I performed a few more quick experiments over the weekend and I’m happy to report further success!

Although the Akai MPK Mini Play has an internal Dream S.A.S. SAM2635 synthezier chip, the Akai microcontroller software blocks full access — a true shame because the SAM2635 firmware is a complete General MIDI/GS implementation.

I modified an Akai MPK Mini Play bringing out four signals from the main printed circuit board (PCB):

• MIDI IN white wire: SAM2635_RX MIDI input
• MIDI OUT black wire: SAM2635_TX MIDI output
• Ground grey wire: Ground for external power
• +3.3V purple wire: +3.3V for external power

My goal is to override the Arm microcontroller and drive the SAM2635 over 5-pin MIDI, completely bypassing the Akai software. If successful, we should be able to use the fully functionality afforded by the SAM2635.

Test 1: UBLD.IT MIDI breakout board

A full test of the UBLD.IT MIDI breakout board is in order before testing the MPK Mini Play. If the UBLD.IT isn’t working, any test with the Mini Play will fail, too.

The UBLD.IT board is a really nice, small MIDI IN and MIDI OUT breakout board. The MIDI OUT side is dirt simple and sends the transmit signal (pin 1 of J2) to the 5-pin DIN connector through a current limiting resistor (R4). Please recall that MIDI is an active LOW current loop. The return pin — the other side of the current loop — is connected to VCC (the positive voltage rail) through a current limiting resistor (R1).

The MIDI IN side provides the opto-isolated MIDI logic signal to the receiving electronics. DIN pins 4 and 5 send the incoming current through the LED in the opto-isolator (U1). The opto-isolator is a 6N138. The LED shines on a photo-transistor. The photo-transistor is essentially a light-controlled pull-down path from the positive voltage rail to ground. Thus, the MIDI logic signal is HIGH when idling and is active LOW. The incoming logic signal is sent off the board through the receive pin (pin 2 of J2).

The test set-up is simple. Jumper the transmit and receive pins (J2). Connect +3.3V and Ground using the Mini Play as a power source. Connect UBLD.IT MIDI IN to the MIDI OUT of any handy keyboard like a Yamaha MODX. Connect UBLD.IT MIDI OUT to the MIDI IN of a second keyboard like a Yamaha Genos. Turn on the Mini Play to apply power to the interface. Play the sending keyboard (MODX) and listen for sounds produced by the receiving keyboard (Genos).

It was gratifying to hear the Genos while playing MODX, i.e., the UBLD.IT is working correctly.

Test 2: Play using an external keyboard

The next test is to drive the Akai MPK Mini Play itself.

Here’s the test set-up. Leave the MODX connected to UBLD.IT MIDI IN. Disconnect UBLD.IT MIDI OUT since we don’t need it. Connect the UBLD.IT pins (J2) to the corresponding Mini Play signals. Receive is SAM2695_RX (white wire) and transmit is SAM2695_TX (black wire). Leave ground and +3.3V connected (grey and purple wires, respectively).

Turn on the Akai MPK Mini Play. Play the MODX. You should hear sounds produced by the SAM2635. Yay! 🙂 Try the MODX MOD and pitch bend wheels.

I believe this design works electronically because MIDI is active LOW and idles in the HIGH state. The pull-down transistor in the opto-isolator pulls the SAM2635_RX node LOW. Normally, you cannot tie two logic outputs together, but the pull-down works in this situation. (Think “wire OR”.) By the way, you shouldn’t play the Mini Play keys or turn its knobs as MIDI signals from the Arm microcontroller will interfere with the MIDI IN logic signal! I recommend turning INTERNAL SOUNDS off when using the 5-pin MIDI IN.

The next step is to try the Dream SAM2635 as a GM/GS synthesizer. Stay tuned.

Copyright © 2022 Paul J. Drongowski

I performed experiments to discover how the Akai MPK Mini Play behaves when controlled through its USB port. I wanted to build on what I had already learned about the MPK Mini Play’s MIDI implementation.

As mentioned in earlier posts, the Mini Play’s host Arm microcontroller communicates with a Dream S.A.S. SAM2635 synthesizer chip:

            Arm       TX  ----> MIDI IN (RX)     Dream 
      Microcontroller                           SAM2635 
           USART      RX  <---- MIDI OUT (TX) Synthesizer 

The Arm sends MIDI messages to the SAM2635 when INTERNAL SOUNDS are turned on, i.e., when the INTERNAL SOUNDS button is lit. When INTERNAL SOUNDS are turned off, the Arm software sends MIDI messages over USB exclusively. Oddly, the Arm sends some MIDI messages over USB when INTERNAL SOUNDS are on — and not always on the expected MIDI channel! Very quirky or buggy, depending upon your point of view.

People who want a 16-channel multi-timbral MIDI module are going to be disappointed. When driving the MPK Mini Play through USB, you really have only two channels: MIDI channel 1 for regular, melody voices and MIDI channel 10 for drums. If you stick to these two channels, you’ll stay out of trouble.

If you venture off the path, however, you’ll be in the deep weeds. The Mini Play responds to NOTE ON and NOTE OFF messages on “keyboard” channels 2-9 and 11-16, playing the default Grand Piano sound. Such joy is misleading, however. Send a program change on channels 2-9 and 11-16, and the program change is redirected to channel 1! Do not send program change to any other channels than 1 and 10 if you value your sanity!

I think someone at Akai envisioned the MPK Mini Play as a two channel MIDI controller with optional synthesis on two channels. Unfortunately, the actual implementation is inconsistent with this model as far as synthesis is concerned. Worse, Akai doesn’t provide any documentation about the MIDI implementation letting punters poke around in the buggy weeds.

All of this is truly a shame knowing that the Dream SAM2635 is a full 16-channel GM/GS synthesizer. What a waste of silicon (AKA “sand”) by Akai!

Here are some additional details about the MIDI implementation such that it is.

The Arm software filters out all System Exclusive (SysEx) messages other than its own. Stated another way, if you send SAM2635 SysEx, it will never be delivered. So, you cannot send GM or GS reset, tweak reverb and chorus parameters, and so forth because these are all done via SAM2635 SysEx.

The Arm software does pass all MIDI Continuous Controller (CC) messages to the SAM2635. RPN and NRPN commands also get through because these kind of commands are formed using CC messages. Here are the MIDI CC messages which appear to work:

    CC#0    Bank Select (MSB) 
    CC#1    Modulation wheel 
    CC#6    Data Entry 
    CC#7    Volume (channel volume) 
    CC#10   Pan 
    CC#11   Expression 
    CC#74   TVF cutoff frequency 
    CC#91   Reverb send level 
    CC#71   TVF resonance 
    CC#93   Chorus send level 
    CC#73   Envelope attack time 
    CC#72   Envelope release time 
    CC#80   Reverb program 
    CC#81   Chorus program 
    CC#91   Reverb send level 
    CC#93   Chorus send level 
    CC#98   NRPN LSB 
    CC#99   NRPN MSB 
    CC#100  RPN LSB 
    CC#101  RPN MSB 

The SAM2635 supports several other MIDI CC message types. Please consult the Dream SAM2635 firmware manual for a complete list.

Pitch bend and Program Change also get through. However, Program Change is always redirected to channel 1.

You can form RPN and NRPN commands via MIDI CC messages. For example, the Arm software controls EQ Low and EQ High using NRPN:

    B0 63 37 B0 62 00 B0 06 48      Equalizer low band 
    B0 63 37 B0 62 03 B0 06 40      Equalizer high band 

In fact, Dream have a whole raft of NRPN commands with MSB 0x37. Consult the Dream SAM2635 firmware manual. Dream call these commands “Special MIDI Controls”.

CC#80 Reverb program and CC#81 Chorus program are the two bright spots in a relatively bleak picture. You can change the reverb type with CC#80:

    00: Room1        01: Room2 
    02: Room3        03: Hall1 
    04: Hall2        05: Plate 
    06: Delay        07: Pan delay 

CC#81 changes the chorus type:

    00: Chorus1      02: Chorus2 
    02: Chorus3      03: Chorus4 
    04: Feedback     05: Flanger 
    06: Short delay  07: Feedback delay 

Since the Arm software blocks all non-Akai SysEx, you cannot tweak the reverb and chorus parameters.

Copyright © 2022 Paul J. Drongowski

I’m happy to report progress with breaking out the internal MIDI traffic within the Akai MPK Mini Play (first generation). Before reading ahead, I recommend reading my previous posts about Akai MPK Mini Play internals:

• Inside the Akai Mini Play Mk1
• Akai Mini Play MIDI mod (1)

The second post, in particular, shows how I brought the internal MIDI signals out from the main printed circuit board.

For the moment, I am focused on the communication between the Arm microcontroller and the Dream S.A.S. SAM2635 synthesis chip:

            Arm       TX  ----> MIDI IN (RX)     Dream 
      Microcontroller                           SAM2635 
           USART      RX  <---- MIDI OUT (TX) Synthesizer 

The Arm microcontroller sends MIDI messages to the SAM2635. For this round of analysis, Mini Play INTERNAL SOUNDS are turned ON. A future post will address the case when INTERNAL SOUNDS are turned OFF.

When INTERNAL SOUNDS are enabled, the key strikes, pad hits, knob turns and joystick movements cause the Arm software to transmit MIDI messages to the SAM2635. These MIDI messages are encoded for and are meaningful to the SAM2635. They are not the same as the messages sent through the USB port, i.e., the user-defined MIDI CCs, etc. (USB messages are sent when INTERNAL SOUNDS are OFF.)

Persistence leads to discovery

At the end of my second post, I noted my failure to monitor the message traffic from the Arm to the SAM2635. I performed additional experiments which eventually paid off.

First, I observed the signal from Arm TX to the SAM2635 MIDI IN (RX) pin using a Gabotronics Xminilab USB oscilloscope and logic analyzer. I tried the Gabotronics protocol sniffer. Unfortunately, the Gabotronics protocol sniffer does not support the standard MIDI baud rate, 31,250 baud. The SAM2635 can communicate MIDI at 31,250, 34,800 and 200,000 baud. The 200,000 baud rate is reserved for USB debug, so it’s use is unlikely in this case. The protocol sniffer displayed HEX gibberish at 34,800.

I switched the Gabotronics to logic analyzer mode and observed the serial data signal. Yes, I found data and the data are repeatable, i.e., hit a C on the keyboard and you get the same bit sequence every time.

I estimated 8 serial bits per oscilloscope time division (256 usec). Do the math and the data rate is 31,250. The MIDI is there; I just need to sense it properly.

I connected the MIDI signal to the MIDI OUT side of the UBLD.IT MIDI breakout board. I sent MIDI from the UBLD.IT to the Genos MIDI IN port. Success! The Akai Mini Play can play notes and change patches on the Genos.

Instead of monitoring the MIDI message stream via MIDI-OX on a Windows PC, I used the LOG display in the MIDI Designer app on iPad. The MIDI interface in this case is an old IK Multimedia iRig MIDI. Success, again! Now I can see the MIDI messages and, indeed, they are targeted for SAM2635.

I can only surmise that my Roland (Edirol) UM-2EX MIDI interface cannot detect the signal transmitted by the UBLD.IT. I couldn’t get a Sparkfun breakout board to work with the UM-2EX either. If I want to use MIDI-OX on PC for monitoring, I’ll need to try another USB MIDI interface (perhaps my old EMU).

SAM2635 message summary

Well, franky, that’s all good news. Here is a short summary of the MIDI messages that I observed. The Akai Mini Play sends a flurry of messages to the SAM2635 when power is first turned on.

Program change also causes a flurry of messages to be sent. Here is an example program change:

B0 01 00                        CC#1 Modulation wheel 
E0 00 40                        Pitch bend (center) 
F0 7F 7F 04 01 00 7F F7         GM Master volume 
B0 0B 7F                        CC#11 Expression 
B0 4A 66                        CC#74 TVF cutoff frequency 
B0 5B 0E                        CC#91 Reverb send level 
B0 47 00                        CC#71 TVF resonance 
B0 5D 00                        CC#93 Chorus send level 
B0 49 00                        CC#73 Envelope attack time 
B0 48 38                        CC#72 Envelope release time 
B0 63 37 B0 62 00 B0 06 48      Equalizer low band 
B0 63 37 B0 62 00 B0 06 48      Equalizer low band 
B0 63 37 B0 62 03 B0 06 40      Equalizer high band 

Some set-up messages are duplicated. I’ve seen this kind of behavior in other synths and sequencers.

Key hits, knob turns, and pad hits produce the following MIDI messages:

Gesture    Example message             SAM2635 Function 
---------  --------------------------  --------------------------- 
Key hit:   90 3C 7F                    Note on, channel 1 
           80 3C 00                    Note off, channel 1 
Filter:    B0 4A 43                    CC#74 TVF cutoff frequency 
Resonance: B0 47 00                    CC#71 TVF resonance 
Reverb:    B0 5B 23                    CC#91 Reverb send level 
Chorus:    B0 5D 00                    CC#93 Chorus send level 
Attack:    B0 49 00                    CC#73 Envelope attack time 
Release:   B0 48 38                    CC#72 Envelope release time 
EQ Low:    B0 63 37 B0 62 00 B0 06 47  Equalizer low band 
EQ High:   B0 63 37 B0 62 03 B0 06 40  Equalizer high band 
Pad hit:   99 2E 6C                    Note on, channel 10 
           89 2E 00                    Note off, channel 10 
Volume:    F0 7F 7F 04 01 00 7E F7     GM Master volume 

No surprises other than volume. I didn’t expect the front panel volume knob to send a GM Master volume message. Volume must be regulated digitally in the synth as opposed to an analog audio pot.

Copyright © 2022 Paul J. Drongowski

Yesterday, I took a quick look inside the Akai MPK Mini Play MIDI controller and synthesizer. I would like to discover a way to drive fully the Dream S.A.S. SAM2635 synthesizer chip without interference by the Arm microcontroller.

My initial investigation concentrated on the set of test points connected to key SAM2695 inputs and outputs:

    Pad Signal        Dream function 
    --- -----------   --------------------------- 
     1  STIN          Serial test input 
     2  STOUT         Serial test output  
     3  SAM2635_RST   Master reset and power down  
     4  SAM2635_TX    MIDI OUT  
     5  SAM2635_RX    MIDI IN  
     6  D GND         Digital ground  
     7  D +3.3V       Digital +3.3V  

If I can read and drive the SAM2635 MIDI IN and OUT, it should be possible to control the Dream synthesizer directly.

First, I need easy access to the relevant test points. I started with a 4-wire ribbon cable with male prototyping pins at both ends. These ribbon cables are easy to find and are generally used to connect a group of signals from one prototyping connector to another. In this case, I cut off the pins at one end with the intention of soldering the wires to the test points on the Mini Play main printed circuit board.

The picture above shows the test points after soldering. The soldering job ain’t pretty, but the joints are functional. The color coded connections are:

    Pad Signal        Dream function  Color 
    --- -----------   --------------  -----  
     4  SAM2635_TX    MIDI OUT        Black  
     5  SAM2635_RX    MIDI IN         White  
     6  D GND         Digital ground  Grey  
     7  D +3.3V       Digital +3.3V   Purple 

After inspection, I took a deep breath and powered up. The Mini Play functioned correctly, thank goodness.

I created a small, unobtrusive notch in the plastic top panel and routed the ribbon cable through the gap above the high C key. If I ever go to sell or trade the Mini Play, I should be able to unsolder and remove the cable while leaving the case visibly intact. As you can see in the picture above, the ribbon cable terminates at four male prototyping pins.

Next up, I assembled a UBLD.IT MIDI breakout board and interface. I bought the UBLD.IT board from Amazon. Unfortunately, the board is now out of stock. Bummer, cuz it’s a nice one.

The UBLD.IT has a 4-pin female header:

    Pin   Function 
    ----  -------------- 
    Xmit  MIDI OUT 
    Recv  MIDI IN 
    Vcc   Positive voltage rail 
    GND   Ground 

I connected the ribbon cable to the UBLD.IT board and connected the UBLD.IT to a Windows PC through a 5-pin Roland MIDI interface.

Power up. No joy. The Mini Play worked properly, but I wasn’t able to monitor the MIDI traffic from the Arm microcontroller to the Dream SAM2635 synthesizer via MIDI-OX. Nor was I able to drive the SAM2635. So much for the quickest and dirtiest solution!

Since nothing smoked and the Mini Play is healthy, it’s time for methodical investigation. First goal, I would really like to monitor the MIDI traffic from the Arm to the SAM2635. It might help me figure out what the heck the Akai software is doing to incoming USB MIDI messages. Once I ace this problem, I’ll try driving the SAM2635, again.

One further observation. I attached SAM2635_RX and SAM2635_TX to a Gabotronics USB oscilloscope. After futzing with the scope controls, I observed what looks like serial data bytes (MIDI) on the SAM2635_RX line. The Arm drives this line when it sends MIDI messages to the SAM2635. Serial bytes are sent in response to note ON/OFF, control knobs, and the joystick. Ooooo, so close!

By golly, there ought to be a way to monitor this MIDI traffic! I conjecture that the logic signal is inverted. Another possibility is a non-standard data rate, that is, something other than 31,250 baud. Probably ought to test the UBLD.IT in isolation, too.

Copyright © 2022 Paul J. Drongowski