Martin Walker]

All sounds good to me Will, but like you I'm surprised that the class 2 medical certification PSU was cheaper

Martin

[Folderol]

Slight change of plan

Two of the PSU mounting holes have pads connected to the mains side Earth. This means floating the chassis would require insulated fittings, and extra space that's in short demand. However, the way everything fits together there is an alternative.

The Pi itself does have insulated mounting holes. Also, the Audio out is on a fully insulated socket, and the USB and Ethernet sockets have clear space round them, so I can hard earth the case, but still float the Pi and all the attachments.

Of course when I started this I completely forgot that I no longer have access to any machine tools, so I'm having to rapidly re-learn old skills

[Martin Walker]

The Pi itself does have insulated mounting holes. Also, the Audio out is on a fully insulated socket, and the USB and Ethernet sockets have clear space round them, so I can hard earth the case, but still float the Pi and all the attachments.

Sounds like a good plan again, and a common scenario in today's electronics - a decent screened enclosure, but no ground loop problems.

Are you going to fit a ground lift switch Will, for those whose remaining gear has two-wire mains cables and therefore no direct earth connection?

Martin

[Folderol]

Are you going to fit a ground lift switch Will, for those whose remaining gear has two-wire mains cables and therefore no direct earth connection?

Martin

Prolly not. I don't think there would be any benefit, and because of the size and shape of the unit, if one was fitted it would have to be a slide switch - a small toggle is too easy to snap off.
Slides are a pig to make apertures for and fit

[The Elf]

I've ordered a Raspberry Pi 3 kit from Amazon that I intend to build with nephew over Christmas (it's our tradition to do a Christmas build). I intend to turn it into a Commodore 64, at least for a days' gaming fun!

[Folderol]

The Pi 3 should handle that with room to spare - in fact it's a better choice than the Pi 4, as you can use a standard HDMI lead.

[Folderol]

Hit a bit of a speedbump

I failed to allow for the height of the relay on the power control PCB, then Murphy saw his chance, when I reached up to get a ruler off the shelf, and accidentally nudged a full reel of solder. This fell off and (with great precision) landed directly on said relay, snapping its pins off.

I bodged it to continue tests on the design - all OK - but obviously need to rebuild with a different board layout and new relay. Of course, I didn't have enough stripboard for a new larger board, so more stuff ordered {sigh}.

[Eddy Deegan]

Phew, I was worried for a brief moment that it was the Pi itself that was going to get damaged by the solder falling while reading that!

When I saw it was a relay I breathed a sign of relief, and although it's a bit of a setback it could have been worse!

Go go Will

[Folderol]

Wounds licked, and a first pic with a paper template in the exact position of the revised PCB will be. The components shown on this are actually the other side of the board.

The white lead is the 12V supply for the monitor. It's sitting in a groove in the box. You can see the HDMI lead for it coming out from under the 'PCB' and that will sit in the other grove. The two small LEDs will indicate 'power applied', and 'running'. The USB audio module is glued to the case and when the Box is closed up plugs into one of the sockets on the Pi.

The lid is the right size - parallax

[BJG145]

Looking good!

[Wonks]

Don't forget to leave some room to put the music in.

[Folderol]

Don't forget to leave some room to put the music in.

Now he tells me.
It's a bit late now, and unfortunately Yoshimi doesn't have a compressor.

[Folderol]

Progress
We now have the new control board which fits correctly and is fully tested.

As you can see I had to lie one cap on its side, otherwise it hit the (unused) USB C power socket. I normally put ICs in sockets, but if I had done that then it would also foul the stuff underneath - there really isn't a lot of space. In future I'll probably see if I can find a deeper 20 way socket, as that's what decides it.

The slit in the board is to ensure the mains pins of the relay are completely isolated from everything else. Also, on the underside, the tracks between the pins have all been removed. I'll probably glue a bit of plastic over the top of this, to provide a little protection for straying fingers.

Some final wiring tomorrow, then hopefully is software funtime

[Eddy Deegan]

The package looks really tidy!

The form factor is perfect for desktop use and such a small package should be accommodatable in a workspace of any size.

Really impressed so far and looking forward to subsequent developments. As a synth man I'm relishing the prospect of Yoshimi goodness in my setup and I anticipate doing some lovely stuff with this

Great work Will

[Folderol]

Another speedbump

The problem with emulation is that you need to be sure that what you're emulating really is equivalent to the real thing.

Up till now, I've been powering the unit from a bench power supply rather than the one installed in the unit. This is so that I can keep a watch on current levels while testing, and also stop the shutdown from actually shutting things down - in case I'd got it wrong and it tried to shutdown while the Pi was running full tilt.

It seemed to be working fine. That was until I watched the current demand while it was going through the emulated shutdown. The power control module was supposed to wait 7 seconds after the Pi itself had gone into standby before disconnecting the supply, but was actually doing so about a second afterwards - and not always quite the same time! I was using one of the GPIO lines to trigger shutdown, setting it 'high' in the knowledge it had to go low when the Pi had stopped - and this had worked perfectly every time when my emulated Pi stopped.

It turns out that the GPIO lines 'disconnect' a long (and variable) time before the Pi completes its tidying up However, the good news is that in the process, I discovered that the internal 3.3V supply does drop when the Pi goes into standby, reliably at that exact point. The result is my nice new PCB is no longer quite so pristine as it was before

It all does work correctly now. The red power light comes on immediately the start button is pressed, and the green 'ready' light only comes on once the desktop is showing. When shutting down, the green light goes off immediately the process starts. Actual power off is 7 seconds after the Pi is completely dormant - at which point the red light goes out.

This delay is to ensure that the Pi and any attachments are all completely settled before being robbed of power.

It's all software from here on. In order to keep background services to the absolute minimum, automounting USB drives is not enabled, nor is networking. However I'm setting up a couple of basic Apps, to enable these when needed.

P.S.
Being in a metal box, wireless and bluetooth wouldn't work anyway.!

[Martin Walker]

Power down and up cycles can be notoriously difficult to second guess, so bravo Will

Don't get too complacent about the software side though - have a read of Eddy's latest bug-hunting post so you can feel the fear and do it anyway

Martin

[The Elf]

[Folderol]

It's ALIVE!!!

A hastily snapped picie here.

As you can see it's running multi part, and if you zoom in you can just about see in the CPU monitor the third CPU is running about 43% (the others are little more than idle).

The MIDI file controlling it is included, and does all the part setup as well.

Apologies for the camera flash - I wanted to get this out as fast as possible.

All the software seems to work correctly, but I'll give it a long run and do some more sanity check through the day.

[Jumpeyspyder]

Nice work Will - it looks great !!

[Eddy Deegan]

It's ALIVE!!!

A hastily snapped picie here.

As you can see it's running multi part, and if you zoom in you can just about see in the CPU monitor the third CPU is running about 43% (the others are little more than idle).

The MIDI file controlling it is included, and does all the part setup as well.

Apologies for the camera flash - I wanted to get this out as fast as possible.

All the software seems to work correctly, but I'll give it a long run and do some more sanity check through the day.

Looking great Will, that monitor seems perfect for the task. VERY much looking forward to putting a Yo-Pi through its paces!

[The Elf]

[Martin Walker]

A further bravo is obviously required, so BRAVO!!!

Martin