Segger J-Link

[kloetpatra]

No, I have no relations to Lauterbach.
Sorry, I had the wrong document opened: https://www2.lauterbach.com/pdf/training_arm_etm.pdf page 19
You are correct, full data tracing requires a FIFO and can overflow the buffer. However it is still non-invasive and does not influence the program execution in any way. Instruction tracing however requires no buffer and low bandwith.

[peter-h]

I did some more quick tests on the chinese boxes.

Both the "V9-G" and the "V12.9410000XX" work to program a STM32F417 with J-Flash 7.20b. Previously I tested J-Flash only with the V9 box and the non-STM chinese chips.

But only if you load a .hex file. If you load the ex-Cube .elf, J-Flash crashes. So I would not call this top quality software

Compared with STLINK the programming is faster because STLINK doesn't appear to check before writing if the data is the same, so if you are doing development, the Segger will often be faster if you have done only small edits, especially if the program code has not moved up/down.

[kloetpatra]

I think that's not the only reason. Do they use the same SWD-CLK speed? What if you compare both doing mass erase before program? I don't consider the ST-LINK a highly optimized flash tool. There are quite some tricks to speed up programming. Most speed gain comes from a flash loader running in the MCUs RAM. Also I've seen flash loaders not using the PLL to speed up the Core-Clock. I am not even sure if ST-Link uses a flash loader in general. Also the verification method has a huge impact. Reading back the complete flash is very time consuming. Best is a block based CRC verificiation using the on-chip CRC module (if present) and only transferring and comparing the checksums. Optimizing the flash process for speed is not so trivial and there are many details which can have quite an effect. I have written some flash loaders myself and a large improvement also consists in simultaneously sending the next data block in to the MCUs RAM while the flash loader is programming.

[peter-h]

It is not just the STLINK; it is also the Cube IDE code which talks to it. I have no idea how that is organised.

Logically one would implement simple RLL encoding, so 900kbytes of 0xFF doesn't have to get sent But that 900k still needs to get written at the target end, and if you do it with a pre-read then that 900k can be done very fast. I am fairly sure Cube+STLINK does that much.

One difference is the "Segger process" emits way more text into the console, and this includes stuff like erasing whole blocks. On the 32F417 these blocks vary from 16k to 128k and to use this feature you obviously need to "scan ahead" at the PC end and if e.g. the top 128k is all 0xFF (quite a likely scenario) then you can just send a block erase command. My boot loader uses these block erase commands and they are fairly quick.

Currently I am loading ~600k of code and if one changes e.g.

int fred=5 to
int fred=6

then the rest of the 600k block does not change, and that is where the Segger is much faster. How relevant that is to reality, I don't know.

[mrthanhapple]

Hi everyone,

I am currently using a genuine SEGGER J-Link BASE for embedded development.
For my work, I would really benefit from some of the advanced features
(e.g. unlimited flash breakpoints, J-Flash, etc.).

Before purchasing a new probe, I would like to ask:
- Has anyone here upgraded a J-Link BASE with additional licenses?
- Are there any official upgrade paths, discounts, or transferable spare licenses?
- Or would you recommend upgrading to J-Link PLUS instead?

I am looking for a legitimate solution and advice from experienced users.

Thanks a lot!

[thm_w]

Hi everyone,

I am currently using a genuine SEGGER J-Link BASE for embedded development.
For my work, I would really benefit from some of the advanced features
(e.g. unlimited flash breakpoints, J-Flash, etc.).

Price is on the website: https://www.segger.com/purchase/pricing/jlink-related/
380 to upgrade from base to plus
500 for jlink base
750 for jlink plus

If your jlink base is an old hardware version (should say on the label), it would likely make sense to buy a new plus instead.