home register FAQ memberlist calendar

Go Back   Turbobricks Forums > Mechanical > performance & modifications > aftermarket engine management

Reply
 
Thread Tools Display Modes
Old 11-04-2015, 05:09 PM   #1
142 guy
Board Member
 
Join Date: May 2014
Location: Saskatchewan, Canada
Default Bosch 280-150-036 and Beck Arnley 158-0438 characterization

This may be of interest if you are running the Bosch 280-150-036 injectors and of particular interest if you are using them on a B20E or F.

I have a mix of injectors on my 1971 B20 E which has been converted to a Megasquirt controller. Two of the original Bosch 280 150-036 injectors and two of the Beck Arnley 158-0438 which B A sells as a direct replacement for the Bosch injector. I decided that it would probably be good to have a set of four matched injectors so I ordered another two of the B A injectors from Rock Auto (at about 1/3 the cost of the replacement Bosch parts).

I also decided that I should probably attempt to verify the flow rates on the B A injectors and attempt to estimate the injector dead time and the sensitivity of the dead time to the supply voltage. To this end, I set up a test rig connected to the car. I fabricated (more like nailed two pieces of wood together) a jig to hold one of the new injectors in place over a graduated cylinder for measurement. I attached my fuel pressure gauge to a T directly above the injector. All of this sat on the ground in front of the car and was connected to the old cold start injector port on the fuel rail with 5 ft of 5/16 fuel line. For the short pulse width tests, I used a 25 cc graduated cylinder for the flow measurements (improved the low volume accuracy) and a 50 cc graduated cylinder for the longer pulse widths.



I connected the injector electrical terminals back to one of the injector plugs and unplugged all of the other injectors on the car. I used the Megasquirt fuel injection controller in injector test mode to deliver a series of 400 test squirts on the injector using the cars fuel supply system running off of the car battery. However, I did not run the test injector from the car battery. I ran the injector off of a separate regulated DC power supply. This allowed me to hold the injector supply voltage steady during the tests and also allowed me to do the tests with different injector voltages so that I could characterize the voltage sensitivity of the injectors. I connected a digital voltmeter directly to the supply resistors for the injectors so that I could confirm the supply voltage at the point where Megasquirt would measure the supply voltage.





I did the most extensive test sets with the injectors operating at 12 v and 14 v taking samples over a range of pulse widths varying from 0.75 ms up to 14 ms with more data points collected in the 2 – 7 ms range because that is where the B20E (at least mine) operates. I did a few tests with injector voltages of 10 volts and 16 volts just so I could confirm the voltage sensitivity of the injectors. All of my tests were done using the cars fuel pump and supply / return piping with a fuel pressure of 38 psi as the 280 ZX FPR I am using has no provision for adjustment. The injectors were fired through the 6 ohm , 1%, 50 W current limiting resistors I use with the Megasquirt (same resistance as used in the Bosch Djet controller) plus the associated resistance of the cars wiring harness. Since the injectors were characterized ‘on the car’ as opposed to a standardized test rig, the results may not be exactly the same as a standardized test set up. The fact that I am using resistors that may be on the smallish size (ohm wise) may result in fairly high peak injector currents which might explain the rather short opening times that I got (more on that later). I did the tests with an ambient air temperature of around 6 C. I make a point of noting the resistance, resistor wattage rating and ambient temperature because the test results indicate that the injector opening times are quite sensitive to operating voltage. If you are using different resistors or the resistors are heating up during the test (changing the resistance value), it will alter the opening times and the results, particularly at the smaller injector pulse widths that I was interested in. If all you were doing was a gross estimate of the average injector flow rate over a period of perhaps 5 or 10 seconds, then this would not be an issue.

For the Beck injectors, I did 3 or 4 tests at each pulse width. The measurement results showed excellent reproducibility between the test results with most of the variance probably resulting from the ability of my eyes to determine the fuel level of each sample in the graduated cylinder. The Bosch Motorsport injector data sheet claims a flow rate of 48.19 lb/hr or about 492 cc/ min at a test pressure of 43.5 psi. Correcting for my 38 psi test pressure I expected that the flow rates should be around 460 cc/min. I was expecting something similar for the Beck injectors; however, a preliminary check of the gross flow based on some of the longer duration pulse widths showed that the injector flow rates were more like 480 – 500 cc/min without consideration of the injector opening time.

One of the things I noted during the test was that the pressure gauge was fluctuating slightly between about 36 and 38 psi. This was not obvious on the short duration tests; but, materialized with the longer pulse widths. This had not been particularly noticeable when the pressure gauge was connected directly to the fuel rail and the engine was operating at idle (2 squirts alternating). However, at idle the pulse widths are short so it may have been present; but, not visible. It might also be the result of the 5 feet of fuel line that I used to connect the test rig to the cold injector port on the fuel rail. So, a word of caution, my test results may be for a fuel pressure of 36 psi instead of 38 psi or something in that range. While we are on the question of fuel supply, if you are using the Megasquirt injector test mode for doing the tests, do not rely on the Megasquirt to energize the pump during the test. I found that with my regulator and pump arrangement, there is enough relaxation in the fuel pressure when the pump turns off that I would not get consistent test results on short duration injector operation unless I manually started the pump before doing the test. If you were doing a long duration flow test this would probably not be an issue.

When I completed the test on the new Beck injector, I then removed the two remaining Bosch injectors on the car and replaced them with the new Beck injectors. Since I had a little time, I decided to try doing a quick test to characterize the flow on the Bosch injector. Because the remaining time was limited, I only did a 14 volt curve and I only did a single sample for each injector pulse width. Because the Bosch 14 V test was done after I completed the Beck 12 V test, the DC power supply had been adjusted and may not be at exactly the same voltage as the Beck 14 V test; however, it would be within +/- 0.1 volt of the Beck test voltage.

First off, I have attached the raw test measurements for both injectors for anyone who is interested. The columns on the far right are the gross fuel measurements for each injector test (the sum of 400 squirts). In the case of the Beck injectors they are the average of the 3 or 4 test measurements. For the Bosch injector they are the single test result. The columns on the left are the calculated values of flow per injector pulse (essentially the gross flow divided by 400). Note that the values are 10-3 ml. So, as an example the Beck injector delivered 0.080 ml for a 10 ms pulse which averages to (60s/min x 0.080 ml/0.010s) 480 ml/min (without consideration of injector opening time).





I plotted the gross fuel delivery measurements for the Bosch and Beck injectors for the 14 volt test results for pulse widths up to 6 ms. I have attached the curve. The test results show that between 2 and 6 ms, the Bosch and Beck flow data lie on pretty much the same curve. The flow on both injectors becomes non-linear at pulse widths below 2 ms with the Beck dropping off marginally slower than the Bosch. This latter effect could be the result of the fact that the Beck was brand new and the Bosch was 30 + years old rather than a characteristic of the injector itself. The injectors display good linearity over the 2 – 6 ms range and could probably operate acceptably down to pulse widths of 1.5 ms. I estimated the average flow rates of both injectors based upon the slope of the flow curve between 2 and 6 ms. Rather than do something like a least squares fit, I just used the actual data points for 2 and 6 ms since the ‘curve’ data points don’t have a lot of scatter. For the Bosch injector the flow rate is 519 ml/min and for the Beck it is 514 ml/min. Significantly higher than the expected flow rates of 460 cc/min (Bosch published data corrected to 38 psi operation). Back in 2011, a member on the Turbobricks forum reported that his cleaned and tested Bosch 280-150-036 injectors came back at 570 ml/min @ 43.5 psi. Adjusting the 519 ml/min 38 psi flow rate for a pressure of 43.5 psi would give a flow rate of 555 ml/min. If in fact my test pressures were 36 psi (referring back to my notes on the pressure gauge fluctuations), then the adjusted flow rate for my test would be 570.5 ml/min, jiving exactly with the previous results posted by the forum member. Clearly the injectors flow much more than Bosch’s published data.

What is really interesting is that if you visually extend the flow curves in a straight line to find where it intercepts the X axis, you will get crossings somewhere below 0.2 ms, which would be the nominal opening time for the injectors using a linear flow curve. If you use the slopes I calculated and the data point at the 2 ms pulse width, you can back calculate where that line intersects the X axis which would also provide an estimate of the injector opening time for the injectors linear operating range. The Beck intersects the line at a calculated value of 0.0152 ms and the Bosch at 0.0150 ms. 0.15 ms would be a good estimate of the opening time for the injectors at an operating voltage of 14 volts.



Megasquirt is programmed on the basis that the injector opening time is defined at 13.2 volts so you need to adjust the opening time for an operating voltage on 13.2 volts. In order to do that, you need the sensitivity of the injector opening time versus operating voltage. That is what I used the Beck Arnley test data at 10 V, 12 V and 16 V for. My calculations were based on the test data at 2 ms and 10 ms because those are the only points I had for 10 V and 16 V operation. I calculated slopes and Y intercepts for the 10 V and 16 V data using the 2 data points that I had and did a least squares fit for the data range of 2 ms to 10 ms for the 12 V and 14 V data set to get the equivalent slopes and Y intercepts. Using these curves I calculated where the curves would intersect the X axis (zero flow)

10 volts 0.606 ms
12 volts -0.22 ms
14 volts -0.01 ms
16 volts -0.076 ms

Note that the 14 V X intercept calculated here does not match the values I calculated previously. That is because the slope for a least squares fit on the 2 – 6 ms data set is not the same as the slope for the least squares fit between 2 – 10 ms. The flow curves are close to linear but not quite linear and small changes in the slope can make a big difference in where the X intercept lies. Since you are attempting to linearize what appears to be a non-linear curve, it is important to linearize it over the expected operating range of the engine. If these injectors were operating at much larger pulse widths than what I am using on my B20E, I expect that the derived injector dead times would be different.

Using the preceding data, the voltage sensitivity of the injector opening time is as follows:

12 10 volts 0.19 ms/volt
14-12 volts 0.12 ms/volt
16-14 volts 0.03 ms/volt

The sensitivities are positive values since they add to the opening time as the injector voltage drops. As you can see and as many other sources report, the opening time versus voltage relationship is highly non-linear. I redid the calculation and calculated the injector opening times that would correspond to a flow of 20 ml/ms for the four different operating voltages.

These values for voltage sensitivity came out at:

12 10 volts 0.19 ms/volt
14-12 volts 0.12 ms/volt
16-14 volts 0.02 ms/volt

If you change the data sets to calculate the least squares fits using the data set from 6 – 2 ms for the 12 V and 14 V tests, you will again get slightly different results. At a flow rate of 20 ml/ms, the calculated 12 – 10 V and 14 – 12 V sensitivity range is very close to the preceding results (0.20 ms/volt and 0.11 ms/volt); however, the 16 – 14 volt value is significantly different. Since Megasquirt only has provision for a single value for voltage sensitivity, I am going to use the sensitivities that correspond to the 14 -12 volt operating range which is where my car is usually running. All the calculated values suggest that 0.12 ms/volt is a reasonable value to use for this operating range. As a note, this applies only to the Beck Arnley injector since I did not do any voltage sensitivity measurements for the Bosch injector.

With a value for the voltage sensitivity, you can calculate the expected opening time for the injectors at 13.2 volts. Given that the inferred opening time at 14 volts is 0.15 ms, the inferred opening time at 13.2 volts would be:

0.15 ms + (14V – 13.2V)x 0.12 ms/volt = 0.25 ms (rounding up)

I concentrated my work primarily on the injector operating range below pulse widths of 8 ms. I did take some measurements up to 14 ms pulse widths. The Bosch seems to maintain a relatively linear flow relationship over the whole 2 – 14 ms range. However, the Beck Arnley measurements at 12 V suggest that the flow curve is turning upwards around 10 ms. I only have two values at these larger pulse widths and nothing for 14 V operation to allow a direct comparison to the Bosch. The caution here is that if the Beck injector is showing significant non-linear behaviour above 10 ms pulse widths, the flow rate, dead time and voltage sensitivity values that I have calculated would all be out to lunch. If you are thinking about using the Beck injector at pulse widths above the 8 – 10 ms range, you would certainly do well to characterize it in this higher flow range to confirm or refute the results that I got.

Use the information or ignore it as you see fit.

Last edited by 142 guy; 11-04-2015 at 05:14 PM..
142 guy is offline   Reply With Quote
Old 11-05-2015, 05:46 PM   #2
B-lennium
Board Member
 
Join Date: Aug 2005
Location: Holland
Default

Thanks for your time man!

This is the first data I have ever seen on the old D-jet injectors! I am surprised to learn the bosch injector is even bigger than I already knew. A opentime of 0.01ms is way faster than what I expected. Did you noticed any differences in tune?

I am currently using a 0,800ms opentime and a fuel pressure of 28psi. its giving me a 1,800ms pulswide @idle and a decent tune. The injectors are just enormous I don't think I ever seen a dutycycle over 45%

My setup:

B30 10,5:1 compression
MS2 extra semi-seq injection (+/- 425deg)
Jbperf injector drivers 3x2 injectors
wastedspark

Last time I checked the millage was 1 liter for 8,9km, stock djet injection would be 1:6 My tuning made a huge improvement!
__________________


Volvo 164e B30e, Megasquirt MSII extra
B-lennium is offline   Reply With Quote
Old 11-05-2015, 07:03 PM   #3
gsellstr
Vintage anti-ricer
 
gsellstr's Avatar
 
Join Date: Oct 2004
Location: Somewhere in a northern California smog bank
Default

I will have to read thru this when I can view the images. I'm also running OE 036's on mine...for now anyway. Sounds like there's a LOT of work into this bit of research...abnormal for the board, but much appreciated!
__________________
RIP
Doug Williams "Mr. Doug" 4/15/2009
Pete Fluitman "fivehundred" 7/14/2013
Mick Starkey "TrickMick" 1/10/14
Mark Baldwin "blue850t5" 7/19/18
Nick Fengler "fengler" 8/6/18
Thomas Fritz "stealthfti" 10/11/18


74 144 B20
http://forums.turbobricks.com/showthread.php?t=224983

90 745Ti
http://forums.turbobricks.com/showthread.php?t=334698

If you need Superpro bushings PM me for price and availability!
gsellstr is offline   Reply With Quote
Old 11-05-2015, 07:44 PM   #4
142 guy
Board Member
 
Join Date: May 2014
Location: Saskatchewan, Canada
Default

Quote:
Originally Posted by B-lennium View Post
Thanks for your time man!

This is the first data I have ever seen on the old D-jet injectors! I am surprised to learn the bosch injector is even bigger than I already knew. A opentime of 0.01ms is way faster than what I expected. Did you noticed any differences in tune?

I am currently using a 0,800ms opentime and a fuel pressure of 28psi. !
I needed to confirm the injectors to do get my tune correct. There are a number of other B20 , B30 MS conversions out there so I figured others would be interested.

I think you miss-read. The opening time I calculated for the injectors was 0.25 ms, not 0.01 ms. Remember that the opening time is where the slope of the injector curve intersects the X (pulse width) axis. Its a theoretical value that Megasquirt uses to calculate the flows in the linear part of the flow curve. Also, what size injector resistors are you using? My 6 ohm resistors may in part be responsible for the short opening time.

As a matter of interest, what are you using for a fuel pressure regulator to run at 28 psi? I wanted to run at 28 psi to make sure my minimum pulse widths did not enter the non linear region below 2 ms. I was using an adjustable Aeromotive FPR; but, was running into problems with the manifold pressure compensation (it would not drop the rail pressure below about 22 psi) and the FPR chattered a lot. Turns out the Aeromotive requires a minimum return line size of 3/8". That's why I am now using a Nissan 280 ZX FPR which is set up for the same 5/16" line as the Volvo has. However, I am not happy about the higher rail pressure.

Yes, my tune went screwy when I entered the new data. I had to return the flow rates to the previous values since I had tuned the VE table on the presumption of a much lower fuel delivery rate and got the engine to run OK on that basis. That VE table now results in grossly lean operation (or a huge amount of EGO correction) if I use the new injector flow rates in the req fuel calculation. I will have to gradually change my injector flow rate in the req fuel calculation and do some intermediate tunes until I get the flow rate in MS up to around the 515 cc/min value.
142 guy is offline   Reply With Quote
Old 11-05-2015, 10:59 PM   #5
bobxyz
Board Member
 
bobxyz's Avatar
 
Join Date: Aug 2014
Location: Boulder CO
Default

Wow! Great write-up!!! What a major effort just to come up with 2 small numbers....

What's your expected operating range for PW, not including the dead time? (How close to the non-linear region are you expecting to run?)

What MS drivers are you using, and what configuration? MS3.0, 2 injectors/bank, all components installed, hi-z mode with pwm off?

I'm surprised by your voltage sensitivity results above 12v. It's a much sharper change than I'd have guessed. I wonder if the MS driver circuits are influencing this? What voltage was the MS running at, car battery of ~13v or lab supply?

I know I've seen it elsewhere, but for convenience when I lookup this thread again, what's the flow rate correction formula for running at a different psi?

Thanks,
Bob
__________________
'85 245glt b21ft aw71 k-jet -> lh2.4
bobxyz is offline   Reply With Quote
Old 11-06-2015, 01:05 AM   #6
142 guy
Board Member
 
Join Date: May 2014
Location: Saskatchewan, Canada
Default

Before I did this test, I was using a flow rate of 460 cc/min and I think my req fuel was at 7.0 or 7.2 ms (I would have to dig up an archived tune to confirm that). My idle pulse widths seemed to be around 2.1 ms and on a log sometimes you would see them dip as low as 1.7 ms during transients. So historically my PW were theoretically in the 2 - 7 range. When I plug in the new numbers for flow rate on the injectors, after everything is fixed up I think the idle PW will stay about the same since I have only changed the value loaded into MS, the injectors remain the same and are delivering the same amount of fuel as before; however, because of the change in the flow rate number in MS it will be delivering less fuel (shorter pulse widths) which is going to cause it to run lean until I can get the VE table fixed up. With a 515 ml/min flow rate I think the req fuel has dropped to 6.8 ms or something like that so my PW should ultimately be in the 2 - 6.8 ms range.

My VE table had always struck me as being a little screwy. In order to hit reasonable AFR values I had very low values for VE in a lot of the table entries because the injectors were delivering a lot more fuel than what Megasquirt was calculating. I think my VE table will end up looking much more 'typical' after I get it tuned with a more realistic injector flow rate.

I am running MS 2, batch fire (for now) with 2 inj per bank and 6 ohm resistors for the injectors.

The Megasquirt controller was run off of the battery with a battery charger maintaining the voltage. The injectors were supplied via the DC power supply. Megasquirt has voltage regulators on its internal power supplies and is relatively unaffected by supply voltage variation up to the point that you exceed the operating range on the regulators which is pretty large. Megasquirt operates the injector transistors in bipolar mode. They are either off or driven to saturation (Vce probably being around 0.2 volts - my guess) and that will be pretty consistent across everybody's MS controllers. If anything is contributing to what seems to be a short opening time it might be my smallish injector resistors which result in a fairly high voltage being applied to the injectors. I have heard of people using 8 ohm resistors which would reduce the injector voltage and perhaps give them an opening time closer to what I would calculate for the 10 volt curve which was around 0.6 ms. As I noted, my results for opening time and voltage sensitivity will be applicable for my set up, although the flow rates should apply to other applications.

The extreme non linearity of the voltage sensitivity is not a total surprise. Check out the Injector Dynamics web site http://injectordynamics.com/ and look at their published performance data for their injectors. See how the offset (their term for opening time) radically changes as the injector voltage changes. My numbers for the change between 10 and 12 volts and 14 and 16 volts may be a little flakey because I have so few data points for the 10 and 16 volt curves. That's the hazard of doing the test, putting everything away or back together and then going and doing the data analysis. You find out that you should probably go back and collect some more data points!

The theoretical correction formula is:

New flow rate = (sqrt (new pressure/ old pressure))* Old flow rate

The theoretical correction is close; but, not perfect. If you go back to the Injector Dynamics web site they provide injector flows for different fuel pressures and you will see that there is about a 1 -2 % difference between their measured data and the theoretical correction. However, that is pretty small considered all the other unknowns and assumptions, so the formula works 'good enough'.

Last edited by 142 guy; 11-06-2015 at 01:38 AM..
142 guy is offline   Reply With Quote
Old 11-07-2015, 07:08 PM   #7
142 guy
Board Member
 
Join Date: May 2014
Location: Saskatchewan, Canada
Default

A little bit of tuning and I have had to really increase the values in the Ve table to deal with the higher flow rate data that I have entered for the injectors. I will post a before and after Ve table when I get some more cells populated with numbers based upon actual operation.

An interesting effect of the 0.25 ms offset is that when lifting off the throttle at the end of an acceleration run, I was getting into super jerky operation and seeing wild swings on the AFR. This problem did not show up when I was using an injector offset of 1 ms. I had my deceleration fuel cut set at 25 % and a log showed that the PW was dropping to 0.70 ms or less on lifting the throttle. With the .25 ms injector offset, the sum of the offset and the calculated PW would be around 1 ms and the injector is definitely running into the non linear area of operation (fuel was probably getting close to completely shutting off if the applied PW were dropping into the 0.7 ms range). I cranked the fuel cut to 50% and then 60% and that seems to have fixed the problem.

The ideal situation would be to reduce my fuel pressure down to something like 30 psi which would increase my pulse widths and hopefully keep the injectors out of that nasty non linear region. However, until I can get an FPR that works with a lower base pressure I am stuck with the Nissan FPR at 38 psi.

Last edited by 142 guy; 11-07-2015 at 07:13 PM..
142 guy is offline   Reply With Quote
Old 11-29-2015, 10:56 PM   #8
bobxyz
Board Member
 
bobxyz's Avatar
 
Join Date: Aug 2014
Location: Boulder CO
Default

I looked up the injector driver circuits used by the MS2 v3.0 PCB (same as v3.57 PCB) and there are 2 different effective injector voltage drive levels used in normal operation, and a third drive level when using low-z injectors in PWM mode.

The MS deadtime value is a measure of the difference between injector opening time and injector closing time. Injector opening times are a function of the injector friction, spring force and the applied opening voltage -- higher voltage causes faster current ramps which causes higher opening forces sooner. The closing time is mostly a function of the spring force and the clipping voltage. Higher clipping voltage results in faster closing times. Closing times are much faster than opening times.

The 3 MS injector drive levels are:
1) Injector Opening Voltage
The MS circuits ground the INJ-x pins, through a MOSFET, to open the injectors. With the other injector pin connected to Vbattery, this applies ~13volts across the injectors (and any added resistors). If the battery voltage goes up, the opening voltage across the injector increases and opening time decreases.

2) Injector Closing voltage
The MS v3.0 pcb circuits include an active clamp circuit that clamps the injector flyback voltage on the INJ-x pins to ~37volts above ground (36v zener plus a couple diode drops). With a ~13v battery voltage, this effectively applies ~24volts (37v clamp-13v batt) across the injectors in the opposite direction. Other MS PCBs may use different clipping levels, resulting in different closing times. If the battery voltage goes up, the closing voltage across the injector decreases slightly (and the clipping voltage stays the same). For instance, going from 13v to 14v battery leaves 37-14= ~23volts closing voltage.

3) PWM mode Hold voltage
In low-z PWM mode, the MS v3.0 pcb circuits include a flyback damping circuit to recirculate the existing current through the injector and "hold" the position. This circuit clamps the INJ-x pins to ~1volt over the MS +12v supply voltage (~1volt comes from a diode drop and Vce,sat of the TIP125 darlington). This mode only operates during PWM mode, when INJ-x pins are not grounded.

There are lots of video tutorials on injector operation. The one below looked pretty good (but I only flipped through it and didn't watch it all). Note: this video describes a linear peak&hold design for low-z injectors, not the PWM design use by MS v3.0 PCB
https://www.youtube.com/watch?v=4mut0Lpg5k4

I'm still curious why the deadtime battery voltage correction changes so much over +12v. Maybe the injector coils are saturating and can't charge any more above 12v?
bobxyz is offline   Reply With Quote
Old 12-04-2015, 12:40 PM   #9
142 guy
Board Member
 
Join Date: May 2014
Location: Saskatchewan, Canada
Default

In the context of the test results, I generally agree with most of your comments.

As an observation, the term injector opening time as applied in Megasquirt is a misnomer. It is really injector offset time which is where the linearized portion of the injector flow line intersects the time axis. Megasquirt assumes a perfectly linear relationship between injector flow and pulse width. If you look at the flow curves that I plotted, the injector 'opening time' or the pulse width time below which no flow occurs is something like 1 msec. Actual opening times are only useful as being an indicator of the absolute minimum PW at which the injector could theoretically operate. In general, based on the data I measured, you should not attempt to operate the Bosch injector below PW of 1.5 msec as the flow curve takes a significant departure from the assumed linear flow relationship below 1.5 msec.


Injector Dynamics probably provides a better (and wordier) discussion of offset.

http://injectordynamics.com/articles...-and-dip****s/

Of particular interest is that some of the injectors that they discuss have offsets very close to opening times and some of the examples have radical differences between the offset and the opening times.

Your comments about the treatment of opening time and closing time may be correct; but, they are perhaps theoretical concepts. The measured flow data incorporates both injector opening time and closing times explicitly in the measured flow results, so it is impossible to separate the two effects based upon the measured flow data. In using a linear relationship for injector flows, Megasquirt ignores the actual injector opening and closing times. The slope of the linear portion of the flow versus pulse width line and the injector offset value establish the linear flow relationship for the injectors.

As an observation, some OEM ECUs now appear to have a two part injector flow curve. For very low pulse widths, they have an offset closer to the injectors actual opening time and a very high initial flow rate and then for longer pulse widths they have a more normal offset and flow rate. This would be equivalent on the Bosch injector of having one straight line flow curve between 1 and 1.5 msec and then a second between 1.5 msec and maximum pulse width. I am not sure why they would be needing to do this unless they are installing super sized injectors.

The Bosch injectors are low impedance injectors. In my test set up they are being energized through 6 ohm resistors so PWM operation does not enter into consideration. I have no empirical data to suggest one way or the other that the use of PWM might alter the injectors performance.

I am not sure that magnetic saturation is the dominant effect that is taking place when looking at the effect of operating voltage on offset time. Consider a hypothetical injector. The pintle has to move a fixed distance from closed to open. The relationship between distance, time and velocity is Velocity = distance/ time.

Go back to the Injector Dynamics website and look at their offset times for their 1000 series injectors.

http://injectordynamics.com/injectors/id1000/

I want to use ID's injector offset data rather than mine because they have more data points and presumably did more (and probably better) data collection to characterize their injector performance at off nominal voltages.

Since the distance the injector pintle travels is fixed, 1/(offset time) becomes a 'measure' of the average pintle velocity while it is opening. If you plot 1/offset time versus voltage, you get a 'measure' of the average opening velocity of the pintle versus the operating voltage over that fixed distance. You will find what looks like an almost linear relationship between average pintle speed and operating voltage. Because speed and time have a reciprocal relationship for a fixed difference, the injector offset times will at some point approach the asymptote. The pintle speed may still be increasing with higher voltage; but, the reduction in offset time is becomes less significant.

I originally made the comment that the opening times versus voltage were highly non linear, probably inferring a higher order relationship. In fact, the relationship may be first order reciprocal with operating voltage.

As an observation, the injectors may always be operating at the point on their B-H curve where they are always in magnetic saturation at any of the reasonable operating voltages.

Last edited by 142 guy; 12-04-2015 at 02:57 PM..
142 guy is offline   Reply With Quote
Old 02-23-2016, 03:58 PM   #10
gross polluter
They see me trollin'
 
gross polluter's Avatar
 
Join Date: Dec 2008
Location: Cold Lake, AB
Default

Bringing this one back up. Recently I've revisited this subject, pretty set in my ways over the last few years with the method(s) I've been using on the dyno to characterize injectors. All of this discussion recently here and elsewhere, I figured it's time to take another look. What motivated me to revisit this topic is the fact that the numbers produced in this thread seem awfully low for such an old injector, and just recently I've been producing the same low numbers with injectors that other's suggest isn't correct. So what gives?

I bought a set of Deka Shorty 80# injectors from YoshiFab recently to replace the LS3 injectors I'm running on the secondary rail in my Rustang. Having no data on these particular injectors I figured now is good as time as any to set up a flow bench and start collecting data. Reading technical documents from Paul Yaw and Greg Banish on the topic, I began to wrap my head around the fact that the methods used in this thread, executed correctly, would yield results that are arguably more correct than any other method.

Without getting too involved in the project just yet, I collected samples at a measured fuel pressure of 43psi, 600 shots at 100ms intervals, 13.4 volts held solid with a lab power supply. I collected samples in 1ms increments from 1ms to 20ms, six samples per increment averaged into one result per increment. MS3X ecu in injector test mode on injector channel 1 off the expansion board. That said, I obtained the following results:



The data collected above puts the injector offset around 309us at 13.4v and 43psi. According to other sources, potentially not credible, the offset is said to be around 1ms at the same operating conditions. In my own experience on the dyno, this value would seem awfully low for such a high flowing injector as well. And this is what made me want to take a closer look at what's going on here.

I decided to take an injector with known operating characteristics and test it as well. See if the results I come up with differ from the characteristics defined in the OEM ecu. Good ol' GM LS3 injector. If you take a look at one of Greg Banish's documents, he tests an LS3 injector and displays his data in the document: https://eficalibrator.files.wordpres...ingdrilled.pdf. Looking at his characteristic chart on page 2 you can see the X intercept in his chart appears to be right around 500us. If I don't get the same data, I think it's safe to say I'm doing something wrong....

13.4v 53psi (couldn't get the regulator to do 58psi) 2000 shots per sample, 6 samples per increment averaged. 1ms increments from 1ms-20ms.


I get an X intercept of 580us from this data, close enough to Greg's results to say that the data I'm collecting is more than likely accurate. With that said, how does our data compare to how the OEM ecu characterizes the injector? Lets find out. Below are images of the injector data taken directly off the exact ECU these LS3 injectors were on.


At 50psi fuel pressure, our pressure delta across the injector is 344.7kpa. Looking at the chart, we can just rough the offset value to be around 735us according to the stock ECU at the same operating conditions. My results come out 155us lower than what the factory ecu defines. Realistically what would such an error account for in the application? I don't know. What I can say, however, is that it's unlikely the error in my own samples is that great, meaning the values I obtained for the Deka 80's compared to what I and others "feel" they should be is more than likely correct.

To further look into this, it would be interesting to replicate the LS3 ecu injector drivers on the bench and run another test. As Bob pointed out earlier in this thread, the electrical characteristics play a role in how the injector acts. I've never had an LS3 ecu open, however, in many other OEM ECU's I've observed, the flyback protection on board is in the form of a simple RC circuit, they don't use diode clamps like MS boards do. At this point, I'm wondering if the 155us difference between my measurements and what the OEM LS3 ecu defines is from the difference in the injector driver circuitry?
__________________
'88 245 - 2JZGTE VVTi | MS3X ECU, MS2 TCU | GT3582R-HTA | Staged simultaneous dual fuel | 590whp 580ftlbs

1968 Furd Rustang fastberk | 2JZGTE VVTi | MS3X ECU | PT6776S | 4L80e Microsquirt TCU Sloppy Transbrake | 3.73 8.8" rear end

1981 Piper Cherokee Archer II | All stock
gross polluter is offline   Reply With Quote
Old 02-23-2016, 06:33 PM   #11
gross polluter
They see me trollin'
 
gross polluter's Avatar
 
Join Date: Dec 2008
Location: Cold Lake, AB
Default

Ok. Here's another really good comparison. This is a characterization of the stock, primary injectors on the 2JZ in the Rustang. Didn't measure fuel pressure, running on the stock FPR for that rail. 13.4v 1200 shots per measurment, 6 measurements per interval. Tested every 1ms from 1ms to 20ms.



Now, lets take a look at the data I came up with after spending a few hours of time on the dyno with this car:


On the flowbench, at 13.4v I came up with an injector offset of 728us.
Interpolating the battery offset table I came up with on the dyno, at 13.4v I came up with an offset of 718us.

Between my method on the Dyno, and using the flowbench method, I only came up with an error of 10us!

I think it's safe to say the numbers we're coming up with on the flow bench are correct numbers. At this point, I think I'd prefer to flow bench injectors and fully characterize them on the bench before throwing them into the application. At least help, and be much easier to develop the voltage offset curve that way.

Last edited by gross polluter; 02-23-2016 at 06:39 PM..
gross polluter is offline   Reply With Quote
Old 02-25-2016, 01:03 AM   #12
bobxyz
Board Member
 
bobxyz's Avatar
 
Join Date: Aug 2014
Location: Boulder CO
Default

Great! Nice work!!! So rare to to see tight data that matches well to independent sources, especially on a forum. I'll need to study the results in more detail this weekend.

If you're now after the second order effects, you could try heating the injectors to operating temperature (heat gun?) before flow testing. The resistance of copper wire increases by ~0.4% per degreeC. Higher resistance would cause a lower/slower current ramp and a longer opening time. I don't know if it would be significant compared to the other factors and injector variation.

Thanks for the write up!
bobxyz is offline   Reply With Quote
Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is On

Forum Jump


All times are GMT -4. The time now is 10:49 PM.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, vBulletin Solutions Inc.