How to protect H-bridge from short?

[Chris Carlen]

Greetings:

I have a H-bridge PWM amp using Apex SA-60. I have implemented an
average current mode control loop, thanks to assistance from Genome,
which makes it possible for the amplifer to safely handle a differential
short circuit fault.

The current sensing is done through two resistors on the low sides of
the bridge, going to a differential amplifier.

Trouble is, this current sense will read zero and thus attempt to push
even more current if there is a short from one H-bridge output leg to
ground.

How can I protect from this? I had considered floating the whole power
amp system, which already has a differential input. Then tie it to
earth ground via a R||C combo. This would be fine except that
if one shorts an output leg, now the input diff amp saturates with
excess common mode input, resulting in an indeterminate command signal
to the PWM amp.

I can't have this because it is a DC amplifier driving a mechanical
system, that if the drive pegs for instance, will result in mechanical
things crashing together.

Thus, current to the load must cease if this short condition occurs.

I am inclining back toward keeping the thing connected to earth ground,
and figuring out some straightforward way to implement short protection.

My thoughts so far:

1. simply fuse each output, after the LC filter, and putting the
voltage sense after the fuses so their resistance doesn't cause voltage
drops.

But the PWM amp might just keep chugging along if one side blows, until
it attempts to reverse the output polarity at which point it will clip.
This isn't ideal.

2. fuse the power supply. The fuse will blow cutting power to the
whole thing if either side shorts. Trouble is, can any fuse blow fast
enough? I don't think so. The rate of current rise is slowed a bit by
the output filters, but still it's on the order of 5us to rise 15A with
my L and supply V.

3. Add an additional current sense on the B+ supply to the H-bridge,
and shut down the amp via the SA-60 DISABLE input if the average current
exceeds 15A with a time constant of something appropriate.

This is probably the best solution, and also would help me if the
amplifier is configured as a straight PWM voltage amp without current
mode, which may be desirable in some cases due to better step response
performance.

But more circuitry to add too something that has already spiraled
somewhat out of control.


Other ways?


Thanks for input.
--
_____________________
Christopher R. Carlen
crobc@bogus-remove-me.sbcglobal.net
SuSE 9.1 Linux 2.6.5

 

[Pooh Bear]

Chris Carlen wrote:

Greetings:

I have a H-bridge PWM amp using Apex SA-60. I have implemented an
average current mode control loop, thanks to assistance from Genome,
which makes it possible for the amplifer to safely handle a differential
short circuit fault.

The current sensing is done through two resistors on the low sides of
the bridge, going to a differential amplifier.

Trouble is, this current sense will read zero

How so ?

Btw - why not position a single sense R so that's it's common to both of the
low side devices ?

Graham

 

[Mac]

On Fri, 27 May 2005 11:40:12 -0700, Chris Carlen wrote:

Greetings:

I have a H-bridge PWM amp using Apex SA-60. I have implemented an
average current mode control loop, thanks to assistance from Genome,
which makes it possible for the amplifer to safely handle a differential
short circuit fault.

The current sensing is done through two resistors on the low sides of
the bridge, going to a differential amplifier.

Trouble is, this current sense will read zero and thus attempt to push
even more current if there is a short from one H-bridge output leg to
ground.

How can I protect from this? I had considered floating the whole power
amp system, which already has a differential input. Then tie it to
earth ground via a R||C combo. This would be fine except that
if one shorts an output leg, now the input diff amp saturates with
excess common mode input, resulting in an indeterminate command signal
to the PWM amp.

I can't have this because it is a DC amplifier driving a mechanical
system, that if the drive pegs for instance, will result in mechanical
things crashing together.

Thus, current to the load must cease if this short condition occurs.

I am inclining back toward keeping the thing connected to earth ground,
and figuring out some straightforward way to implement short protection.

My thoughts so far:

1. simply fuse each output, after the LC filter, and putting the
voltage sense after the fuses so their resistance doesn't cause voltage
drops.

But the PWM amp might just keep chugging along if one side blows, until
it attempts to reverse the output polarity at which point it will clip.
This isn't ideal.

2. fuse the power supply. The fuse will blow cutting power to the
whole thing if either side shorts. Trouble is, can any fuse blow fast
enough? I don't think so. The rate of current rise is slowed a bit by
the output filters, but still it's on the order of 5us to rise 15A with
my L and supply V.

3. Add an additional current sense on the B+ supply to the H-bridge,
and shut down the amp via the SA-60 DISABLE input if the average current
exceeds 15A with a time constant of something appropriate.

This is probably the best solution, and also would help me if the
amplifier is configured as a straight PWM voltage amp without current
mode, which may be desirable in some cases due to better step response
performance.

But more circuitry to add too something that has already spiraled
somewhat out of control.


Other ways?


Thanks for input.

As I recall, a simple way to detect a short circuit fault is this: the
voltage across the high-side device (or low for that matter) should be
very low when that device is on. If this is NOT the case, you need to shut
the amp down very quickly.

You need to allow reasonable time for the device to turn on, of
course.

If this doesn't make sense, then maybe you could draw a simplified version
of your output stage, and indicate where you are envisioning the short?

In any event, you can always add high-side AND low-side current sensing.
Maybe you could somehow even employ a GFI unit to automatically switch off
some portion of the circuitry when the upper and lower currents don't
match.

--Mac

 

[Terry Given]

Mac wrote:

On Fri, 27 May 2005 11:40:12 -0700, Chris Carlen wrote:


Greetings:

I have a H-bridge PWM amp using Apex SA-60. I have implemented an
average current mode control loop, thanks to assistance from Genome,
which makes it possible for the amplifer to safely handle a differential
short circuit fault.

The current sensing is done through two resistors on the low sides of
the bridge, going to a differential amplifier.

Trouble is, this current sense will read zero and thus attempt to push
even more current if there is a short from one H-bridge output leg to
ground.

How can I protect from this? I had considered floating the whole power
amp system, which already has a differential input. Then tie it to
earth ground via a R||C combo. This would be fine except that
if one shorts an output leg, now the input diff amp saturates with
excess common mode input, resulting in an indeterminate command signal
to the PWM amp.

I can't have this because it is a DC amplifier driving a mechanical
system, that if the drive pegs for instance, will result in mechanical
things crashing together.

Thus, current to the load must cease if this short condition occurs.

I am inclining back toward keeping the thing connected to earth ground,
and figuring out some straightforward way to implement short protection.

My thoughts so far:

1. simply fuse each output, after the LC filter, and putting the
voltage sense after the fuses so their resistance doesn't cause voltage
drops.

But the PWM amp might just keep chugging along if one side blows, until
it attempts to reverse the output polarity at which point it will clip.
This isn't ideal.

2. fuse the power supply. The fuse will blow cutting power to the
whole thing if either side shorts. Trouble is, can any fuse blow fast
enough? I don't think so. The rate of current rise is slowed a bit by
the output filters, but still it's on the order of 5us to rise 15A with
my L and supply V.

3. Add an additional current sense on the B+ supply to the H-bridge,
and shut down the amp via the SA-60 DISABLE input if the average current
exceeds 15A with a time constant of something appropriate.

This is probably the best solution, and also would help me if the
amplifier is configured as a straight PWM voltage amp without current
mode, which may be desirable in some cases due to better step response
performance.

But more circuitry to add too something that has already spiraled
somewhat out of control.


Other ways?


Thanks for input.


As I recall, a simple way to detect a short circuit fault is this: the
voltage across the high-side device (or low for that matter) should be
very low when that device is on. If this is NOT the case, you need to shut
the amp down very quickly.

You need to allow reasonable time for the device to turn on, of
course.

If this doesn't make sense, then maybe you could draw a simplified version
of your output stage, and indicate where you are envisioning the short?

In any event, you can always add high-side AND low-side current sensing.
Maybe you could somehow even employ a GFI unit to automatically switch off
some portion of the circuitry when the upper and lower currents don't
match.

--Mac

In the IGBT world this is called "desat" detection, IOW detecting that
the IGBT has pulled out of saturation. I typically use a Vcesat
threshold of 10V, but thats because IGBTs tend to desaturate completely
when these sorts of faults appear (so Vcesat = Vbus). Transient thermal
impedance is your friend - many IGBTs are rated for 10x rated current
for 10us under these conditions. A good thermal model helps here. 30us
into a desat, kaBOOM. You would need to check the desat behaviour of
your FET, but the +ve Rdson tempco helps here (many hundreds of kW peak
pulse power....)

other techniques:
- DCCT in DC bus, with both +Vdc and -Vdc running thru the CT (dont
forget to run the -Vdc wire in the opposite direction so the DM currents
dont cancel). OTTOMH I dont see why an ordinary CT wont do the trick, as
we are looking for a transient (and a big one, too). Beware the extra
inductance . A company I worked for built tens of thousands of small
AC motor controllers like this.

- current sense resistor in low side of DC bus cap. wont help much if
short is to Earth, and a current path exists w/o passing thru cap

- current sense resistor in output lead (can float cct with upper gate
drive) and, say, an opto to feed back the "oh shit" command. I know a
guy who patented this in the early '80s. A DCCT makes this a doddle,
*but* wont protect against a half-bridge failure (where your resistors
will). Again, many tens of thousands of motor controllers exist that do
this.

Cheers
Terry

 

[Terry Given]

Chris Carlen wrote:

Greetings:

I have a H-bridge PWM amp using Apex SA-60. I have implemented an
average current mode control loop, thanks to assistance from Genome,
which makes it possible for the amplifer to safely handle a differential
short circuit fault.

The current sensing is done through two resistors on the low sides of
the bridge, going to a differential amplifier.

Trouble is, this current sense will read zero and thus attempt to push
even more current if there is a short from one H-bridge output leg to
ground.

How can I protect from this? I had considered floating the whole power
amp system, which already has a differential input. Then tie it to
earth ground via a R||C combo. This would be fine except that
if one shorts an output leg, now the input diff amp saturates with
excess common mode input, resulting in an indeterminate command signal
to the PWM amp.

I can't have this because it is a DC amplifier driving a mechanical
system, that if the drive pegs for instance, will result in mechanical
things crashing together.

Thus, current to the load must cease if this short condition occurs.

I am inclining back toward keeping the thing connected to earth ground,
and figuring out some straightforward way to implement short protection.

My thoughts so far:

1. simply fuse each output, after the LC filter, and putting the
voltage sense after the fuses so their resistance doesn't cause voltage
drops.

But the PWM amp might just keep chugging along if one side blows, until
it attempts to reverse the output polarity at which point it will clip.
This isn't ideal.

2. fuse the power supply. The fuse will blow cutting power to the
whole thing if either side shorts. Trouble is, can any fuse blow fast
enough? I don't think so. The rate of current rise is slowed a bit by
the output filters, but still it's on the order of 5us to rise 15A with
my L and supply V.

even semiconductor fuses wont protect your silicon. What they will do,
though, is prevent the serious damage that occurs when you short the
national grid thru your (now dead) box of tricks. minimises fire risk,
explosive potential etc.

poor mans ultra-fast 400A fuse: use a 10L plastic bucket & lid. Fill it
with de-ionised water. On the lid, mount a couple of 100mm long ceramic
standoffs, so they are underwater. use ordinary 5A fuse wire. The water
sucks the heat out of the fuse wire, allowing it to carry much, much
higher currents than otherwise. But, if current is too high (adjust
empirically) the water turns to steam around the wire, thermal
resistance skyrockets and the fuse vaporises. A buddy of mine taught me
this trick, which he used when developing an SCR V/Hz controller for his
PhD (and had no budget for $300 fuses, three at a time)

3. Add an additional current sense on the B+ supply to the H-bridge,
and shut down the amp via the SA-60 DISABLE input if the average current
exceeds 15A with a time constant of something appropriate.

This is probably the best solution, and also would help me if the
amplifier is configured as a straight PWM voltage amp without current
mode, which may be desirable in some cases due to better step response
performance.

But more circuitry to add too something that has already spiraled
somewhat out of control.


Other ways?


Thanks for input.

 

[Chris Carlen]

bigcat@meeow.co.uk wrote:

shame the whole shebang isnt running off a psu volt reg with inbuilt i
limit. Very simple.

No, the power supply is limited to 12A, and my amp's surge current is
15A. The power supply responds on the order of 100us, not 1us or so in
which things can go to hell within the H-bridge. That is because there
is a large capacitor bypassing the PWM amp bus. A short to ground of
one output leg can cause damaging currents to flow much more quickly
than this before the power supply limits. In fact the damaginc current
could develop in ns if it weren't for the output filter inductors. They
will buy me about 5us. Thus, I think I need some active current sensing
and comparator looking at the high side switch currents.


Thanks for the response.


Good day!


--
_____________________
Christopher R. Carlen
crobc@bogus-remove-me.sbcglobal.net
SuSE 9.1 Linux 2.6.5

 

[Chris Carlen]

Pooh Bear wrote:

Chris Carlen wrote:

Greetings:

I have a H-bridge PWM amp using Apex SA-60. I have implemented an
average current mode control loop, thanks to assistance from Genome,
which makes it possible for the amplifer to safely handle a differential
short circuit fault.

The current sensing is done through two resistors on the low sides of
the bridge, going to a differential amplifier.

Trouble is, this current sense will read zero

How so ?

Btw - why not position a single sense R so that's it's common to both of the
low side devices ?

Because a short from one leg to ground shunts current away from the low
side switches. It's the high-side switches that will see the fault current.

Also a single sense resistor wouldn't allow implementation of bipolar
current sensing, rather returning the absolute value of the load current.

Thanks for input.


Good day!



--
_____________________
Christopher R. Carlen
crobc@bogus-remove-me.sbcglobal.net
SuSE 9.1 Linux 2.6.5

 

[Pooh Bear]

Chris Carlen wrote:

Pooh Bear wrote:

Chris Carlen wrote:

Greetings:

I have a H-bridge PWM amp using Apex SA-60. I have implemented an
average current mode control loop, thanks to assistance from Genome,
which makes it possible for the amplifer to safely handle a differential
short circuit fault.

The current sensing is done through two resistors on the low sides of
the bridge, going to a differential amplifier.

Trouble is, this current sense will read zero

How so ?

Btw - why not position a single sense R so that's it's common to both of the
low side devices ?

Because a short from one leg to ground shunts current away from the low
side switches. It's the high-side switches that will see the fault current.

I see you're concerned about a short to ground rather than across the load.

In which case you can detect the high side current and level-shift the signal.

Also a single sense resistor wouldn't allow implementation of bipolar
current sensing,

Indeed but why would you need to do that ?

rather returning the absolute value of the load current.

Graham

 

[colin]

"Chris Carlen" <crobc@BOGUSFIELD.sbcglobal.net> wrote in message
news:d77pif0vtd@news4.newsguy.com...

Greetings:

I have a H-bridge PWM amp using Apex SA-60. I have implemented an
average current mode control loop, thanks to assistance from Genome,
which makes it possible for the amplifer to safely handle a differential
short circuit fault.

The current sensing is done through two resistors on the low sides of
the bridge, going to a differential amplifier.

Trouble is, this current sense will read zero and thus attempt to push
even more current if there is a short from one H-bridge output leg to
ground.

How can I protect from this? I had considered floating the whole power
amp system, which already has a differential input. Then tie it to
earth ground via a R||C combo. This would be fine except that
if one shorts an output leg, now the input diff amp saturates with
excess common mode input, resulting in an indeterminate command signal
to the PWM amp.

I can't have this because it is a DC amplifier driving a mechanical
system, that if the drive pegs for instance, will result in mechanical
things crashing together.

Thus, current to the load must cease if this short condition occurs.

I am inclining back toward keeping the thing connected to earth ground,
and figuring out some straightforward way to implement short protection.

My thoughts so far:

1. simply fuse each output, after the LC filter, and putting the
voltage sense after the fuses so their resistance doesn't cause voltage
drops.

But the PWM amp might just keep chugging along if one side blows, until
it attempts to reverse the output polarity at which point it will clip.
This isn't ideal.

2. fuse the power supply. The fuse will blow cutting power to the
whole thing if either side shorts. Trouble is, can any fuse blow fast
enough? I don't think so. The rate of current rise is slowed a bit by
the output filters, but still it's on the order of 5us to rise 15A with
my L and supply V.

3. Add an additional current sense on the B+ supply to the H-bridge,
and shut down the amp via the SA-60 DISABLE input if the average current
exceeds 15A with a time constant of something appropriate.

This is probably the best solution, and also would help me if the
amplifier is configured as a straight PWM voltage amp without current
mode, which may be desirable in some cases due to better step response
performance.

But more circuitry to add too something that has already spiraled
somewhat out of control.


Other ways?

why does it need to be earthed ? if its just safety could you get away with
earthing the high side V instead of the low side ? that would apear to be
the simplest solution.

Colin =^.^=

 

[Ian Stirling]

Chris Carlen <crobc@bogusfield.sbcglobal.net> wrote:

bigcat@meeow.co.uk wrote:
shame the whole shebang isnt running off a psu volt reg with inbuilt i
limit. Very simple.


No, the power supply is limited to 12A, and my amp's surge current is
15A. The power supply responds on the order of 100us, not 1us or so in
which things can go to hell within the H-bridge. That is because there
is a large capacitor bypassing the PWM amp bus. A short to ground of
one output leg can cause damaging currents to flow much more quickly
than this before the power supply limits. In fact the damaginc current

One relatively easy way is a faster current limit.
Something like a 12A FET, with sufficient voltage rating, and Vgs set
to the right level to get it to intrinsically current limit.

Add a feedback loop to short the Vgs capacitor out, if the voltage across
the FET exceeds 20V, and you're done.

(Of course, you need to be sure this FET is either on, or off)

 

[Chris Carlen]

Pooh Bear wrote:

Chris Carlen wrote:

Pooh Bear wrote:
Chris Carlen wrote:

Greetings:

I have a H-bridge PWM amp using Apex SA-60. I have implemented an
average current mode control loop, thanks to assistance from Genome,
which makes it possible for the amplifer to safely handle a differential
short circuit fault.

The current sensing is done through two resistors on the low sides of
the bridge, going to a differential amplifier.

Trouble is, this current sense will read zero
How so ?

Btw - why not position a single sense R so that's it's common to both of the
low side devices ?
Because a short from one leg to ground shunts current away from the low
side switches. It's the high-side switches that will see the fault current.

I see you're concerned about a short to ground rather than across the load.

In which case you can detect the high side current and level-shift the signal.

Also a single sense resistor wouldn't allow implementation of bipolar
current sensing,

Indeed but why would you need to do that ?

DC servo amp, not AC coupled audio. Need 4-quadrant control.


--
_____________________
Christopher R. Carlen
crobc@bogus-remove-me.sbcglobal.net
SuSE 9.1 Linux 2.6.5

 

[Chris Carlen]

bigcat@meeow.co.uk wrote:

Chris Carlen wrote:
bigcat@meeow.co.uk wrote:
shame the whole shebang isnt running off a psu volt reg with inbuilt i
limit. Very simple.

No, the power supply is limited to 12A, and my amp's surge current is
15A. The power supply responds on the order of 100us, not 1us or so in
which things can go to hell within the H-bridge. That is because there
is a large capacitor bypassing the PWM amp bus. A short to ground of
one output leg can cause damaging currents to flow much more quickly
than this before the power supply limits. In fact the damaginc current
could develop in ns if it weren't for the output filter inductors. They
will buy me about 5us. Thus, I think I need some active current sensing
and comparator looking at the high side switch currents.


Thanks for the response.


Good day!


If you monitor anything other than the current out of the reservoir
cap, you need multipoint monitoring to cover all possible problems. If
you monitor only i out from the reservoir, all overcurrent fault modes
are covered.

Yes, I had reached that conclusion as well when I should have been
sleeping last night.


Thanks for the input.



--
_____________________
Christopher R. Carlen
crobc@bogus-remove-me.sbcglobal.net
SuSE 9.1 Linux 2.6.5

 

[Pooh Bear]

Chris Carlen wrote:

Pooh Bear wrote:
Chris Carlen wrote:

Pooh Bear wrote:
Chris Carlen wrote:

Greetings:

I have a H-bridge PWM amp using Apex SA-60. I have implemented an
average current mode control loop, thanks to assistance from Genome,
which makes it possible for the amplifer to safely handle a differential
short circuit fault.

The current sensing is done through two resistors on the low sides of
the bridge, going to a differential amplifier.

Trouble is, this current sense will read zero
How so ?

Btw - why not position a single sense R so that's it's common to both of the
low side devices ?
Because a short from one leg to ground shunts current away from the low
side switches. It's the high-side switches that will see the fault current.

I see you're concerned about a short to ground rather than across the load.

In which case you can detect the high side current and level-shift the signal.

Also a single sense resistor wouldn't allow implementation of bipolar
current sensing,

Indeed but why would you need to do that ?

DC servo amp, not AC coupled audio. Need 4-quadrant control.

I don't see the issue. Consider audio as slowly moving DC if you like.

Either there's an overcurrent issue or there's not. Since it's PWM you'll see the
overcurrent anyway,

Graham