This page lists important changes to the firmware. Check this before downloading a firmware that is meant for production
When updating the board with latest firmware, READ EVERYTHING BETWEEN OLD AND NEW REVISION and check if any of the changes affects your configuration. Take action if necessary. For example, if you used an 1.0.13 firmware, and upgrading to (eg. 1.0.23), when reading through all changes for intermediate releases, you will see (at 1.0.14) that fuelcut_min_kpa must be adjusted reasonably. If you skip 10 mins to read through, you'll sit in the cold car for hours and wonder why the engine does not start and notice injector pulsewidth is 0.
Developers: please update GenBoard/Manual matrix when configuration semantics change. Ask on that page (in the relevant matrix element; make a line if necessary and start to put in links you find) for questions about config variables.
You can report problems/bugs on this page: IssueReports
Downloadable firmware releases
Download
- [1.0.23] Warning: you might need to replace MegaTune's vemsv3.ini with vemsv3.ini in this package (at least in MegaTune r027).
- [1.0.18] - tested
- [1.0.14 with precompiled 12x12 size maps]
- [1.0.13]
Although it has a compiled vems.hex binary (=> no compiling needed by the enduser), at the minimum check my_make if using it. It has MSNS option defined, that should work with MegaTunix or MegaTune. vems.hex and my_make files should probably move to subdirs (several meaningful different my_make options - feel free to compile + package)
Changelog - from 20041124 release
- multimap support has been added. You can now save and load different maps on the fly. This will load and save the lambda, ve, rpm, kpa and boost tables.
- SIPR support has been added for PSITuner
- a small patch that allows tps_low > tps_high configuration for silly TPS sensor that reads higher voltage for lower throttle opening
- AlphaN
- engine.batt displayed on LCD mlp06
- BoostControl low freq PWM configurable
- injector PW reading changed to 10 usec on LCD
- MegaTunix stores VE to both ve_ref and learnt VE table
- send the real ("learnt" if learning is enabled) VE table to MegaTunix not the ve_ref table (that can be confusing to the unaware)
Since 1.0.44 - [1.0.44 experimental prelease]
- ego.correction initialized to start from 100% (no leaning for a split second when WBO2 just warms up and ego starts to act that caused hickup)
Since 1.0.43 - [1.0.43 experimental prelease]
- tach output tach_div changed to allow precise fine-tune of the output. 32 (hexa 20) is the default, results in 1 tach pulse for each event. For a 5cyl equipped with a 6cyl tach gauge use tach_div=1B (decimal 27: 6 * 27 /32 / 5 = 1.0125 that is 1.25 % error); TODO: check if it's not 16 (hexa 10) in this case: use 0D (decimal 13) for the 5cyl for now and change in new release.
- note: >100% injector duty code rewritten because of a misunderstanding. This should be good (but not well tested), but testing didn't reveal problem with old code either (only found some strange unreasonable config eg. injpwmt=24msec injpwm=6% duty that - although does what it should, can confuse the unaware).
- 1.0.43 has better initial sync for multitooth N-2 or N-1 wheel with camsync (note that it needs lucky missing-tooth and camsync relative position to work always well with in 1.0.42 and before).
Multitooth advanced filter changed - neglect if you don't use it (but read if you like primary_trigger=09)
- toothrel_normal=Nn #tg_prev * n/16 <= tg <= tg_prev * (1+N/16)
- toothrel_missing=Mm #tg_prev * (1.25 + m/8) <= tg <= tg_prev * (2.5 + M/8)
Multitooth advanced filter OLD semantics - before 1.0.40
If the tg (time-gap) between 2 teeth compared to tg_prev (previous time-gap)
- if( tg_prev * (1-n/16) <= tg <= tg_prev * (1+N/16)) than it's considered normal tooth
- if( tg_prev * (1.5 + m/8) <= tg <= tg_prev * (2.5 + M/8)) than it's considered missing tooth
Multitooth advanced filter NEW semantics - since 1.0.41
If the tg (time-gap) between 2 teeth compared to tg_prev (previous time-gap)
- if( tg_prev * n/16 <= tg <= tg_prev * (1+N/16)) than it's considered normal tooth
- if( tg_prev * (1.25 + m/8) <= tg <= tg_prev * (2.5 + M/8)) than it's considered missing tooth
As you see, for the missing-tooth condition, original 1.5+m/8 changed to 1.25+m/8 ... Also, the "n" changed, that is the lower nibble of the toothrel_normal. In new, recommended values:
- toothrel_normal=85 # (decimal 132) for either N-1 or N-2
- toothrel_missing=D6 # (decimal 214) for N-2 and 22 (decimal 34) for N-1
Since 1.0.36 - ALS made nicely configurable (variables not collide with knock2)
- It is extremely important to not leave any random data in the ALS variables (because it might retard heavily or add much fuel, etc...) Adjust the variables even if ALS is not used: see "disable" on the bottom of MembersPage/GaborRacz/NewAlsLaunchAndOthers
Since 1.0.34 - beware that this was accidentally compiled without AUDITRIGGER option (in my_make), that is without 135-tooth support. We'll compile 1.0.35 soon.
- some variables dropped that you'll not feel sorry for;
- if your iac_integral_speed was significantly lower than 0xFF (255) and your iac PID is already tuned fine: using a factor of 256/iac_integral_speed :
- scale up iac_integral_limit_dec
- and iac_integral_limit_inc, and scale down iac_pid_ki
- earlier iac_ign_advance_change and iac_ign_retard_change merged to a common iac_ign_slope
- awec allows 3 times longer afterstart than earlier awc (unit is 6 crank rotations vs. previous 2 crank rotations). Note: decay is slower in the first half of the afterstart
- new als variables: see MembersPage/GaborRacz/NewAlsLaunchAndOthers for how to configure or disable
Beware, [1.0.32] .. 1.0.35 is not released. Only a compiled snapshot. BMW dualpwm is believed to be fixed. my_make has ALS, SOFTLAUNCH and other enabled in a way that doesn't collide (cheers!) with misc2_output options (like in 1.0.31: therefore 1.0.31 is not recommended even for testing). Requires in-depth megatune (vems.ini) knowledge (it should be a 20 mins vemsv3.ini job for someone knowledgable, plus 80 mins testing!) Don't use it unless you are testing and absolutely know what you are doing.
- I drove around with 1.0.33 and it seems to work fine. I used an old MegaTune (r028 or so that was already installed) - the variable offsets are unchanged (except for the meaning of some previously unused or "of little use" variables). Others drove around with 1.0.34.
BEWARE: 1.0.30 is not released - no known problems so far (been used on several cars for weeks now). BMW dualpwm air-valve users made an improvement that should go into 1.1.0 (or 1.0.31, but soon time for 1.1.0 number). Not suitable for recompile, but EasyTherm provides hexpatch utility (and manual method) for changing temp-factors.
- iac_conf bit5=1 (add 0x20 == decimal 32) enables dualpwm. Configure PWM-iac (precise-idle, but stepper disabled), and set this bit. The iac_sol_channel is PWM-ed normally (like a PWM valve) and the next GenBoard/Manual/DigitalOut channel (see GenBoard/Manual/DigitalOut/Table) is PWM-ed inverted. This is useful for the BMW type valve, see MembersPage/GergelyLezsak/IdleControl and MembersPage/GergelyLezsak/IdleControl/Firmware for more info.
- iac_speed is removed from the upper bits of iac_conf, removing the double bit(4) allocation. iac_afterstart_steps config variable is removed and replaced by iac_speed, pwm frequency is now adjustable (as is stepper speed in steps/second) from config through adjustment of iac_speed variable (higher iac_speed means lower frequency / steps/second)
- iac_pwm_frequency = 1 / ((4 + iac_speed & 0x3F) * 4uS * 256) = xxx Hz
- iac_stepper_speed = 1 / ( (1 + iac_speed & 0x3F) * 1024uS) = xxx Steps / S
- the default tables (that anyone can change for himself, see EasyTherm) changed slightly: airdenfactor -40C (first=0 and last=0xFF position) will no longer assume faulty sensor.
- 0V on MAP sensor (which normally does not happen) will no longer mean 100kPa (limphome). But kpafac=00 (otherwise stupid value, since 2kPa MAP sensor range is meaningless ;-) will mean 101 kPa so anyone can trigger limphome from keyboard if he knows what he's doing.
since [1.0.29] - Experimental, use at your own risk
- InputTrigger/AudiTrigger seems to be working on the batch with one known problem: must be rebooted if trigger is lost (switch off ignition for a second if engine stops). Care is needed to provide clean input signals, eg. not injecting primary trigger into secondary trigger input (via capacitive coupling). mdd04 provides binary debug with appr 5+3 bytes per camrot, while mdd0c is verbose binary debug with appr 273 bytes per camrot. Capture and publish binary logs (with notes of how they were taken, eg. RPM) in case of any problem. mdd00 switched auditrigger debugging off.
since 1.0.23
- shiftcut_conf default=00 (old behavior, disables igncut and shiftcut stuff)
- bit4=1 enables fuelcut when shiftcut is triggered
- bit1=1 enables spark-cut for shiftcut_time * 4 msec time when shiftcut is triggered
- bit0=1 enables spark-cut at rev_limit - 200 RPM (ignition based redline instead of fuel based)
- shiftcut_channel specifies mcp3208 input channel: default 0xFF disables shiftcut; 0..7 enables shiftcut for configurable time (shiftcut_time * 4msec) after the given mcp3208 input channel (default is 06 for mcp3208 channel06, that is mcp3208pin7) is pulled to ground
- Now the temp calibration tables (airdenfactor, matfactor, thermfactor) live on dedicated flash-pages, so can be changed independently from the rest of the firmware.
- Either tweaked in the vems.hex file before uploading the whole vems.hex
- or in the flash. However, this requires a new bootloader, as unfortunately the old bootloader does not support erasing just one page :-( (only the whole application-part of the flash)
since 1.0.18
- fixed tps acceleration enrichment. It left infinitely small time for user to press TPS in the last couple releases: effectively disabled tps accel enrichment. The sampling is now synchronized to the tps acceleration enrichment calculation, so result should be more deterministic than earlier.
- note that interpolation of the n[] table (ignition advance) was not fixed because no problem was found: it tested perfect!
since 1.0.17
- boostcontrol configured with MAP-target considers the scaled MAP (not the MAP sensor input voltage)
- boost_conf bit3:2 00,01,10,11 means 20,40,60,80% reference position for the solenoid. 40% or 60% recommended if otherwise no idea. (TODO: someone check the generated assembly code for overflow at the 8*8 bit multiplication - result should be 16bit).
- small firmware bugfix: from megatune config.primep was altered instead of element 0 of the b[] table (boostcontrol-target)
- the new (1.0.14) fuelcut_kpa is renamed to fuelcut_min_kpa. Meaning is unchanged, read below
- new variable to prevent overboost (useful until the boostcontrol PID is tuned) fuelcut_max_kpa=FF to disable; fuelcut_max_kpa=F0 to cut fuel above MAP > 240 kPa (assuming the default MY_CONF += -D KPA_CONFIG_RESOLUTION=1; naturally same meanse 480 kPa for MY_CONF += -D KPA_CONFIG_RESOLUTION=2)
- tach_channel=FF to disable
- tach_divider=.. is not used. Use FF for now, I think 00 will be used finally to disable completely
since 1.0.14:
- 8,10,12 or 16 cyl support with multitooth wheel
- Compiled with 12x12 maps. Known (optical) bug: h-table is apparently 12 element wide on LCD
- fixed the boostcontroller bug, that effected boost-target (calculated from the TPS and t[] and b[] table). If you are planning to use boost control, get this bugfix.
- the firmware intro has been re-introduced to the LCD code and tested. This is an optical feature but tells the user what firmware revision they are running.
- some boostcontroller items have been added to the display
- changed airdenfactor to not lean any more above MAT > 60C (actually, it enriches above MAT > 60C with appr. +0.1%/C ). Below MAT < 50C the universal gas-law is followed (which is appr -0.33%/C). A small plateu is applied from appr. MAT=50..60C
- added lowmap_fuelcut function: config.fuelcut_min_kpa=0E recommended (fuelcut under 14 kPa) or up to 12 (fuelcut under 18kPa). The hysteresis is compile-time (3kPa, and applied as config.fuelcut_kpa + LOWMAP_FUELCUT_HYSTERESIS). Those who use KPA_CONFIG_RESOLUTION=2 (for extreme boost; rare) know that every MAP-kpa related config values must be multiplied by 2 (so max range is 510kPa and 09 means 18 kPa): fuelcut_kpa is no exception. fuelcut_kpa=00 gives old behavior (disables lowmap_fuelcut), but why would anyone do that ? It is extremely difficult to tune injopen, injocfuel, kpaoffs and other config variables that become significant at low-MAP for nice engine operation at MAP=0..10 kPa . The result was engine-stumbling when you suddenly get your foot off the throttle - the iac reference positions, config.iac_integral_limit_dec and overrun_fuelresume can be tuned to eliminate the stumbling by ensuring MAP > 10kPa always holds at RPM < overrun_fuelresume, but appropriate fuelcut_min_kpa value ensures fuel is cut even if those crazy-low MAP values - and therefore short pulsewidths - are reached. This works so well that config.tpsdq related decel-fuelcut seems quite useless, scheduled for demolition in the near future.
since 1.0.13 rc5:
- Tested with MegaTune r021, the high resolution MAP and RPM shows correctly in megatune.
- New config variables for advanced multitooth filter: config.toothrel_normal=0x8A recommended (for either N-1 or N-2 wheels). config.toothrel_missing=0xB5 (for N-2, eg. 60-2) and config.toothrel_missing=0x00 (for N-1, eg. 36-1). Actually, these only matter if the advanced multitooth filter is enabled by config.primary_trigger bit3=1 and multitooth is configured bit1=0 (thus primary_trigger=09 instead of 01). If using the values recommended earlier: primary_trigger=01 for multitooth, or primary_trigger=FF/FE for "coil-type", the new (toothrel_normal and toothrel_missing) variables do not effect operation at all. For multitooth (either with the advanced filter ON or OFF ), watch the wheel_error counter (W.. on the top of LCD page 01: mlp01), if it keeps increasing, the trigger setup must be examined. A few counts during startup (cranking) is normal.
since 1.0.13:
- camsync improvements
- resync if VR trigger has noise during cranking
- unfortunately rc5 was compiled with decreased trigger resync-noise-suppression (mostly used at first-crank) values (since than reverted, to decrease chance of firstcrank misfire in case of noisier setups)
since 1.0.12:
- better MegaTune compatibility (vemsv3.ini r16 or newer)
since 1.0.11:
- kpa configuration resolution can be changed to 2kPa (but is 1kPa by default) for very high boost engines: add MY_CONF += -D KPA_CONFIG_RESOLUTION=2 to my_make
- Note1: changes also needed in MegaTune vemsv3.ini
- Note2: all the following config variables change to max 510kPa range (2kpa unit):
- k[] table
- ego_maxmap
- config.inj_stage2_start_map
- tpsdot rates
- hybrid_rpm_a
- hybrid_rpm_m
- some kpa text on LCD will become 2 kPa (eg. the axis in mtt menu)
- misc1out_minmap
- misc1out_maxmap
- misc2out_minmap
- misc2out_maxmap
- ve_learn_kpa_scale (this is not in 2kPa units, but it needs to be doubled ?)
- restart not needed after adjusting wbo2_nernstdc_target (from menu) to take effect.
- iac integral starts at max value, not 0 (higher iac position => faster startup without throttle)
- better wbo2 leave_warmup condition (leaves warmup and starts normal heater control earlier if sensor was warm for some reason; eg. because of mde02)
- after mdd02 command, fuel is cut if throttle is pressed (above tps_thresh). This can be used to calibrate WBO2 free-air calibration with WOT and fuelcut (or as a minimalistic "antitheft"). mdd00 changes back to normal behaviour.
- comm.c changed for recent r008 MegaTune (more variables logged)
- in the r1.0.11 minor lcd_display changes: "RPM:" string was addded (back), and fixed the scale of free-air o2 on LCD
- in the r11 there is a new feature on LCD page00 (mlp00): La:... shows Lambda (like "0.95" or "1.02") or O2% (like "20.7" - note that the dot moves to different position in this case!) or undefined (??.?)
- note: there was a small problem with 02 display, so in the r11_v2 in lcd_print_current_o2() a small fix was made to lcd_print_XX_X( t32 >> 8 );
- note: 'mde02' command must be used to switch on WBO2 unless engine is running (trigger is applied). This is intended this way.
since 1.0.9: camsync with coil-type primarytrigger (not restricted to multitooth). VERY IMPORTANT: USE secondary_trigger=02 if you don't use camsync, otherwise (eg. with commonly used FF value, that works no more) you won't get spark nor fuel nor automaticly activated WBO2 !!!!!
Remember:
- WhatEverYouDoYouDoItAtYourOwnRisk
- GenBoard/Manual/License/WBSensor applies to the release, since it has WBO2 enabled in my_make
2005-02-20
- committed 2x20 LCD support. This is identical to 4x20 except that it has twice the number of pages. Every second page is just the bottom 2 lines of the 4x20 version.
2005-02-08
- stable revision of 2004/11/24 16:00 was tagged as STABLE_1_0. This was done in order to separate alpha quality code found in HEAD with tested production quality code we should be running. Each revision tag will only be modified with bugfixes.
2004-10-31
- BoostController (old variables not changed) and improved injector-opening rampup control. injocfuel semantics changed, see GenBoard/Manual/Config/InjectorOpening
2004-10-27
- Improved idle control. Ignition advance is adjusted to maintain a constant idling rpm. To disable this feature, set iac_ign_advance_change=00, iac_ign_retard_change=00, iac_ign_advance_limit=00, iac_ign_retard_limit=00, iac_ign_threshold=ff.
2004-10-10
- speed-density precision bumped up greatly. The first chance ever for the diy-ers to do precise fueling for turbo engines. The old semantics we used earlier (inherited from motorola-megasquirt) used 4% (!) steps at 25kPa MAP (and 1% steps at 100kPa). This was the last remained inferior signal precision in the whole fuel-calculation (we made everything nice in the meantime, eg. ego correction. this waited somehow...). Now this signal is 11 bit internally (1/20 %).
- set config.kpafac to your MAP range [2kPa].
- 2.5 bar MAP sensor: 7D
- 1.15 bar MAP sensor: (hexa) 39
- 5.1 bar MAP sensor (impressive boost !!): FF (max value)
Note that this makes GenBoard/VerThree the first diy ECM that supports boost pressure higher than 150 kPa without hack. (this was the main reason for the change: a racecar runs at 200..240 kPa boost and they asked us to make precision superior)
2004-09-17
- eventqueue changed in some ways:
- more reasonable inlining
- faster reaction
- better readability
CVS head, with eventqueue.c 1.96 (or earlier 1.94, but not 1.95 experiment) is good for production.
Played with eventqueue, make it simpler and faster. For readability and marketing reasons; because some people are concerned about parameters that cannot be measured in engine output, only on oscilloscope. At 15000 RPM measured 6/20 degree average delay in the ign signal, this is now lowered to about 4 / 100 degree (could be even better with further changes). This is so close to theoretical optimum 0.000000 sec that it means less than 0.3C degree in intake temps (MAT).
2004-09-12
- comm.c tabledump improved so it can send out the 16x8 table (if someone forgot what he sent in :-), someone please test mct command (write here) -> mct command verified, j/l/n/r 16 wide, h/k 8 wide - DB . If VE_TABLE_FIX is defined in my_make (what a bad name !!!) the 16 bit learnt VE table is outputted in j[x]=xxxx xxxx xxxx ... format, possibly 16 values on a line (if VE_SIZE_RPM=16).
2004-08-27
- config.cwl, cwh, injpwmt and primep can be configured to higher than 25.5 msec : see GenBoard/Manual/Config/CoolantTemperature
2004-08-25
- knock sensor code is now allowed to modify ignition advance. Unless you have mounted a knock sensor, set config.knock_conf = 00 AND config.knock_default_retard = 00 such that the actual ignition advance, as specified in the n-table, is unaffected.
2004-08-08
- the multitooth branch initialized engine.igncount properly => cranking was long for coil-type setup. (it was immediately obvious from TimingLight table tests, so I guess everyone uses multitooth trigger). Anyway, anyone with coil-type setup should upgrade.
2004-07-28
- synchronized o2() to WBO2 calc_afr(). This allows setting faster EGO convergence, without the risk of divergent behavior for bad (speed_limit =) config.ego_pid_kp; config setting.
- maybe we could make calc_afr() even more frequent, with increasing the frequency of nernst sampling or just running the pump controller more often. Someone could experiment with it.
2004-07-27
- multitooth setup improved.
- Dave will test it soon (and document the new variable), and than it worths to upgrade (with multitooth setup).
- preparations done for time from last possible pulse
2004-07-25
- change config.tooth_wheel_twidth2 to include the config.tooth_wheel_twidth1 as well (add tooth_wheel_twidth1 at upgrade if you had it configured before);
- 2 * tooth_wheel_twidth1 for a 36-1 and
- 3 * tooth_wheel_twidth1 for a 60-2
2004-07-07
variable intake actuator (simple RPM-based switch) and WOT actuator (that switches at WOT, but only above a certain RPM) for NOS made easily configurable.
Both the RPM and the output channel is configurable. On GenBoard/VerThree you can select a free channel, which can be a 100A FET if you like (like IRL2203, but the connector cannot handle that current). Even several channels (this is only for the injector FETs) can be selected to switch at the same time, according to GenBoard/Manual/DigitalOut
Set all these config variables to FF if you don't need the feature:\n
uint8_t act_wot_rpm; // WOT switch RPM threshold (FF=disable) uint8_t act_wot_channel; // WOT switch output channel selector (P259_0) uint8_t act_rpm_rpm; // RPM switch RPM threshold (FF=disable) uint8_t act_rpm_channel; // RPM switch output channel selector
Iac.conf is extended to select between a ON/OFF idle air valve, and a PWM controlled type by unsetting/setting bit 3 of iac.conf
2004-05-26
Lambdacorr table is now unconditionally used. If no l-table was used before, then add the following (stoich) table to tables.txt\n
### lambdacorr: afr=14.7 ### l[0]=38 38 38 38 38 38 38 38 l[1]=38 38 38 38 38 38 38 38 l[2]=38 38 38 38 38 38 38 38 l[3]=38 38 38 38 38 38 38 38 l[4]=38 38 38 38 38 38 38 38 l[5]=38 38 38 38 38 38 38 38 l[6]=38 38 38 38 38 38 38 38 l[7]=38 38 38 38 38 38 38 38
The lambdacorr-values are a multiplied enrichment: \n
lambdacorr = (l+200)/256
You can see that 56 (hexadecimal 38) means a 1.0 (stoich) lambdacorr value.
Plan: Status indicators in megatune\n
;for the following 3 bits, they should indicate if the switch is activated, not function
indicator = {status1 & 0b00000001}, "ALS Off", "ALS Btn On", white, black, blue, black
indicator = {status1 & 0b00000010}, "Launch Off", "Launch On", white, black, blue, black
indicator = {status1 & 0b00000100}, "Shift-cut Off", "Shift-cut On", white, black, red, black
;for the next 2, they should indicate active function (WBO2 sampling, IAC PID control)
indicator = {status1 & 0b00001000}, "WBO2 Off", "WBO2 On", red, black, green, black
indicator = {status1 & 0b00010000}, "IAC Off", "IAC On", white, black, green, black
If someone can come up with more bits here, feel free to add them to firmware and i'll commit the needed lines in vemsv3.ini
Remove this bit when firmware part is implemented
//Emil
Discussion about a limphome feature:
GenBoard/UnderDevelopment/FirmwareChanges/LimpHome
Release system discussion:
GenBoard/UnderDevelopment/FirmwareChanges/TestingAndReleases