Another GL1100 Megasquirt EFI project

This is where discussions of EFI systems, Dual Carbs, Single Carb mods and plans can be discussed and shared.

If you are working on or have a problem with stock carbs then your questions do not belong here, please post them in the Goldwing Tech Discussions Forum.

Moderators: sunnbobb, Neil, Forum Moderators

Post Reply
thesteve
Zinc Member
Zinc Member
Posts: 40
Joined: Fri Sep 04, 2015 10:25 am
Location: Phoenix AZ

Re: Another GL1100 Megasquirt EFI project

#46

Post by thesteve »

I just set some basic noise filtering up and went for another ride. I set it to discard any tach pulses which happen 3.8ms after the previous valid pulse. It's definitely better, but I'm still getting some spikes. They've moved from the 5000 RPM range up to 6500 RPM and above. I'll try the more advanced filters tomorrow. The FSM says the dwell time ranges from 0.2 to 0.4ms, so I should be able to set 0.2ms as the minimum pulse required to register as an ignition event. That should take care of it. If it doesn't I'll probably need a larger capacitor on the tach input circuit.

After reviewing my datalogs tonight I noticed that my VE values are getting incredibly high. They're now in the low 200% range. I think the fuel injector size value I've got is way off, which is making my reqfuel half of what it should be. Do you know what size the stock GL1200 injectors are? I found somewhere (source is eluding me) that they were 280cc/min, but that appears to be nearly double what they actually are if my VE table is to be believed. I'm running 4 squirts alternating, which should give each bank 2 squirts per cycle.
User avatar
wingrider
SUPER BIKER!!!!
SUPER BIKER!!!!
Posts: 3724
Joined: Thu May 24, 2012 12:15 pm
Location: Maple Lake, MN USA

Re: Another GL1100 Megasquirt EFI project

#47

Post by wingrider »

I believe Irishcarbomb had said the 1985 injectors were 280cc/26.7lb per hour....so I would say that is correct.
2002 1800 Wing
2001 Valkyrie Interstate
1978 1200 EFI Wing with Motorvation Coupe Royale sidecar
1977 1200 Wing cafe/bobber project
1974 Suzuki GT550
thesteve
Zinc Member
Zinc Member
Posts: 40
Joined: Fri Sep 04, 2015 10:25 am
Location: Phoenix AZ

Re: Another GL1100 Megasquirt EFI project

#48

Post by thesteve »

That must be where I saw it. I read through every single EFI conversion thread here before beginning, and 280cc is definitely what I have written in my notes. What gets me is that the built in calculator uses engine displacement and injector size/count to make a number for the time it takes to get you a theoretical 14.7:1 AFR burn at 100% load, and for my other projects it's been pretty accurate. Here it's looking like it's half the size it needs to be. So either my injectors are closer to 140cc, my engine was somehow enlarged to 2200cc, or there's something weird going on elsewhere in the settings that I can't find. The VE table is an efficiency percentage, so at 100% it should be delivering that exact number that was calculated. Mine are above 200% now, so it's more than doubling that theoretical max.
thesteve
Zinc Member
Zinc Member
Posts: 40
Joined: Fri Sep 04, 2015 10:25 am
Location: Phoenix AZ

Re: Another GL1100 Megasquirt EFI project

#49

Post by thesteve »

I have an idea regarding the tach spikes. Here's a diagram of the GL1100 ignition system:

Image

I have a 1983, so mine is the one on the right. Thinking back, my Katana had no tach spikes once I switched to the coil trigger. It also didn't have the resistor limiting the charging speed of the coils. I'd be willing to bet that on an oscilloscope, the voltage on the wire between the coil and spark unit on an 80-81 would show pretty much a square wave. It'd charge nearly as fast as it would discharge, giving a nice square wave for the MS to interpret.

The 82-83 have that large resistor between the coil and positive power. That should limit the charging speed, and therefore make the voltage increase slower. This would mess up the rising edge of the signal, causing an extended moment where the signal is at the crossing point. Any fluctuation in the voltage during this period could cause the computer to think that the coil has fired again.

This drawing shows what I'm talking about:

Image

Both graphs show voltage over time, with the coils firing 3 times. The falling edge is the spark module grounding the coil and firing a spark. The rising edge is the spark unit cutting off the spark and the bike's electrical system recharging the coils. The peak voltage is the point at which the coil becomes saturated, and the flats at the bottom indicate the spark module grounding the coil and holding it (dwell time). The red line is the crossover point. Every time the voltage goes above then below this point, the computer sees that a coil has fired.

The top graph shows the approximate voltage as seen by the computer for an 80-81 Goldwing ignition. With no charge limiting resistor, the voltage rises almost instantly.

The bottom graph shows the approximate voltage for the 82-83 Goldwing. The resistor causes voltage to rise much slower in comparison. The voltage lingers around the crossover point (red line) much longer. If that voltage sags under that line after going above it for any time (we're talking microseconds here, 1/100s of milliseconds) then the computer counts it as a valid tach pulse. This is such a short period of time and such a high sensitivity that the charging system itself can cause issues as the rotor passes individual coils in the stator, causing an incredibly brief drop in voltage.

The solution is noise filtering. The Megasquirt 2 can do two types of filtering: time-based rejection and pulsewidth-based acceptance.

Time-based rejection will ignore any pulses that come in too quickly. Our engines have a redline of 8000RPM. Being a four stroke four cylinder, there are two ignition events per revolution of the engine, or 16000 sparks per minute. Divide by seconds in a minute and we have 267 sparks per second at redline. That's the most this engine will ever see, so we can assume any tach pulses coming in faster than 267/second are fraudulent. 1000ms (1 second) split into 267 events gives us 3.7ms between events at redline. Therefore, after a valid tach pulse comes in we can assume that nothing useful will be picked up for a minimum of 3.7ms, and anything that is can be rejected.

Pulsewidth-based acceptance only accepts a tach pulse if the pulse has been held low (below that red line in the graphs) for a certain amount of time. This is measured by the Megasquirt in microseconds. The manual for the Goldwing says that dwell time ranges from 0.2 to 0.4 milliseconds, or 200-400 microseconds. Therefore, any pulse shorter than 200us should be regarded as noise and ignored.

Now here's something funny. While typing this up, I needed to grab some definitions and field names from Tunerstudio so I started up the laptop and had a look. I noticed I had left the "Ignition Input Capture" on Rising Edge instead of Falling Edge. Looking at my graph above it's obvious to see that Falling would work much better than Rising for this bike. Filtering is still probably needed due to the resistor. The possibility of coil voltage dropping could still cause an event during the rising edge.
User avatar
robin1731
Membership Admin
Membership Admin
Posts: 21724
Joined: Mon May 29, 2006 8:31 am
Location: Decatur, Indiana

Re: Another GL1100 Megasquirt EFI project

#50

Post by robin1731 »

I think you're on the right track. Especially with this, "Looking at my graph above it's obvious to see that Falling would work much better than Rising for this bike".
1976 Goldwing Super Sport
1985 Honda Elite
1976 KZ900 Dragbike
1992 ZX7 Dragbike (KZ900 style motor w/NOS)
and a rotation of various purchases
Randakk approved Carb Rebuilder
thesteve
Zinc Member
Zinc Member
Posts: 40
Joined: Fri Sep 04, 2015 10:25 am
Location: Phoenix AZ

Re: Another GL1100 Megasquirt EFI project

#51

Post by thesteve »

Went for a quick ride after my previous post. The tach spikes are gone with the filtering tightened. I have a different issue now but with an easy solution. At 7300 RPM, the engine cuts out. The logs show the indicated RPM dropping from 7300 to 3650, and holding there until I decrease speed. So one of my filters is actually too strict, and it's ignoring every other tach pulse once I hit 7300 RPM.

Best guess here is the time based rejection works in a different way than I expected. If it's basing its rejection period on the end of the trigger instead of the beginning, the 3.7ms time I mentioned above is too low. With a 0.2-0.4 dwell time, the Megasquirt may not start that 3.7ms timer until the end of the dwell period. Decreasing the timer to 3.5ms (3.7ms minus the 200usec dwell) or lower should fix that. Assuming a 400usec dwell, the safest time would actually be 3.3ms.

With how well the "pulsewidth based acceptance" filter works, I may be able to get good results by just disabling the rejection period filter. Setting it to falling edge didn't make much difference in comparison. The rising edge on the coil charging is much more ambiguous in its timing, but it doesn't really matter when the fuel is injected. Spark timing has to be precise. Not that it matters here, because you can't use the coil trigger for spark control. Falling edge for this engine on a fuel only install is just best practice; it shouldn't affect rideability.
User avatar
irishcarbomb
Silver Member
Silver Member
Posts: 605
Joined: Fri Jan 04, 2013 6:58 pm
Location: Humble, TX

Re: Another GL1100 Megasquirt EFI project

#52

Post by irishcarbomb »

Lots of good info in here, nice work.

So all your VE table values are 200+? I have mine set to 2 squirts per cycle alternating with 332cc injectors and this is what my tables look like. Maybe the 4 squirts is whats causing the high values?
Attachments
ve.jpg
ve.jpg (59.12 KiB) Viewed 431 times
1983 GL1100 Fuel Injected Turbo Build Thread
thesteve
Zinc Member
Zinc Member
Posts: 40
Joined: Fri Sep 04, 2015 10:25 am
Location: Phoenix AZ

Re: Another GL1100 Megasquirt EFI project

#53

Post by thesteve »

With the reqfuel value the built in calculator came up with (7ms total or so), I was hitting 200s around the same areas yours are showing 110. It changes the number based on injector count and squirts/cycle so it shouldn't skew it that badly. What's your reqfuel value? To get my VE numbers in the same range you have my reqfuel is 14ms, or 7ms squirted twice per cylinder per revolution.
User avatar
irishcarbomb
Silver Member
Silver Member
Posts: 605
Joined: Fri Jan 04, 2013 6:58 pm
Location: Humble, TX

Re: Another GL1100 Megasquirt EFI project

#54

Post by irishcarbomb »

My reqfuel is 5.5
1983 GL1100 Fuel Injected Turbo Build Thread
thesteve
Zinc Member
Zinc Member
Posts: 40
Joined: Fri Sep 04, 2015 10:25 am
Location: Phoenix AZ

Re: Another GL1100 Megasquirt EFI project

#55

Post by thesteve »

Interesting. I'll try lowering my squirts per cycle and see if my numbers lower a bit.
User avatar
irishcarbomb
Silver Member
Silver Member
Posts: 605
Joined: Fri Jan 04, 2013 6:58 pm
Location: Humble, TX

Re: Another GL1100 Megasquirt EFI project

#56

Post by irishcarbomb »

Just re-read your post, where I'm hitting 110 is over 10psi of boost...the left table is idle to 100kpa (no boost) and the right is 100-200kpa (anything in boost). I've tuned to 8psi which is 150kpa on the ve table. I did some reading about reqfuel and squirts per cycle and it sounds like more squirts per cycle changes the duty cycle of the injector. So to achieve 4 squirts per engine cycle you may be needing that much extra on the ve table, same results just higher numbers and the injectors are working harder by "firing" more (my interpretation).
1983 GL1100 Fuel Injected Turbo Build Thread
socrace
Titanium Member
Titanium Member
Posts: 334
Joined: Sun Aug 16, 2009 11:46 pm
Location: Springfield, Ill.

Re: Another GL1100 Megasquirt EFI project

#57

Post by socrace »

Sounds like you're making good progress! Will contine to follow along..
1981 Goldwing Standard - efi'd, other stuff
User avatar
Toehead
True Blue Steel Biker
True Blue Steel Biker
Posts: 2139
Joined: Fri Jul 15, 2011 3:10 pm
Location: Mendon, MA

Re: Another GL1100 Megasquirt EFI project

#58

Post by Toehead »

Have you sorted the tach input?

If not, you can make a very simple circuit with a blocking diode and a pull-up resistor which will give a clean 12V square wave when triggering off of the coils. Let me know if you are interested in the diagram. I'm using it on my bike and knock on wood, no issues with noise.
2012 Golf TDI, fast and frugal oil burner
1980 CX500C, All around road warrior (Sold)
1975 TS125 (Sold), old smoker
1977 naked GL1000, Highway cruiser
'81 GL1100 14,000 (reconditioned and sold)
'81 GL1100 50,000 miles (under construction)
2X 1981 CM400 beginner bike (sold)
1980 CB750C (sold)
1994 Virago 750 (Wifey's bike)
'76 GL1000 Sulphur Yellow (Megasquirted efi build) http://www.ngwclub.com/forum/viewtopic.php?f=47&t=59774
User avatar
ericheath
Honored Life Member
Honored Life Member
Posts: 9580
Joined: Wed Jul 29, 2009 11:20 am
Location: Winnipeg, Manituba

Re: Another GL1100 Megasquirt EFI project

#59

Post by ericheath »

You guys keep writing this foreign language. I think I'm starting to get it. I'll Toeheads fix is what I needed for my aftermarket gauges whose tach jumped around so badly.
Whatever I suggest here should be given ample time for a moderator to delicately correct. I apologize in advance.
77 WING, 1200 engine with 77 heads, cams, gl1100 foot pegs, Magna V65 front end, 764A carbs, [-gone Suzuki M109 monoshock--, replaced with gl1100 shocks] gl 1200 swing arm, gl1500 final drive, wheel and rear brakes Valkyrie seat, Meanstreak tank, Sportster pipes, Power Arc ignition off crank.
77 Wing. black
83 Wing, in pieces
"Continuing education is important even if the subject matter is fairly useless (as in this case)."---Greg Foresi
thesteve
Zinc Member
Zinc Member
Posts: 40
Joined: Fri Sep 04, 2015 10:25 am
Location: Phoenix AZ

Re: Another GL1100 Megasquirt EFI project

#60

Post by thesteve »

So the tach spikes are completely gone. I changed the trigger to falling edge and set reasonable values for the filtering, and it's 100% fixed via software. No extra circuitry needed. The Megasquirts have had a few different revisions and they've got most of the necessary hardware filtering in place already on the main board.

My latest issue is my fuel feed line. The hose going from the tank to the pump has a sharp bend in it, and under heavy load the pump pulls that bend into a tight kink and the engine dies until you shut the pump off. I need to find a rigid metal elbow to reroute that line properly.

I haven't really been playing with this much, so there haven't really been any major updates. Apart from that kinked line it runs great; I've made several full power pulls to redline and my fueling is pretty close. I'd like to get the startup and the warmup enrichment a bit closer so that it "just works", but right now it's quite rideable already.
Post Reply
  • Similar Topics
    Replies
    Views
    Last post

Return to “Unique Carburetion”