Fuel Pump speed control
Discussion
How useful would an entirely automatic and calibration free(see note) fuel pump control system be?
If you placed a fuel flow sensor in the return line from the pressure regulator, such as:
http://www.flowmeters.co.uk/pdf/turbine_meters/900...
Then the system could be entirely closed loop and self calibrating, it would just aim to always maintain a small amount of return flow (as measured by the flow meter). As the engine used more fuel, the return flow would reduce, and the system would automaticaly add extra pump speed to return to the targeted amount of return flow.
Luckily the PCB i have designed can easily interface with a hall effect flow sensor !
(Note: the only calibration might be some fine tuning of the PID values to get perfect control of flow rate, but i most cases a pesimistic system that always kept a certain minimum flow (i.e. you don't care in the short term if you get a little more return flow) would probably be possible with a std set of control terms)
If you placed a fuel flow sensor in the return line from the pressure regulator, such as:
http://www.flowmeters.co.uk/pdf/turbine_meters/900...
Then the system could be entirely closed loop and self calibrating, it would just aim to always maintain a small amount of return flow (as measured by the flow meter). As the engine used more fuel, the return flow would reduce, and the system would automaticaly add extra pump speed to return to the targeted amount of return flow.
Luckily the PCB i have designed can easily interface with a hall effect flow sensor !
(Note: the only calibration might be some fine tuning of the PID values to get perfect control of flow rate, but i most cases a pesimistic system that always kept a certain minimum flow (i.e. you don't care in the short term if you get a little more return flow) would probably be possible with a std set of control terms)
Sounds like a reasonable plan. It would require the fitment of an FPR again though.
Might also simplify plumbing for people who for example wanted to use a small diameter existing breather line for a return or something.
or in general, it would mean return line wouldnt need to be big at all.
Might also simplify plumbing for people who for example wanted to use a small diameter existing breather line for a return or something.
or in general, it would mean return line wouldnt need to be big at all.
I will have to do some testing to confirm (and calibrate) the response of the system (set the overal gain of the control loop) but i doubt there will be an issue. However, it probably will be a "different" response to the existing system, that probably currently has a fairly large pressure spike on "stab" tip ins (FPR slams shut as MAP jumps up to atmospheric when WOT) and all the fluid flowing through the pre-regulator system stops, and transfers it's dynamic head to the static pressure.
As a result i would expect to check the engines transient fuelling with a different fuel delivery system fitted of any kind. It's just possible that at low rpms, a full stab-in from trailing throttle might require a slight adjustment to the transient fuelling gains in your ems system.
As a result i would expect to check the engines transient fuelling with a different fuel delivery system fitted of any kind. It's just possible that at low rpms, a full stab-in from trailing throttle might require a slight adjustment to the transient fuelling gains in your ems system.
Running a few numbers through Matlab, (with some admittidly guestimated values for things like pump rotor inertia) suggests that a control loop of 500hz (easily achieved, normally run at 1kHz) and with a typical system volume (haven't yet included "springyness of the fuel delivery hose, but this is probably fairly small with aeroquip etc), the longest response and settling time is in the order of 80 to 120ms, with some harmonics of the pressure surge taking 250-400ms to damp out (but these are small).
Ultimately i will build an instrumented pressure test rig and confirm the results.
( i bet that if you stuck a fast response pressure sensor into a typical fuel system you would be shocked by how much dynamic and pressure wave stuff is going on. Luckily these effects tend to average themselves out, and off course the injectors flow rate varriation is SQRT2 better anyway)
Ultimately i will build an instrumented pressure test rig and confirm the results.
( i bet that if you stuck a fast response pressure sensor into a typical fuel system you would be shocked by how much dynamic and pressure wave stuff is going on. Luckily these effects tend to average themselves out, and off course the injectors flow rate varriation is SQRT2 better anyway)
Max_Torque said:
Running a few numbers through Matlab, (with some admittidly guestimated values for things like pump rotor inertia) suggests that a control loop of 500hz (easily achieved, normally run at 1kHz) and with a typical system volume (haven't yet included "springyness of the fuel delivery hose, but this is probably fairly small with aeroquip etc), the longest response and settling time is in the order of 80 to 120ms, with some harmonics of the pressure surge taking 250-400ms to damp out (but these are small).
Ultimately i will build an instrumented pressure test rig and confirm the results.
( i bet that if you stuck a fast response pressure sensor into a typical fuel system you would be shocked by how much dynamic and pressure wave stuff is going on. Luckily these effects tend to average themselves out, and off course the injectors flow rate varriation is SQRT2 better anyway)
Fantastic stuff, I've really enjoyed reading this thread. I did some work with the WRC looking at GPS data that was run at around 10Hz. I've also looked at other systems running to around 50Hz, it's fantastic to see people talking in the 500Hz range.Ultimately i will build an instrumented pressure test rig and confirm the results.
( i bet that if you stuck a fast response pressure sensor into a typical fuel system you would be shocked by how much dynamic and pressure wave stuff is going on. Luckily these effects tend to average themselves out, and off course the injectors flow rate varriation is SQRT2 better anyway)
A1!
Please carry on posting your thoughts and developments I'm sure there's a heap of other lurkers watching this thread.
The Excession said:
Fantastic stuff, I've really enjoyed reading this thread. I did some work with the WRC looking at GPS data that was run at around 10Hz. I've also looked at other systems running to around 50Hz, it's fantastic to see people talking in the 500Hz range.
A1!
To keep the costs down i am only going to use an 8bit microcontroller, but even these run at 20MHz these days, and as long as you keep away from floats or doubles (stick to integer numbers) and try to reduce any divide / multiply routines to Bitwise manipulation, then you can easily run 1kHz PID loops without issue.A1!
Power per ££ has gone through the roof recently thanks to the proliferation of mobil devices, and it can be cheaper to use a 32bit uC than an 8 bit'er (although generally a free tool chain for the software development outweighs the parts cost at low volumes.
The electrical schematic is complete, the first pcb layout is done:
I need to idiot check the component placements and footprints, and double check the pin-outs and routing, then i will get a few manufactured (first batch is expensive (due to the tooling costs), so they need to be right first time or it gets really expensive!)
If you would like to offer your car / fuel system as a trial run then i'm sure we could come to some sort of arrangement ;-) (what EMS system are you running with? the std MBE?)
I need to idiot check the component placements and footprints, and double check the pin-outs and routing, then i will get a few manufactured (first batch is expensive (due to the tooling costs), so they need to be right first time or it gets really expensive!)
If you would like to offer your car / fuel system as a trial run then i'm sure we could come to some sort of arrangement ;-) (what EMS system are you running with? the std MBE?)
Given the power ratings etc involved.
I guess if you wanted, you could run 3 x pumps directly off the two outputs ? It's unlikely ever to exceed the combined 40A rating ?
What have you decided upon for the primary control for the pumps ? or is it still wide open ?
Given I only use a single filter before my 044's due to tank design. I cant help but think this would be a good idea. The over voltage would also offer scope for a bit more ooomph if it was actually required.
I guess if you wanted, you could run 3 x pumps directly off the two outputs ? It's unlikely ever to exceed the combined 40A rating ?
What have you decided upon for the primary control for the pumps ? or is it still wide open ?
Given I only use a single filter before my 044's due to tank design. I cant help but think this would be a good idea. The over voltage would also offer scope for a bit more ooomph if it was actually required.
I have a half finished 2.5kW power supply sitting waiting (50A at 50V or 100A at 25V) to be completed, so if i get this done i will use it with a couple of 044's and a lift pump to try out some control strategies. At the moment, for most people, i think the "returnflow control" might be the best option. Will order a fuel flow turbine and have a play and see what works best.
Stress testing at the full 800 odd watts will probably have to be into a water cooled load resistor
, which will enable me to sort the temperature detrate functions etc
The way i have designed the pcb, it's pretty flexable, so can do all sorts of strats with just a code re-write ;-)
Stress testing at the full 800 odd watts will probably have to be into a water cooled load resistor
, which will enable me to sort the temperature detrate functions etc
The way i have designed the pcb, it's pretty flexable, so can do all sorts of strats with just a code re-write ;-)
Max_Torque said:
I have a half finished 2.5kW power supply sitting waiting (50A at 50V or 100A at 25V) to be completed, so if i get this done i will use it with a couple of 044's and a lift pump to try out some control strategies. At the moment, for most people, i think the "returnflow control" might be the best option. Will order a fuel flow turbine and have a play and see what works best.
Stress testing at the full 800 odd watts will probably have to be into a water cooled load resistor
, which will enable me to sort the temperature detrate functions etc
The way i have designed the pcb, it's pretty flexable, so can do all sorts of strats with just a code re-write ;-)
other query was overload protection.Stress testing at the full 800 odd watts will probably have to be into a water cooled load resistor
, which will enable me to sort the temperature detrate functions etc
The way i have designed the pcb, it's pretty flexable, so can do all sorts of strats with just a code re-write ;-)
Will this be built in, or via external fuses ? Obviously no relays involved now ?
Because there is closed loop current control on all 3 channels, it will not be possible to "Overload" the controller as such, a dead short on the ouput of the controller will be limited to 20A max (transiently, i.e for less than approx 10ms, overload currents of upto 60A per channel will be tollerated before the device shuts down the offending channel).
However, it's probably sensible to protect the wiring to the controller in an automotive environment so i think it will be a good idea to fit a 60A fuse or circuit breaker in the main power feed cable.
However, it's probably sensible to protect the wiring to the controller in an automotive environment so i think it will be a good idea to fit a 60A fuse or circuit breaker in the main power feed cable.
Gassing Station | Engines & Drivetrain | Top of Page | What's New | My Stuff