injectors - [56k beware page 2]

[jharvey]

I don't fully understand the over voltage feature of the PIP. My understanding is that the over voltage will turn on the gate allowing it to drain some current, when enough energy has gone through, it's voltage will drop such that it stops current. Basically going from a 0 ohm resistor gradually to a high impedance resistor.

The above schematic with clamping diode should be quite similar to this feature in the PIP. Note the oscillations that occur under 36v. That can be a good source of noise, but can also be blocked by a small cap across Vd and gnd. Perhaps a bypass resistor would work as well. That might allow for the primer feature as well. Say a 5 watt 70 ohm across Vds. This would prime the injector as well as prevent oscillations.

Energy stored by the coil will be dissipated by the PIP. So paying attention to heat sinking will be important. Hmm, perhaps we should think about temp feed back?

I kind of blew away the clamped schematic above, I'll see if I can recreate it.

[jharvey]

Seems QUCS doesn't like the clamping diode as much as other components. I'd bet that's because I'm using a normal diode as a zener. Takes it a bit to get through the calcs for the clamped part. Any how, back to injectors.



I limited the clamp diode to 17.5 volts to keep a consistent voltage scale with a max of 18V like the other above post's. I didn't see significant change when I had this at 50V.

Note if you have 1 ohm for Rconn2, it didn't get through the calc in half an hour. I think the clamping get's stretched out significantly when that happens. So I'd say this circuit is likely more prone to increasing injector on time, due to a bad connections. So don't do that, use the dielectric lubricant like your supposed to.

So I guess it's time to add this to the schematic(s)? Any thoughts? These are the options I think should be added that aren't shown in the above.

-- I think it would be good to add a 0 ohm resistor at Iinj allowing for the option of a series current measurement, this would be used for prototyping simply a place to add an amp meter if we want it.
-- I think we should add the hall for current sensing, I think this should stay after pro-typing, and can be used as a diagnostics/setup tool.
-- I think I like the 70 ohm resistor for prime and anti oscillation.

Also, the pics on sheet 2 are less then 300k. Big size, but fairly small file size.

[Fred]

Mate, the voltage ringing is a non-issue for noise provided that current is low. Current produces radiated noise, not voltage. Hence the internal clamp on the FET is all that's required. If you allow the voltage to go higher the current is by definition lower when it clamps (energy is fixed).

As for thermal feedback, they are thermally clamped too.

Basically, if you use an autoprotected FET driven straight of the CPU through a single resistor (perhaps with rail clamps to protect the cpu in case of bad connections) they are pretty much indestructible and provide the best injector behaviour possible within reason.

If you were to PWM them for Low Z it heat would become an issue, that is where the external components come in : keeping the heat out and sharing the work load.

This great research for if/when we decide Low Z on board is a goer, but not required for a basic functional High Z setup IMO.

I'd like to see Jeans and GearHeads comments on this too as they probably know more about it than I do.

Fred.

[jharvey]

Mate, the voltage ringing is a non-issue for noise provided that current is low.

You're probably right, however, in the above spice it shows light blue as the ringing signal, the right legend shows current on the order of several amps. However I suspect the spice isn't entirely accurate in that case, and I would expect the real world ringing to be much less.

Using the PIP may act like a very fast switch, if it's not slightly damped. Perhaps the injector will damp it enough, perhaps not. For a proto, I don't think it will hurt to have a cap included, it doesn't need to be installed.

Some things to keep in mind, the 40 ms 1000 points spice examples only samples at about 25Khz. Higher frequencies are ignored unless I want to tell it to do more points. I have to admit I'm surprised that 10 pF registered at all.

[Fred]

Sure, one cap won't hurt anything layout wise, but if that's all you mean to include, what are all the other components? Simulated deficiencies in the board and traces? Perhaps this is just a case of mis communication :-)

I have another question too, why are there two 7.5 Ohm resistors in the path from FET to power feed? Is the simulated injector a 7.5 Ohm one or a 15 Ohm one? If 15, why split it like that, does it make a difference? I can't see how it would.

Fred.

[jharvey]

Many of the parts are correcting ideal assumptions for more realistic assumptions. I split the inductor resistance because the resistance is spread through the entire wire that makes the inductor. Splitting it makes it feel more real. I originally had one resistor in series, but split it as a troubleshooting technique when I thought QUCS was having trouble keeping track of nodes. Turned out the problem was elsewhere and was simply my ignorance. It can probably model the inductor w/ single resistor. I'm just to lazy to fix it now. Besides it makes people wonder if I'm doing something special with it set like that.

I should have another box in the bottom like I have for the inductor and battery, because I'm using the diode, cap, and MOSFET as a model for the PIP device. But if I did that it might show consistency in my approach.

Some values are to look at what might happen with certain conditions, like a bad injector connection, or a bad battery connection. Crf was required because when the diode would clamp QUCS would jump off into the weeds. I think in the real world, we might see similar, but as RF not errors in software. 10pF appears to be as small as I can go in software. 10pF might be the natural capacitance of the real world parts anyhow.

I was thinking of a schematic something like this. It's not done yet, mostly the hall sensor might not be right. I tossed it in there, then noticed, I don't have a clue how it might work, so I'm learning now then I'll go do a test circuit.



There's a good chance the 5K isn't needed, but I'm used to doing that to prevent the current in the FET from creating a voltage that will keep it on. I think the PIP takes care of that, but if it's there, then folks can perhaps use other chips.

[Fred]

The 5k isn't required because the CPU will pull up and down actively in both directions :-)

Another thing : Your injector is being "powered" with two grounds in that pic ;-) Might wanna cut the lower trace and connect it to battery +ve :-)

Also, the current sense will need a resistor to measure V drop across won't it? I'm not sure where that belongs, but either in the top trace or the bottom one before the FET.

Fred.

[jharvey]

Another thing : Your injector is being "powered" with two grounds in that pic Might wanna cut the lower trace and connect it to battery +ve

I smart. I guess I should have looked at that at least slightly before I posted it. Heres a slightly more refined schematic.



The hall sensor is kind of like an NPN transistor, that's why I took the NPN symbol, however it's not an accurate symbol. The hall doesn't have a connection for the base, instead it's connected to V+. Also the pins aren't numbered right. V+ is pin 1, V- is pin 2 and the collector is pin 3. The hall will saturate the base when it has a magnetic field, making it work as a switch to gnd.

I breadboarded a quick circuit last night. I gave the hall 12V for V+, gnd for V-, the ran 12V to a resistor then the LED then to the hall collector pin. When I moved a fridge magnet near it, the LED would light up. So yeah, it will work with DC. To get the magnetic field for the current of the injector, we'll need a small coil. Not sure what will be required yet, I've played some with wrapping a couple turns of wire on a nail, but then blew my wall-wart, so I really should free up my bench supply and use that instead. I suspect the coil will be in the nH range.

I tossed in the LED for kicks. When the LED is on, the injector is off. So a red LED might make sense.

The hall current sense can still use some work. I haven't predicted R values for that circuit yet. We need to remember 50V (plan for 60) input voltage, with the hall at a max of 30V. So the first resistor is simply a voltage divider limiting the voltage the hall will see. Probably wants a filtering cap as well, to keep the spikes down. Also also doesn't want a - voltage for the input. Max reverse is 1V.

I think we might want that Zener or some such device to keep current sense below 5V. There might be a more elegant way to do it, but I haven't thought of it yet. The NPN was drawing 4ma when I had the LED on it. Once we have a value for that initial R, then we can select a real value for the 600R that will keep the current at around 2 to 5ma.

Any how, that's what I've got for now. I'm sure the hall circuit can be done much better.

[gearhead]

Woah! This is interesting, but it seems we may be going the way of Yamaha Amps of the 80s and 90s. Lots of complex circuitry which measures and simulates well, but misses the simple task we are trying to do. Interesting thought to using PWM to decrease opening time but, as was posted earlier, the goal is not to have a zero opening plus closing time (dead time), but a dead repeatable value for that time. That any ElectroMechanical device/system has some dead time and, due to its mechanical nature, will also have a variation in this time and a standard deviation of this time. Spending valuable resources (time) to determine the 'perfect circuit' without a feedback loop of real world testing may not bring us the results we are looking for. I tend to agree with Fred that we need a large enough (current capable) FET to pop the injectors open as quickly as practical. As for a current trap for the spike when the injector is released, a fast, low impedance trap is what we need. If this is merely the FET or a 'flyback' circuit as in MS speak, that is fine, to me, as well. Just that it allows the spike to be quickly shunted.

Am I missing something?

Gearhead

[Delta]

I'm going to have a play with your circuit and have a look, but I think you will find that in real life some of your components will melt. Using a zener to clamp the inductive spike from the injector WILL NOT WORK. A 25mH inductor being turned of by a fet or mosfet with roughly 1amp going through it will generate in excess of 1kV, the current sunk by the zener will be quite high, on the order of an amp at its breakdown voltage. Even though this power dissipation is transient, its a repetitive event, and you will burn it unless you have a very very high wattage rating on the zener.

I'll have a play with the simulations I have of my solenoid driver circuits with feedback, and put in your values for injector inductance capacitance and resistance and see what I can come up with. The other thing you have neglected is that you are driving your fets with only 5V while their Vds is at least 12v, this will put it in the triode region and hence it will act as a voltage controlled resistor, and hence disipate a fair amount of power. You need to bring your driving voltage up near Vds to put it in switching range and hence not fry the fet/mosfets. The circuit I usually use for this type of application is a voltage controlled mosfet current mirror with current limiting and feedback. This means that even if the injector frys and it goes shortcircuit it won't kill the fets.

btw, distributed LR circuits are generally only used to model large inductors, and if you were doing a distributed network it would have to have the L split also.

Gimmie a few hours to have a play, and a few more to simulate your circuit and I'll report back.