Resistor pulse handling?

[Chris Carlen]

Greetings:

I wish to sample a raw 75V 125kHz H-bridge output. The two H outputs
will go to a differential amplifer/attenuator to create a single ended,
low-level sample of the output voltage. Trouble is, I don't want the
raw signals running along my PCB from the high powered department to the
low level zone and radiating spikes all over the place. Thus, I'd like
to slow the edges just a bit before routing them long distances.

So use an RC. Trouble is, for instance a 1.25MHz cutoff filter with a
500 ohm resistor could potentially dissipate 11.25W in the ressitor on
transitions, even though the RMS current is only 17.7mA.

The question is, can a simple 0.25W 1210 resistor handle the transients?

If I make the resistors larger, then they contribute to errors in the
differential amp gain by being close to its input resistance. Well just
make them a part of the input resistors you say? Ok, but then doesn't
the AC CMRR suffer if the caps aren't matched?

Looking at the Panasonic ERJ 1% series, they allow only 2.5x rated power
for short term overload. But I suspect this isn't really the applicable
spec for this situation. There is nothing in the datato help one figure
out the device's tolerance to repetitive VERY short term overload, where
average power is still well within the continuous rating.


Comments appreciated.


Good day!



--
_______________________________________________________________________
Christopher R. Carlen
Principal Laser/Optical Technologist
Sandia National Laboratories CA USA
crcarleRemoveThis@BOGUSsandia.gov
NOTE, delete texts: "RemoveThis" and "BOGUS" from email address to reply.

 

[Mike Monett]

Chris Carlen wrote:

Greetings:

I wish to sample a raw 75V 125kHz H-bridge output. The two H outputs
will go to a differential amplifer/attenuator to create a single ended,
low-level sample of the output voltage. Trouble is, I don't want the
raw signals running along my PCB from the high powered department to the
low level zone and radiating spikes all over the place. Thus, I'd like
to slow the edges just a bit before routing them long distances.

[...]

Comments appreciated.

Good day!

Christopher R. Carlen
Principal Laser/Optical Technologist
Sandia National Laboratories CA USA
crcarleRemoveThis@BOGUSsandia.gov
NOTE, delete texts: "RemoveThis" and "BOGUS" from email address to reply.

Chris,

Why not use a passive divider to get down to 50 ohms, and skip the diff amp? You
might miss some important stuff with a low-pass filter.

This allows you to look at either output individually to see if one side has a
problem such as distortion, oscillation, risetime, etc. You can connect both
outputs to the scope and use the diff amp in the scope and add them together.

A 10k resistor into 50 ohms will give plenty of bandwidth to see oscillations
well into the VHF region. With 50% duty cycle, you would only need 1/2 watt.

I find every time I throw information away, it turns out to be just the piece I
need to solve a problem

Mike Monett

 

[Mike Monett]

Chris,

Another thought - with an H bridge, you definitely don't want both sides
on at the same time. It would be difficult to monitor this with a single
diff amp, but easy with separate outputs.

Mike Monett

 

[Robert Baer]

Chris Carlen wrote:

Greetings:

I wish to sample a raw 75V 125kHz H-bridge output. The two H outputs
will go to a differential amplifer/attenuator to create a single ended,
low-level sample of the output voltage. Trouble is, I don't want the
raw signals running along my PCB from the high powered department to the
low level zone and radiating spikes all over the place. Thus, I'd like
to slow the edges just a bit before routing them long distances.

So use an RC. Trouble is, for instance a 1.25MHz cutoff filter with a
500 ohm resistor could potentially dissipate 11.25W in the ressitor on
transitions, even though the RMS current is only 17.7mA.

The question is, can a simple 0.25W 1210 resistor handle the transients?

If I make the resistors larger, then they contribute to errors in the
differential amp gain by being close to its input resistance. Well just
make them a part of the input resistors you say? Ok, but then doesn't
the AC CMRR suffer if the caps aren't matched?

Looking at the Panasonic ERJ 1% series, they allow only 2.5x rated power
for short term overload. But I suspect this isn't really the applicable
spec for this situation. There is nothing in the datato help one figure
out the device's tolerance to repetitive VERY short term overload, where
average power is still well within the continuous rating.


Comments appreciated.


Good day!



Since you definitely are not considering carbon composition

resistors, then decades of experience with them is not useful.
Perhaps the following compromise can be made.
1) start with a 100 ohm resistor from the source to node 1 that has a
(filter) capacitor to ground; set the capacitor value to ten percent of
the original calculated value. Peak pulse dissipation is fairly low, as
there is less energy in the harmonics.
2) from node 1 to node 2 use a 200 ohm resistor and at node 2 use a
capacitor that is 40 percent of the original calculated value. Do not
know energy (spectrum) content of what flows thru resistor thru
capacitor, but some whiz could do the math. I would dare say that
energy may be in the same ballpark.
3) from node 2 to node 3 use a 200 ohm resistor and at node 3 use a
capacitor that is 80 percent of the original calculated value. Do not
know energy (spectrum) content of what flows thru resistor thru
capacitor, but some whiz again could do the math. Guess in similar ballpark.
The idea is to dissipate the fastest stuff first, and do the rest
later; spreading the energy in steps.
The values may need adjustment, but the idea is given for evaluation,
not as an exact solution.
**
Alternate method: use a ferrite bead in series with the source to the
bypass node; ther aer SMT versions as well as leaded versions.

 

[Chris Carlen]

Mike Monett wrote:

Chris Carlen wrote:
Greetings:

I wish to sample a raw 75V 125kHz H-bridge output. The two H outputs
will go to a differential amplifer/attenuator to create a single ended,
low-level sample of the output voltage. Trouble is, I don't want the
raw signals running along my PCB from the high powered department to the
low level zone and radiating spikes all over the place. Thus, I'd like
to slow the edges just a bit before routing them long distances.

[...]

Comments appreciated.

Good day!

Christopher R. Carlen
Principal Laser/Optical Technologist
Sandia National Laboratories CA USA
crcarleRemoveThis@BOGUSsandia.gov
NOTE, delete texts: "RemoveThis" and "BOGUS" from email address to reply.

Chris,

Why not use a passive divider to get down to 50 ohms, and skip the diff amp? You
might miss some important stuff with a low-pass filter.

This allows you to look at either output individually to see if one side has a
problem such as distortion, oscillation, risetime, etc. You can connect both
outputs to the scope and use the diff amp in the scope and add them together.

Perhaps I should have explained more. I agree with you that a 10k:50
ohm divider would provide a nice diagnostic port. But the purpose of
the diff amp is specifically differential to single ended conversion in
order to implement a voltage mode control loop around the PWM amplifier
H bridge.

A 10k resistor into 50 ohms will give plenty of bandwidth to see oscillations
well into the VHF region. With 50% duty cycle, you would only need 1/2 watt.

Indeed.

I find every time I throw information away, it turns out to be just the piece I
need to solve a problem

Thanks for the input Mike.

Good day!


--
_______________________________________________________________________
Christopher R. Carlen
Principal Laser/Optical Technologist
Sandia National Laboratories CA USA
crcarleRemoveThis@BOGUSsandia.gov
NOTE, delete texts: "RemoveThis" and "BOGUS" from email address to reply.

 

[Chris Carlen]

Robert Baer wrote:

Since you definitely are not considering carbon composition resistors,
then decades of experience with them is not useful.

I am aware of their pulse handling capability. Trouble is tolerance.

Hmm. Perhaps it would work anyway, since the carbon component of the
total resistance into the diff amp would be a very small one.

However, I have discovered that my fears regarding screwing the
performance of the diff amp are unwarranted, and that I can use a larger
SMT resistance that is still within it's surge rating.

Perhaps the following compromise can be made.
1) start with a 100 ohm resistor from the source to node 1 that has a
(filter) capacitor to ground; set the capacitor value to ten percent of
the original calculated value. Peak pulse dissipation is fairly low, as
there is less energy in the harmonics.
2) from node 1 to node 2 use a 200 ohm resistor and at node 2 use a
capacitor that is 40 percent of the original calculated value. Do not
know energy (spectrum) content of what flows thru resistor thru
capacitor, but some whiz could do the math. I would dare say that
energy may be in the same ballpark.
3) from node 2 to node 3 use a 200 ohm resistor and at node 3 use a
capacitor that is 80 percent of the original calculated value. Do not
know energy (spectrum) content of what flows thru resistor thru
capacitor, but some whiz again could do the math. Guess in similar
ballpark.
The idea is to dissipate the fastest stuff first, and do the rest
later; spreading the energy in steps.
The values may need adjustment, but the idea is given for evaluation,
not as an exact solution.

Interesting method. Can't afford the space for a lot of parts though.

Alternate method: use a ferrite bead in series with the source to the
bypass node; ther aer SMT versions as well as leaded versions.

Good to keep in mind as well.


Thanks for the reply.


Good day!





--
_______________________________________________________________________
Christopher R. Carlen
Principal Laser/Optical Technologist
Sandia National Laboratories CA USA
crcarleRemoveThis@BOGUSsandia.gov
NOTE, delete texts: "RemoveThis" and "BOGUS" from email address to reply.

 

[Ban]

Chris Carlen wrote:

Mike Monett wrote:
Chris Carlen wrote:
Greetings:

I wish to sample a raw 75V 125kHz H-bridge output. The two H
outputs will go to a differential amplifer/attenuator to create a
single ended, low-level sample of the output voltage. Trouble is,
I don't want the raw signals running along my PCB from the high
powered department to the low level zone and radiating spikes all
over the place. Thus, I'd like to slow the edges just a bit before
routing them long distances.

[...]

Comments appreciated.

Good day!

Christopher R. Carlen
Principal Laser/Optical Technologist
Sandia National Laboratories CA USA
crcarleRemoveThis@BOGUSsandia.gov
NOTE, delete texts: "RemoveThis" and "BOGUS" from email address to
reply.

Chris,

Why not use a passive divider to get down to 50 ohms, and skip the
diff amp? You might miss some important stuff with a low-pass filter.

This allows you to look at either output individually to see if one
side has a problem such as distortion, oscillation, risetime, etc. You
can connect both outputs to the scope and use the diff amp in
the scope and add them together.

Perhaps I should have explained more. I agree with you that a 10k:50
ohm divider would provide a nice diagnostic port. But the purpose of
the diff amp is specifically differential to single ended conversion
in order to implement a voltage mode control loop around the PWM
amplifier H bridge.

A 10k resistor into 50 ohms will give plenty of bandwidth to see
oscillations well into the VHF region. With 50% duty cycle, you
would only need 1/2 watt.

Indeed.

I find every time I throw information away, it turns out to be just
the piece I need to solve a problem



Thanks for the input Mike.

Good day!

I would recommend the THS4502 or AD8132 true differential amplifiers. I used
one in a similar application and they are fast enough to follow the rise and
fall times of a H-Bridge.
The inverted output can come very handy for delta-sigma or integrator stages
of higher order or any other analog signal processing. The differential
output can also drive a high-speed ADC directly.
--
ciao Ban
Bordighera, Italy

 

[Mike Monett]

Chris Carlen wrote:

[...]

Perhaps I should have explained more. I agree with you that a 10k:50
ohm divider would provide a nice diagnostic port. But the purpose of
the diff amp is specifically differential to single ended conversion in
order to implement a voltage mode control loop around the PWM amplifier
H bridge.

Ha! That's a completely different problem

You may have mentioned it in another post - have separate RC filters on
each side going to ground, that feed the inputs of the op amp. Then slow
the loop bandwidth down enough to follow the average output.

This is similar to the problem of handling fast pulses in pll error
amplifiers.

[...]

Thanks for the input Mike.

Thanks for the explanation!

Good day!
Christopher R. Carlen
Principal Laser/Optical Technologist
Sandia National Laboratories CA USA
crcarleRemoveThis@BOGUSsandia.gov

Mike Monett

 

[John - KD5YI]

Chris Carlen wrote:

Greetings:

I wish to sample a raw 75V 125kHz H-bridge output. The two H outputs
will go to a differential amplifer/attenuator to create a single ended,
low-level sample of the output voltage. Trouble is, I don't want the
raw signals running along my PCB from the high powered department to the
low level zone and radiating spikes all over the place. Thus, I'd like
to slow the edges just a bit before routing them long distances.

So use an RC. Trouble is, for instance a 1.25MHz cutoff filter with a
500 ohm resistor could potentially dissipate 11.25W in the ressitor on
transitions, even though the RMS current is only 17.7mA.

The question is, can a simple 0.25W 1210 resistor handle the transients?

If I make the resistors larger, then they contribute to errors in the
differential amp gain by being close to its input resistance. Well just
make them a part of the input resistors you say? Ok, but then doesn't
the AC CMRR suffer if the caps aren't matched?

Looking at the Panasonic ERJ 1% series, they allow only 2.5x rated power
for short term overload. But I suspect this isn't really the applicable
spec for this situation. There is nothing in the datato help one figure
out the device's tolerance to repetitive VERY short term overload, where
average power is still well within the continuous rating.


Comments appreciated.


Good day!

I know the following link is not exactly what you're looking for, but I hope
is some information you could use. Vishay uses a MELF-style resistor for an
example in the document but they go on to say it is applicable to other
styles and in carbon film.

See the PDF document at the bottom of the page titled "Pulse Load Handling
for Fixed Linear Resistors."

http://www.vishay.com/resistors-discrete/case-0603-and-larger/surface-mount/related#technt

Good luck.

John

 

[John Larkin]

On Tue, 10 May 2005 15:42:07 -0700, Chris Carlen
<crcarleRemoveThis@BOGUSsandia.gov> wrote:

Greetings:

I wish to sample a raw 75V 125kHz H-bridge output. The two H outputs
will go to a differential amplifer/attenuator to create a single ended,
low-level sample of the output voltage. Trouble is, I don't want the
raw signals running along my PCB from the high powered department to the
low level zone and radiating spikes all over the place. Thus, I'd like
to slow the edges just a bit before routing them long distances.

So use an RC. Trouble is, for instance a 1.25MHz cutoff filter with a
500 ohm resistor could potentially dissipate 11.25W in the ressitor on
transitions, even though the RMS current is only 17.7mA.

The question is, can a simple 0.25W 1210 resistor handle the transients?

If I make the resistors larger, then they contribute to errors in the
differential amp gain by being close to its input resistance. Well just
make them a part of the input resistors you say? Ok, but then doesn't
the AC CMRR suffer if the caps aren't matched?

Looking at the Panasonic ERJ 1% series, they allow only 2.5x rated power
for short term overload. But I suspect this isn't really the applicable
spec for this situation. There is nothing in the datato help one figure
out the device's tolerance to repetitive VERY short term overload, where
average power is still well within the continuous rating.


Comments appreciated.


Good day!

Sounds fine to me.

John

 

[Terry Given]

Chris Carlen wrote:

Greetings:

I wish to sample a raw 75V 125kHz H-bridge output. The two H outputs
will go to a differential amplifer/attenuator to create a single ended,
low-level sample of the output voltage. Trouble is, I don't want the
raw signals running along my PCB from the high powered department to the
low level zone and radiating spikes all over the place. Thus, I'd like
to slow the edges just a bit before routing them long distances.

So use an RC. Trouble is, for instance a 1.25MHz cutoff filter with a
500 ohm resistor could potentially dissipate 11.25W in the ressitor on
transitions, even though the RMS current is only 17.7mA.

The question is, can a simple 0.25W 1210 resistor handle the transients?

If I make the resistors larger, then they contribute to errors in the
differential amp gain by being close to its input resistance. Well just
make them a part of the input resistors you say? Ok, but then doesn't
the AC CMRR suffer if the caps aren't matched?

Looking at the Panasonic ERJ 1% series, they allow only 2.5x rated power
for short term overload. But I suspect this isn't really the applicable
spec for this situation. There is nothing in the datato help one figure
out the device's tolerance to repetitive VERY short term overload, where
average power is still well within the continuous rating.


Comments appreciated.


Good day!

Hi Chris,

see if you can find a copy of Philips ACM2 databook, they have a lot of
pretty pictures of peak pulse power rating for various smt resistors. an
RC01 1206 resistor can handle a single 10W pulse 4us long, and
repetitive 3W pulses 1us long and 1ms apart. A PRC201 1218 will do 10W
for 10us every 10ms, continuously.

IRC make a range of seriously impressive parts, with a Joule rating -
stay within the Joule rating, and the average power rating, and you are
OK. CHP1/8 would eat your task, as would any MMA0204 part (roughly 1206
sized MELF-like package)

the problem is simply thermal. smt resistors have bugger all thermal
mass, and so cant reliably take large pulses.

not all wirewound resistors are created equal, either . Vitrohm make
great parts. I have done a lot of peak pulse power testing of WW R's,
the cheap ones have a wound former sitting in a U channel in the ceramic
case, with stuff poured over the top. Large pulses can be seen as a red
glow through the back of the ceramic - air gap = localised hotspot =
sudden/gradual failure. Better parts ensure the wound former is fully
encapsulated, removing air gaps and extracting the heat better.

Cheers
Terry