Page: 0 1
Refactoring for performance
Date: 26/8/2016
Tags: mc2
A few weeks into actually using the MC2 I realized that the response time of the expression pedal was quite poor, running in the 100-200ms range. After doing some tests I worked out that it was because of the way messages were passed from the hardware thread that received the events, like analog to digital reads for the expression pedal, to the user interface thread, where they would be processed and sent back out the hardware thread to the Axefx. The problem with this was the the Raspberry PI is quite slow (or at least the version I'm using) and the user interface update speed was lagging. If the user interface is being updated, it's not processing messages. Hence the delay. But fixing it wasn't straight forward, because all the user interface code and Axefx code were all jumbled up like this:

So I took all the code that relates to the Axefx, and all the code that updates the user interface and separated them out into their own source code files. I called the Axefx specific code the "logic". Now I had 3 separate bits of code:
  • The user interface code. Responsible for updating the screen. Changing settings, like IA button actions etc.
  • The Axefx logic code. Which responds to MIDI inputs, converts hardware events to outgoing MIDI etc.
  • The hardware interface code. This is responsible for talking to all the hardware, both inputs and outputs.
Now I originally had 2 threads. And I decided that unless I really had to I would stay with those 2 threads, but the logic had to move from the UI thread, into the hardware thread. This meant that responding to MIDI and hardware events was decoupled from the user interface updating. Like this:

So hardware events, like the expression pedal value changing, would by processed immediately in the same thread and an outgoing MIDI message could be sent almost immediately, even if the UI was busy doing something. The latency was hugely improved. It's got to the point that it no longer has an impact musically. It seems to respond instantly.

In doing this I realized that I could hide all the details of each block of code behind a message passing interface. Which is excellent for thread safety. There was a little shared data that I wrapped in a mutex. But all in all the design feels nice and clean.

In fact I loved the design so much I'm rewriting i.Ftp to use the same architecture. Oh yeah, and because it was kinda badly written and wouldn't port to modern UI systems that don't have thread safe UI APIs.
(0) Comments | Add Comment

Negative ADS1115 readings
Date: 1/8/2016
Tags: mc2
I've been using an ADS1115 to convert the analog voltage from the expression pedal input, which is just a potentiometer really, to a 16 bit value. On the weekend I had a gig and both at the rehearsal and the set the expression pedal would flip to 127 (the maximum) when in the minimum position. It was also behaving erratically, the full range of output values was compressed into a very short throw of physical movement. I didn't technically need it for the gig so I just unplugged it.

So today I thought I'd get to the bottom of it. After switching on the logging and reproducing the issue it was apparent that a lot of "plugged" / "unplugged" events were happening. This was due to the values coming from the ADS1115 were bouncing between very low values (say 0 - 100) to very high values (65530 - 65535). I started by putting the multi-meter in voltage mode over the input pin of the ADS1115 and ground and found that the lowest voltage was about 0.1V which is, well, correct for the minimum position. So that led me to believe it was solely a software issue. Looking at the code that does the read, I found it was assuming the 16bit value was a unsigned integer. But that fact that the value was jumping to 65500 made me wonder if maybe the ADS1115 was giving me a signed value instead. So I changed the code to assume signed behaviour and now the values all seemed to follow each other as the pedal moves back and forth. Then checking with the data sheet:
The ADS1113/4/5 provide 16 bits of data in binary twos complement format
Ugh, ok so I should read that more carefully. Still it going slightly negative like that even with a positive voltage is a bit weird to me. Oh well, I'm just going with max(0, AdcReadValue) for the moment and it seems to be working just fine now.
(0) Comments | Add Comment

MC2 April Progress
Date: 3/4/2016
Tags: mc2
So here is a scary idea... I'm playing next weekend in my live band. Should I use the MC2 for the first time? If I get a bunch of stuff finished I don't see why not, although I'll definitely have the MC1 packed in case it doesn't go smoothly. The main hurdle will be the expression pedal ports which I haven't really done any work towards other than getting the ADS1115 chip to show up on the I2C bus. Nothing like a dead line to get you moving.

Running progress:
  • 1/4
    Wired up the rest of the encoders, and their buttons. Implemented the software side to read all their state changes. I now have 6 encoders fully working.
  • 2/4
    Fixed the button handling to only do something of the "down" event. Lowered the maximum brightness of the Neopixels a lot... otherwise they are almost white instead of the selected colour. This also uses less power, which is nice.
  • 3/4
    Implemented software button debouncing to remove duplicate up/down events. Also added "held" buttons events that fire if I button is held down for more than 1200ms. This is useful for hiding advanced features. I might use it to access the Y settings of Axefx2 blocks connected to Instant Access buttons for instance.
  • 4/4
    Implemented tuner support on the higher resolution screen of the MC2. Yay for not having to make custom text characters to display a line.
    Hooked up the encoder output to the Axefx2's EXTERN 1-8 CC messages. Limiting the range to 0 - 127 and adding a little acceleration if the user is turn the dial a lot. I've managed to get one twist of the dial to go from 0 to 127 and yet retain some fine control when moving slowly. It seems to be reasonably good transitioning from non-accelerated to accelerated.
    Bought some stereo switched insulated jacks for implementing 2 expression pedal ports. The system supports 4 but I only need one at the moment. Also bought some plastic knobs for the encoders, just till I work out what ones I really want. Ideally I'd get something that is suitable for recessed shafts. Most knobs cater to raised shafts, where the nut is above the panel surface.
  • 5/4
    Drilled and deburred holes to mount the MIDI and expression pedal jacks on the case.
    Soldered together the expression port jack and hooked it up to the ADS1115 chip. The stereo socket is a DPDT switch as well which allows me to detect when there is no expression pedal plugged in. Basically I set +3.3V for that condition, and start about +3V for maximum expression.

    Wrote the corresponding code to read the value from the ADS1115 chip over the I2C bus. This was pretty straight forward. I keep a min and max value to auto-calibrate to the pedal.
    And that technically means the MC2 is out of Alpha into Beta! All the basic functionality works. However there is a lot of road testing to go :D
  • 6/4
    Added a row of labels under the IA config buttons to say what the buttons are currently configured as.
    Added an continuous parameter mimic control to display graphically the position of things like encoders/expression pedals. This looks like a circle, with a line showing the "amount" radially, just like the Axefx2 parameter display:

    Fixed some display bugs when switching presets.
    Added a script to start the MC2 application on startup. Obviously I want the program to start automatically when I'm using the system live.
    Added a button on the setup page to shutdown Linux. There is a confirmation dialog before that happens. Now I can start and shutdown the whole thing without a keyboard attached. Excellent!
    Took the plastic off the touchscreen... oooooooooooh! ;)
  • 7/4
    Added a MIDI disconnect event to the Raspberry Pi hardware layer. Basically if it doesn't see any MIDI for 3 seconds it flicks over to "disconnected" mode. Normally the tempo messages keep this from happening. There is no easy way at a hardware level to see if MIDI cable has been unplugged.
    Going to start "testing" things by doing my personal rehearsal during the day. Then tonight I have the full band rehearsal.
    Fixed the select scene 'n' IA button selecting scene 'n+1'.
    Removed the nice highlights on the buttons to improve rendering speed. Currently the GUI thread is running the controller code, and message passing is limited by the redrawing of the screen. This is not ideal and I will fix it at some point but I just want it to work ok for the weekend. So things like the expression pedal send a message from the hardware IO thread. And it's laggy because of the screen updates. At some point I'll need to break out the profiler and find out where the time is going. Most likely there are a number of CPU bottlenecks in the graphics and text handling code.
    Changed the expression pedal reads to happen every 8ms instead of 16. Again in an attempt to improve latency. I can probably improve that again by reading up on all the ADS1115 modes. I think it can do like 800 samples a second but I'm getting a lot less than that ATM.
  • Band Rehearsal
    There were no major issues at the rehearsal. It all worked pretty much as I wanted it to. The only minor issue is that if I configure a bunch of IA buttons the text on the labels is pushing the layout off the right edge of the screen:

    So I'll have to fix that in the layout code. One problem is that the edge of the touch screen doesn't really work, so if things are too near the edge you can't activate them. The second issue is that is just looks bad. Hahah.
    I'm really liking the layout of the hardware buttons. Having the IA buttons along the near edge is much better for triggering things during the song. And the tap tempo on the corner is also better than the location on the MC1. Being able to dive in and setup a new IA button quickly was also really useful (I decided I needed some Phase90 in my life last night).
    I also noticed a text clipping bug when text is off the right edge of the screen it wraps around to the left edge. Hmmm should fix that sometime.
  • 8/4
    Fixed the layout bug by adding some CSS to the text label cells:
    max-width: 59px;
    The text overflows a bit for larger labels, but at least the position of the buttons doesn't move.
  • 10/4
    Live play testing!
    Yes I took it out to play live this weekend. It went pretty well overall. For the most part everything worked as well as I had hoped. The main glitch was at the very end of the first set the buttons stopped responding for a little while. Touch screen was still good so I just shut it down and restarted for the next set. It didn't affect my performance though. After rebooted it seemed to settle in ok and the 2nd set went without any problems. The expression pedal lag was still there obviously but I worked around that. There may be some display issues with the IA LEDs. I thought I noticed some block IAs in the wrong state after scene changes. But I was pretty busy just getting on with playing rather than analyzing the IA states. Something to follow up later.
(0) Comments | Add Comment

MC2 March Progress
Date: 1/4/2016
Tags: mc2
After when seems like forever I bring you an update on the MC2 project. The good news is that most of the hardware is now working, including MIDI Input/Output, the Neopixel chain, the I2C bus to the GPIO expander chips, which in turn connects to the buttons and encoders. In short I can now use it to change presets on the Axefx which is like the bare minimum functionality.

Some of the issues I've had to sort out:
  • Very slow compile times on the Raspberry Pi
    I'm only using a Pi B (version 1, rev 2) so it's got a single core running at 700MHz which frankly is molasses slow. So I decided that it was worth getting a cross compiler working on something with more grunt. It took days of mucking around, and 3 different builds, copying off almost the entire Pi's worth of headers and libraries, but it now works. More importantly it's fast. My main dev box (3.8ghz i5 /w 16GiB of ram) chews through the code very quickly, and then copies the binaries down to the Pi over the network.
  • PiTFT touch screen not working correctly
    The behaviour I was seeing was the cursor stuck in the top left, and the x & y axis' were swapped. I ended up building SDL from source and then inserting a pile of logging during the mouse driver setup only to discover that it was using the default mouse driver instead of the touch screen driver that I had configured in the environment. It turns out that as I'm running the code as root (sudo) the environment is different than my normal user account. Ok, so now I just set the right env vars in my C code before initializing SDL. Which forces it to do the right thing on any account.
  • I2C bus not working / unreliable
    This came down to adding 1000uF electrolytic over the 3.3v supply line, and also adding 0.15uF decoupling caps over the +V/GND pins of the GPIO expander chips. Also I had to change the pull up resistors connected to the SDA+SCL data lines from 4.7k to 10k.
  • Neopixel chain stopped working
    Eventually I found a cold solder joint on the ground pin output. Only after disassembling it completely, including the button harness and taking the plastic backing plane out.
  • MIDI input not working / garbled
    This has actually taken most of the week to sort out. The hardware for the input side seemed ok, and I even did see some incoming MIDI at one point. Initially it was garbled and I tweaked the baud rate with system settings. It seemed to work for a while then stopped again. I got very frustrated with not being able to see the logic signals in the circuit so I went and bought a Bus Pirate which can read both the UART and do logic analysis. I tested all the hardware I could in isolation. I went over the traces like 6 times. I tried the optical isolater chip in my MC1 and everything appeared to be good. I even pulled the diode out and tested it in isolation. I tested the Rx+Tx pins of the Pi by hooking it up to the Bus Pirate and had the Pi talking to the PC. Then I tried minicom and saw the incoming Tempo messages from the Axefx! Ah, so the hardware IS good, and thus I started looking at the software side again. It turns out that in more recently Linux kernels you can see custom baud rates in code. Most of the people getting MIDI working on the Pi are doing system config settings. But it's a hack. Using an API to set the speed to the exact right value is way better. Now the existing code snippets kinda get you there, but for me I had to merge the old code and new code to get something that works for me. This is the final result:
            int MidiHnd = open("/dev/ttyAMA0", O_RDWR | O_NOCTTY | O_NDELAY);
            if (MidiHnd > 0)
                struct termios2 tio;
                ioctl(MidiHnd, TCGETS2, &tio);
                tio.c_cflag = BOTHER | CS8 | CLOCAL | CREAD;
                tio.c_iflag = IGNPAR;
                tio.c_oflag = 0;
                tio.c_lflag = 0;
                tio.c_ispeed = 31250;
                tio.c_ospeed = 31250;
                ioctl(MidiHnd, TCSETS2, &tio);
    The MIDI output was an easy fix: just turn the 4011 logic chip the right way around. And boom: working! Hahaha... in my defense it has a little stamped circle at the other end of the chip. Which is sometimes used to mark the "1" pin. Grrrr.
Oh and here is some video showing it working:
(0) Comments | Add Comment

Got the NeoPixels working
Date: 3/3/2016
Tags: mc2

I've got around to working on the MC2 again. Last I got up to was getting the NeoPixels working and for some reason none of them would light up. To isolate the issue I used the bread board to step the data signal from 3.3v up to 5v. As you can see it works great. Now I know the NeoPixels themselves, and their connections are good I can get to work finding the issue on the peripheral board.
(0) Comments | Add Comment

Mc2 Progress #4
Date: 12/10/2015
Tags: mc2
The time comes to stop mucking around with software and start building the hardware. I'm no where near as good with the hardware things. Software is easy and fine... so how much did I break it? Turns out a fair bit. Lets take a look.

First I started with an Inkscape file that described all the positions of the components and specifically the places where I have to drill or cut. I printed it out in A3 and glued it down to the top surface of the chassis.

I started drilling out the holes and all the switch / neopixel holes went fine. Then came the LED rings and the paper started tearing everywhere.

So I left the LED ring holes for a while and worked on the screen rectangle. I bought this tool called a "nibbler" to cut the hole. You drill a hole in the area to be removed, and insert the nibbler into the hole and start cutting. It can cut in a curve so I used that to align to the border.

Then I came back and finished all the LED ring holes. However my accuracy was terrible and I don't feel all those holes are good enough.

I've decided that this chassis is a bit of a write off. I have since gone and bought a new chassis (they are about $60 AUD inc shipping so not too bad). And now I'm looking for someone that can do CNC drilling.

Moving on I tried test mounting some of the components, starting with the screen and Pi.

Seems to fit quite nicely. So I tried to bend the holes in to accommodate the countersunk machine screws that I had bought. That bent the metal around the hole too much and deformed the chassis around the screen area. In hind sight I should've used a large drill bit instead of mechanical force. So because of the deformations in the metal when I tightened up the screws around the screen I heard a faint cracking sound. Yup I broke the screen just a little.

That's just fantastic! :( Another thing I have to re-order. Similar price too... bah. There is some good news thought... the screen still works:

Although I haven't had a chance to test the touch screen part.

The PCBs and Neopixels need to be mounted behind the front of the chassis, so I went and bought some rectangles of 3mm perspex:

The switches go through the perspex and keep it in place. I've started soldering some 3 wire ribbon cable between the Neopixels, not shown in this photo. It's slow and tedious work :( And the various glue electronics to interface the Pi with the buttons, encoders, expression pedal ports and neopixels is all mounted to that as well.

(0) Comments | Add Comment

MC2 Progress #3
Date: 23/9/2015
Tags: mc2
The last week or so has been spent a twisty maze of fonts, colour spaces, CPU endianess and cross platform C++. Anyway I won't regail you with all the fun I've had but I will show you some screenshots.

Firstly because I have RGB LEDs for each Instant Access switch, I need a way for the user to select a colour:

It remains to be seen how easy it is to use on the real hardware. But colour selection works either by hitting one of the palette entries or pushing the Red, Green and Blue sliders around.

Also for entering details like the CC number and other textual information I probably need a touch screen keyboard:

This is roughly copied off my Android phone's touch keyboard. Without Swype unfortunately ;)

The screen to select a scene change for an IA button:

And the page where the user picks an Axefx block:

(0) Comments | Add Comment

MC2 Progress #2
Date: 14/9/2015
Tags: mc2
This week I've been working on the screen layouts. A fair bit of mucking around with CSS and resource file formats was required to get it all working but here is what it looks like now:

Working from the hand drawn ideas I had last week I'm mocking all the dialogs up in the graphical editor:

Also in the mean time I'm working out if I can hide the boot time of the Raspberry Pi behind a custom splash screen. A static image is easily doable but something with a progress spinner might be a bit tricky.
(0) Comments | Add Comment

MC2 Build
Date: 28/8/2015
Tags: mc2
So the other day I was at rehearsal and the MC1 that built started to have some issues with the External control knobs (along the top). They were jumping all over the place and making it annoying to use. I did get around to pulling all the circuitry around those pots off and doing it properly later. But it occurred to me that I'm using gear that I have no backup for. There is no 2nd controller to swap out if this one dies at a gig.
And because I'd rather custom build one out of parts I [mostly] have lying around that means building the MC2.
For starters the MC1 uses a tiny micro-controller with no USB or network or any modern connectivity at all. I have an old XP computer JUST to reflash the damn thing (needs a REAL parallel port). For any new system I wanted an easy way to do software development and ethernet/USB connectivity. And because I had a Raspberry Pi already I settled on that as a platform to run everything. I already had a little touchscreen to go with it and some RGB led rings with encoders that I built for a different project that I never got around to. So with all those parts just lying around I got to making some bits to glue it all together. Here's a family photo:

1) The Raspberry Pi. This is the "old" B rev 2 version... which has the slower processor.
2) The touchscreen. This interfaces with the SDL api using the Linux framebuffer. Touch events show up as mouse clicks. It's all quite straight forward to be honest. The only gotcha is you HAVE to set the video mode back to text if you crash... otherwise you're stuck in video mode without a working console.
3) LED rings + encoders: These I built for someone else but was too slack to finish that project. They basically are an extension of the Mayhew LED ring except that I'm running Red/Green/Blue LEDs instead of just one colour. It also is designed to fit in a 1U rack box, so they are nice and compact. I'll have 3 pairs of these across the top of the box, giving me 6 independent continuous controllers. Just like the old MC1 box, but importantly I'll be able to update their "position" using software when the preset changes.
4) Neopixel chain: Each foot switch on the controller will have an associated LED. I settled on the Neopixels because they wouldn't require a lot of support circuitry and there is now a software library to make them work on the Raspberry Pi. Which I've used successfully on a bread board.
5) MIDI in and out ports. This is a direct implementation of the circuit used in the MC1. All I really have to do is connect it up to the Raspberry Pi's UART pins and then program the right frequency. *fingers crossed*
6) This is a bunch of extra GPIO (general purpose input/output) for connecting the foot switches and encoders up to. It runs on the I2C bus from the Pi so it's easy to chain lots of things and address them in software. (I love software solutions).
7) Some analog inputs for the expression pedals. I'll be able to have up to 4 connected at once.
8) This is the break out board for the expression pedals. It connects to (7) via a 4 wire bus and basically sets up the right voltages for the 3 pins on the socket. It allows me to detect when the pedal is connected and disconnected by limiting the valid ranges of input voltage when the pedal is connected to less than 0-3.3v...
That's an overview of the hardware side. But what about the software? Well I already have a lot of cross platform C++ code that will come in handy. Firstly I have written a library called LGI that allows the same code to run on Windows, Mac and Linux without much changes. So why not extend that to SDL on the Raspberry Pi? So that's what I did this month. Which gives me access to a whole bunch of little controls like, edit boxes, check boxes, labels, buttons, table layout controls, graphical design tools etc. Stuff I've already written.
But I still needed to work out the general look and feel of the user interface. Check it out:

Basically I'm drafting up the flow of buttons and the layout of the controls to get a feel for what I need to write in the software. I have this control that does layout for me, kinda like a HTML table in that it works out the spacing of everything based on the size of the content. And I can style all the controls with CSS which will make for a nice consistent look over the product.
I expect though it'll be some months before I get this all working. Only a few parts of this are tested and none of it has been integrated into a working system. I'll try and post updates when I can.
(1) Comment | Add Comment