Testing Mosfets in Parallel

[EdV]

My group has been tasked with designing a test for a board that has ten
mosfets in parallel. No jumpering allowed. The designers say that
adding jumpers and bus bars make the circuit oscillate.

The first thing I want to test is that all of the parts are actually
stuffed with the correct part(this sort of stuff happens to us).
Perhaps driving the mosfets into the linear region with a modest load
and then measuring their individual temperatures with an array of
thermistors.

Any thoughts?

This array is capable of sinking 100 amps and there are two arrays on
each board. Building fixtures to test operate them at full load is
kind of cost prohibitive.

Thanks much,
Ed V.

 

[John Larkin]

My group has been tasked with designing a test for a board that has ten
mosfets in parallel. No jumpering allowed. The designers say that
adding jumpers and bus bars make the circuit oscillate.

Sounds like they have no gate resistors. Bad news... edge of
oscillation already.

The first thing I want to test is that all of the parts are actually
stuffed with the correct part(this sort of stuff happens to us).
Perhaps driving the mosfets into the linear region with a modest load
and then measuring their individual temperatures with an array of
thermistors.

Any thoughts?

This array is capable of sinking 100 amps and there are two arrays on
each board. Building fixtures to test operate them at full load is
kind of cost prohibitive.

Thanks much,
Ed V.

The problem with the linear thing is that current sharing will be
poor, so fet temps will be all over the place. You could pulse them at
a modest duty cycle, enough to warm them up, and use the new Fluke IR
viewer.

Can you probe the board and measure voltage drops in traces? That
would tell you if all the fets are conducting, and (approximately) how
equally.

John

 

[Ian Stirling]

The problem with the linear thing is that current sharing will be
poor, so fet temps will be all over the place. You could pulse them at
a modest duty cycle, enough to warm them up, and use the new Fluke IR
viewer.

Can you probe the board and measure voltage drops in traces? That
would tell you if all the fets are conducting, and (approximately) how
equally.

Or do the magnetic thing, with an array of chip magnetometers at
appropriate points.

 

[John Larkin]

My group has been tasked with designing a test for a board that has ten
mosfets in parallel. No jumpering allowed. The designers say that
adding jumpers and bus bars make the circuit oscillate.

The first thing I want to test is that all of the parts are actually
stuffed with the correct part(this sort of stuff happens to us).
Perhaps driving the mosfets into the linear region with a modest load
and then measuring their individual temperatures with an array of
thermistors.

Any thoughts?

This array is capable of sinking 100 amps and there are two arrays on
each board. Building fixtures to test operate them at full load is
kind of cost prohibitive.

Thanks much,
Ed V.

Why not just test the board functionally, and ignore internals? If the
designers made it untestable, that's what they deserve.

John

 

[kell]

My group has been tasked with designing a test for a board that has ten
mosfets in parallel. No jumpering allowed. The designers say that
adding jumpers and bus bars make the circuit oscillate.

The first thing I want to test is that all of the parts are actually
stuffed with the correct part(this sort of stuff happens to us).
Perhaps driving the mosfets into the linear region with a modest load
and then measuring their individual temperatures with an array of
thermistors.

Any thoughts?

This array is capable of sinking 100 amps and there are two arrays on
each board. Building fixtures to test operate them at full load is
kind of cost prohibitive.

Thanks much,
Ed V.

You could keep your cost down to a couple hundred dollars.
Go to your local auto supply store.
Buy a truck or car battery and a couple dozen 2.5 ohm 50 watt
automobile ballast resistors.

 

[Harry Dellamano]

The problem with the linear thing is that current sharing will be
poor, so fet temps will be all over the place. You could pulse them at
a modest duty cycle, enough to warm them up, and use the new Fluke IR
viewer.

Can you probe the board and measure voltage drops in traces? That
would tell you if all the fets are conducting, and (approximately) how
equally.

John

Hi John,
Tell us more of this Fluke IR viewer, like a part number. Can you measure
junction temps?
Regards,
Harry

 

[John Larkin]

Hi John,
Tell us more of this Fluke IR viewer, like a part number. Can you measure
junction temps?
Regards,
Harry

Well, there is the Fluke web site. They have a nice looking thermal
imager now for $9995 or something, about a third of the price of
competing products. I *really* need an excuse to buy one.

John

 

[Winfield Hill]

EdV wrote...

My group has been tasked with designing a test for a board that
has ten mosfets in parallel. No jumpering allowed. The designers
say that adding jumpers and bus bars make the circuit oscillate.

I assume this is a switched MOSFET design, and not a linear
one (if it's linear, send it back to the drawing board, now!).

Folks often overlook the ease of measuring voltage drops along
PCB traces if high currents are flowing. So, run your switch
at full load current, get out your 5-1/2 digit multimeter set
on its 200mV scale, lsb resolution only one thin uV.

Measure away.

 

[Harry Dellamano]

Well, there is the Fluke web site. They have a nice looking thermal
imager now for $9995 or something, about a third of the price of
competing products. I *really* need an excuse to buy one.

John

So at 2' you can resolve 0.27" spot size. Your intent is to scan an
operating PCB and resolve temps on SMD. Sounds great, let's see some color
scans before we buy.
Cheers,
Harry

 

[Winfield Hill]

Winfield Hill wrote...

EdV wrote...

I assume this is a switched MOSFET design, and not a linear
one (if it's linear, send it back to the drawing board, now!).

Either redesign the circuit, or change the installed parts to
lateral "audio" MOSFETs, like the Hitachi -> Renesas parts,
"Power MOSFETs for General Amplifier" on their website.

 

[John Miles]

Winfield Hill wrote...

Either redesign the circuit, or change the installed parts to
lateral "audio" MOSFETs, like the Hitachi -> Renesas parts,
"Power MOSFETs for General Amplifier" on their website.

Hey, Win, clean out your picovolt.com mailbox! I'm getting lots o'
bouncing action from it, and eBay will hose your account if that address
is the one associated with it.

Apologies for the misuse of public bandwidth...

-- jm

 

[Winfield Hill]

John Miles wrote...

Winfield Hill wrote...

Hey, Win, clean out your picovolt.com mailbox! I'm getting lots o'
bouncing action from it, and eBay will hose your account if that
address is the one associated with it.

It's not, my Harvard email is the one. I thought that was clear.
How would anyone even know about my old home email address, which
I haven't used in over two years? Yes, I do have the picovolt.com
domain, but I long ago stopped using its mailbox, because it became
inundated with over 100MB of spam per day! I mean, HELP! So now I
struggle with one single spam-blasted mailbox. With all my tools
to help fight spam, half of my 300 daily Harvard email messages are
correctly determined as spam and shunted aside, leaving me only 75
spam emails per day to hand identify. I then have 75 real messages
to address. When I come in to work each weekend... And after a few
weeks of vacation, sheesh! Over 5000 messages, and 2500 to study.
Anyway, at home I seek some peace. No email for 25 months now.

 

[John Miles]

It's not, my Harvard email is the one. I thought that was clear.
How would anyone even know about my old home email address, which
I haven't used in over two years? Yes, I do have the picovolt.com
domain, but I long ago stopped using its mailbox, because it became
inundated with over 100MB of spam per day! I mean, HELP! So now I
struggle with one single spam-blasted mailbox. With all my tools
to help fight spam, half of my 300 daily Harvard email messages are
correctly determined as spam and shunted aside, leaving me only 75
spam emails per day to hand identify. I then have 75 real messages
to address. When I come in to work each weekend... And after a few
weeks of vacation, sheesh! Over 5000 messages, and 2500 to study.
Anyway, at home I seek some peace. No email for 25 months now.

Ah, that explains it! Thanks; now I have to wonder where I got the
picovolt.com address. Probably by taking your eBay handle too
literally. I wasn't sure you got my reply on the Turbo option question,
which I cc'ed to both harvard.edu and picovolt.com.

Anyway, all six packages shipped out today; let me know if you did not
get the tracking # notification from UPS. Sorry for the threadjacking!

-- jm

 

[Fred Bartoli]

Well, there is the Fluke web site. They have a nice looking thermal
imager now for $9995 or something, about a third of the price of
competing products. I *really* need an excuse to buy one.

Hey, so I volonteer for you to buy me one

 

[EdV]

Maybe an array of hall effect sensors would work.

 

[EdV]

Functionally testing the board requires a 200 amp 12 V supply and a 24
KW load which still doesn't really tell you if there is a FET missing
or stuffed with something else that is not conducting.

Regarding the "deserve" thing. My performance is rated on how much
warranty return I prevent and prettty much nothing else.

 

[Harry Dellamano]

So at 2' you can resolve 0.27" spot size. Your intent is to scan an
operating PCB and resolve temps on SMD. Sounds great, let's see some color
scans before we buy.
Cheers,
Harry

Ok, so I only need one for about two hours maybe once every three months.
I'll pitch in $1K for 10% usage. Get another nine dependable folk and you
get yours free, just have to manage it's whereabouts. Got to see some scans
first.
Regards,
Harry

 

[John Larkin]

Functionally testing the board requires a 200 amp 12 V supply and a 24
KW load which still doesn't really tell you if there is a FET missing
or stuffed with something else that is not conducting.

Regarding the "deserve" thing. My performance is rated on how much
warranty return I prevent and prettty much nothing else.

Even when engineering feeds you a bad or untestable design?

John

 

[John Larkin]

Hey, so I volonteer for you to buy me one

Ah. Humor.

John

 

[kell]

Functionally testing the board requires a 200 amp 12 V supply and a 24
KW load which still doesn't really tell you if there is a FET missing
or stuffed with something else that is not conducting.

Regarding the "deserve" thing. My performance is rated on how much
warranty return I prevent and prettty much nothing else.

I hope that 24 KW was a typo.