Help diagnose scope signals

Thread Starter

DC_Kid

Joined Feb 25, 2008
1,072
looking for some feedback to what i am seeing on my scope. let me set the scene.

its a driver circuit for driving fuel injectors. i made a bunch of dummy injectors which match exactly that of a real injector, yields 14.4ohms(dc) and 5.6mH inductance (two 27ohm wirewounds in parallel, then in series with the inductor). this driver gets its drive signal from a external source and will conduct current when the signal goes to ground. some parts from the scope was exactly what i was expecting, but other parts i am not yet sure of and will require me to do further investigating, but just wanted some feedback 1st to help steer my investigation. in this test i drive the circuit using my freq gen 60Hz @ 50% duty. voltage supply is 14.4v to ground. my signal drives a small npn and i tap off the collector to get the ground signal i need. this npn turns on/off very well into 300kHz range, so i dont see this causing any switching issues..... my application only runs from 0-60Hz. the pics stay the same regardless of drive frequency (other than their period of on/off, etc).

(click pics for bigger pic)
this 1st pic shows the switching. red is the signal driving my circuit. yellow is the wire coming from the fuel injector to the fet. fuel injector always has supply voltage on it on one side and i simply switch a nFet of the other side to control the current. as red goes low the fet conducts and current flows. when red goes high things turn off and i get a flyback spike clipped by 34v zeners.


this next pic shows some timing events
t1= 0.22ms, i dont know why the injector is still conducting current (or is it??). this may be because of the 1st drive stage i have using a pFET. the pFET may not turn off as fast as it turns on. as you can see in the above pic things turn on real fast.
t2= 0.116ms, simply looks like the magnetic field starting to collapse and voltage is generated until it reaches zener voltage.
t3= 0.2ms, looks like the time it takes for inductor mag field to decay.
t4= 1.54ms, a period of some ringing. not really sure what exactly is ringing down to supply voltage.


here's the ring up close. rings at 30.3kHz. hmmm, osciallation, but from where? perhaps some C in the inductor? not really sure.


 
Last edited:

SgtWookie

Joined Jul 17, 2007
22,230
What are you using for the P-ch MOSFET? And how are you driving the gate? You really need to move a good bit of current for the gates of some power MOSFETs.

Looks like you need some snubbers to take care of the ringing.
 

Thread Starter

DC_Kid

Joined Feb 25, 2008
1,072
1st drive stage is a pair of Vishay BS250KL pFET's in parallel. these basically "flip" the ground signal to power the gates of nFET's. the nFET's are a pair of IRF1404's in parallel. i have pairs for redundancy purposes.

there are 6 channels of load being driven. the pics are of a single load. the loads will be driven from 0-80% duty cycle. 100% duty cycle yields 1amp per channel. all channels are driven in parallel.

what would you suggest for snubber? 1000pf ceramic?
 

jpanhalt

Joined Jan 18, 2008
11,087
A little bit of knowledge is dangerous. To prove that point, let me pass along the "cookbook" procedure from Marty Brown's book on power supplies (p. 146). His example is to design a series RC snubber for a low-side mosfet and an inductive load.

1) Measure the oscillation period of your ringing in the un-snubbed circuit.
2) Put a small capacitor across the load. Choose a capacitor that increases the period to three times the original period. Another source suggested a starting test value of about 2 to 3X the mosfet gate capacitance.
3) The resistor, R, can then be calculated as: R= 1/(2*pi*F*C); where F is the original ringing frequency and C is the capacitor value found in step 2.

As stated initially, this procedure is from Marty Brown's book. I throw it out as something recommended by an expert and for discussion.

John
 

Thread Starter

DC_Kid

Joined Feb 25, 2008
1,072
my bad, was late yesterday and i was getting sleepy.

so yes, i understand that a RC will damp that ringing, but i'm more interested in knowing where the "RLC" oscillator is and why it starts after zener clamping. the "RL" piece i suspect is the dummy injector i have, but where is the "C"? i'll assume the "C" is a net "C" from the nFET's and zener diodes after they stop anvalanching. does this sound about right?

and when you say "across the load" is that:

+v---------load--------FET---------gnd
...........|___cap___|

or

+v---------load--------FET---------gnd
.........................|
.........................|________cap____gnd


i was looking at some other snubber docs out there. one was http://www.st.com/stonline/products/literature/an/6785.pdf
 
Last edited:

nomurphy

Joined Aug 8, 2005
567
First off, you need to get as close to reality as possible. That means getting rid of the long scope wires and scope grounds. Place the scope tip directly on the component and use a GND wire that is an inch or less to a nearby ground.

You must make sure you have sufficient wire gauge and trace thickness to handle the currents. Also, adding a ferrite bead in series with the output can help.

yields 14.4ohms(dc) and 5.6mH inductance
Is it correct that the inductance is 5.6mH and not 5.6uH? Driving 5.6mH is quite a bit, and would explain excessive ringing. Snubbers may be necessary, but tend to waste power. You could try a reverse diode across the inductor. Other versions would be the RC snubber already mentioned, or a RCD snubber (diode/cap/res) that is often used with transformer primaries in switching supplies. Do not use only a capacitor as a snubber.

Hitting the gate hard enough to overcome gate capacitance and turn the FET on/off quickly is also important. Of the FET's that will work within the design's parameters, choose the one with the lowest gate capacitance.
 
Last edited:

Thread Starter

DC_Kid

Joined Feb 25, 2008
1,072
it is mH, not uH. fuel injectors have large inductance. the scope probes are, in this test, as close as they can get to the signals being watched. everything is grounded to a common ground rod on the board. my dummy injector, just like a real injector, are iron core inductors. a real injector has the steel needle going through the coil core. reality is probing on the wires since the device will be connected to things via "long" wires, etc. board traces are sufficiently sized to handle the currents. at minimum, the trace width (1oz copper) for each channel carrying 1amp for each injector is 0.25" wide, and as they merge together leading into the fet's the copper becomes more like a fill area, etc.

diode across the inductor (injector) is not a good idea for fuel injectors. letting the field die too fast causes the injector pintle to slam close too fast which causes pintle bouncing and damage over time. selecting zener of proper voltage helps tailor the closing action of the injector pintle, and 34v is most common for the high Z injectors, and it protects the switching power semiconductor (fet in this case) from the flyback voltage. zener from inductor output to ground, etc.
 
Last edited:

SgtWookie

Joined Jul 17, 2007
22,230
Take a look at the total gate charge for the IRF1404s. Up to 200nC. You're using two in parallel, so double it. That'll take a lot of current if you want them to "snap" on and off.

Look at the capacitance numbers. Output effective capacitance is 1540pF. Double it, since you have two in parallel. However, I suspect it's even higher than that - look at the numbers:
Vgs=0, Vds=32, Coss=1490pF
Vgs=0, Vds=25, Coss=1680pF
Vgs=0, Vds=1, Coss=6630pF
Notice how with a lower Vds, C is increased?
Anyway, if the only inductance in your circuit is 5.6mH (which it isn't, because of your wire runs, circuit board traces, etc) then your total Coss is around 5900pF (2*2950pF). But 5900pF is a good starting point. Try a 10nF cap; that should get you around where you want to be.

But really, consider removing one of those 1404's before proceeding. You're at maximum overkill for Id at the moment, but your gate and output capacitance are becoming problematic.

If you really want to run two N-ch MOSFETS in parallel, try swapping in a couple of IRFZ24s for the IRF1404s. Each will carry 17A, and has 1/10 the gate charge and output capacitance of the IRF1404. Rds(on) is certainly higher than the 1404's, but it's quite small in comparison to your 14.4 Ohm load (<0.5%) and your gate drivers will be happy about their lightened load.
 
Last edited:

jpanhalt

Joined Jan 18, 2008
11,087
You will see snubbers across the mosfet and load, across just the mosfet, and across just the load. Brown's specific example for the RC snubber is an inductive load, and he shows the snubber just across the load (the first option you drew). Later in the text, he shows other configurations.

Here is another very good discussion, but not cookbook: http://www.cde.com/tech/design.pdf The first example in that link (Figure 4) is just across the switch. Other examples in that link are across the load and switch (mosfet). The discussion includes much more theory too.

While I like empirical solutions (e.g., Brown's book), one of the great advantages of this forum are the expert members who can speak from experience and theory. I am not in that group and simply hoped my post would help get the discussion going.

John
 

nomurphy

Joined Aug 8, 2005
567
in this test, as close as they can get to the signals being watched.
I think you misunderstand, I'm not talking about "long" connecting wires (which is why you should add a ferrite bead), I'm talking about the probe wires and ground. In these instances, long ground wires (> 1 inch) to a scope probe easily cause a misrepresentation of the signals being probed. What you see on the scope as far as ringing amplitude and frequency are not real, they are exacerbated by the probe wires.

For example, with fast clock signals, even the input impedance of the scope probe makes a difference and one may need a FET probe to get relatively accurate signal info. Because, what may look like ringing or over/undershoot on a clock line can actually be due to poor implemetation of the probe and its use.

If you want to diagnose the problem properly, it needs to be represented properly, which means placing the scope tip (by hand if necessary) at the very point of interest and using a very short probe GND wire.

Attached are a couple design guidelines for snubbers. Although the guidelines are for flyback designs, you should be able to pick up a few pointers.
 

Attachments

Caveman

Joined Apr 15, 2008
471
This ringing is common in circuits of this type. You are basically creating a boost type switching circuit. It is caused by a combination of all the capacitances at the junction. These include FET transistor capacitance, inductor interwinding capacitance, and trace capacitance.

Since it is after the energy in the inductor has been absorbed by the zener, it likely doesn't have much energy in the ringing, ie. it won't radiate much. You should still snub it to reduce noise, but it is likely not a major problem.

IGNORE THIS IF YOU DON'T KNOW WHAT YOU ARE DOING
If you are experience enough to know how not to hurt yourself, touching things can give hints as to where to place the snubber capacitors.
 

Thread Starter

DC_Kid

Joined Feb 25, 2008
1,072
Take a look at the total gate charge for the IRF1404s. Up to 200nC. You're using two in parallel, so double it. That'll take a lot of current if you want them to "snap" on and off.

If you really want to run two N-ch MOSFETS in parallel, try swapping in a couple of IRFZ24s for the IRF1404s. Each will carry 17A, and has 1/10 the gate charge and output capacitance of the IRF1404. Rds(on) is certainly higher than the 1404's, but it's quite small in comparison to your 14.4 Ohm load (<0.5%) and your gate drivers will be happy about their lightened load.
thanks so much for this info. z24's are also half the price. i will do a tad of research on these z24's (and others) and then swap out the 1404's with a alternative. i'll save the 1404's for a much higher current module i am working on. seems that the 1404 has 10x less Rds but 10x more C than the z24's. is this a typical relationship between Rds and Ciss in mosfet world ??

as for on/off, switching "on" looks like it is very very fast. its the small t1 delay that looks like a switching "off" problem. i need to probe the pFET's along with the switching signal to see if they are the problem, etc.

i'll take all this good info, combine it, make some changes, and voila, a new and improved device.
thanks again to all.
 
Last edited:

Thread Starter

DC_Kid

Joined Feb 25, 2008
1,072
in the Flyback Design Notes pdf posted by nomurphy, page 2, what do they mean by "snubber connected to top of sense resistor, not to ground"? i do use a sense resistor, but 0.05 ohms.

my circuit looks like this:

+v-----inductor-----sense resitor------fet------gnd

so the snubber should go (doesnt make sense to me)

+v-----inductor-----sense resistor------fet------gnd
........................|____--cap--R--___|

why not
+v-----inductor-----sense resistor------fet------gnd
................................................|__--cap--R---gnd
 

Thread Starter

DC_Kid

Joined Feb 25, 2008
1,072
It means, place the snubber across the FET D-S junction. Typically, the sense resistor is between the Source and ground.
hmmm, in my circuit i placed sense resistor between load and drain. i did it this way so that i could use a pair of fet's in parallel to handle all the load from all 6 channels at the same time vs. having a single fet for each channel. i need (wanted) redundancy in the fet switching, so doing 2 per channel added more parts to the board, etc. i may rethink my design, possibly using 2 less robust fets per channel and placing sense resistor between source and gnd.

i'll try to fix my op-amp issue that i have with my current pcb. if that turns out to be pita i'll redesign it.

thanks
 

SgtWookie

Joined Jul 17, 2007
22,230
[late post, but some of it is still relevant]
Seems to me that your sense resistor needs to be between your MOSFET and ground.

That's how it's done with H-bridges. Look up the datasheet for an L298 or L297.

It would eliminate the problem of having severe out-of-range voltage levels across Rsense when your FET stops conduction, and the inductive load starts ringing.

Then you can put your snubbers between the FET output and ground. If you stick with dual IRF1404's, you'll need in the vicinity of 7.5-10nF for a cap, then calculate your resistor from there.

If you go with dual IRFZ24's, you'll probably need a cap 1/10 that size.
 

Thread Starter

DC_Kid

Joined Feb 25, 2008
1,072
yes, looks like a redesign in required. the changes will require me to use two fets per channel, but they can be smaller fet's, more on the 2-5amp range as opposed to the hefty Id that 1404's can support.

as for snubber. with sense resistor between fet source and ground the snubber should be placed across the fet drain-to-source (well, according to that pdf, post #10, etc).
 

SgtWookie

Joined Jul 17, 2007
22,230
If the snubber is connected drain to source, you'll cause Rsense to be subjected to the brunt of the snubbing! This will cause your instrumentation amplifier inputs to swing wildly.

But if the snubber is between drain and ground, you may see a small amount of swinging due to the capacitive coupling between the in and out (D vs S) of the MOSFET, but nothing like you would the other way around.

As the Id rating of the MOSFET decreases, the resistance will increase. You probably would want to stay with a Rds(on) that's pretty low, perhaps under 2% of Rload. That's one of the reasons why I suggest the IRFZ24, as it's around 0.4% of Rload - besides having 1/10 the capacitance, inexpensive, and widely available. Having this low Rds(on) will help a great deal with keeping things cool.

As the IRFZ24 has been around for a while, I'm sure there are other alternatives that would work as well or better, and perhaps even be less expensive. I don't see it being obsolesced for a few years yet, though.
 
Last edited:

Thread Starter

DC_Kid

Joined Feb 25, 2008
1,072
what's the consensus to using just a single fet per channel to switch the load on/off vs. using a pair in parallel for redundancy purposes? i know semiconductors are "robust" devices when implemented within their design parameters, but, i like redundancy.... but do i really need it?
 

jpanhalt

Joined Jan 18, 2008
11,087
What's the most common failure mode? Can they fail in another mode?

I vote for single. I am not sure redundancy helps, if you blow a gate to be short-circuit to D or S.

John
 
Top