I have read that the GM solenoid is has optimal pwm at 19.5hz

 

so meaning it can distinguish between on and off ~20 times a second?

 

This might be where he read that:

http://forums.turbobricks.com/showthread.php?t=71071

 

Actually it was honda-tech lol

 

http://fluids.ingersollrand.com/IS/product.aspx-en-4102

Looks like the response time of the ingersol rand (perrin/grouppe-s) valves are all in the order of 10-12ms (.010-.012) (PN# P251SS-012-D ) for all of their DC valves. Thats more like 83-100hz. Also interesting that their rated operating range is +-10%... good thing we don't use a resistor for those!

 

Just looked for the MAC 36 series information. They brag about how high speed their valve is but provide no data on response time.

 

Complete spec sheet of my the ingersoll rand.

Performance Specifications:
Port Size-NPT 1/8” NPT
Media Air or Inert Gas
Operating Pressure 3-Way, 0 to 125 PSI
4-Way, Vac to 125 PSI
Ambient Temperature Range 32 to 125 F (0 to 50 C)
Cv Factor .144
Coil Rated Voltage 120VAC (50/60Hz); 12, 24 VDC
Allowable Voltage Fluctuation + or - 10% of Rated Voltage
Coil Insulation Type Class B Rated, 100% Duty Cycle
Power Consumption DC 4.5 Watts
Electrical Entry 24” Lead Wire (22 AWG)
Plug-In DIN Connector
(Industrial Micro Type C)
Manual Override Yes, Top of Coil, Non-Locking
Materials Seals; Buna-N, Coil: Acetal
Body; Aluminum, Brass and Stainless
Response Time (On/Off) .012/.010 (DC), .012/.020 (AC) Sec.
Max. Cycle Rate 2700 (DC), 1875 (AC)
SCFM @ 100 PSIG >10
Leak Rate (Max. Allowed) 4cc/Min. @ 100 PSIG
Lubrication None Required, Factory Pre-Lubed
Weight 3-Way; .26 lbs (116g)
4-Way; .28 lbs. (128g)

 

Very interesting! Maybe this is why Mitsubishi use twin solenoid setup for JDM IX and after? Hope better resolution in control can result in finer boost control. Thank you for your efforts!

 

I have done a little bit of disassembly.
The EVO ECU uses 8 MHz clock multiplied to 16 MHz peripheral clock.
This clock will be divided by 8 (=2 MHz) for the compare match timer 0.
2 MHz = 0.5 µs clock
The match counter is loaded with 0x9BF (2495 in decimal)
ROM:0000B79C mov.w @(h'110,pc), r10 ; [0000B8B0] = h'9BF ; 2495 dec
ROM:0000B79E mov.l @(h'118,pc), r11 ; [0000B8B8] = CMCOR0

So the irq routine will called every 0.5µs*2496 = 1248µs
In this irq routine the counter for the boost control solenoid count to 48 for a whole cycle. 1248µs*48= 59904µs= 59.9ms = 16.7Hz

This value is not so far away from 19.5Hz

Have anybody check this with a scope?

With this 60ms duty the correction timer value (found by mrfred and OEM value of 10 =100ms ) should not be to low (<7) because this would be shorter than one cycle.

 

yeah a rough log finds that the ECU runs through the 48 about 20 times a second...

I have mixed feelings about the correction interval. if you set it too high then the ECU TRY to fix the under/over boost condition... thus mine is set to 1.

I think smaller changes more often is better than bigger changes less often...

 

Good stuff. I finally had a chance to measure the response time of the Prodrive BCS that I picked up. While its not the stock BCS, it is interesting to look at its performance relative to the stock boost control system.

I tested the Prodrive BCS in series with a variable resistor. The circuit voltage (not the BCS voltage) was set to 14 V. I measured the latency of the BCS at 1, 5, and 10 ohm values for the series resistor. There was no air pressure on the BCS when I tested it. The latency values could be higher when there is air pressure in the BCS line.

The latency will determine the useful range of pulse widths where the BCS can be operated. The limiting factor is the fraction of the total possible pulse width that is consumed by the latency of the BCS. For instance, the stock pulse width is apparently about 60 ms. For the example of a 10 ohm resistor in series with the Prodrive BCS, the latency of the BCS consumes roughly 11% of the available pulse width, resulting in a true maximum duty cycle of 88%. Just to be clear, this duty cycle is not the duty cycle in the ECUFlash BWGDC vs RPM table. It is the true maximum duty cycle of the system. When the BWGDC is set to 100%, the true duty cycle will be 88% in this example because the ECU always shuts off the pulse at the end of the count down that JoeBee mentioned.

The fun part is looking at how this relates to tephra's pulse width routine. He's now running two loops of 50 count. If I am understanding correctly, this results in reducing the pulse width by a factor of two, to 30 ms. Now the BCS latency becomes much larger... 24% of the available pulse width. The maximum true duty cycle drops down to 76%. And if the pulse width is decreased by another factor of two, then the true maximum duty cycle drops to 50%. This would explain why tephra was only able to maintain 19 psi with the four loops of 50 count.

One other interesting tidbit to consider is the effect of resistance on the BCS latency. Its very minimal! This is because instantaneous resistance of the BCS is quite high in comparison to the series resistance when the circuit is closed. So, adding a 10 ohm resistor has very little effect on the operation of this BCS at least for the stock pulse width. There is some value to using a smaller resistor or no resistor with a shorter pulse width. It would depend on how much the pulse width is shortened. Probably true for the GM BCS as well.

 

Not quite right on the 2x50.

Let me explain this way:

We know the following:
1) the stock code loops between 0 and 47, ie 48 loops.
2) the stock code runs through 48 loops roughly 20 times a second
3) from the above we can roughly say that there are 1000 loops per second.

In my modified routine I have 2 counters, 1x "outer loop" and 1x "inner loop", each counter has a max, and the must multiple up to 100.

ie if the outer loop has a max of 4 then the inner loop must have a max of 25, so when logging the two loop variables you would see something like this (if you could log fast enough)

OUTER/INNER
0/0
0/1
0/2
0/3
...
...
0/23
0/24
1/0
1/1
1/2
...
1/23
1/24
2/0
...
...
...
3/23
3/24
0/0

Now because of the latency of the GM3port when I ran this my boost dropped, this must be because of the increase in "open" time because of the latency.

So for the GM3port I ran an outer max of 2 and an inner max of 50 - still with a total of 100, ie:

0/0
0/1
0/2
...
0/48
0/49
1/0
1/1
...
1/48
1/49
0/0 etc etc

So the more outer loops we have the more times per second the BCS will switch on/off, so the more "deadtime" there is.

Why not just have 1 loop of 100? Well the less loops the less frequency = less smooth boost.

 

That's what I was thinking. What is the time duration of your 50 count loop? Still ~50-60 ms or is it approx ~25-30 ms?

 

just slightly more than the original 48 count.. (because its 50 vs 48)

so, ~1000 a second, so 50ms...

 

How is the 2 x 50 setup different than stock if the maximum pulse width is still ~50 ms?