Vida CEM swapping

[RickHaleParker]

I didn't mean the noise over the voltage, i meant the voice in timing. You have an scope, can you measure the time between sending and receiving, lines should stay at 2.5V for 60 to 90us. Also a logic analyzer performes very well for this task.

The scope is a 20 Mhz Analog CRO. I have never tried using it for digital signals. It might be time for me to get a DSO and/or a logic analyzer. I did RF communications. Digital is just a hobby.

I still think using the Teensy's interrupts would be a better approach to detecting Latency. That way the Teensy would tell us when events happen rather then trying to ferret out the events from a messy signal. Using the interrupts may get more consistent Latency times. Which would make detecting the correct Byte as simple as: If Latency > X+Y then Byte = true.

If I was something other then a C grasshopper. I would try the interrupt approach myself.

I do recall using hardware interrupts to time events on the old C= 64 and C= Amiga. Worked real good and was blazing fast.

[T5Luke]

The cheapest (around 10 dollars) usb analyzers are enough to analyze the bus. A anser for wrong or right code should be relatively constant.

Interrupts also have its cons, on most cpus they have a lot of overhead, they backup registers to stack and call other routines before the main task begins. So lots of cpu cycles get lost just calling an interrupt, also it is possible time is not the same for each interrupt calls. I don't know how they are handled by arduino, especially i dont know how they are handled by teensy. Normally a simple poll loop should be more precise. The arduino functions loose lots of cpu cycles just reading a pin (around 50 cycles on standard arduino), it could be less then 10 with direct port access, i dont know how much cycles get lost at teensy for just on digital read.

[vtl]

The original idea was to detect latency as low as like 6 cycles of MCU base frequency (30 MHz for Renesas). No interrupts in the world would allow doing this, other than on CPU clocked at like 10+ GHz. Later it turned out the CAN bus cycle is always aligned at CAN clock frequency, not MCU clock. So such high the precision is not required in that case.

I've tried reading directly, but it didn't help, because the initial algorithm was misfit for the task.

Agree with T5Luke, a counterfeit $11.xx Saleae Logic clone is enough for the task. I use it all the time. When I become rich I'll buy a real one, just as a thanks to Saleae team.

[RickHaleParker]

The cheapest (around 10 dollars) usb analyzers are enough to analyze the bus. A anser for wrong or right code should be relatively constant.

Interrupts also have its cons, on most cpus they have a lot of overhead, they backup registers to stack and call other routines before the main task begins. So lots of cpu cycles get lost just calling an interrupt, also it is possible time is not the same for each interrupt calls. I don't know how they are handled by arduino, especially i dont know how they are handled by teensy. Normally a simple poll loop should be more precise. The arduino functions loose lots of cpu cycles just reading a pin (around 50 cycles on standard arduino), it could be less then 10 with direct port access, i dont know how much cycles get lost at teensy for just on digital read.

It looks like the $10.00 USB Logic analyzers are meant to work with certain software packages. Is there a open source application that will work with the $10.00 USB Logic analyzers?

How about we let the interrupt method be an alternative idea in case we cannot get the reliability needed. Nobody will know if or how well it works until somebody tries it. I do think using the interrupt method would eliminate the cycles between Request and Received, TX ready and TX sent as variables.

[T5Luke]

Anybody has Dice, CODE and likes to try?
Only tested 1 time on bench and it worked. CEM allows only 100 write cycles so i dont like to do endless tests.



There are some small things i need to fix, if tests work 3 times i will upload here for all...

[lukas743]

Anybody has Dice, CODE and likes to try?
Only tested 1 time on bench and it worked. CEM allows only 100 write cycles so i dont like to do endless tests.



There are some small things i need to fix, if tests work 3 times i will upload here for all...

does it work in old P2 CEM? MY02. I have CEM on bench and working PIN

[T5Luke]

No at the moment only CEM from MY 2005 or newer, teensy can only generate code for this at the moment, when we crack the earlier CEM this way i will also modify for this...

[RickHaleParker]

Updated CEM list 4/24/2021

P1 CEM Part numbers.
8690719 | 8690720 | 8690721 | 8690722 | 30765471 | 30728906 | 30765015 | 31254317 | 31327215 | 31254749 | 31254903 | 31296881

P2 CEM-B (Brick shaped 1999-2004 with K-line ) Part numbers.
8645716 | 8645719 | 8688434 | 8688436 | 8688513 | 30657629 | 9494336 | 9494594 | 8645171 | 9452553 | 8645205 | 9452596 | 8602436 | 9469809 | 8645200 |

P2 CEM-L ( L shaped and marked L 2005 - 2014 ) Part numbers.
30682981 | 30682982 | 30728542 | 30765149 | 30765646 | 30786475 | 30786889 | 31282457 | 31314468 |

P2 CEM-H ( L shaped and marked H 2005 - 2007 )
30786476 | 30728539 | 30682982 | 30728357 | 30765148 | 30765643 | 30786476 | 30786890 | 30795115 | 31282455 | 31394157 | 30786579

Notes
30759114 listed as a L shaped P2 CEM unable to determine if it is a CEM-L or CEM-H.

[RickHaleParker]

Agree with T5Luke, a counterfeit $11.xx Saleae Logic clone is enough for the task. I use it all the time. When I become rich I'll buy a real one, just as a thanks to Saleae team.

What software are you using with the Saleae Logic clone? All the clones I looked at don't say anything about included software.

[T5Luke]

They work by original software.