Looking for a small, inexpensive ($25-30) ARM development board (say 120-180 MHz ARM Cortex-M4 with FPU, 512kB-1MB of FLASH and 256 KByte of RAM)? Then have a look at the Teensy 3.5 and Teensy 3.6 by PJRC/Paul Stoffregen:
The only problem? It is not possible to debug it. At least not in the traditional sense. This article is about how to change the board to use it with any normal SWD debugging tool, e.g. Eclipse and the Segger J-Link.
The tinyK20 we produced at the university for student projects are great: small, inexpensive, can be used for many projects, and can be both debugged with normal SWD debuggers, and can be even used as an SWD debug probe. However, with an NXP K20 (ARM Cortex-M4) running at 50 MHz, 128 KByte of FLASH, and 16 KByte of RAM, it is not the most powerful board. So something more powerful would be cool!
The tinyK20 has the option of a micro SD card on the bottom side. Maybe Paul had the same idea, and Hackaday produced an article back in August 2016 about two new boards: the Teensy 3.5 and 3.6. The Teensy boards are great (thanks, Paul!), but only if using Arduino libraries and the Arduino IDE, which never has been an option for me.
Instead, I want to use normal debugging tools like a J-Link. For the Teensy 3.1, I have found a way to do this (see Hacking the Teensy V3.1 for SWD Debugging) and we used that for several Teensy boards. But that wiring was not really easy, so we better used the tinyK20 as boards.
SWD Debug Signals on Teensy 3.5/3.6
Now it seems that Paul finally has added an easier way to do SWD debugging (several asked for that in the forums/community).
“For these new Teensy boards, I added a provision to use the debug signals. See the comments and bottom-side photo on the Kickstarter page. There’s a pin you pull low to tell the bootloader chip to tri-state those signals. Drive that pin high or just disconnect (it has a weak pullup) to return to Teensy’s normal mode.”
And from the Kickstarter:
“You’ll be happy to hear those test point are on the bottom side. Look for the pads labeled “DD” and “DC”, for Debug Data and Debug Clock. There’s not a lot of room on the bottom side, so had to abbreviate. You might also notice the “DE” pin, for Debug Enable. This pin as a weak pullup. The idea is you pull it low to tell the MKL02 chip to tri-state the debug pins, so you can control them with whatever debug adaptor you use. Just disconnect or drive the pin high to return the Teensy to its normal functionality.”
Well, that sounds great!
Below are the schematics of the Teensy 3.5/3.6 from https://www.pjrc.com/teensy/schematic.html:
An NXP Kinetis KL02Z acts as the onboard interface and bootloader between the Arduino IDE and the ARM Cortex-M4F on the board. I have marked the relevant signals for debugging below (3.3V sense, Reset, GND, DD (Debug Data), and DC (Debug Clock)). The DE (Debug Enable) is close to G (GND):
So all that I have to do is to connect DE and G and I should be able to debug it, right?
- Reset: Optional, but good if the debug probe can force a reset to the device. There is no pull-up on the reset line, as the KL02Z seems to use an internal pull-up.
- 3.3V: I need that signal routed to the debug probe, as, for example, the SEGGER J-Link is sensing the voltage to determine the target device voltage to use the correct voltage levels, plus it can sense if the device is powered. Unfortunately, Paul has not put that signal to the row of ‘debug’ singals (DD, DC, DE, G)
- DD: Debug Data, JTAG_TMS, SWD_DIO
- DC: Debug Clock, JTAG_TCLK, SWD_CLK
- DE: Debug Enable, to be pulled down to tristate the KL02Z debug signals
- G: Ground
Adding a Debug Connector
The following picture shows the location of the debug signals on the bottom side of the Teensy 3.5 and 3.6:
I added a 2×4 SMD pin header and a 1×5 pin row on the bottom side:
I’m using a SEGGER J-Link to connect to the board. The Segger J-Link has the advantage that I can easily connect the signals. The pin out is available on https://www.segger.com/interface-description.html:
Below are the connected the signals to the Segger J-Link:
- 3.3V to pin 1 (VTref)
- Reset to pin 15 (RESET)
- G to pin 4 (GND)
- DE to pin 6 (GND)
- DD to pin 7 (TMS)
- DC to pin 9 (TCK)
It Does Not Work?
So I tried it out, and I was not able to connect to the board. Looking at the signals, it was clear that despite pulling the DE signal low, the Kinetis KL02 keeps sending something to the K64F, making it impossible to take over with the J-Link:
It turned out that pulling down the DE signal is still not supported, from https://forum.pjrc.com/threads/42728-Debugging-strategies from 04-24-2017:
It’s *still* not implemented. Why a year late? Work on Teensy 3.5 & 3.6 (released October 2016) and the USB host library for 3.6 (first release March 2017) pushed that and many other project goals well into 2017.
So what now? Wait again maybe a year or forever? Luckily, I have found a way, but it is the ‘hard’ way.
WARNING: This modifies the Teensy board, and without a doubt you lose any warranty doing that board change!
Pulling the reset line LOW to keep the KL02Z in reset does not work, as the reset line is shared between the KL02Z and the MK66F/MK64F. As the KL02Z on the board is not of any use for me anyway, the decision was to remove it from the board.
Removing the KL02Z from the board is a bit tricky, so we're going to go through how Christian Di Battista (thanks!) helped me get the hardware modification done.
Use something heat resistive and place the board on it:
To get the device off more easily, use Flux Gel:
Apply the Flux Gel on the pins around the KL02Z:
Use a hot air station or anything suitable:
Then heat up the KL02Z until it can be removed. Because it has a bottom ground plate, it might take a while until it can be removed.
Video of the process:
With the KL02Z removed:
There is only one thing missing: The reset line needs now a pull-up to 3.3V. One option would be to add an SMD pull-up resistor (e.g. 4.7K) somewhere on that KL02Z footprint. Another simple, and easier, option is to add a pull-up to the header near the SD card:
Simplified J-Link Connection
With this, the GND line to DE is not needed anymore and can be removed:
If your debug probe does not do any target voltage sensing (as the J-Link does), you don’t need that red Vdd wire. The Reset wire is optional too (but highly recommended), because it is used to reset the target if the processor does not respond on the SWD connection.
And with this, the J-Link SWD protocol is able to take over the ARM Cortex-M4F:
Finally, I can debug the Teensy with Eclipse, GDB, and standard SWD debug tools:
The Teensy is a great board, and Paul Stoffregen does a really great job. But I really don’t understand why he makes it so hard to use his boards for debugging? Maybe others are used to printf() style debugging as a hobby, but to me, this is not the adequate way to develop for an ARM Cortex-M.
To use normal SWD debugging tools with the Teensy 3.5 and 3.6, it requires a hardware change. Because the DE (Debug Enable) signal does not work, it requires removal of the KL02Z microprocessor from the board. And because the board does not allow the standard SWD debug header, I have to create a custom cable. It works fine, but it is not ideal. I really wish Paul would have provided a normal 2×5 SWD debug header or footprint (ideally on the top side). That would make his boards even more useful.