Prev: IBIS
Next: book on synchros/resolvers
From: Johnny5 on 19 Apr 2010 19:57
I am trying to manage large inductive transients from an ignition coil
being discharged via igbt. Gate driven via mcu. Transients are huge
but duration is minimal, using bat54's with HIGH series limiters and
SMBJ5's everywhere. LM2940 regulator with rf caps on the front end. I
am not sure if it is -.3V is punching it or not.

SP724AHTG scr array vs CDSOT23-0502B tvs. Any comments would be
greatly appreciated.

Thanks/
From: miso on 20 Apr 2010 19:27
On Apr 20, 5:43 am, Johnny5 <dirtylogicdesi...(a)gmail.com> wrote:
> On Apr 20, 1:05 am, "m...(a)sushi.com" <m...(a)sushi.com> wrote:
>
>
>
> > On Apr 19, 4:57 pm, Johnny5 <dirtylogicdesi...(a)gmail.com> wrote:
>
> > > I am trying to manage large inductive transients from an ignition coil
> > > being discharged via igbt. Gate driven via mcu. Transients are huge
> > > but duration is minimal, using bat54's with HIGH series limiters and
> > > SMBJ5's everywhere. LM2940 regulator with rf caps on the front end. I
> > > am not sure if it is -.3V is punching it or not.
>
> > > SP724AHTG scr array vs CDSOT23-0502B tvs.  Any comments would be
> > > greatly appreciated.
>
> > > Thanks/
>
> > You say so much while saying so little. ;-) Find the junction you are
> > forward biasing, and then stop doing that.
>
> > Since you didn't even provide a schematic, all I can guess is the IGBT
> > turns off, the collector flies like a bat out of hell, and cap couples
> > a signal into the uP. I really can't see why a resistor isn't helping
> > under those circumstances since it should keep the chip beneath the
> > trigger current if the resistor is large enough. Now I'm assuming you
> > are clamping to the rails with those schotky diodes, but clamping to
> > the positive rail won't do much good since a regulator with a pass
> > device simply gives way when pulled above the target voltage. That is,
> > LDO only source, not sink. So my guess is the cap coupling is pulling
> > the LDO output high, and the schokty is going along for the ride.
> > Perhaps a transient absorbing diode at the LDO output will help. Note
> > it would be the cap at the LDO output that is critical here if the
> > problem is due to cap coupling from the flying node.
>
> > You could test the uP latchup levels. It's quite easy to do. Inject
> > current until it latches. You need to put a resistor across the supply
> > so that it will maintain the supply voltage as you forward bias diode
> > to the positive rail. If you want to test latch up to 200ma, then you
> > need to insure this resistor is pulling that much current with some
> > margin. It is quite possible the part you are using is prone to latch
> > up. I can't imaging any uP these days not being on epi, but who knows.
>
> Thanks for your input, the igbt is clamped internally. The mcu is a
> 10f200.

Again, there is no circuit provided, but a flying node generally needs
to fly for the circuit to work. A flying node will couple current if
there is a cap path. I don't see how a clamp in the IGBT helps since
it is the uP that is latching, thus it is a node on the uP that needs
clamping.

What did you measure for the latch-up on the chip? Microchip doesn't
have a corporate spec for latch-up limits on the website (in my brief
search), but typically you expect a digital part to exceed 100ma,
usually way more. Analog parts might be as poor as 40ma. These are
room temp specs. Get the part warm, beta increases, and it is easier
to latch.

Who knows, perhaps the part has lousy latch-up.
From: Johnny5 on 21 Apr 2010 11:22
On Apr 20, 6:27 pm, "m...(a)sushi.com" <m...(a)sushi.com> wrote:
> On Apr 20, 5:43 am, Johnny5 <dirtylogicdesi...(a)gmail.com> wrote:
>
>
>
> > On Apr 20, 1:05 am, "m...(a)sushi.com" <m...(a)sushi.com> wrote:
>
> > > On Apr 19, 4:57 pm, Johnny5 <dirtylogicdesi...(a)gmail.com> wrote:
>
> > > > I am trying to manage large inductive transients from an ignition coil
> > > > being discharged via igbt. Gate driven via mcu. Transients are huge
> > > > but duration is minimal, using bat54's with HIGH series limiters and
> > > > SMBJ5's everywhere. LM2940 regulator with rf caps on the front end. I
> > > > am not sure if it is -.3V is punching it or not.
>
> > > > SP724AHTG scr array vs CDSOT23-0502B tvs.  Any comments would be
> > > > greatly appreciated.
>
> > > > Thanks/
>
> > > You say so much while saying so little. ;-) Find the junction you are
> > > forward biasing, and then stop doing that.
>
> > > Since you didn't even provide a schematic, all I can guess is the IGBT
> > > turns off, the collector flies like a bat out of hell, and cap couples
> > > a signal into the uP. I really can't see why a resistor isn't helping
> > > under those circumstances since it should keep the chip beneath the
> > > trigger current if the resistor is large enough. Now I'm assuming you
> > > are clamping to the rails with those schotky diodes, but clamping to
> > > the positive rail won't do much good since a regulator with a pass
> > > device simply gives way when pulled above the target voltage. That is,
> > > LDO only source, not sink. So my guess is the cap coupling is pulling
> > > the LDO output high, and the schokty is going along for the ride.
> > > Perhaps a transient absorbing diode at the LDO output will help. Note
> > > it would be the cap at the LDO output that is critical here if the
> > > problem is due to cap coupling from the flying node.
>
> > > You could test the uP latchup levels. It's quite easy to do. Inject
> > > current until it latches. You need to put a resistor across the supply
> > > so that it will maintain the supply voltage as you forward bias diode
> > > to the positive rail. If you want to test latch up to 200ma, then you
> > > need to insure this resistor is pulling that much current with some
> > > margin. It is quite possible the part you are using is prone to latch
> > > up. I can't imaging any uP these days not being on epi, but who knows..
>
> > Thanks for your input, the igbt is clamped internally. The mcu is a
> > 10f200.
>
> Again, there is no circuit provided, but a flying node generally needs
> to fly for the circuit to work. A flying node will couple current if
> there is a cap path. I don't see how a clamp in the IGBT helps since
> it is the uP that is latching, thus it is a node on the uP that needs
> clamping.
>
> What did you measure for the latch-up on the chip? Microchip doesn't
> have a corporate spec for latch-up limits on the website (in my brief
> search), but typically you expect a digital part to exceed 100ma,
> usually way more. Analog parts might be as poor as 40ma. These are
> room temp specs. Get the part warm, beta increases, and it is easier
> to latch.
>
> Who knows, perhaps the part has lousy latch-up.

Thank you for your help again,
I am not familiar with the term "flying node" or its methodology.
Could you elaborate?
It makes sense, I have yet to find an app note for the implementation.
My google searches have given me nothing.
I will be measuring currents asap.

This was the only doc found from microchip.

AN595
Title: Improving the Susceptibility of an Application to ESD-Induced
Latch-up

Thanks
From: Johnny5 on 21 Apr 2010 11:25
On Apr 20, 11:58 pm, ehsjr <eh...(a)nospamverizon.net> wrote:
> Johnny5 wrote:
> > I am trying to manage large inductive transients from an ignition coil
> > being discharged via igbt. Gate driven via mcu. Transients are huge
> > but duration is minimal, using bat54's with HIGH series limiters and
> > SMBJ5's everywhere. LM2940 regulator with rf caps on the front end. I
> > am not sure if it is -.3V is punching it or not.
>
> > SP724AHTG scr array vs CDSOT23-0502B tvs.  Any comments would be
> > greatly appreciated.
>
> > Thanks/
>
> Can you opto isolate?
>
> Ed


Next move is to scrap it and go opto but I am very limited on space.
SOP-4 will be the max real estate we can use. Height off the pwb is
an issue as well.
Thanks /
From: miso on 21 Apr 2010 18:31
On Apr 21, 8:22 am, Johnny5 <dirtylogicdesi...(a)gmail.com> wrote:
> On Apr 20, 6:27 pm, "m...(a)sushi.com" <m...(a)sushi.com> wrote:
>
>
>
> > On Apr 20, 5:43 am, Johnny5 <dirtylogicdesi...(a)gmail.com> wrote:
>
> > > On Apr 20, 1:05 am, "m...(a)sushi.com" <m...(a)sushi.com> wrote:
>
> > > > On Apr 19, 4:57 pm, Johnny5 <dirtylogicdesi...(a)gmail.com> wrote:
>
> > > > > I am trying to manage large inductive transients from an ignition coil
> > > > > being discharged via igbt. Gate driven via mcu. Transients are huge
> > > > > but duration is minimal, using bat54's with HIGH series limiters and
> > > > > SMBJ5's everywhere. LM2940 regulator with rf caps on the front end. I
> > > > > am not sure if it is -.3V is punching it or not.
>
> > > > > SP724AHTG scr array vs CDSOT23-0502B tvs.  Any comments would be
> > > > > greatly appreciated.
>
> > > > > Thanks/
>
> > > > You say so much while saying so little. ;-) Find the junction you are
> > > > forward biasing, and then stop doing that.
>
> > > > Since you didn't even provide a schematic, all I can guess is the IGBT
> > > > turns off, the collector flies like a bat out of hell, and cap couples
> > > > a signal into the uP. I really can't see why a resistor isn't helping
> > > > under those circumstances since it should keep the chip beneath the
> > > > trigger current if the resistor is large enough. Now I'm assuming you
> > > > are clamping to the rails with those schotky diodes, but clamping to
> > > > the positive rail won't do much good since a regulator with a pass
> > > > device simply gives way when pulled above the target voltage. That is,
> > > > LDO only source, not sink. So my guess is the cap coupling is pulling
> > > > the LDO output high, and the schokty is going along for the ride.
> > > > Perhaps a transient absorbing diode at the LDO output will help. Note
> > > > it would be the cap at the LDO output that is critical here if the
> > > > problem is due to cap coupling from the flying node.
>
> > > > You could test the uP latchup levels. It's quite easy to do. Inject
> > > > current until it latches. You need to put a resistor across the supply
> > > > so that it will maintain the supply voltage as you forward bias diode
> > > > to the positive rail. If you want to test latch up to 200ma, then you
> > > > need to insure this resistor is pulling that much current with some
> > > > margin. It is quite possible the part you are using is prone to latch
> > > > up. I can't imaging any uP these days not being on epi, but who knows.
>
> > > Thanks for your input, the igbt is clamped internally. The mcu is a
> > > 10f200.
>
> > Again, there is no circuit provided, but a flying node generally needs
> > to fly for the circuit to work. A flying node will couple current if
> > there is a cap path. I don't see how a clamp in the IGBT helps since
> > it is the uP that is latching, thus it is a node on the uP that needs
> > clamping.
>
> > What did you measure for the latch-up on the chip? Microchip doesn't
> > have a corporate spec for latch-up limits on the website (in my brief
> > search), but typically you expect a digital part to exceed 100ma,
> > usually way more. Analog parts might be as poor as 40ma. These are
> > room temp specs. Get the part warm, beta increases, and it is easier
> > to latch.
>
> > Who knows, perhaps the part has lousy latch-up.
>
> Thank you for your help again,
> I am not familiar with the term "flying node" or its methodology.
> Could you elaborate?
> It makes sense, I have yet to find an app note for the implementation.
> My google searches have given me nothing.
> I will be measuring currents asap.
>
> This was the only doc found from microchip.
>
> AN595
> Title: Improving the Susceptibility of an Application to ESD-Induced
> Latch-up
>
> Thanks

Without a circuit, what I think you have is a coil that you are
"charging" with the switch. Turn off the switch, the node goes flying
upward, hence flying node. Quite a few hits for "flying node".

 | 
Pages: 1
Prev: IBIS
Next: book on synchros/resolvers