Injector Control Options

[jharvey]

To me it seems like these are the primary features that might be handy. Most desired is on top.

Industrial temperature range or better
Hight Z injector
Low Z injector
PWM Peak/Hold
diagnostics feed back
changeable profiles

Are there other injector drive features that would be nice?

I'm tempted to say this say this is most desired for 6 cyl's. I think that 8 cyl's will want 2 banks any how, so we can probably keep the part count down a bit. The goal would be a 0 to 5 V input from the MCU with nearly no delay to an injector profile output. It would be more important to maintain a constant propagation delay, then to have a short delay.

One thing that has me wondering, is if a live or near live uH measurement could be taken. Perhaps if a 100khz couple mV signal was passed to the injector, then you could received, filtered and energy detected. You should be able to see a variation in uH when the pin moves in the injector. By seeing this variation in uH, you can know when the pin has hit full open or full closed. You would also be able to see if you are driving it so hard that it bounces when it hits the physical stop. Which might allow dampening before it hits the end of travel. Perhaps this meter can offer some insight into measuring uH on the fly.

http://khup.com/download/1_keyword-lc-m ... -meter.pdf

I also know that with ham Radio it's common measure impedance with a bridge device. Kind of a balancing act like a wheat stone bridge, but with inductance. You have an inductor of know impedance, and you connect your antenna. When you cancel a beat frequency, you know your are spot on with your impedance match. Perhaps that could also offer a live reading of uH.

This is why I was thinking PSoC. I think the uH sensor circuit will require small analog signal generation and receiving. Measuring uH is mostly for initial tuning, and perhaps diagnostics. So one measurement circuit switched across several injectors would be fine. A PSoC allows for that kind of analog switching via software, and is the only chip I know that includes both analog and digital components in one low cost package.

Any how, those are just a couple thoughts to ponder. Off to bed I go.

[Fred]

Also, Sean did say that I was free to contribute to this forum while using good judgement.

Cool :-)

All - Since there appears to some interest in developing a Low-Z Injector board, should a new message thread be started? My initial thoughts hardware wise, would be to just use the cheapest Freescale S12 part that will do the job (S12 only because you are already working with the family). Use a protected MOSFET (to provide the high Vclamp at turn-off) and use something like the active clamp to BAT circuit that Mega uses to moderate the current decay in the inductor during PWM. I know, not much of a "spec" but maybe a starting point for discussion.

I see no reason to use freescale stuff for an app like that. I'd go for something common, cheap, available, and with the right specs. There are many options.

- Huff (by the way Fred, what's a "hoon"? :)

To hoon is to drive fast, or loudly. A hoon is someone that does that. It's an old term, from people your age :-) And from NZ/Oz and maybe the UK?

This 20/4 amp talk seems to be random. Jean uses normal BJTs as per the datasheet. They have inherent V drop and heat output, which isn't too bad... till you have 8 of them. Any 20 amp parts would likely be FETs which that chip can't drive.

I'd be looking to have the code/pcb work as follows:

Allow the primary micro to control the pulsewidth directly through an AND gate.
Use the secondxary micro to watch the primary micro's output.
The secondary micro would hold all of its pins in the ON state such that the propogation of the primary micro's signal was immediate.
The secondary micro, upon sensing the primary one come on, would control the injector current as per its configuration.
Firmware for the micro should be loadable easily with cheap tools/parts.
Configuration of the micro should be done over SPI or I2C by the primary micro (FreeEMS)

Configurables:

PWM duty/frequency for opening
PWM duty/frequency for hold
Opening time: fixed, battery dependant, measured or controlled by the primary micro via I2C/SPI in real time.

There are two aspects to this project:

Controllnig the P&H behaviour
Disipating the heat and noise generated

Thoughts?

Fred.

[DaWaN]

Thoughts?

I think your idea pretty much covers it.

Personally I think it might be cool if we add even more features to this specific micro which the current Freescale main controller cannot provide.
For instance stuff like a wideband lambda sensor or knock sensing etc.
But it is important that the FreeEMS will always work without this extra micro, so you can choose to have the extra micro with the extra features or go without it to keep costs down.

In the end it basically depends on costs I guess, if the P-H idea can be made very cheap then you problably just stick to a single micro for the PH part, if it turns out that for instance a high end PSoC is not much cheaper than a bone-stock AVR then we might want to integrate more stuff in the extra micro. We should not go wild in the amount of microcontrollers and IC's on the board as it drives up costs and make development and updating harder (as you have to compile and program 2 microcontrollers instead of one).

But first things first, I would start with getting all the requirements for the whole PH circuit, then we can look futher

[TonyS]

Clamp good, snubber diode/device bad. Flyback terminoligy, common in MS land but shows a mechanic vs an engineer. Snubber is a better term.

Not sure I understand your comments. "Flyback" or "inductive kick" are both terms used to describe the voltage generated when the current flow in an inductor (injector) is abruptly cut off. You can "clamp" (or "snub" if you prefer) this voltage (limit the magnitude) in a number of ways. One common way is to connect a diode "backwards" across the inductor. This is "good" if you are trying to maintain the current flow in the inductor (injector) for as long as possible (what you would want during the PWM "hold" portion"). Another way is to let the "flyback" voltage climb to a higher "clamp" level before allowing current flow to continue and dissipate the energy in the inductor. This is "good" if you want to stop the current flow in the inductor (injector) as soon as possible (what you would want when turning the injector off). For Peak and (PWM) Hold injectors, you really need two "clamps", one that comes into play during the "hold" period and one that is used at injector turn-off (this looks like how the Mega works from what I can tell from their schematics).

Allow the primary micro to control the pulsewidth directly through an AND gate.
Use the secondxary micro to watch the primary micro's output.
The secondary micro would hold all of its pins in the ON state such that the propogation of the primary micro's signal was immediate.

Just curious, how much "time" is being saved by using this scheme as opposed to just feeding the secondary micro directly from the primary micro?

Also -
Is this to be an "add-on" board (remote?) or is to be a "populate parts" option on the main board?
What is the "range" of injector impedances that are being considered? 16 Ohms for high-Z is what I seem to remember as a max, but what is the "practical" minimum for a low-Z?
Which low-Z injectors are popular with "hoons"?

- Huff

[jharvey]

Allow the primary micro to control the pulsewidth directly through an AND gate.

I believe the features you are looking for with the AND gate include a redundant isolation voltage barrier, such that voltage spikes have more trouble making it back to the MCU increasing it's robust nature. As an example, if the FET cracks and the OV protection fails, you still have a second layer of defense before the brain gets toasted. The other feature is a hard configurable polarity, such that 5v from the MCU can either mean injector on or off.

TonyS I think we are both nearly on the same page.

Clamp is typically a good term, it indicates dissipation of energy by limiting amplitude. Clamp doesn't really concern itself with why it's amplitude is high, just that it is high.

Over Voltage protected FET's will voltage clamp. They can often be found in a 40 to 70 voltage range, then another range around 300V. For injectors, it's better to keep the clamp voltage low. So use the 40v to 70v devices. For ignition you probably want that 300 v range. By voltage clamping you dissipate energy with a variable load such that the voltage is fixed. This variable load is typically the drive silicon, so the clamped energy is dumped in the form of heat at the ECU or driver, not the injector. Because the clamping load is variable and the voltage is fixed, the energy decay is linear, and more precise.

When clamping current, you will want at least two levels. Full current and hold current are by far the most important, but there are transition points that can be of interest as well. So the use of clamp can be a bit vague about about what you are working with.

Snub is also typically a good term. It indicates you are dissipating short energy spikes caused by an inductive or capacitive loads. A snubber diode, or snubber cap will not dissipate the energy in a linear fashion.

The snubber diode will typically dissipate the energy based on a fixed resistance in the system, and allow the voltage generated by the inductive spike to flow across it. For a snubber diode on an injector, this typically uses the injector resistance as the load and causes the heat to be absorb by the injector, self heating the injector slightly. The decay is exponential and the off time tolerance is a bit larger and harder to control than with a clamped device.

The use of a snubber cap will dissipated the energy quickest. However the cap would be at least the size of your fist per injector, so it's not physically practical. Also we don't need quick. We need either good enough, or precise.

Picture coming to a stop with your car, you can simply kill the engine and let it dissipate the energy bringing you to a stop. Or you can use the brakes. The brakes prevent heat in the engine bay, and allow for a more precise control over when you will come to rest. So I recommend that approach. However, you can technically use the other engine kill method, which doesn't require brakes.

Flyback indicates you are only dissipating energy by killing the engine, and freewheeling your energy. It's a term that became popular by the MS crowd because it had a good feel to the mechanically minded folks. It would be kind of like saying a Jacobs brake. It's a very limited scope and only really applies to a very specific application, that's not typically found in most applications. Because of this limited scope and common misuse, I'll point out it's deficient.

Any how, these point are shed color as far as I'm concerned, I like keeping focus on design, and moving forward.

[Fred]

Just curious, how much "time" is being saved by using this scheme as opposed to just feeding the secondary micro directly from the primary micro?

You would be feeding the secondary micro directly, and the injector driver directly, too. It just changes the whole ball game to do it this way. If you ONLY feed the secondary micro directly, you've got a lot of accuracy to maintain and you have to do things fast and right. You'll always have ISR latency to deal with, or polling period delays or something. You're taking an accurate output and making it inaccurate, no matter how good your code and secondary cpu are. By doing it as I suggested, you change the game to one of just observing and current limiting, which is simple for software, performant (as much as the main cpu is, anyway) and allows your CPU to be shit, cheap and free of all the peripheral modules that you would need if you didn't do it this way.

Is this to be an "add-on" board (remote?) or is to be a "populate parts" option on the main board?

I don't think it matters much, does it? If we come up with a great topology and some software for a cheap cpu to do this, then it can be implemented as part of a design, or as an external box, the latter of which I would always prefer due to heat and noise concerns.

What is the "range" of injector impedances that are being considered? 16 Ohms for high-Z is what I seem to remember as a max, but what is the "practical" minimum for a low-Z?
Which low-Z injectors are popular with "hoons"?

I have some sard 800cc units here that measure 3 ohms static resistance. There is discussion in this thread or another about possible minimums.

Fred.

[Fred]

Allow the primary micro to control the pulsewidth directly through an AND gate.

I believe the features you are looking for with the AND gate include a redundant isolation voltage barrier, such that voltage spikes have more trouble making it back to the MCU increasing it's robust nature. As an example, if the FET cracks and the OV protection fails, you still have a second layer of defense before the brain gets toasted.

No, i don't consider a failing fet to be a significant risk, and didn't mean that at all. I said it for the reasons in my previous post, mainly that you cant get the same accuracy if you dont run it direct (through the and) to the fet. But also that you can use a cheap shitty cpu to do this (part of the goal) if you dont have to control things so precisely.

The other feature is a hard configurable polarity, such that 5v from the MCU can either mean injector on or off.

No, injectors can always be considered to be low side drive. Having it configurable polarity would actually break my idea anyway, i think. Certainly, its a bad and pointless idea in this case, unlike ign.

Fred.

[TonyS]

Is this to be an "add-on" board (remote?) or is to be a "populate parts" option on the main board?

I don't think it matters much, does it? If we come up with a great topology and some software for a cheap cpu to do this, then it can be implemented as part of a design, or as an external box, the latter of which I would always prefer due to heat and noise concerns.

The reason I asked is because I initially assumed that an external box application was a given, but when I read "Configuration of the micro should be done over SPI or I2C by the primary micro (FreeEMS)" I wasn't sure, as these protocols are more typically used for intra-board communications, while board - board communications tend to use more environmentally "robust" protocols (485, CAN,...).

-Huff

[jharvey]

Allow the primary micro to control the pulsewidth directly through an AND gate.

I believe the features you are looking for with the AND gate include a redundant isolation voltage barrier, such that voltage spikes have more trouble making it back to the MCU increasing it's robust nature. As an example, if the FET cracks and the OV protection fails, you still have a second layer of defense before the brain gets toasted.

No, i don't consider a failing fet to be a significant risk, and didn't mean that at all. I said it for the reasons in my previous post, mainly that you cant get the same accuracy if you dont run it direct (through the and) to the fet. But also that you can use a cheap [shizzle] cpu to do this (part of the goal) if you dont have to control things so precisely.

I see what you're saying, the AND gate is simply a place for a MCU to monitor and modify a signal, you suggest the MCU does not direct drive the injector. This speaks well for the PSoC as it includes an on-board AND so no external components would be needed for such a feature. Can you pencil up a draft sketch of how you see the sub MCU working with the AND gate?

[jharvey]

Might as well repost this fellows page, lots of good injector info.

http://sonic.net/~mikebr/ecm_555/inj_inductance.html