Injector Control Options

[AbeFM]

Uh huh, I don't see much in the way of fast switching. I did see some pulsing on the +12V line, but it was more around 1 kHz than ten times it. In fact, I didn't really understand that.

I'm not sure the power feed wires are good enough, I should probably run power and ground from somewhere other than Jean's signal interceptor board (but MAN isn't it handy?), but I'm not seeing stuff drag down except during the 'ramping up' part of the valve opening.

Also, as I said before, a lot of voltages aren't going to zero where I'd expect.

Fred, when you said hold currents of "8/2" what did you mean? 4:1 open to hold? 8 amps open, 2 hold?

Anyway, 6-7 amps is a LOT for a hold current.

[jharvey]

I just took a minute and searched linear.com for "current regulator switching" and found this chip

LTC3810-5

Does 1-6 amps input voltage up to 60V and about 95% efficient. I would suspect this would reduce the dissipated power quite a bit, and better yet it has a data-sheet telling us how to use it.

[Delta]

6-7amps is way too high for hold current - the design should only go to 4/1.2 for peak and hold respectively. Either the resistor values calculated and tested in spice are wrong (the ones around the dual mosfet) or perhaps having only a 0.01 ohm resistor means the voltage levels are too low at the comparator inputs. I'm concerned a little by the continuing voltage after turn off - does the injector continue ot dribble or is it perhaps thermal noise from the hot resistor?

[jharvey]

Can you measure the resistance of the .01R? Likely done by measuring both the input current and voltage drop across the resistor. Also remember the typical meter is around 5% to 10% accurate anyhow. If you have the scope, that's usually below 5% with a good eye.

The current feedback is high gain, and is very susceptible to small variations. I've always had to either use very expensive laser trimmed resistors, or had to manually trim them at assembly. There are many tolerance build up issues the op-amp gain varies as well as the FET gain. The only one you can adjust is the resistor. This is one of the reasons why I'm not a huge fan of the sensing resistor, it's requires manual tuning, and no two people will tune it the same.

Delta can you do a quick check of the variations if you changed the resistor +/- 1% and +/- 10%?

[Delta]

Can you measure the resistance of the .01R? Likely done by measuring both the input current and voltage drop across the resistor. Also remember the typical meter is around 5% to 10% accurate anyhow. If you have the scope, that's usually below 5% with a good eye.

The current feedback is high gain, and is very susceptible to small variations. I've always had to either use very expensive laser trimmed resistors, or had to manually trim them at assembly. There are many tolerance build up issues the op-amp gain varies as well as the FET gain. The only one you can adjust is the resistor. This is one of the reasons why I'm not a huge fan of the sensing resistor, it's requires manual tuning, and no two people will tune it the same.

Delta can you do a quick check of the variations if you changed the resistor +/- 1% and +/- 10%?

I chose those values (0.01 and 100 in combination with the ones around the twin fets) because even if it was around 10% out it would only move hold current to around 1.5A max and 1A min. The simulations confirm that. I think the problem here is either:

A) The resistor values around the N/P Mosfet pair are wrong for what ever reason.

or more likely

B) The sense resistor is too small and hence the voltages we're working with are too small at the input of the comparator (there is always a small offset voltage at the differential input). Looking at the datasheet for the LM19xx (god I wish i could remember all these numbers of the top of my head) they use a 0.1ohm resistor - and I'd make it 5W instead of the 2W version they use - this means at the bottom of the N/P pair we need a 1k resistor instead of the 100 as well. I'd say what is happening is the noise in the circuit is either holding the feedback to low or pumping up the input at the + of the comparator as we are only dealing with 10's of mV. If we change to 0.1 and 1k we are dealing with 100's of mV which is much better.

The gain of the comparator really doesn't have much to do with anything in this case - the + input is attached to a function that turns on for a certain time at a level defined by the resistors then drops to another level. So with the 100ohm its 40mV then 10mV - i9f the input at the - terminal is below this then the difference is amplified (yes by the gain) Only a few volts are needed to turn the mosfet on - so the gain only has to be around 1000 max for it to function properly - which is very low. Once the mosfet turns on its possible to draw lots of current, so it easily possible to get the required offset voltage at the input - say its 2-3mv so you need 43mV at peak and 13mv at hold - to start turning it off. So gain is not the problem at all - unless you think I'm missing something glaring (share please).

Its much more likely that like I stated earlier - we are dealing with 10mV at the inputs, which is easily within the range of noise in a circuit like this. I should really have thought about this earlier.

Is it possible to try with a 5W 0.1ohm under the main MOSFET and 1k 1% under the N/P pair. If that generates a weird peak/hold current then its time to think about the resistors around the N/P pair.

[AbeFM]

Just comparing to my favorite chip:

LM324N:
Input Offset Voltage: 3mV (9 max)
Input Offset Current: 50nA

TL082C:
Input Offset Voltage: 3mV (13 max)
Input Offset Current: 100pA

No, no smoking guns there. Certainly it's hard to get a great number out with so little voltage. Well, I know we're getting 60 mV across the 0.01 ohm, so is taking 10% out of it (6mV) a big deal?

Anyway, while it could be componant variation (I seem to remember using 1220 ohms where I was supposed to have 1.2k?), I don't think so. I have a nice four wire measurer at work I could test the 0.01 on, no problem...

[AbeFM]

Well, I have:
4 0.25ohm (5W)
4 1omh (25W)
2 0.001ohm (25A)

And a few 10 ohm high precision, low power resistors, and a 4.7ohm as well. Oh, and I think a 250 watt 1 ohm and a 350 watt 10 ohm or some such sillyness.

The question - could I use either flavor of 250 mohm resistors, just use a 1 ohm, or?

I think at home I have some 0.050's as well, left over from the MS build.

Anything I should try with what I've got?

[jbelanger]

Delta is suggesting using 0.1 and 1K instead of 0.01 and 100 so you could try with your 0.05 and 500 (or 510 since 500 isn't standard).

Jean

[AbeFM]

Mainly what I was worried about - is 510 better than 470? Those I have a lot of. And I'll just measure them till I find one <1%...

[jharvey]

Hmmm, how long of a fone wire do you need for .01? From http://www.powerstream.com/Wire_Size.htm I get 1000ft 22awg at 52.9R, so isn't that about 2.2 inches? The table claims it's good for max 7 amps. Perhaps your phone could be the next resistor?

Yup, that's for the freaks out there.