canadian
sy in progress
some interesting news from another list i'm on:
Bruce Bowling, the co-creator of the Megasquirt, had the following to say:
>>>>>>>>>>>>>>>>>>>>>>>
Yes, we have been (in secret, of course) working on a combined MegaSquirt,
MegaJolt Lite (and support for MegaJolt), and wideband (Bosh LSU) single
board setup for
the last 3 months.
We have been hearing the requests over the last two years for various
upgrade and enhancements. Since the MS is starting to become stable and
withstanding the test of time it is time to take the next step.
We were going to put out just an ignition board as a companion to MS, but
several outstanding individuals have taken the existing MS board an adapted
it to ignition - with phenomenal success! In fact, with the MegaSpark
having the EDIS function in the code, it has become a very significant
ignition setup.
What we are working on has the following enhancements:
1) Processor. When MS was started almost three years ago, the best bang
for the money was the HC08 family (also Atmel and Microchip). In three
years, looking at the next step range of processors, the family that stands
out is the HC12/HCS12 - more on this on the next email...
2) Connector. The current MS connector works well. But, it can be a little
tedious to solder in large wires. We have been shaking the trees for
connector suppliers, and it looks like we can get the AmpSeal series of
connectors - these are sealed automotive-grade connectors.
3) Automotive-grade. Everything is automotive-grade, including processor,
connector, etc.
4) Existing software. The goal of choosing a processor is to try to keep
the current established code operational. With the HC12 family, it is
extremely easy to modify existing assembly code for this processor - it
amounts to about a two hour effort, without taking advantage of the HC12
architecture. And, if people want to code in C, there is a good GNU port
available, as well as Metrowerks Codewarrior. See the next email.
5) Ignition - Since this is a hardware upgrade with a easy migration path
for existing software, codes like MegaSpark, Tomtek, EDIS, etc can work on
this platform. And, the logic will be there to interface to full-up
MegaJolt.
6) Wideband O2 - It is quite clear that people want to tune to mixtures
other than 14.7:1. So, the circuitry to handle the Bosch LSU is on-board.
This is the current R&D effort right now. Al and myself have spent
literally thousands of dollars on good wideband testers, including the
coveted ETAS LA3 Lambda Meter. We have been setting up a WBO2 flow bench in
order to measure the WB circuit and compare against the LA3. Controlling
the Bosch sensor is quite different compared to the NTK unit, especially
the heater circuit. We will be able to get real data comparisons with real
test equipment, which will help work out things like feedback transient
response, etc.
We are making up the flow bench right now - we are collecting exotic items
like Swagelock solenoid vales, Nupro metering valves, Venturi mixers,
bottles of gas mixtures (including hydrogen, CO, etc), etc.
This is the same thing that Garfield was doing with EGOR, but, unlike
Garfield, we will share everything we find out to everyone. This by itself
will be really educational - worth the money that Al and myself have
invested.
7) Form factor. The box must be smaller than the current MS box. To do this
we must go to surface mount for everything.
8 ) Cost. Target cost is just a little more than the current MS setup. The
connector is quite expensive (~20.00) but the other components are roughly
the same cost as the current MS box.
Now, before everyone asks for circuits, etc - right now the schematics are
in a notebook and the subsections are breadboarded - there is no complete
schematics right now, but tested subsections. This will change shortly.
Oh, yes, it is called "UltraMegaSquirt" - what else...
More on the processor - the MC9S12A256 device from Motorola:
http://e-www.motorola.com/webapp/sps/site/prod_summary.jsp?code=MC9S12A256B&
nodeId=0162468636K100
A totally awesome part. It is a true automotive processor with automotive
temperature range. Its original use was for ABS braking systems - zillions
of these are used in today's cars. Being that it is for braking, not only
is it tested and rated for automotive applications, it has an extended
testing for latchup due to OEM safety requirements (see the specs in the
electrical section of the datasheet). It is a TRUE 16-bit processor with a
TRUE 16-bit data bus. The bus runs at 25 MHz - this is true instruction
clock speed, not a crystal speed. This particular part has 256K of flash -
a 512K version will be out in a few months (that's 1/2 MEG of on-board
flash). It has an extended timer channel (8 of them) - one can see that it
has a lot of the functionality that was needed in ABS braking (i.e.
transmission control). And the usual 10-bit ADCs, UARTs - even CAN versions.
The 64-K version of this part (all HC12/HCS12's are drop-in replacements)
cost only $1.20 more than the current MC68HC908GP32 we are using right now.
Read this link below :
http://e-www.motorola.com/brdata/PDFDB/docs/AN1284.pdf
Pay particular attention to page 8 - the table of hardware math functions.
The really big thing about this part is the hardware EDIV function - this
is a "hardware divide" function - 32 bits by 16 bits. This part does this
in 11 clock cycles!!!!! Why is this important? Because to determine RPM to
high accuracy a divide with many bits is required. More RPM resolution
requires a divide with more bits. You can synthesize a divide up with a
bunch of op-codes (RISC machines tend to do this since many lack hardware
divide, especially of this magnitude), but these take clock cycles - lots
of them. This part does this in 11 clock cycles - and these are coming at
25 million ticks a second! It only makes sense - this is an ABS-targeted
part - so it needs to monitor wheel motion and compute RPM (or change in
RPM) extremely fast. BTW - the most expensive operation in the MS code
(time-wise) is the computation of the 16 x 16 divide in order to determine
slow RPMs - the HC08 only has a 16 x 8 hardware divide - but when it cones
time to compute a slow RPM a 16x16 is required. And this yields the RPM to
the resolution of 100 RPM that we have today. Take a look at the MegaSpark
code to see what it takes for high-bit divides.
Yes, this is fast at 25 MHz, but it is a CISC machine, meaning many clocks
cycles per instruction - right? Look at page 8 again and see the reference
that "there are more than 50 opcodes which execute in one clock cycle".
Also look at the instructions like "WAV" and "TBL" - think how much they
will help things like filtering and table lookup...
Finally, the op-codes are very similar between the HC08 and the HC12 - so
similar in fact that the whole MS assembler code could be converted in a
few hours - the changes are of order like converting "LDA" to "LDAA" and
converting to the new timer calls. One may want to use the TBL function to
do the VE table lookup as well. And, of course, using the nice hardware
divide...
Tools? There are:
- Free assemblers, like the AS12, MCUez, Metrowerks, and low cost P&E.
- A good GNU toolchain port, including GDB and GCC. Many of the robot
people are using this on their boards.
- Metrowerks Codewarrior. In fact, there is a free version that is limited
to 8-K compile, but unlimited assembly and part download size (get this now
while it is still available from Metrowerks.com). And, the simulator is
absolutely phenomenal!
- The Background Debug Mode is one of the best I have used - you can
actually look at variables and registers while the code is executing on the
target. A real upgrade compared to the MON08.
- And, there is a D-Bug12 monitor code which can be put in the chip
(available free from Motorola). Many people include this with their final
application - you enter the mode via a few pin logic values at bootup
(sorta like the bootloader in the current MS). You can download S-record
code just like we can right now, but you can also look at memory and
registers, step code, upload and execute user code into RAM (for testing),
etc.
So, here is a processor that is three times faster than what we have today
(and what we have right now is plenty fast) at true 16-bit, has up to 512K
of on-board flash, is automotive temperature grade, can easily run existing
MS assembler code with a little touchup, has gobs of free tools available,
and costs roughly the same as what we are paying for the current HC08.
http://www.bgsoflex.com/megasquirt.html
Bruce Bowling, the co-creator of the Megasquirt, had the following to say:
>>>>>>>>>>>>>>>>>>>>>>>
Yes, we have been (in secret, of course) working on a combined MegaSquirt,
MegaJolt Lite (and support for MegaJolt), and wideband (Bosh LSU) single
board setup for
the last 3 months.
We have been hearing the requests over the last two years for various
upgrade and enhancements. Since the MS is starting to become stable and
withstanding the test of time it is time to take the next step.
We were going to put out just an ignition board as a companion to MS, but
several outstanding individuals have taken the existing MS board an adapted
it to ignition - with phenomenal success! In fact, with the MegaSpark
having the EDIS function in the code, it has become a very significant
ignition setup.
What we are working on has the following enhancements:
1) Processor. When MS was started almost three years ago, the best bang
for the money was the HC08 family (also Atmel and Microchip). In three
years, looking at the next step range of processors, the family that stands
out is the HC12/HCS12 - more on this on the next email...
2) Connector. The current MS connector works well. But, it can be a little
tedious to solder in large wires. We have been shaking the trees for
connector suppliers, and it looks like we can get the AmpSeal series of
connectors - these are sealed automotive-grade connectors.
3) Automotive-grade. Everything is automotive-grade, including processor,
connector, etc.
4) Existing software. The goal of choosing a processor is to try to keep
the current established code operational. With the HC12 family, it is
extremely easy to modify existing assembly code for this processor - it
amounts to about a two hour effort, without taking advantage of the HC12
architecture. And, if people want to code in C, there is a good GNU port
available, as well as Metrowerks Codewarrior. See the next email.
5) Ignition - Since this is a hardware upgrade with a easy migration path
for existing software, codes like MegaSpark, Tomtek, EDIS, etc can work on
this platform. And, the logic will be there to interface to full-up
MegaJolt.
6) Wideband O2 - It is quite clear that people want to tune to mixtures
other than 14.7:1. So, the circuitry to handle the Bosch LSU is on-board.
This is the current R&D effort right now. Al and myself have spent
literally thousands of dollars on good wideband testers, including the
coveted ETAS LA3 Lambda Meter. We have been setting up a WBO2 flow bench in
order to measure the WB circuit and compare against the LA3. Controlling
the Bosch sensor is quite different compared to the NTK unit, especially
the heater circuit. We will be able to get real data comparisons with real
test equipment, which will help work out things like feedback transient
response, etc.
We are making up the flow bench right now - we are collecting exotic items
like Swagelock solenoid vales, Nupro metering valves, Venturi mixers,
bottles of gas mixtures (including hydrogen, CO, etc), etc.
This is the same thing that Garfield was doing with EGOR, but, unlike
Garfield, we will share everything we find out to everyone. This by itself
will be really educational - worth the money that Al and myself have
invested.
7) Form factor. The box must be smaller than the current MS box. To do this
we must go to surface mount for everything.
8 ) Cost. Target cost is just a little more than the current MS setup. The
connector is quite expensive (~20.00) but the other components are roughly
the same cost as the current MS box.
Now, before everyone asks for circuits, etc - right now the schematics are
in a notebook and the subsections are breadboarded - there is no complete
schematics right now, but tested subsections. This will change shortly.
Oh, yes, it is called "UltraMegaSquirt" - what else...
More on the processor - the MC9S12A256 device from Motorola:
http://e-www.motorola.com/webapp/sps/site/prod_summary.jsp?code=MC9S12A256B&
nodeId=0162468636K100
A totally awesome part. It is a true automotive processor with automotive
temperature range. Its original use was for ABS braking systems - zillions
of these are used in today's cars. Being that it is for braking, not only
is it tested and rated for automotive applications, it has an extended
testing for latchup due to OEM safety requirements (see the specs in the
electrical section of the datasheet). It is a TRUE 16-bit processor with a
TRUE 16-bit data bus. The bus runs at 25 MHz - this is true instruction
clock speed, not a crystal speed. This particular part has 256K of flash -
a 512K version will be out in a few months (that's 1/2 MEG of on-board
flash). It has an extended timer channel (8 of them) - one can see that it
has a lot of the functionality that was needed in ABS braking (i.e.
transmission control). And the usual 10-bit ADCs, UARTs - even CAN versions.
The 64-K version of this part (all HC12/HCS12's are drop-in replacements)
cost only $1.20 more than the current MC68HC908GP32 we are using right now.
Read this link below :
http://e-www.motorola.com/brdata/PDFDB/docs/AN1284.pdf
Pay particular attention to page 8 - the table of hardware math functions.
The really big thing about this part is the hardware EDIV function - this
is a "hardware divide" function - 32 bits by 16 bits. This part does this
in 11 clock cycles!!!!! Why is this important? Because to determine RPM to
high accuracy a divide with many bits is required. More RPM resolution
requires a divide with more bits. You can synthesize a divide up with a
bunch of op-codes (RISC machines tend to do this since many lack hardware
divide, especially of this magnitude), but these take clock cycles - lots
of them. This part does this in 11 clock cycles - and these are coming at
25 million ticks a second! It only makes sense - this is an ABS-targeted
part - so it needs to monitor wheel motion and compute RPM (or change in
RPM) extremely fast. BTW - the most expensive operation in the MS code
(time-wise) is the computation of the 16 x 16 divide in order to determine
slow RPMs - the HC08 only has a 16 x 8 hardware divide - but when it cones
time to compute a slow RPM a 16x16 is required. And this yields the RPM to
the resolution of 100 RPM that we have today. Take a look at the MegaSpark
code to see what it takes for high-bit divides.
Yes, this is fast at 25 MHz, but it is a CISC machine, meaning many clocks
cycles per instruction - right? Look at page 8 again and see the reference
that "there are more than 50 opcodes which execute in one clock cycle".
Also look at the instructions like "WAV" and "TBL" - think how much they
will help things like filtering and table lookup...
Finally, the op-codes are very similar between the HC08 and the HC12 - so
similar in fact that the whole MS assembler code could be converted in a
few hours - the changes are of order like converting "LDA" to "LDAA" and
converting to the new timer calls. One may want to use the TBL function to
do the VE table lookup as well. And, of course, using the nice hardware
divide...
Tools? There are:
- Free assemblers, like the AS12, MCUez, Metrowerks, and low cost P&E.
- A good GNU toolchain port, including GDB and GCC. Many of the robot
people are using this on their boards.
- Metrowerks Codewarrior. In fact, there is a free version that is limited
to 8-K compile, but unlimited assembly and part download size (get this now
while it is still available from Metrowerks.com). And, the simulator is
absolutely phenomenal!
- The Background Debug Mode is one of the best I have used - you can
actually look at variables and registers while the code is executing on the
target. A real upgrade compared to the MON08.
- And, there is a D-Bug12 monitor code which can be put in the chip
(available free from Motorola). Many people include this with their final
application - you enter the mode via a few pin logic values at bootup
(sorta like the bootloader in the current MS). You can download S-record
code just like we can right now, but you can also look at memory and
registers, step code, upload and execute user code into RAM (for testing),
etc.
So, here is a processor that is three times faster than what we have today
(and what we have right now is plenty fast) at true 16-bit, has up to 512K
of on-board flash, is automotive temperature grade, can easily run existing
MS assembler code with a little touchup, has gobs of free tools available,
and costs roughly the same as what we are paying for the current HC08.
http://www.bgsoflex.com/megasquirt.html