injectors - [56k beware page 2]

From DIY contraptions to sophisticated FreeEMS-specific designs! Plus general hardware development!
User avatar
Delta
LQFP112 - Up with the play
Posts: 111
Joined: Fri Jul 25, 2008 8:04 pm
Location: Perth, WA, Australia

Re: injectors - [56k beware page 2]

Post by Delta »

jharvey wrote:Have you noticed the discharge times for the spice on pg 3 of this thread? The autoFET drops like a rock, because it drops so quickly, the off time tolerance is tighter, so more accurate control.

I know that freewheel (snubber, fly-back, suppressor, catch, TVS, ect) component is very common. However it really rolls over from the relay days, when there were no other options. I think the autoFET shows newer tech, increasing quality. If the dissipated power for the autoFET is really a fraction of a watt, that would be something. I think there is a good chance it really is that low. Most FET's get hot because this spike can put you in a tranient zone, where you're simply a resistor across the battery. However, this isn't a normal FET, it appears it will make that transient period much shorter then normal.

I'll be eagerly waiting for Fred's reply.
Its new tech but its not magic. Its control circuitry that monitors the voltage across the mosfet and opens the gate if it gets too high...I can build that myself with discrete components, its not new, its just integrated so less external components can be used.

http://www.datasheetcatalog.org/datashe ... Xutyrr.pdf

If you look at the circuit in fig 5 for testing switching times of inductive loads it uses either a fast recovery diode, or another fet to shunt the inductive spike. Sure it will work without, but its not a good idea. Just because they have incorporated a system that prevents bad design from destroying the device, doesn't mean you should then ignore good design.
User avatar
Fred
Moderator
Posts: 15431
Joined: Tue Jan 15, 2008 2:31 pm
Location: Home sweet home!
Contact:

Re: injectors - [56k beware page 2]

Post by Fred »

Delta wrote:I suppose being an electronic engineer I tend to look at the worst case scenario hahha
That's a very good thing :-)

Thanks for the link to that sheet. So, can we safely concluded that two resistors and one fast diode with an autofet is the way to go? what affect does the diode have on closing times for the injector? shunting the spike away at circa 15V means more current holding the injector open longer doesn't it? Can we get fancy to protect the FET (as a precaution and because it's the right thing to do) and use something other than a diode to allow faster closing? Are there any suitable options that we can reference on the web or in a text book to avoid issues with lawyers?

BTW, prioritise your thesis above and before this website. I'd own a couple of houses by now if I'd spent more time in lectures and less drinking beer and learning abut girls ;-)

Fred.

PS, I had a PM from another member that said "I ran a single VNP5N07 for a year or 2 driving 2 hi-Z inj's with no heatsink at all, <snip> never got hot at all." However if it's a hack to use them like that, then we should do it properly, whatever that is :-)
DIYEFI.org - where Open Source means Open Source, and Free means Freedom
FreeEMS.org - the open source engine management system
FreeEMS dev diary and its comments thread and my turbo truck!
n00bs, do NOT PM or email tech questions! Use the forum!
The ever growing list of FreeEMS success stories!
User avatar
Delta
LQFP112 - Up with the play
Posts: 111
Joined: Fri Jul 25, 2008 8:04 pm
Location: Perth, WA, Australia

Re: injectors - [56k beware page 2]

Post by Delta »

Yeah, my thesis takes priority, and yep like I said, with high Z it will be 'ok', but the question is for how long. With low Z i'd assume its not alright, and it would be better if the hardware is future proof. Fast recovery diode is the way to go, it will probably add around 100-150nS to switch off time, but the clamp takes roughly this time also, so its pretty much same same. You might be able to do it quicker with the second mosfet, but I doubt it. Can try looking at the waveform with and without recovery diode on an oscilloscope if your really worried about it??
User avatar
jharvey
1N4001 - Signed up
Posts: 1607
Joined: Tue Jun 10, 2008 5:17 pm

Re: injectors - [56k beware page 2]

Post by jharvey »

Delta wrote:If you look at the circuit in fig 5 for testing switching times of inductive loads it uses either a fast recovery diode, or another fet to shunt the inductive spike.
That's a nice data-sheet. I like the graphs and extra data. I like it so much, I... um... started to hump it :oops: I guess at times my attraction can be for a different type of silicone :roll: Any how I'll back away from the bad humor now.

Fig1 un-clamped inductive load test, shows it as a direct drive. Fig 1 is basically what Fred purposed and I drew up. I recommend adding the current feed back, especially on a first draft. I feel that the $.50 hall sensor is cheap enough, I'm sure the inductor is going to cost less then $.50 as well. You don't have to populate these parts if you don't want that feature. I want them because I feel the system should be a bit more closed loop. Their footprints will be quite small, and I consider PCB cost next to null.

The difference in off times from using a ISC is because your dissipating the heat in the injector with a fixed resistance, so power dissipated decreases as the voltage potential decreases. However the FET's OV protection dissipates a constant power until it's gone. Both ways require proper heat sinking, so I don't think I'd really consider one vs the other good or bad design. The biggest issue that might exist, and could be considered bad design, is the ringing. That can still exist because you have the LRC allowing for a tank circuit. Guess what, here's a picture of spice.

Image

I've removed Rprime and left it with a 10ohm Rconn2. Notice the ringing. That ringing is actually dissipated mostly by Rconn2. I'm currently running a sim for Rconn2 at .1 ohm. It's having some trouble because of the ringing. With out the dissipation of Rconn2, I'm sure it rings all the way through and rings rail to rail. Granted it's V ringing, and it's a small amount of power but I still don't like it. In real life the leakage current might even take care of most of it. Here it is with Rprime 1K.

Image

I can change the values for low impeadance and see what that produces. Looks like 10mH is about right for the inductance. Any one know what the typical resistance value might be? If you have one handy, put it on your DMM and let me know what it claims for ohms. My guess, is that unclamped it's also good for low impeadance.

Engineer you say, so you've also referred to a horse as a sphere to simplify math? BS in Electro-Mechanical and AS in Mechanical here.
User avatar
jharvey
1N4001 - Signed up
Posts: 1607
Joined: Tue Jun 10, 2008 5:17 pm

Re: injectors - [56k beware page 2]

Post by jharvey »

Looks like my first post in this forum included the same link twice. I edited one of those links. It now links to the measurements this guy took of a high and low impedance injector.

Interesting stuff. Notice the needle bounce and variations in inductance based on needle movement.
User avatar
Delta
LQFP112 - Up with the play
Posts: 111
Joined: Fri Jul 25, 2008 8:04 pm
Location: Perth, WA, Australia

Re: injectors - [56k beware page 2]

Post by Delta »

Low Z injectors are between 2.8 and 4 ohms. Generally for feedback I'd use a 0.1ohm resistor between Vs of the fet and ground. It tracks the current exactly, and you can then use an opamp with 10x gain to get amps in volts. I'm assuming even with peak and hold we will only see 4A and the logic on the micro is 5V? if its 3.3V then perhaps 5x gain?
User avatar
jharvey
1N4001 - Signed up
Posts: 1607
Joined: Tue Jun 10, 2008 5:17 pm

Re: injectors - [56k beware page 2]

Post by jharvey »

Why type when you can draw?

Image
User avatar
jharvey
1N4001 - Signed up
Posts: 1607
Joined: Tue Jun 10, 2008 5:17 pm

Re: injectors - [56k beware page 2]

Post by jharvey »

Hey look at the power dissipated by the FET for low impedance. I'll round the V drop up to 4V and the current peak up to 5A for 15ms. So power draw for that 15ms is about 1.6 watts, and with a 100ms period it's about .24 watts average. Not as low as .07, but still quite low. Low enough that perhaps heat sinking isn't exactly required.
User avatar
Fred
Moderator
Posts: 15431
Joined: Tue Jan 15, 2008 2:31 pm
Location: Home sweet home!
Contact:

Re: injectors - [56k beware page 2]

Post by Fred »

Why do you guys *want* to know what current there is flowing? Once the initial spike is over it will be resistance vs voltage vs current as always and that can easily be determined before hand.

The xdp can run in either 3.3 or 5v mode. It's running in 5v mode for us though.

What will you gain by knowing what current flow there is on an injector?

You are either going to need a multiplexer or X ADC channels for X injectors to measure and know this. OEMs do stuff like this for all sorts of LONG term reliability and emissions reasons over the LONG term too. I just don't see the practical benefit in having it. How could you utilise that information? Who will write that "icing" code?

"If in doubt, leave it out!"

Seems like a good way to go.

Having some stated goals for this particular board design might be a goer too.

Something like :
  • Reliable
  • Low noise by design
  • Lowest component count possible
  • Drives a single high Z injector per FET only (supports 4 5 or 6)
  • Doesn't include any ignition drivers
  • Does include to92 or similar FETs/BJTs to drive external ignitors safely (supports 4 - 12)
  • Does include as many as possible spare FETs for driving accessories (to92 and t0220) (at least one to220 for the fuel pump)
  • X by Y physical dimensions
  • No feedback from outputs (assume outputs work as designed, tested and installed)(
  • Uses TA XDP card
etc.

Obviously that's my list (or some of it), and yours may be and is welcome to differ, but it would be good to know what the target is so the discussions aren't for no point while one person thinks one thing and the other thinks something else if you see what I mean.

So, if we assume the "one inj per fet" and "lowest comp count" and "no feedback" and "high Z only", what do we need? 3x resistors, 1x cap, 1x fast diode, 1x autoFET ?

If you add in "low Z" then you need a driver chip on the card too.

Thoughts?

Fred.
DIYEFI.org - where Open Source means Open Source, and Free means Freedom
FreeEMS.org - the open source engine management system
FreeEMS dev diary and its comments thread and my turbo truck!
n00bs, do NOT PM or email tech questions! Use the forum!
The ever growing list of FreeEMS success stories!
User avatar
Delta
LQFP112 - Up with the play
Posts: 111
Joined: Fri Jul 25, 2008 8:04 pm
Location: Perth, WA, Australia

Re: injectors - [56k beware page 2]

Post by Delta »

If those are the goals then yep thats exactly what you need, no more no less. I only usually have current sense in these circuits because I want to determine if a fault has occurred - but I believe some of the protected fets pull gate to ground if short circuit/overheat, so you can set pin, read pin....if its not high, then its a fault. The other reason I've had a feedback resistor is if I'm using the driver circuit in triode mode to drive gearbox solenoids. Since battery voltage changes and to keep shift consistent you need consistent line pressure during gear change as controlled by a solenoid I needed to know that the duty cycle I set was actually achieving the correct current through the solenoid. In this case its not needed.

I'd be using IGBT's for ignition drive. Mosfets for on/off/pwm channels. Depending on cost, I'd use these new protected fets for all your general outputs + injectors, as they won't die if someone stuffs up their wiring.

The other thing is you want to be able to use left over channels for general outputs...or perhaps build a stackable system, ie drivers come in sets of 4 and can be used for ANY on/off/pwm output, so any channel can be assigned any job. So you have a 4cyl and want minimal setup with sequential inj + ign, you buy 1x4driver board 1x 4 ign board. If you have 12cyl and want drivers for ISC, Boost control, water injection, nos, pump control + seq inj + ign - then you get 3x 4 ign board, 5x 4 on/off/pwm board etc.

Or am I mad??

Just an idea I suppose.
Post Reply