The “modifications” you need to do depend on what you want to accomplish.
The minimum modification (let’s call this “Level 0”) doesn’t require any soldering at all (at least not to the recorder): You can just get some plugs and wires to intercept the traffic between the front panel and the main controller on the back of the deck unit, and maybe the traffic between the main controller and the deck control circuit board. I bought the connectors to do that from Digikey; I don’t remember what the types of connectors are but I got them at Digikey.
If you want to do anything that’s more exciting than controlling the deck and the front panel, you will need to do some soldering to tap into the L3 bus and the subband I2S bus. Let’s call this “Level 1”. You still won’t be modifying anything in the recorder, really, but you’ll have to do some soldering: If you attach some wires to the I2S subband bus, you can intercept the PASC data coming from and going to the tape, but you can’t inject your own data. You can also tap into the L3 bus to get the SYSINFO/AUXINFO data coming from and going to the tape. These are not too hard and are low-impact: the recorder will keep working the way it does, and the modifications are still easily reversible.
If you want to be able to record your own PASC/MP1 stream to tape without using the SPDIF or analog input, you will need to inject data into the DRP chip. For that, the SBDA line of the DRP needs to be disconnected from the SFC chip. Let’s call this “Level 2”. It looks like this should be easy to do (you only need to lift one leg of a chip, and it’s on the corner of the chip and near the edge of the circuit board) but I haven’t done this yet.
To inject your own SYSINFO and AUXINFO data on a tape (and create your own prerecorded tapes with ITTS) will require that you take over the L3 bus as well as the I2S bus (let’s call this “Level 3”). The only way I think this is possible is to take the main controller offline, so you would also have to make it possible for your project to disconnect the front panel SPI bus and the deck controller bus (which is an async port at 38400bps). Additonally it will be necessary to pull a few lines low or high to keep the chips working instead of going to standby mode when the main controller goes into sleep mode. As I discussed in my presentation in Eindhoven in June 2019 (see the DCC Museum Youtube channel), I expect that it’s possible to take the main controller offline by injecting a “Stand By” command into the front panel SPI bus while the recorder is in playback/recording mode, and holding the chips in active mode while you do it. It will also be necessary to disconnect the deck control bus so the main controller can’t stop the deck before it goes offline. All of this is very complicated and obviously I’m not ready to go that far yet.
So far, I’ve been using the “SAMC21N Xplained Pro” by Atmel because it has a lot of serial and SPI ports which are really useful for analysis. But I’ll probably switch to the Teensy 3.5 because it’s much cheaper, smaller and has an I2S port (actually I think it has two I2S ports with a shared clock) and there’s a great audio library that even makes it possible to let the Teensy 3.5 appear to a PC as a USB audio device.
Whatever hardware you want to get, make sure it’s 5V tolerant: The SAMC21N is 5V tolerant, but the cheaper more ubiquitous SAMC21 (without “N”) and SAMD21 are not! Similarly, the Teensy 3.5 is the only Teensy that’s powerful enough and has enough I/O for our project and is 5V tolerant.
(Topic moved to Technology because this is mostly about hacking recorders, not normal use cases )