So I am trying to control the valves on a Briggs & Stratton 3.5 HP engine by using electromagnetic solenoids as linear actuators. They are triggered by a signal from the Maple, which goes through an opto isolator, MOSFET driver, and to an NMOS which is protected by two 200V Zener diodes in series with each other (400V effectively) but in anti-parallel with the NMOS. I am thinking about using rocker-roller arms to magnify both the speed the valve travels and the distance it opens. Ex. the solenoid moves 1/20th of an inch in 1 ms, so with an arm that creates a 10:1 ratio, I can get a half inch movement in the same 1 ms. The problem I face is the return spring. I can place it on the solenoid so that force is not also multiplied, but it will still impact performance. Is there an easy way to use a solenoids back EMF to reverse polarize it and cause it to slam shut?
Solenoids
(20 posts) (3 voices)-
gbulmer
ModeratorSilntknight - I can't think of a way to get the back-EMF to do work. It is already doing work resisting the current driving the solenoid. (It feels a bit like perpetual motion otherwise)
Could you operate two solenoids in a push/pull configuration (a bit like the cylinders of a boxer engine)?
(With a bit of ingenuity, you could probably use a lower strength spring too.)Anyway, I am wondering if you could use some large capacitors to store energy generated in the second solenoid, and make the second solenoid act as a damping spring. The only advantage over a mechanical spring would be the flexibility to adjust its effect, and then the ability to use it to drive the movement in the opposite direction with more force.
Just a thought.
GBPosted 4 years ago # -
Silntknight
ModeratorCan you think of a way to configure a push-pull solenoid setup that is controlled by an NMOS? I am thinking one neodymium core with two separate coils. The anode of the coil is at the "center" (area between both coils) and moves outward, so that the magnetic fields they generate are opposite each other. What I need to figure out is how I can use an NMOS to control the circuits. I guess I could use an NMOS and PMOS in a CMOS setup... Each one controls a specific coil. That would actually be pretty cool.
Posted 4 years ago # -
gbulmer
ModeratorThe setup you describe sounds like a single unipolar stepper motor, and as Wikipedia says, commutation only needs to connect half the coil to ground. That can be handled with NMOS FETs only (or NPN bipolar transistors).
If that is okay for you to make, skip the rest of my comments. (It would be more than I've facilities to fabricate.)
I am assuming:
1. a good way to get the solenoid to move back and forth is by applying voltage in either direction, i.e. drive the solenoids forward or backwards with current flowing in either direction.2. a double-throw relay might not be adequate to change the direction of current flow (i.e. it is too slow, or current capacity or arching might be a problem). I am not 'into' relays, so this may be too conservative.
If those two assumptions are appropriate, then there is still a way to drive current in either direction using NMOS FETs only. But, it is a bit more complex electronically than using both NMOS and PMOS.
Before I go on, is this relevant? If it is can you confirm or deny my two assumptions?
(full disclosure: I am not a member of the LeafLabs staff)
Posted 4 years ago # -
Silntknight
ModeratorUnless I can get a logic IC that switches two signals (ex. for input pin Vs, one output is Vs and the other is _Vs (Vs with a line above it, inverted)), then a CMOS is the type of circuit I want. I only have a limited number of signal pins (space constriction) so each solenoid needs to be controlled by a single signal pin. While the stepper motor idea would work, it is unideal.
Assumption 1: I believe you are correct. The only concern I would have is the back emf from polarizing the (essentially) inductor one way, then suddenly reversing it.
Assumption 2: This is correct. A relay is far too slow.
I am willing to get some PMOSs and more Zener diodes or even TVS devices and make a CMOS circuit. I am thinking something like the linked:
http://i1221.photobucket.com/albums/dd470/Hari_Ganti/InductorSetup.jpg
*Side note, if I have a MOSFET driver that can drive the NMOS with 3.3V, is is necessary to have an optoisolator with a 3.3V low side and 12V high side. The outputs go into the MOSFET driver, which then send 12V to the gate. Is the opto really necessary?
Posted 4 years ago # -
gbulmer
ModeratorUnless I can get a logic IC that switches two signals (ex. for input pin Vs, one output is Vs and the other is _Vs (Vs with a line above it, inverted)) ...
I'm not sure I understand. That is an inverter, or am I missing something?
If you don't want to use an ordinary integrated circuit (e.g standard 74xx or 4000 series, or even a 74LVC1GU04DCKTG4), you could make it from two transistors, or two MOSFETs, and a few resistors.While the stepper motor idea would work, it is unideal.
The point about the stepper motor link to wikipedia is it explains how to drive a similar electro-mechanical system using only N-channel MOSFETs, i.e. no need for complementary (N-channel and P-channel) MOSFETs.
I was not suggesting you use a stepper motor.Assumption 1: I believe you are correct. The only concern I would have is the back emf from polarizing the (essentially) inductor one way, then suddenly reversing it.
Why? What is the inductance? Have you measured or characterised it?
Depending on budget, MOSFETs with resistances of a few milliohm are available.I think the optoisolator is a useful insurance policy while building a prototype. There is no isolation between a Maple and the host PC, and you may well want to gather data from the Maple while experimenting.
Posted 4 years ago # -
Silntknight
ModeratorI just realized that I could simply use a quad op amp as an inverter for controlling each pair of NMOSs.
I'm not sure I understand what you mean when you say that I can get low resistance NMOSs when you are talking about the solenoid inductance. I do not have a way to measure it, but I think I found the answer to my question anyway.
Right now I am thinking of two solutions:
1. The drawing I made shows what I am thinking of currently. The MOSFETS would have 400V worth of Zeners anti-parallel to them for back EMF protection. The MOSFET driver goes low when it is off, so they should function as a CMOS.
2. Use only NMOS and an inverter IC. Ironically, I accidentally ordered inverting MOSFET drivers, so this would actually solve many problems. If I do this, can both NMOS in each pair share the same Zener diodes? I'll link a schematic of what I mean.
http://i1221.photobucket.com/albums/dd470/Hari_Ganti/NMOSOnlyInductorSetup-1.jpg
Posted 4 years ago # -
mikep
MemberI think the schematic for the second solution is wrong.
You have basically both Solenoids connected to each other and both MOSFET Drains connected to each other.
You need to have it like this:
MOSFET1:
- Source goes to ground
- Gate goes to TC4426 - Out 1
- Drain goes to Solenoid1MOSFET2:
- Source goes to ground
- Gate goes to output of the inverter (and the inverter input goes to TC4426 - Out 1)
- Drain goes to Solenoid2That also means you need 2 protection diodes, one from Drain1 -> Ground and one from Drain2 -> Ground
Also, in many circuits I am used to, the protection diode is actually placed across the inductive load (solenoid) itself and not across the MOSFET, but the end effect is the same, it limits the voltage that appears on the drain of your MOSFETs.
http://www.softwareforeducation.com/electronics/notes/AS/mosfet/switch.php
BTW, the circuit "Drive a MOSFET from a PIC Chip" in the link above shows how to use an transistor instead of an inverter, to achieve the same effect:
- Maple Pin = HIGH --> Transistor = conducting --> MOSFET = OFF (Solenoid OFF)
- Maple Pin = LOW --> Transistor = not conducting --> MOSFET = ON (Solenoid = ON), since the MOSFET gate is pulled high through the 10k resistor.Some more general info (Coil guns - yes :)):
http://www.coilgun.info/theoryinductors/inductivekickback.htmPosted 4 years ago # -
Silntknight
ModeratorGoing back and looking at it, it is incorrect. I was hoping to not need more Zeners, but adding more is only a minor problem. I actually built a test circuit like what you specified, but I don't have an inverter, so I used an op amp without any gain.
In the final design, I might go with using a transistor. If so, it would be placed where the op amp is currently (and connected like in the diagram, but such that the output is inverted).
http://i1221.photobucket.com/albums/dd470/Hari_Ganti/TestNMOSCircuit.jpg
*I am having a problem with this circuit. I am not sure why, but for some reason LED 1 does not light up when the opto cathode is grounded. I checked the voltage across the drain and source with the LED connected, and it read 4V, so the LED should have turned on. I am not sure why it didn't. LED 2 reacted correctly.
Placing Zeners across the load would be difficult because this is an automotive application, so the fuel injector and solenoid are close to the engine and far from anywhere I'd be placing a PCB. I understand that they are usually placed there, but it would be more difficult to do so.
Posted 4 years ago # -
mikep
MemberWhat is the voltage on "Out_1" in both cases?
Also, what color is LED 1? If you measured 4V across the drain and source of the MOSFET it means it is not fully ON, the gate voltage might not be high enough to fully saturate the MOSFET.
With a 6V supply voltage and 4V across the MOSFET that leaves 2V for the LED + Resistor, and some LEDs (green or blue) need more voltage than that (that's why the question about the color).
In any event, you want to make sure your MOSFETs are fully conducting, not more than 1V across source/drain in the ON state and 6V across source/drain in the OFF state.A couple more suggestions:
1. put a 10k or so resistor from the emitter of the optocoupler to ground so that the TC4426 input is pulled low when the Optocoupler LED is off.
2. you can just connect OUT_1 of the TC4426 to IN_2 of the TC4426 and use OUT_2 to drive the second MOSFET directly. That way you can save the OpAmp and the transistor, since the TC4426 is already an inverter.Posted 4 years ago # -
Silntknight
ModeratorWith the opto cathode connected to +3.3V (positive lead indicated first):
Vce = 6.58V, Vin_1 = 0V, Vout_1 = 6.68V, V(op amp inverting) = 4.12V,
V(op amp out) = 1.62V, Vled_2 = 2.41V (across cathode/anode, lights up), Vled_1 = 0V (across cathode/anode, doesn't light up), Vds(led_2) = .03V, Vds(led_1) = 4.93V.*Note, during the test, LED_1 suddenly lit up. Pushing the MOSFET around causes it to light... It would appear the tabs are electrically "hot" even though the datasheet did not say that, and that when LED_1 lit up, it the MOSFET tabs were touching...
With the opto cathode connected to Gnd (positive lead indicated first):
Vce = .03V, Vin_1 = 6.67V, Vout_1 = 0.03V, V(op amp inverting) = 2.14V,
V(op amp out) = 4.25V, Vled_2 = 0V (across cathode/anode, doesn't light up),
Vled_1 = 1.75V (across cathode/anode, doesn't light up), Vds(led_2) = 4.99V,
Vds(led_1) = 4.92V.The LEDs are green. The MOSFETs are FDP12N50-ND. They have a gate saturation of 3V min, 5V max. Both MOSFETS turned on with a test voltage of 3.3V, so I'm not sure that is the problem.
For the suggestions:
1. I did not show this (I forgot) but I have a 22K pull-down resistor connected to In_1.
2. I need that output for another NMOS pair. I am trying to control both NMOS with a single control signal.I am going to try using a transistor-based inverter instead of the op amp and see if that makes a difference.
Posted 4 years ago # -
gbulmer
ModeratorI'd like to add a couple of extra points to mikep's comments.
1. First, as explained at wikipedia ...
For a center-tapped coil, you don't need complementary (P-channel and N-Channel) MOSFETs, N-Channel is enough.
They are both sinking current to ground, neither is connected to the high-voltage rail, so there is no need for P-Channel MOSFETs.
(mikep was probably reading the circuit this way)2. If you put the output of the TC4426 into a MOSFET gate via an inverter, you will lose many of the benefits of using a TC4426, such as of the current source, protected outputs.
It would be better to use drive the inputs of TC4426's, then use the outputs to directly drive each N-Channel MOSFET gate. The TC4426 inputs could be driven directly and via an inverter (of some form).3. Better yet, I see that a TC4428 has one inverted and one non-inverted output, so it would drive the two MOSFETs in anti-phase, which is what you want, if you can only spare one signal to drive the solenoid coils.
I'd think more about having overlap or a dead time between energising the two coils.
(Edit: Sorry if this seems out of order, I wrote it after mikep's first post.)
Posted 4 years ago # -
Silntknight
Moderator1. I agree. I am thinking of using only NMOS at this point.
2. I was able to make the transistor-based solution work. I am not sure about how much delay I will be adding by taking the output of a MOSFET driver and splitting the signal.
3. The cost of the drivers isn't an issue (shipping will cost more than the parts...), but the problem is space. I'd really rather not redesign the board again. Although, I guess it might be the best solution over all.
If nothing else, I did learn a lot about sanity checks when I was building test circuits.
Posted 4 years ago # -
gbulmer
Moderator... I am not sure about how much delay I will be adding by taking the output of a MOSFET driver and splitting the signal.
The TC4426 and TC4428 both contain two MOSFET gate drivers.
The suggestion is to use each drive output to drive a MOSFET.Further there is no need for an inverter. Using mikep's suggestion of driving TC4426 input 2 from output 1, or using a TC4428, the inverter goes away, so the circuit area is smaller.
As mikep explains, the optoisolator emitter signal should be connected to ground via a resistor.
I think, if you can get another signal in their, and control the energising of the two coils seperately, there would be more flexibility to control the valve movement.
Posted 4 years ago # -
Silntknight
ModeratorFirst, I did have a pull-down resistor connected to the MOSFET driver gate, which I forgot to illustrate for some reason.
Second, I will either take the suggestion of taking out_1 and connecting that to in_2 or just get the TC4428. I was able to redesign the circuit last night to incorporate that kind of change. I am thinking of getting the TC4428, but what are your opinions? I know either will work, but one is likely to be better for some reason.
As for getting more signals, that would add a lot of complexity and require more time than I have. I understand that it is ideal, but this solution should work with the one signal. Thanks.
Posted 4 years ago #
Reply »
You must log in to post.