DIYEFI.org Forum

So, we are designing a board and it's turning into something pretty cool instead of the dead basic thing I had imagined for the first cut, SO, about injector control....

I think we can all agree (to some extent) that the best approach for high Z is to "just" use an autoFET maybe with a couple of basic parts around them to keep it safe and happy.

However, low Z ...

Given that we've agreed to include ignition drivers on board to keep the innevitable people who want that happy, surely we should take the same "lets do it right from the start" approach to injectors too in the knowledge that some people WILL ask for low Z driving from the same case.

There are a few options, but first I'd like to point something out : an injector of 1 Ohm will be drawing 15 Amps+ during opening. This means some serious grunt to drive them OR a slow opening injector riding the current limit during its opening time.

1. Don't support them
2. P&H chip in PWM mode with a FET that can push out 15+ Amps (VNP20N07 perhaps)
3. P&H chip in linear mode with a BJT that can handle 15+ Amps (no idea what?)
4. Ghetto approach of using a low current limit FET (5 Amps or so) and just allowing it to saturate the injector slowly (defeats the purpose)
5. Other options?

1 is easy
2 produces a LOT of noise in the case and probably requires some sort of effort to clean that up
3 produces a LOT of heat in the case and is susceptible to damage from shorts etc
4 doesn't actually provide the P in P&H at all and is pretty much a waste of time, but it DOES allow low Z to be used at all, which could be enough to satisfy some users. 4 is actually equivalent to 1 but with larger heat dissipation demands like 3.

Any further suggestions and methods?

Comments and thoughts on some philosophy behind this?

Bare in mind, that P&H is not required under about 200hp/litre so is a bit of a minority affair anyway. Having said that, this entire site is a minority affair in the first place (we actually WANT to build our own stuff by hand).

Can anyone do the math to show just how much heat will need to be dissipated by 6 devices working at 100% duty and show whether or not we can reasonably get rid of that heat in a 4" x 6" x 2" case without excessive rise in internal case temperature?

It's nearly 4am, so I'm off to bed soon.

I look forward to reading your thoughts on this.

Fred.

4am, that's when I get up for work.

I elect 1, but half hearted. I think the 8 amp PIP can work, but we might need to beef it up in some places to handle the current. I've seen creative use of solder to make traces thicker, allowing for more current. Perhaps we can plan for high imp, but leave the traces bare. Then those that need low imp, can solder these specific traces giving them the current capability, while not gobbling real estate for the average Joe.

Low-Z injectors are usually much more than 1 Ohm. They are usually in the 2 to 3 Ohms range but there are some that are under 2 Ohms. Even then, you will not want to let the injector current go to 15A (or whatever the current would be from battery voltage divided by the injector impedance). And the current will not instantly go to this maximum value even if you apply the full battery voltage: it will rise with a certain slope depending on the inductance. Also, you probably don't want to let the current go to the maximum value because it will heat up the injector uselessly since the opening will occur at a lower current.

Most low-Z injectors are rated at 4A/1A peak/hold current. So you won't need FETs or BJTs that are rated at much more than 5A assuming your injector control is correctly done (either through PWM or some linear control) and/or you have auto-protected devices (such as the VNP5N07). I'm using TIP122s for my p&h boards which are rated at 5A and they are fine (as long as they have a correct heat sink).

As for the heat generated, assuming 15V and 2 Ohm low-Z injectors being held at 1A, by simplifying thing you can see that with 15V you need 15 Ohms to have a 1A current. Since the injector is 2 Ohms that means the BJT has the equivalent of 13 Ohms which means that the power dissipated is 13W (Power = RxIxI = 13x1x1) per injector. For 6 injectors, you have 78W which requires a good heatsink and air flow. However, this is really a worse case scenario because you have 100% duty cycle: in a real application not only will you have a lower duty cycle but you will also have the peak phase which will generate very little heat because the BJT will be fully saturated. However, this power dissipation is dependent on the injector impedance.

So from this, I think it is logical to leave this option for an external box. Not only are the board space requirements somewhat higher but the heat management is significantly more demanding. However, this might not be an issue if an enclosure is chosen with this in mind.

Jean

jharvey wrote:Perhaps we can plan for high imp, but leave the traces bare. Then those that need low imp, can solder these specific traces giving them the current capability, while not gobbling real estate for the average Joe.

That's another interesting idea. I wonder though, will it act as a trap for those upgrading injectors and not realising that they need to thicken the board? I guess standard assembly instructions could be clear about it and urge users to do it anyway such that if they upgrade later they don't have an issue. Otherwise it might end up working much like your fuse trace idea :-)

jbelanger wrote:Low-Z injectors are usually much more than 1 Ohm. They are usually in the 2 to 3 Ohms range but there are some that are under 2 Ohms.

Sure, but such designs should be based on the toughest conditions imaginable unless they cause the design to become impractical. I was just going for worst case values :-)

Even then, you will not want to let the injector current go to 15A (or whatever the current would be from battery voltage divided by the injector impedance). And the current will not instantly go to this maximum value even if you apply the full battery voltage: it will rise with a certain slope depending on the inductance. Also, you probably don't want to let the current go to the maximum value because it will heat up the injector uselessly since the opening will occur at a lower current.

OK, but where are some of these specs? I'd like to see the specs for the ones that are close to 1 Ohm. We don't want the situation where the driver is configured for 4 Amp peak and the injector needs 6 or 10 etc. Obviously this is on a per injector basis and probably best determined with a dual scope showing current and the output from a transducer on the body of the injector.

For example, if an injector needs 6.5 A to open ASAP and we feed it 4 A then the opening point will move from pale blue to dark blue. This is bad. It's not so much the reduction in current that matters, but rather the shortened time of higher current. If you pull the feed to it prematurely then it could become very dozy to open or intermittently not open.



Do we have any empirical evidence of current vs opening time curves?

As for the heat generated, assuming 15V and 2 Ohm low-Z injectors being held at 1A, by simplifying thing you can see that with 15V you need 15 Ohms to have a 1A current. Since the injector is 2 Ohms that means the BJT has the equivalent of 13 Ohms which means that the power dissipated is 13W (Power = RxIxI = 13x1x1) per injector. For 6 injectors, you have 78W which requires a good heatsink and air flow.

How much heat would be generated in a PWM setup?

However, this is really a worse case scenario because you have 100% duty cycle: in a real application not only will you have a lower duty cycle but you will also have the peak phase which will generate very little heat because the BJT will be fully saturated.

It may be a worst case scenario, but it is definitely a real world scenario for some vehicles that are setup marginally (like mine at the moment). Once I get a decent set of ratios behind my engine it will be shifting from 7500 back to 5000 with 1000/2500 = 40% of the time spent at 100% (and NO opening period of low dissipation) and the rest spent close to 100%. In that case the worst case is realised. We do stuff around with cars so that we can drive them fast afterall, and driving fast = high duty :-)

So from this, I think it is logical to leave this option for an external box. Not only are the board space requirements somewhat higher but the heat management is significantly more demanding. However, this might not be an issue if an enclosure is chosen with this in mind.

I'd tend to agree unless we can use those chips in PWM mode to keep the heat manageable. If we can, then I think it's a goer (as it will buffer the CPU and provide good switch on current to the FET too (24mA they say versus the 4mA or so that a CPU pin can safely push with a full 5v differential though a 1k resistor or so.

Another member PMed me to say he's working on a solution for this, so we'll see what he comes up with. Perhaps it will be suitable, perhaps not. I told him you are the P&H MAN so he should allow your all-knowing eyes to be cast over it and pass judgment :-)

Time to fix those smilies!

Fred.

Fred wrote:

jharvey wrote:Perhaps we can plan for high imp, but leave the traces bare. Then those that need low imp, can solder these specific traces giving them the current capability, while not gobbling real estate for the average Joe.

That's another interesting idea. I wonder though, will it act as a trap for those upgrading injectors and not realising that they need to thicken the board? I guess standard assembly instructions could be clear about it and urge users to do it anyway such that if they upgrade later they don't have an issue. Otherwise it might end up working much like your fuse trace idea

Sounds interesting. As for potential problems, explicit instructions should be good enough. In any case, if you try to be foolproof "they" will just build a better fool.

Fred wrote:

jbelanger wrote:Low-Z injectors are usually much more than 1 Ohm. They are usually in the 2 to 3 Ohms range but there are some that are under 2 Ohms.

Sure, but such designs should be based on the toughest conditions imaginable unless they cause the design to become impractical. I was just going for worst case values

Agreed.

Fred wrote:

Even then, you will not want to let the injector current go to 15A (or whatever the current would be from battery voltage divided by the injector impedance). And the current will not instantly go to this maximum value even if you apply the full battery voltage: it will rise with a certain slope depending on the inductance. Also, you probably don't want to let the current go to the maximum value because it will heat up the injector uselessly since the opening will occur at a lower current.

OK, but where are some of these specs? I'd like to see the specs for the ones that are close to 1 Ohm. We don't want the situation where the driver is configured for 4 Amp peak and the injector needs 6 or 10 etc. Obviously this is on a per injector basis and probably best determined with a dual scope showing current and the output from a transducer on the body of the injector.

For example, if an injector needs 6.5 A to open ASAP and we feed it 4 A then the opening point will move from pale blue to dark blue. This is bad. It's not so much the reduction in current that matters, but rather the shortened time of higher current. If you pull the feed to it prematurely then it could become very dozy to open or intermittently not open.



Do we have any empirical evidence of current vs opening time curves?

The specs should be available from the injector manufacturer and/or vendor. All the petrol injectors I've seen have been either 4A/1A or 2A/0.5A peak/hold rated with the latter being obsolete ones. For the 4A/1A ones, there was quite a range of impedance values. And all the commercial peak/hold driver boxes I've seen are rated at 4A/1A. That doesn't mean there aren't other ratings because I haven't seen that many injectors and I haven't done an extensive research on injectors and driver boxes.

However, I have seen different ratings for LPG and CNG injectors with most values being 4A/1.5A for very low impedance and one case where a single priming pulse required 8A.

So it shouldn't be a question of measuring but just getting the correct specs.

Fred wrote:How much heat would be generated in a PWM setup?

I haven't done any computations but from the LM1949 datasheet (which can be driven in PWM mode) the heat generation saved in the BJTs is almost all transferred to the flyback Zener diode (by the way the flyback term is not MS talk but from the datasheet) which makes heat management more problematic if anything.

Fred wrote:

However, this is really a worse case scenario because you have 100% duty cycle: in a real application not only will you have a lower duty cycle but you will also have the peak phase which will generate very little heat because the BJT will be fully saturated.

It may be a worst case scenario, but it is definitely a real world scenario for some vehicles that are setup marginally (like mine at the moment). Once I get a decent set of ratios behind my engine it will be shifting from 7500 back to 5000 with 1000/2500 = 40% of the time spent at 100% (and NO opening period of low dissipation) and the rest spent close to 100%. In that case the worst case is realised. We do stuff around with cars so that we can drive them fast afterall, and driving fast = high duty

Still, you should spec your injectors with some margin simply to be able to get a reliable fuel delivery. Once you get over about 80% duty cycle (depending on injector and engine speed) you enter in non-linear region where the injectors are not fully closing (or not at all) with a somewhat random fuel amount when going from a fully closing injector between pulses to a fully open injector. Also if you really need 100% duty cycle then what happens when the atmospheric and engine conditions are such that you have more air mass going into the engine? It sound like a recipe for running lean and melting your expensive engine parts.

As for designing the board, I agree that this is what should be used for the worse case scenario.

Fred wrote:Time to fix those smilies!

Fred.

I almost didn't noticed that you had.
Nice to have more space AND not being distracted by the bouncing words.

Jean

jbelanger wrote:The specs should be available from the injector manufacturer and/or vendor. All the petrol injectors I've seen have been either 4A/1A or 2A/0.5A peak/hold rated with the latter being obsolete ones. For the 4A/1A ones, there was quite a range of impedance values. And all the commercial peak/hold driver boxes I've seen are rated at 4A/1A. That doesn't mean there aren't other ratings because I haven't seen that many injectors and I haven't done an extensive research on injectors and driver boxes.

Excuse the skeptic in me, but I'd be interested in making them open as fast as possible rather than operating them to spec. Spec might work just fine, but driving it a little harder could aid performance somewhat/slightly. Provided it's only for circa 1ms you could just about do anything you like to them if you keep the hold current correct and they won't overheat.

I haven't done any computations but from the LM1949 datasheet (which can be driven in PWM mode) the heat generation saved in the BJTs is almost all transferred to the flyback Zener diode (by the way the flyback term is not MS talk but from the datasheet) which makes heat management more problematic if anything.

Interesting. Why is it that we must catch the flyback in that way, and at what voltage? Would we be better off using IGBT's for the injectors too so that we didn't have to catch the spikes? That should make them close as fast as possible by allowing a rapid field collapse right? I ran across an article on flyback catching that was interesting here : http://www.priorartdatabase.com/IPCOM/000006968/ So catching it at a low voltage removes noise and protects the drivers from high voltages, but what does it do to the injector close time. I asked this before and someone told me I was worrying about a non-issue. I guess thats right. What do you think of the SCR based circuit? Seems like a good idea. Would it work for PWM too?

Still, you should spec your injectors with some margin simply to be able to get a reliable fuel delivery.

It is/was a temporary arrangement that I'm not happy with, but it is probably fairly typical of various installs to be running in that region sometimes.

Once you get over about 80% duty cycle (depending on injector and engine speed) you enter in non-linear region where the injectors are not fully closing (or not at all) with a somewhat random fuel amount when going from a fully closing injector between pulses to a fully open injector.

I disagree with this. At least on my setup the only time I got a rock solid AFR was when they were maxed out. At lower duties you could see waver in the AFR from the single shot batch fire nature of it. The transition from duty to ON was dead smooth. The 80% arbitrary figure is a good one to stick to, but mainly for the reason you mention below. If you run them fully on, then they are being cooled by the fuel and are 100% consistent in their flow. The only thing that could possibly cause an issue is if the behaviour of the field post switch off interfering with switch on again. I don't think that is a realistic issue though. Anyway, no point arguing about whether it works or not as A it does work and B we don't want to run like that anyway...

Also if you really need 100% duty cycle then what happens when the atmospheric and engine conditions are such that you have more air mass going into the engine? It sound like a recipe for running lean and melting your expensive engine parts.

You are spot on the money there. A guy I know had same engine, similar turbo, same injectors and went out one chilly night and melted his stuff. I was running less boost than him and it was cold when I had 100% duty and rich enough that some leaning would still be OK. Again, I don't like it, but it is a realistic usage case for some people.

As for designing the board, I agree that this is what should be used for the worse case scenario.

Cool :-)

Nice to have more space AND not being distracted by the bouncing words.

Yeah, that was getting to me too. Now we have the best of both worlds :-)

I wonder if I could modify the html or php or css to keep that smiley box narrow even when the screen is wide?

Fred.

Fred wrote:Excuse the skeptic in me, but I'd be interested in making them open as fast as possible rather than operating them to spec. Spec might work just fine, but driving it a little harder could aid performance somewhat/slightly. Provided it's only for circa 1ms you could just about do anything you like to them if you keep the hold current correct and they won't overheat.

But the thing is that you can't drive them harder than by providing full battery voltage which is what is done in a peak/hold driver during the peak phase (whether PWM or linear). And if the spec says that the injector is open when the current has reached X amps then it will probably be open significantly before that. So pushing X+Y amps through the injector will just heat it up with no benefit (other than not heating up the driver). And there's no other way than having a longer peak phase to provide the additional current so that won't lower the opening time.

I think that either I'm not getting what you mean or you're not getting what I mean.

Jean

I've almost got a design together for this. Perhaps it will be suitable, perhaps not, depends what everyone wants.

P&H drivers apply full battery voltage for around 4-5ms then pull current down to 1-1.5A for the rest of the injector pulse. Generally this is done internally by using a comparater to drive the gate. An internal reference voltage is applied to one side and a feedback resistor is used below the mosfet to compare values. The feedback line is driven through an RC circuit so it takes a certain amount of time to switch to current controll depending on the current being driven through the Injector.

I'm almost there with my design, it uses discretes instead of a driver chip and is relatively simple. A twin DRC snubber circuit is being used for spike suppression and a single RC circuit is being used for ring suppression.

A little more tweaking and I'll post up a diagram.

jbelanger wrote:And if the spec says that the injector is open when the current has reached X amps then it will probably be open significantly before that but could take longer to open that specified/current could rise more quickly than specified/etc

Fixed it for ya :-) Hence skeptic (paranoid?)

I think that either I'm not getting what you mean or you're not getting what I mean.

Usually it's me failing to communicate, but occasionally the other too :-)

Delta wrote:P&H drivers apply full battery voltage for around 4-5ms

Jesus, no way, even saturated injectors are fully open by around 1ms that's just going to fry something.

I'm almost there with my design, it uses discretes instead of a driver chip and is relatively simple.

What is the advantage of this approach over a chip that is known to perform well and is therefore a predictable choice? Not necessarily from me, but I think you will face a similar backlash to if you say suggested that your v3 board would use a VR conditioner based on discretes instead of an lm1815. Many people would argue the chip is a better call. Perhaps your solution thumps the lm1949, but I think a few people might not like it. Just warning you about such people.

A little more tweaking and I'll post up a diagram.

Looking forward to it :-)

Fred.

Fred wrote:

Delta wrote:P&H drivers apply full battery voltage for around 4-5ms

Jesus, no way, even saturated injectors are fully open by around 1ms that's just going to fry something.

I'm almost there with my design, it uses discretes instead of a driver chip and is relatively simple.

What is the advantage of this approach over a chip that is known to perform well and is therefore a predictable choice? Not necessarily from me, but I think you will face a similar backlash to if you say suggested that your v3 board would use a VR conditioner based on discretes instead of an lm1815. Many people would argue the chip is a better call. Perhaps your solution thumps the lm1949, but I think a few people might not like it. Just warning you about such people.

A little more tweaking and I'll post up a diagram.

Looking forward to it

Fred.

Sorry should have clarified - applies full battery voltage and limits to 4-5A for 4-5ms then drops the limit to 1-1.5A. You will actually find that most low Z injectors at around 4ohms will not make the full 4A.

The distinct advantage of using discrete components is that you can CHANGE anything you like about the peak and hold setup. By adjusting 2 resistors you can change the peak/hold current limits. By adjusting a resistor + capacitor you can change the peak time. etc etc. This means the drivers will work off a single pin, and can drive both low Z and high Z injectors with the same layout, and could conceivably drive any new type of injector even if it requires say 8A/2A hold or direct injection where the peak current are very high etc.