Any beefy coil driver chips?

Jim Thompson · Aug 2, 2009

Actually I must ramp the coil current to the hilt the instant the rail
comes up. In cases like this the rail is the switched line and also the
only power source. Wanting an extra DC power line to come in is almost
like renting a backhoe and starting to trench

I have to go through their stuff one more time, maybe they really have
something useful. Usually Supertex comes to the rescue in such cases but
not this time. They have only the HV9901 and it wouldn't buy me that
much in space savings versus discretes. OTOH, cost is not the driving
factor on this design, within reason, meaning no chip design. Ain't that
nice for a change?

Sometimes I envy guys like Jim but I had that luxury only twice: You
shop around for HV processes and then piece it all together on your
computation machine. After I did a design on a 60V process I was so spoiled.

HV ASIC processes are not cost-effective. You're better off using a
low voltage process plus discretes... unless you're the government and
only size matters ;-)

It's probably been close to 20 years since I last used a 30V process
(not counting specialty processes with HV open drain or collector
devices... like the DMOS processes now available).

...Jim Thompson
--
| James E.Thompson, P.E. | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona 85048 Skype: Contacts Only | |
| Voice

Jim Thompson · Aug 2, 2009

Need to drive a >30V (several amps) coil but back off once it's on to
avoid a cook-out. This needs to be PWM'd because the drive electronics
can't get hot at all. So I looked at coil driver chips. DRV103 and the
like. Many can't stomach more than 32V which is borderline for me and
the saturation voltage respectively Rdson values, well, yawn. Nothing to
write home about.

Did I miss a really good chip? Else I'll just roll my own.

The inductance spec seems to jump around from ~1mH to 10mH... maybe
you have an ampere-turn spec... maybe ?

Joerg · Aug 2, 2009

Jim said:
[...]

Sometimes I envy guys like Jim but I had that luxury only twice: You
shop around for HV processes and then piece it all together on your
computation machine. After I did a design on a 60V process I was so spoiled.

Click to expand...

HV ASIC processes are not cost-effective. You're better off using a
low voltage process plus discretes... unless you're the government and
only size matters ;-)

It's probably been close to 20 years since I last used a 30V process
(not counting specialty processes with HV open drain or collector
devices... like the DMOS processes now available).

This was 1997/98: 60V Delco process, grand total with masks, first
silicon, my flights, but without my engineering time somewhere around
$120k. However, I realize that NRE has gone up quite a bit since then.

When we fired it up for the first time lo and behold it worked. A major
pool party over here was called for and held, booze and food provided by
yours truly.

Man, I wish I could get such a deal just one more time

Joerg · Aug 2, 2009

Jim said:
The inductance spec seems to jump around from ~1mH to 10mH... maybe
you have an ampere-turn spec... maybe ?

Where did the 1mH come from? Maybe I am getting old ...

It's likely a tad over 10mH. How large this "tad" is remains to be seen
since the mechanics of the relays structure still need to be designed.
Again, this is outside my influence, although we could fudge with the #
of turns versus wire diameter a bit. But not to worry, I'll get it done.
Was just hoping there'd be a solution in a can maybe from the world of
diesel rail injection or whatever.

Jim Thompson · Aug 2, 2009

Where did the 1mH come from?

You spoke of 1 Ohm and "ramping" to 3A, in a small fraction of a msec.
Suppose that "small fraction" was 100usec, 24V implies 800uH.

Maybe I am getting old ...

Certainly exhibiting Alzheimer-style symptoms ;-)

It's likely a tad over 10mH. How large this "tad" is remains to be seen
since the mechanics of the relays structure still need to be designed.
Again, this is outside my influence, although we could fudge with the #
of turns versus wire diameter a bit. But not to worry, I'll get it done.
Was just hoping there'd be a solution in a can maybe from the world of
diesel rail injection or whatever.

Jim Thompson · Aug 2, 2009

Jim said:
Jim said:

[...]

Sometimes I envy guys like Jim but I had that luxury only twice: You
shop around for HV processes and then piece it all together on your
computation machine. After I did a design on a 60V process I was so spoiled.

Click to expand...

HV ASIC processes are not cost-effective. You're better off using a
low voltage process plus discretes... unless you're the government and
only size matters ;-)

It's probably been close to 20 years since I last used a 30V process
(not counting specialty processes with HV open drain or collector
devices... like the DMOS processes now available).

Click to expand...

This was 1997/98: 60V Delco process, grand total with masks, first
silicon, my flights, but without my engineering time somewhere around
$120k. However, I realize that NRE has gone up quite a bit since then.

When we fired it up for the first time lo and behold it worked. A major
pool party over here was called for and held, booze and food provided by
yours truly.

Man, I wish I could get such a deal just one more time

Joerg · Aug 3, 2009

Jim said:
Jim said:

[...]

Sometimes I envy guys like Jim but I had that luxury only twice: You
shop around for HV processes and then piece it all together on your
computation machine. After I did a design on a 60V process I was so spoiled.
HV ASIC processes are not cost-effective. You're better off using a
low voltage process plus discretes... unless you're the government and
only size matters ;-)

It's probably been close to 20 years since I last used a 30V process
(not counting specialty processes with HV open drain or collector
devices... like the DMOS processes now available).

Click to expand...

This was 1997/98: 60V Delco process, grand total with masks, first
silicon, my flights, but without my engineering time somewhere around
$120k. However, I realize that NRE has gone up quite a bit since then.

When we fired it up for the first time lo and behold it worked. A major
pool party over here was called for and held, booze and food provided by
yours truly.

Man, I wish I could get such a deal just one more time

Click to expand...

Richard Davis?

Doesn't ring a bell but it's over 10 years now.

Jim Thompson · Aug 3, 2009

Jim said:
Jim said:

Jim Thompson wrote:
[...]

Sometimes I envy guys like Jim but I had that luxury only twice: You
shop around for HV processes and then piece it all together on your
computation machine. After I did a design on a 60V process I was so spoiled.
HV ASIC processes are not cost-effective. You're better off using a
low voltage process plus discretes... unless you're the government and
only size matters ;-)

It's probably been close to 20 years since I last used a 30V process
(not counting specialty processes with HV open drain or collector
devices... like the DMOS processes now available).

This was 1997/98: 60V Delco process, grand total with masks, first
silicon, my flights, but without my engineering time somewhere around
$120k. However, I realize that NRE has gone up quite a bit since then.

When we fired it up for the first time lo and behold it worked. A major
pool party over here was called for and held, booze and food provided by
yours truly.

Man, I wish I could get such a deal just one more time

Click to expand...

Richard Davis?

Click to expand...

Doesn't ring a bell but it's over 10 years now.

Joerg · Aug 3, 2009

Jim said:
You spoke of 1 Ohm and "ramping" to 3A, in a small fraction of a msec.
Suppose that "small fraction" was 100usec, 24V implies 800uH.

Not this small of a fraction but even that would not be completely
impossible. If someone says "impossible" that really gets my
productivity up

Think outside the box, that's what the world is paying us for. As I said
in another post there are ways to goose a coil. For example with a very
zippy boost converter, meaning one that really steps on it the
microsecond the light turns green. It would not be my first one. My
latest one was so tiny that I forgot it was there for a minute since it
all was connected to an innocent 12V bench supply. It promptly bit me.

Certainly exhibiting Alzheimer-style symptoms ;-)

Ahem. Grumble. Don't joke about this stuff, several such cases to take
care of in the neighborhood :-(

If it's a "relay" why the speedy "ramping" requirement? It's going to
bounce for some time anyway.

Not this one

Joerg · Aug 3, 2009

Jim said:
Jim said:

Jim Thompson wrote:
[...]

Sometimes I envy guys like Jim but I had that luxury only twice: You
shop around for HV processes and then piece it all together on your
computation machine. After I did a design on a 60V process I was so spoiled.
HV ASIC processes are not cost-effective. You're better off using a
low voltage process plus discretes... unless you're the government and
only size matters ;-)

It's probably been close to 20 years since I last used a 30V process
(not counting specialty processes with HV open drain or collector
devices... like the DMOS processes now available).

This was 1997/98: 60V Delco process, grand total with masks, first
silicon, my flights, but without my engineering time somewhere around
$120k. However, I realize that NRE has gone up quite a bit since then.

When we fired it up for the first time lo and behold it worked. A major
pool party over here was called for and held, booze and food provided by
yours truly.

Man, I wish I could get such a deal just one more time
Richard Davis?

Click to expand...

Doesn't ring a bell but it's over 10 years now.

Click to expand...

Last I heard he was running the ASIC group at DELCO. He was one of my
young bucks when I was at Motorola ;-)

Ok, then I probably do not know him. We did not use their design folks,
just the foundry.

Old buck sometimes must see that young buck also make good stuff

Jim Thompson · Aug 3, 2009

Jim said:
Jim said:

[snip]

When we fired it up for the first time lo and behold it worked. A major
pool party over here was called for and held, booze and food provided by
yours truly.

Man, I wish I could get such a deal just one more time
Richard Davis?

Doesn't ring a bell but it's over 10 years now.

Click to expand...

Last I heard he was running the ASIC group at DELCO. He was one of my
young bucks when I was at Motorola ;-)

Click to expand...

Ok, then I probably do not know him. We did not use their design folks,
just the foundry.

Old buck sometimes must see that young buck also make good stuff

Jim Thompson · Aug 3, 2009

Not this small of a fraction but even that would not be completely
impossible. If someone says "impossible" that really gets my
productivity up

Think outside the box, that's what the world is paying us for. As I said
in another post there are ways to goose a coil. For example with a very
zippy boost converter, meaning one that really steps on it the
microsecond the light turns green.

Microsecond? Show me ;-) "Goose" is what my "Nagle-solution" does.

It would not be my first one. My
latest one was so tiny that I forgot it was there for a minute since it
all was connected to an innocent 12V bench supply. It promptly bit me.

Ahem. Grumble. Don't joke about this stuff, several such cases to take
care of in the neighborhood :-(

"Use it or lose it" seems to fit quite well, at least in my
observations.

Not this one

Joerg · Aug 3, 2009

Jim said:
Jim Thompson wrote:

Click to expand...

[...]

Not this small of a fraction but even that would not be completely
impossible. If someone says "impossible" that really gets my
productivity up

Think outside the box, that's what the world is paying us for. As I said
in another post there are ways to goose a coil. For example with a very
zippy boost converter, meaning one that really steps on it the
microsecond the light turns green.

Click to expand...

Microsecond? Show me ;-) "Goose" is what my "Nagle-solution" does.

Memory issues setting in? SCNR, but I did mention 120usec before in this
here thread. When you step on the accelerator of a car it won't be at
60mph in a microecond either

Anyhow, not at liberty to share that converter but I sent you another, a
regular run-of-the-mills buck being ready at around 70usec. Not bad, eh?
You have mail.

"Use it or lose it" seems to fit quite well, at least in my
observations.

Not always. I knew folks who were very smart engineers and then ...

Moving target. Tiring. Time for the vino ;-)

I design a lot for moving targets. Very normal business.

Just came out of the pool. Time for the cerveza.

Fred Bartoli · Aug 3, 2009

Jim Thompson a écrit :

If it's a "relay" why the speedy "ramping" requirement? It's going to
bounce for some time anyway.

With the right driving curve (and polarized) you could drive the hell
out of it and then make it land gently...

Joerg · Aug 3, 2009

Fred said:
Jim Thompson a écrit :

With the right driving curve (and polarized) you could drive the hell
out of it and then make it land gently...

Bingo!

Fred Bartoli · Aug 3, 2009

Joerg a écrit :

Bingo!

With a well placed shorted turn you may even slow down the arrival
enough without having to change the drive current...

Jim Thompson · Aug 3, 2009

Bingo!

Joerg · Aug 3, 2009

Winfield said:
Hah! No, this isn't a problem. Use a small mosfet, with Ciss
no more than 100pF, so the flying cap, Cboot, need not be
larger than say 470pF. Use a small cap across the IRS2153
internal 15V zener, say 2.7nF. If the input-power resistor is
680 ohms (nominal 10 to 15mA of current), then the start-up
time, including charging Cboot, is under about 3us. This
means that you can very quickly assert the ON state.

That is a nice trick. It doesn't even have to be that fast. Several ten
usecs for start-up would be ok. Compared to SMPS converters this stuff
is all relaxingly "slow".

Hmm, I wonder, might there be any bounce in the 30V
input drive, better add a diode in series with the 680 ohms.

Note, you can run the rest of your control circuitry from the
15V power tap of the 2153 chip. Hah, a cmos 555 still has
a chance to play a role, for your 1 to 5ms high-power timing.
With a diode, its output can hold up the CT pin.

Abs max says 20mA so even at half that value there should be plenty to
spare. For some reason they only give a typical supply current but with
Iq max at 1mA plus whatever the oscillator causes that leaves plenty for
some CMOS logic. Of course they stated the typical current at a highish
timing resistance where you can't get it above 20kHz. If you ever need
something like this with more nifty power supplies the HV9901 has a 2nd
regulator in there that can be set to the levels of modern uC and is
nicely tied to a bandgap reference in there. Only 1mA but that suffices
for a little MSP430 at a low clock speed. But the comparator offset
specs are horrid, I am wondering if they just were ultra-conservative.

Well, it's fun to talk about these things. But if we get down
to brass tacks, and get serious, I have to say I'm not happy
with the IRS21531 chip's long 0.6us dead time during PWM.
While it may be appropriate for a 200 to 600V application,
it's far too long at low voltages, and means the lower power
FET's body diode will always be conducting during the dead
time, with its ugly reverse-recovery-time snap-off issue. OK,
add a Schottky diode across the lower mosfet. That'll fix it.

That's a common trick in synchronous bucks but so far I always got away
without the Schottky.

Joerg · Aug 3, 2009

couldn't you just switch the low side of the coil and not need
really use Cboot ?

You can, but it makes current sense a bear. Unless your chip has an
integrated sampling/gating with proper FET spike blanking built in. Some
do but the IRS2153 doesn't.

Joerg · Aug 3, 2009

Winfield said:
So, if the two mosfets are in an soic-8 package (lots
to choose from) you're talking about three 8-pin soic
packages and a smattering of discrete parts. A small
PCB, with two input pins for the coil power signal, and
two output pins (the two negative IN and OUT pins are
wired together) for the coil, and you're all set.

Getting three SO-8 packages on there will already be a challenge but the
smattering of discretes is what really can do this in. The HV9901 looks
good so far except for the huge offset voltages, of course. The fact
that it doesn't come any smaller than SO-16 isn't exactly helping either.

Moore's Lobby Podcast

Menu

Categories

Platforms

Content

Connect With Us

Network

Any beefy coil driver chips?

Any beefy coil driver chips?

Jim Thompson

Jim Thompson

Joerg

Joerg

Jim Thompson

Jim Thompson

Joerg

Jim Thompson

Joerg

Joerg

Jim Thompson

Jim Thompson

Joerg

Fred Bartoli

Joerg

Fred Bartoli

Jim Thompson

Joerg

Joerg

Joerg

Similar threads