Searching \ for '[EE] blown diodes' in subject line. ()
Make payments with PayPal - it's fast, free and secure! Help us get a faster server
FAQ page: www.piclist.com/techref/index.htm?key=blown+diodes
Search entire site for: 'blown diodes'.

Exact match. Not showing close matches.
PICList Thread
'[EE] blown diodes'
2006\04\21@060952 by Wouter van Ooijen

face picon face
part 1 1402 bytes content-type:text/plain; (decoded 7bit)

Next problem with my motor driver hardware :(

(gif of relevant part attached)

My customer confronted me with a unit in which at least one of the
supression diodes (D7, 1N5819) was totally blown. I think both FETs and
the other diode were also dead, but with less visible damage. After
replacing the two diodes and the two FETs the unit seems to work OK
again.

Vmotor = 24V, 5A supply, motor stall current ~ 300 mA (normal operation
50..100 mA) , there is stall detection that switches the motor of after
~ 20 ms. There is reverse-polarity protection in the circuit (P fet in
the power line). As far as the customer could tell the problem (very bad
smell) occured while the unit was not touched (so reverse polarity is
unlikely, but for instance a short in the cable to the motor might be
possible). The wiring to the motors does contain a gnd wire, but no
Vmotor wire.

I can imgaine various errors (like shorting the motor lead to ground)
that could destroy the FETs (P fet first, when that one fails short the
N fet will follow soon). But what failure scenario (except reverse
power, but that is very unlikely) would blow the top (and/or bottom)
diode?

Wouter van Ooijen

-- -------------------------------------------
Van Ooijen Technische Informatica: http://www.voti.nl
consultancy, development, PICmicro products
docent Hogeschool van Utrecht: http://www.voti.nl/hvu



part 2 5003 bytes content-type:image/gif; (decode)


part 3 35 bytes content-type:text/plain; charset="us-ascii"
(decoded 7bit)

2006\04\21@062900 by Vasile Surducan

face picon face
Wouter, we have talk about this several times on this list and on others.
1N5819 has only 40V breakdown voltage. It's less than 2x24V requested
by any engineering design requirements. Also the motor load is 5A,
while your maximum supressed current is 1A (25A/8mS maximum maximorum
peak current).
Change diode with at least 100V one and check if your maximum peak
current through diodes tooks more than 2-3mS. If yes, choose a 2A
diode.

greetings,
Vasile


On 4/21/06, Wouter van Ooijen <spam_OUTwouterTakeThisOuTspamvoti.nl> wrote:
{Quote hidden}

> -

2006\04\21@063850 by Luisizan

flavicon
face
Hi Wouter and all.

Perhaps spikes and reverse voltage of diodes and transistors could be
the problem. Thats if an strange voltage incoming o misconexion is not
possible.

In my controls I dont use Schottky diodes at this application. Reverse
voltage of semiconductors is always 4-6 times the supply. Also use an rc
network at the output.

Best regards

Luis Izquierdo

Wouter van Ooijen escribió:

{Quote hidden}

2006\04\21@065432 by Wouter van Ooijen

face picon face
> Wouter, we have talk about this several times on this list
> and on others.
> 1N5819 has only 40V breakdown voltage. It's less than 2x24V requested
> by any engineering design requirements.

you might have a point here

> Also the motor load is 5A,

no, the power supply is 5A, the *stall* current of this motor (there is
a heavier motor too, this is the smaller one) is only 300 mA.

> while your maximum supressed current is 1A (25A/8mS maximum maximum
> peak current).
> Change diode with at least 100V one and check if your maximum peak
> current through diodes tooks more than 2-3mS. If yes, choose a 2A
> diode.

which diode would you suggest for 2A/100V?

I think your points are probably valid, but not the root cause of this
problem. The circuit contains two H-bridges, and this problem occured in
one leg of the H bridge for the smaller motor!

Oh, and you are quick responder: I don't even see my original question
on the list :)

Wouter van Ooijen

-- -------------------------------------------
Van Ooijen Technische Informatica: http://www.voti.nl
consultancy, development, PICmicro products
docent Hogeschool van Utrecht: http://www.voti.nl/hvu


2006\04\21@070309 by Wouter van Ooijen

face picon face
> In my controls I dont use Schottky diodes at this
> application. Reverse
> voltage of semiconductors is always 4-6 times the supply.
> Also use an rc
> network at the output.

I am a digital guy, how should I calculate that RC network?

Wouter van Ooijen

-- -------------------------------------------
Van Ooijen Technische Informatica: http://www.voti.nl
consultancy, development, PICmicro products
docent Hogeschool van Utrecht: http://www.voti.nl/hvu


2006\04\21@071021 by Howard Winter

face
flavicon
picon face
Wouter,

On Fri, 21 Apr 2006 12:54:31 +0200, Wouter van Ooijen wrote:
>...
> I don't even see my original question on the list :)

Nor do I - 40 minutes after Vasile's reply, I've seen a total of 4 replies, but still not the original!

Cheers,


Howard Winter
St.Albans, England


2006\04\21@080004 by olin piclist

face picon face
> (gif of relevant part attached)

Several things:

What is R119 doing there?  It looks like it's just asking for trouble.  If
whatever is driving the left end of R112 fails open, or is at high impedence
during power up or reset or whatever, Q7 will be turned on.  Sounds like a
bad idea.

Your message says D7 is a 1N5818, which is a common Schottky diode.  The
schematic shows it as being BA159, which I'm not familiar with, but it's
shown as an ordinary silicon diode, not a Schottky.

Why the separate diodes anyway?  The FETs already contain body diodes
exactly in parallel with D7 and D8.  If D7 and D8 really are Schottky (then
they should be shown as such) then I can see the point was probably to
reduce the on voltage and gain the much faster reverse recovory time of the
Schottky.

If D7 and D8 are indeed 1N5818, then you need to think about power
dissipation due to reverse leakage, at least at maximum temperature.  If I
remember right (your job to check), 1N5818 can have 20mA (yes, milliamps) of
leakage over the full operating range.  20mA x 24V = 480mW.  That will
probably fry them if already at the max ambient limit.  If not, you are
still in danger of thermal runaway even starting at normal temperatures.
Once they get a little warmer they leak a little more, making them a little
warmer, etc.

> My customer confronted me with a unit in which at least one of the
> supression diodes (D7, 1N5819) was totally blown. I think both FETs and
> the other diode were also dead, but with less visible damage. After
> replacing the two diodes and the two FETs the unit seems to work OK
> again.

You can't go by what looks visibly most blown to determine what the original
failure was.  Something conducted when it shouldn't have, but it is
impossible to tell from all four being dead what happened first.

> Vmotor = 24V, 5A supply, motor stall current ~ 300 mA (normal operation
> 50..100 mA)

Worrying about the extra voltage drop due to silicon diodes versus Schottkys
at this low current isn't worth it.  I'm really starting to think Schottky
D7 and D8 are more of a problem than a solution.  What happens when you just
remove them?

> there is stall detection that switches the motor of after
> ~ 20 ms.

Wow, that's really short.  I'm surprised your motor picks up significant
speed in 20mS.  However, this is a different problem.  An overly aggressive
stall current shutdown shouldn't have caused the problem you are seeing.
Hmm, unless...  Does the stall shutdown completely shut down the motor for a
time, or does it limit the current somehow?  If the latter, this is probably
kicking in every time the motor is started, possibly causing high
dissipation somewhere depending on how it's implemented.

> but for instance a short in the cable to the motor might be
> possible).

This again comes down to exactly how over current is detected and handled.
You definitely don't want to go into linear mode on over current.  I've seen
some bad designs that do exactly that.  They appear to function for a while
because over current never happens.  But when it does, the unit basically
destroys itself.

> But what failure scenario (except reverse
> power, but that is very unlikely) would blow the top (and/or bottom)
> diode?

Again, I don't see why those diodes are there in the first place.  However,
one easy way to blow them would be to drive the motor externally.  All the
resulting generator power ends up being dumped into those diodes.  One of
them failing short then causes a cascade of failures.


******************************************************************
Embed Inc, Littleton Massachusetts, (978) 742-9014.  #1 PIC
consultant in 2004 program year.  http://www.embedinc.com/products

2006\04\21@080234 by olin piclist

face picon face
> Wouter, we have talk about this several times on this list and on others.
> 1N5819 has only 40V breakdown voltage. It's less than 2x24V requested
> by any engineering design requirements.

But, assuming the other side of the motor is driven by a similar half H
bridge, there is still never more than the 24V Vmotor (plus a coulple of
diode drops) accross any of the diodes.


******************************************************************
Embed Inc, Littleton Massachusetts, (978) 742-9014.  #1 PIC
consultant in 2004 program year.  http://www.embedinc.com/products

2006\04\21@084809 by Wouter van Ooijen

face picon face
part 1 4548 bytes content-type:text/plain; (decoded 7bit)

> What is R119 doing there?  It looks like it's just asking for
> trouble.  If
> whatever is driving the left end of R112 fails open, or is at
> high impedence
> during power up or reset or whatever, Q7 will be turned on.  
> Sounds like a
> bad idea.

That's part of the driver for the FETs. I can post that part of the
circuit too, but I think it is unlikely to be part of the problem.
Anyway: more circuit attached. Maybe it will show me a fool, but in that
case I wille end up as a wiser fool :) Bottom right side shows the 2 x
1R 'carbon' resistors that are no replaced by a 5W 0R47 wire-wound.

> Your message says D7 is a 1N5818, which is a common Schottky
> diode.  The
> schematic shows it as being BA159, which I'm not familiar
> with, but it's
> shown as an ordinary silicon diode, not a Schottky.

I never updated the circuit. 1N5819 is what I use.

> Why the separate diodes anyway?  The FETs already contain body diodes
> exactly in parallel with D7 and D8.  If D7 and D8 really are
> Schottky (then
> they should be shown as such) then I can see the point was probably to
> reduce the on voltage and gain the much faster reverse
> recovory time of the Schottky.

I assumed Schottky's to be better (= faster) at surpressing the spikes
than the body diodes.

{Quote hidden}

I'll have to check that. The total circuit (two H-bridges + 18F252 at
5V) draws 40..70 mA at rest.

> You can't go by what looks visibly most blown to determine
> what the original
> failure was.  Something conducted when it shouldn't have, but it is
> impossible to tell from all four being dead what happened first.

true, but I try to get all information I get from a failure!

> Worrying about the extra voltage drop due to silicon diodes
> versus Schottkys
> at this low current isn't worth it.  I'm really starting to
> think Schottky
> D7 and D8 are more of a problem than a solution.  What
> happens when you just remove them?

I don't think it will make any immediate difference. (I don't have a
motor unit right now, so I can't test).

> Wow, that's really short.  I'm surprised your motor picks up
> significant speed in 20mS.

No, it is meant to detect a stalled motor being started.

> However, this is a different problem. An
> overly aggressive stall current shutdown shouldn't have caused the
problem you
> are seeing.

I just mentioned it to rule out a long stall condition.

> Hmm, unless...  Does the stall shutdown completely shut down
> the motor for a
> time, or does it limit the current somehow?

Audible alarm, and shutdown untill the user button that caused the motor
to run is released, and the motor is started only when after that a
(same or different) button is pressed. (Actual implemntation: shutdown
first, then audible alarm).

> This again comes down to exactly how over current is detected
> and handled.

It is a PWM with very limited use of PWM: start at 100% to retreat from
a stall (end of a screw), after ~200 ms back to 70%.

> Again, I don't see why those diodes are there in the first
> place.  However,
> one easy way to blow them would be to drive the motor
> externally.  All the
> resulting generator power ends up being dumped into those
> diodes.

Is that true? I would expect all the power to be transferred to the
power lines. There is a big (3W) 36V zener there that would get the
power, if the actual power supply does not accept power back (in this
case it was a power supply which probwbly would not, normal opartion is
with an accu which would).

Anyway, I'll probably try without the diodes. Saves some board area and
assemblt cost :)

Wouter van Ooijen

-- -------------------------------------------
Van Ooijen Technische Informatica: http://www.voti.nl
consultancy, development, PICmicro products
docent Hogeschool van Utrecht: http://www.voti.nl/hvu



part 2 12997 bytes content-type:image/gif; (decode)


part 3 35 bytes content-type:text/plain; charset="us-ascii"
(decoded 7bit)

2006\04\21@100515 by Alan B. Pearce

face picon face
>That's part of the driver for the FETs. I can post that part of the
>circuit too, but I think it is unlikely to be part of the problem.

Actually I think that is precisely where your problem is. The bottom FET
will switch faster than the top one in the totem pole, because you have a
capacitor on the gate of the top FET, but not the bottom one. Hence when the
top one is on, bottom off, and you swap states there will be significant
shoot through current while the top one turns off.

2006\04\21@103857 by Wouter van Ooijen

face picon face
> Actually I think that is precisely where your problem is. The
> bottom FET
> will switch faster than the top one in the totem pole,
> because you have a
> capacitor on the gate of the top FET, but not the bottom one.
> Hence when the
> top one is on, bottom off, and you swap states there will be
> significant
> shoot through current while the top one turns off.

IIRC (this circuit was designed over a year ago) the shootthrough is
minimal, *because* the capacitor is there. I can't reproduce the logic
now, but without the cap it did not work well.

Wouter van Ooijen

-- -------------------------------------------
Van Ooijen Technische Informatica: http://www.voti.nl
consultancy, development, PICmicro products
docent Hogeschool van Utrecht: http://www.voti.nl/hvu


2006\04\21@113251 by Luisizan

flavicon
face
Hi Wouter and all:

Wouter van Ooijen escribió:

>I am a digital guy, how should I calculate that RC network?
>
>  
>

I use the controls in an industrial very electricaly noisy environment.
I've seen sometimes an rc of 12r and 1nF similar to other controls that
some welding controls use, but I've not calculated these values,  they
depend of the PWM frecuency.

Tired to see sophisticated circuits failing I'm testing the oulput
circuit that Unitrode propose in the UC3637 application note, adapted to
really work with 0-15 pwm pulses (UC3637 generated) and added four BYW29
diodes. As yours, this circuit allows to run at full power (speed) with
a continous ON signal. Charge pump circuits don't allow this. It seems
to work ok.

Best regards

Luis Izquierdo

2006\04\21@115625 by olin piclist

face picon face
Wouter van Ooijen wrote:
> Anyway: more circuit attached.

I assume you set only one of RC0 or RC3 high depending on direction, then
put the PWM signal on RC1?  I didn't notice this before, but what are C7 and
C8 supposed to accomplish?  These will slow down switching of the top FETs,
which I guess is OK if you are reversing direction infrequently.  Still, I
don't see any advantage from them.

A possible issue is the gate voltage of the lower FETs when they are
supposed to be off.  Aren't the ULN2003 darlington drivers?  If so, figure
at least 1V when on.  Then add in the voltage from the divider formed by
R119 and R112, which is another 580mV.  The lower FET gates will therefore
be at 1.6V when they are supposed to be off.  I haven't looked up IRFR024,
so I don't know if this is a problem or not.  These kinds of FETs are often
poorly specified in this region.  They may have substantially more leakage
at 1.6V than at 0V.  Unless they were specified for some maximum current or
minimum resistance at 1.6V or higher, I would be little nervous although
probably has nothing to do with your problem.

Do you realize that the top zeners D18 and D15 will never come on given 24V
supply?  R123 and R124 form a divider so that there can not be more than 12V
accross R123 even if IC3-6 would drive all the way to ground.

You will have to leave considerable dead time when switching direction.  The
limiting factor is not the FETs, but the way they are driven.  R123 and C7
have a time constant of 5uS, and the effective gate capacitance could make
this considerably larger.  Then you want to wait probably 3 time constants.
Just from quick calculations without looking up the details, it looks like
you need a dead time of at least 100uS on switching direction.  Making this
too short could cause damage eventually.

> I assumed Schottky's to be better (= faster) at surpressing the spikes
> than the body diodes.

Yes, they would be.  Hopefully you have at least some capacitance accross
the motor right by the motor.  You don't want too much else there will be
switching transient currents as the cap is charged and discharged each PWM
pulse, but you want some just to slow down the voltage edges.  With a
reasonable cap by the motor and another on the board for additional
protection, the body diodes should be able to react fast enough.  They are
usually about the speed of the FET.

> No, it is meant to detect a stalled motor being started.

What I meant is that just an ordinary stopped motor will likely look stalled
in the first 20mS.  Unless you've got an exceptionally agile motor, it's
probably not going to pick up significant speed from a dead stop in the
first 1/50 second.

>> All the
>> resulting generator power ends up being dumped into those
>> diodes.
>
> Is that true? I would expect all the power to be transferred to the
> power lines.

Yeah, I said that wrong.  I meant to say all the *current* will go thru
those diodes.  Most of the power will be dumped back onto the supply as you
say.


******************************************************************
Embed Inc, Littleton Massachusetts, (978) 742-9014.  #1 PIC
consultant in 2004 program year.  http://www.embedinc.com/products

2006\04\21@130643 by olin piclist

face picon face
Wouter van Ooijen wrote:
> IIRC (this circuit was designed over a year ago) the shootthrough is
> minimal, *because* the capacitor is there. I can't reproduce the logic
> now, but without the cap it did not work well.

Then you are not leaving enough dead time in the firmware (I assume a PIC is
driving the RC0-RC3 signals).  C7 and C8 should not be necessary if the
proper timing and control of the RC0-RC3 lines is realized.


******************************************************************
Embed Inc, Littleton Massachusetts, (978) 742-9014.  #1 PIC
consultant in 2004 program year.  http://www.embedinc.com/products

2006\04\21@130746 by Wouter van Ooijen

face picon face
> Tired to see sophisticated circuits failing I'm testing the oulput
> circuit that Unitrode propose in the UC3637 application note,
> adapted to
> really work with 0-15 pwm pulses (UC3637 generated) and added
> four BYW29 diodes.

BYW29!! for what maximum motor current was your circuit?

Wouter van Ooijen

-- -------------------------------------------
Van Ooijen Technische Informatica: http://www.voti.nl
consultancy, development, PICmicro products
docent Hogeschool van Utrecht: http://www.voti.nl/hvu


2006\04\21@133231 by Wouter van Ooijen

face picon face
> I assume you set only one of RC0 or RC3 high depending on
> direction, then put the PWM signal on RC1?

correct, PIC is 18F252, C1 is one of the PWM outputs

> I didn't notice this before, but what are C7 and
> C8 supposed to accomplish?  These will slow down switching of
> the top FETs, which I guess is OK if you are reversing direction
> infrequently.

I do switch direction infrequently. But I can realy reproduce why they
are there. I dawns me that they might have been required only in a
previous vesrsion, where I PWMed both FETs.

> Still, I don't see any advantage from them.

You recall me mentioning 'a wiser fool'?

> A possible issue is the gate voltage of the lower FETs when they are
> supposed to be off.

I don't think the ULN will drop 1V at 10 mA, but I will check.

> Do you realize that the top zeners D18 and D15 will never
> come on given 24V
> supply?  R123 and R124 form a divider so that there can not
> be more than 12V
> accross R123 even if IC3-6 would drive all the way to ground.

I have experimented somewhat with that divider, but yes, with these
values the zeners are not needed, for reasonable values of the power
supply. (Wouter: make me a motor driver for 24V. One year later: it will
also be used on an accu, with the charger attached. sigh...)

> Just from quick calculations without looking up the details,
> it looks like
> you need a dead time of at least 100uS on switching
> direction.  Making this
> too short could cause damage eventually.

100uS direction switching is not very probable, to put it mildly. (PWM
is ~ 30 kHz == 30us!) The control loop is at ~10ms, and the control is
by the user pressing keys and/or reading from a gravity sensor. But I
will still look into those capacitors, every component off is one less
to solder and one less that could fail.

> Yes, they would be.  Hopefully you have at least some
> capacitance accross the motor right by the motor.

At the moment none. I experimented with 1 nF but could see no effect at
all.

If I use a cap there, how do I determine the value (at type of cap)? Or
should it be an R-C?

> With a reasonable cap by the motor and another on the board

again, which values (or which method to calculate the values)?

{Quote hidden}

But anyway, you said that for the case that the motor was spun and
forced to act as dynamo. That is almost impossible, the motors are
geared down and then drive a worm. When I once had to force the worm
something went 'crack' in the gearbox.

As often (always?) the piclist has given me considerable feedback on my
circuit to think about, even though I am not sure the culprit of this
particular failure is identified (but the reverse-leakage of the diodes
is a thing to check, especially because they are next to the FETs and
could get heated).

Wouter van Ooijen

-- -------------------------------------------
Van Ooijen Technische Informatica: http://www.voti.nl
consultancy, development, PICmicro products
docent Hogeschool van Utrecht: http://www.voti.nl/hvu


2006\04\21@143741 by Peter

picon face


On Fri, 21 Apr 2006, Wouter van Ooijen wrote:

>> In my controls I dont use Schottky diodes at this
>> application. Reverse
>> voltage of semiconductors is always 4-6 times the supply.
>> Also use an rc
>> network at the output.
>
> I am a digital guy, how should I calculate that RC network?

You want to slow the pulses from the motor lines to within the switching
speed of your semiconductors. E.g. the line to the motor is a
transmission line and when your circuit chops the current a very fast
spike appears. Depending on how good (!) the wiring is (= high Q) this
can exceed the supply voltage 10-20 times if uncontrolled. The first
move is to slow it down. A series RC in parallel with the motor directly
at the controller can do this, you can calculate it to consume say 3% of
your motor power. Calculate R so it takes ~5% of the motor inrush
(assume it is in parallel with the motor) and then calculate C at the
working freq. (lowest = 50% duty c. pwm) to be at the 3dB corner with
that R. This is not exact but it should solve your problem quickly.

Peter

2006\04\21@150439 by olin piclist

face picon face
Wouter van Ooijen wrote:
> 100uS direction switching is not very probable, to put it mildly. (PWM
> is ~ 30 kHz == 30us!)

Yikes!  That's way too high for this circuit.  Your drivers are slow.  I bet
the FETs are spending a significant fraction of their time in transition.
Are they getting warm?  There is no reason they need to be, but I bet they
are at full load.

This is a motor with a big clunky chunk of iron.  It's not going to do
anything meaningful in a 1mS, let alone 30uS.  It would operate fine at a
few 100Hz PWM.  The only reason for higher speed is to reduce audible
whining.  There is no way to guess what frequencies will be audible and
objectionable with a specific motor.  You just have to try a few.  Hopefully
there is a frequency you can use in the 500Hz to 5KHz range that is
acceptable.

> The control loop is at ~10ms, and the control is
> by the user pressing keys and/or reading from a gravity sensor.

Yes, but what happens when the user hits the button to change direction and
the motor is currently being driven in the other direction.  The firmware
needs some special case timing.  You can't just turn off RC0, turn on RC3,
and resume PWM.  Unless the firmware deliberately shuts everything down and
waits 100-200uS before turning things on for the other direction, you can
get both top and bottom FETs on the same side on at the same time for 10s of
uS.

> At the moment none. I experimented with 1 nF but could see no effect at
> all.

Well you probably wouldn't see the difference in an obvious way.  I bet
you're not meeting CE RF emissions limits though.  The main problem is that
the motor looks mostly like an inductor, and can produce very fast voltage
edges.  These can be so fast that the clamping diodes don't have time to
react.  They eventually will, but that can cause stress that might
eventually cause failure.

> If I use a cap there, how do I determine the value (at type of cap)? Or
> should it be an R-C?

First you have to understand the reasons for the cap and the reasons for
limiting its value.  The reason is to slow down voltage edges to that the
rest of the circuit can deal with them properly.  The reason that limits the
cap's value is that it will be charged and discharged every PWM pulse, and
you don't want excessive currents thru the FETs just because of the cap.

1-10nF sounds good, but I suppose you want something more rigorous.  You
said that the maximum intended current is 300mA and you've got a 24V supply
and you are using a 33uS PWM period.  Therefore just to get a ballpark, what
size cap would slew 24V in 10uS at 300mA?

300ma * 10uS / 24V = 125nF

I'm not saying this is the value you want to use, but it provides some
perspective.  125nF would be too burdensome as it would take 1/3 of maximum
current just to charge/discharge the cap.  So in the end we get back to what
felt right in the first place, 1-10nF.  If you want a fixed answer, put 2nF
right at the motor as close to its terminals as possible, and another 2nF on
the board at the other end of the cable.  This will make a big difference in
RF emissions if nothing else.

So let's work backwards and do a sanity check.  If you're switching the full
300mA and your FET body diodes can turn on in 5nS (just picking a number,
you're job to look it up), then there can be a voltage excursion of

 300mA * 5nS / 4nF = 375mV

before the FET diode starts conducting.  That doesn't sound too bad.

Now let's sanity check the current.  Charging the caps to 24V requires

 24V * 4nF = 48nC

This would happen every PWM pulse, or at 30KHz that will add

 48nC * 30K/S = 1.4mA

to the apparent motor current drain, which sounds like no problem at all.
Also keep in mind that the PWM doesn't actually drive both directions each
pulse.  In reality the voltage on the motor will not go to 0 each pulse.  A
bigger issue is that the cap current happens immediately and will therefore
go thru the FET as it is still transitioning from off to on.  The cap
current will therefore heat the FET more than the same motor current, which
is spread out over the PWM pulse.  This is another reason to get the PWM
frequency lower.

> As often (always?) the piclist has given me considerable feedback on my
> circuit to think about, even though I am not sure the culprit of this
> particular failure is identified (but the reverse-leakage of the diodes
> is a thing to check, especially because they are next to the FETs and
> could get heated).

So the FETs are getting hot?  They really shoudn't be given the voltages and
currents.  There is no excuse for the FETs getting anything more than
slightly warm to the touch.  I'm thinking now the original problem was
caused by voltage stress on one or more componets due to high dV/dt and the
finite time for the diodes and other parts of the circuit to react and clip
the voltage.  This plus possible pass thru during reversal are exactly the
kind of stresses that build up over time and eventually cause a part to
fail.  The FETs getting warm is also a bad sign.  I don't know how warm, but
that adds to the stress.

So here are some concrete recommendations:

1 - Get rid of D5, D6, D7, D8.

2 - Get rid of C7, C8.

3 - Implement guaranteed 200uS dead time on reversal in the firmware.

4 - Add 10nF cap right accross the motor leads as close to the motor as
possible.

5 - Add another 10nF at the other end of the cable where it enters the
board.

6 - Reduce the PWM frequency to 500Hz unless this produces objectionable
whining.  Increase as necessary, but hopefully not more than a few KHz.  If
you can't keep the frequency low, decrease R119 and R120 (like maybe 1Kohm)
and lose R111 and R112 altogether.


******************************************************************
Embed Inc, Littleton Massachusetts, (978) 742-9014.  #1 PIC
consultant in 2004 program year.  http://www.embedinc.com/products

2006\04\21@150831 by Luisizan

flavicon
face
Hi Wouter and all:

Wouter van Ooijen escribió:

>>Tired to see sophisticated circuits failing I'm testing the oulput
>>circuit that Unitrode propose in the UC3637 application note,
>>adapted to
>>really work with 0-15 pwm pulses (UC3637 generated) and added
>>four BYW29 diodes.
>>    
>>
>
>  
>
Sorry, I mean 0 to 15Volts PWM pulses. The original circuit is a nice
heater for motor supplies greater than  UC3637 supply.


>BYW29!! for what maximum motor current was your circuit?
>
>  
>

6-8 Amps continous, but my customer reported that the board is running
with adjusted current limits over 15 amps, not the day long, only 8
hours a day, but the board still is not blown after some weeks. Mosfets
are IRF5210 and IR540. Supply is 24Volts AC rectified and smoothed. The
thing is a submerged arc welding "vertical" seam follower, I mean the
motor is continously reversing.

Best regards

Luis Izquierdo

2006\04\21@202939 by Wouter van Ooijen

face picon face
> > 100uS direction switching is not very probable, to put it
> mildly. (PWM
> > is ~ 30 kHz == 30us!)
>
> Yikes!  That's way too high for this circuit.  Your drivers
> are slow.  I bet
> the FETs are spending a significant fraction of their time in
> transition.
> Are they getting warm?

in normal operation, no

I can't test them for long on a loaded motor of the real system (the
mechanism would wear out quickly) so I used a bunch of lamps for a few
hours, IIRC 2A when at 100%, running at 70% PWM. IIRC the fets were
somewhat warmer too the touch, but not uncomformtably (I would guess no
more than 45 C).

> It would operate fine at a few 100Hz PWM.  
> The only reason for higher speed is to reduce audible
> whining.  There is no way to guess what frequencies will be
> audible and objectionable with a specific motor.  You just have to try
a
> few.  Hopefully there is a frequency you can use in the
> 500Hz to 5KHz range that is acceptable.

I preferred an unaudible frequency :)

> Yes, but what happens when the user hits the button to change
> direction and
> the motor is currently being driven in the other direction.

I don't think this is possible, but more from accident than from design.
But I will check.
 
>> If I use a cap there, how do I determine the value (at type
>> of cap)? Or should it be an R-C?

(snip calculations)

thanks (peter too)

> So let's work backwards and do a sanity check.  If you're
> switching the full
> 300mA and your FET body diodes can turn on in 5nS (just
> picking a number,
> you're job to look it up),

5ns for the N, 13 for the P

>> (but the reverse-leakage
>> of the diodes
>> is a thing to check, especially because they are next to
>> the FETs and
>> could get heated).
>
> So the FETs are getting hot?

Not in normal operation, but while thinking about how those diodes could
have blow I did consider a short from a motor wire to ground or power.
This might have heated the FETs, heated the diodes, and caused them to
blow from leakage current. But the leakage is specified as 10 mA @ 100C,
=> 0.24W, with 80C/W (case-ambient) that does not seem like a problem to
me.

> I'm thinking now the original problem was
> caused by voltage stress on one or more componets due to high
> dV/dt and the
> finite time for the diodes and other parts of the circuit to
> react and clip
> the voltage.

I think all recommendations are OK, but I fear it still is not the real
problem. There are two such H bridges in my controller, IIRC (no
hardware here now) one for 300 mA stall, the other for 2.5A stall. This
failure occured in the 300 mA bridge, yet the 2.5 one is the one that is
used most. (Yes, I shall re-do all calculations for the 2.5A side!).

Wouter van Ooijen

-- -------------------------------------------
Van Ooijen Technische Informatica: http://www.voti.nl
consultancy, development, PICmicro products
docent Hogeschool van Utrecht: http://www.voti.nl/hvu


2006\04\21@204358 by Wouter van Ooijen

face picon face
as for the ULN not getting the gate of the N fet low enough: Am I
correct that the parameter for this is the Gate Treshold Voltage?

GTV is 2V minimum, VCEsat (ULN2003) is 1.1 V (at 100mA, lower than I
would have assumed), + 1/40 * 24V = 0.6V, so that should be safe,
although some more margin would be more comfortable. Removing R111/112
would do that.

Wouter van Ooijen

-- -------------------------------------------
Van Ooijen Technische Informatica: http://www.voti.nl
consultancy, development, PICmicro products
docent Hogeschool van Utrecht: http://www.voti.nl/hvu


2006\04\21@204733 by Wouter van Ooijen

face picon face
more questions...

Should I have a decoupling cap on the power lines near the H bridge? If
so, I think I can work out a calculation of the value. But should it be
a special type?

Wouter van Ooijen

-- -------------------------------------------
Van Ooijen Technische Informatica: http://www.voti.nl
consultancy, development, PICmicro products
docent Hogeschool van Utrecht: http://www.voti.nl/hvu


2006\04\22@052308 by Vasile Surducan

face picon face
On 4/21/06, Wouter van Ooijen <.....wouterKILLspamspam@spam@voti.nl> wrote:
> > Wouter, we have talk about this several times on this list
> > and on others.
> > 1N5819 has only 40V breakdown voltage. It's less than 2x24V requested
> > by any engineering design requirements.
>
> you might have a point here
>
> > Also the motor load is 5A,
>
> no, the power supply is 5A, the *stall* current of this motor (there is
> a heavier motor too, this is the smaller one) is only 300 mA.

Well, then my point is wrong (for the current), except maybe if you
change the sense of the rotor without stoping it for a while ?
Or if the rotor have a high load on his shaft which is creating large inertia.

The self induction can kill the suppresing diodes, but usually the
killing comes from current flow (as max peak power dissipation on
diode) and not because of voltage. However schottky diodes aren't too
good at any huge spikes, just because the junction has low parasitic
capacitance and short spikes are jumping happy through it.

>
> > while your maximum supressed current is 1A (25A/8mS maximum maximum
> > peak current).
> > Change diode with at least 100V one and check if your maximum peak
> > current through diodes tooks more than 2-3mS. If yes, choose a 2A
> > diode.
>
> which diode would you suggest for 2A/100V?

With these new informations 1A should be enough, but choose a bigger voltage
for diodes.

>
> I think your points are probably valid, but not the root cause of this
> problem. The circuit contains two H-bridges, and this problem occured in
> one leg of the H bridge for the smaller motor!

The poorless diode crash the first one, but it could be any of the four.
Check if your clockwise and anticlockwise movement of the motor are
eating both the same current. I bet there is a difference and could be
a bigger one.

>
> Oh, and you are quick responder: I don't even see my original question
> on the list :)

Don't worry, you're compensating the hardware guy side by your
software guy side, which is huge evolved, you deserve fast answers,
even not entirely usefull.

greetings,
Vasile

2006\04\22@083021 by Howard Winter

face
flavicon
picon face
Wouter,

On Sat, 22 Apr 2006 02:29:38 +0200, Wouter van Ooijen wrote:

{Quote hidden}

But lamps are a really bad representation of a motor!  They don't have the inductance, they don't generate
back-EMF, and when their power is shut off they don't turn into generators and send voltage the wrong way into
the electronics (think of Fleming's two hands...).

I think you need to set up a test rig with actual motors, even if they don't have the actual live mechanism
connected, but some sort of dummy load such as a flywheel.

Cheers,


Howard Winter
St.Albans, England


2006\04\22@103227 by Vasile Surducan

face picon face
On 4/22/06, Wouter van Ooijen <wouterspamKILLspamvoti.nl> wrote:
> more questions...
>
> Should I have a decoupling cap on the power lines near the H bridge?

Absolutely.

If
> so, I think I can work out a calculation of the value. But should it be
> a special type?

It must have as low ESR as possible and cut all noise frequency
generated by the motor switching (including heavy mechanical load
balancing which have very low frequency -if any).
Choose one tantalum capacitor 47uF-150uF (usually highest voltage have
lowest ESR, so the package is large) with quite large voltage compared
with the supply. (ie for a 24V supply, minimum 50V capacitor- 63V to
100V better). Chose one X7R 1uF capacitor for medium frequencies and
another 0.1uF for highest frequencies. For those 1206 could be enough.
Connect them in parallel.
If your choose was right, those capacitors must remain cold at the
ambient temperature. If you have *very large* switching spikes, these
capacitors would be slighty warm (the first one heated is the
tantalum).

Vasile




>
> Wouter van Ooijen
>
> -- -------------------------------------------
> Van Ooijen Technische Informatica: http://www.voti.nl
> consultancy, development, PICmicro products
> docent Hogeschool van Utrecht: http://www.voti.nl/hvu
>
>
> -

2006\04\22@104126 by Vasile Surducan

face picon face
On 4/21/06, Alan B. Pearce <.....A.B.PearceKILLspamspam.....rl.ac.uk> wrote:
> >That's part of the driver for the FETs. I can post that part of the
> >circuit too, but I think it is unlikely to be part of the problem.
>
> Actually I think that is precisely where your problem is. The bottom FET
> will switch faster than the top one in the totem pole, because you have a
> capacitor on the gate of the top FET, but not the bottom one. Hence when the
> top one is on, bottom off, and you swap states there will be significant
> shoot through current while the top one turns off.

 I'm afraid Allan has right. You have large timing difference in your
FETs N and P driving for half bridges.  I didn't look at the
datasheet, but the RC (where R is equivalent resistor in the gate
circuit and C is the equivalent Cis + your capacitors)
must be identical for both FETs.

Vasile


>
> -

2006\04\22@104539 by Wouter van Ooijen

face picon face
> But lamps are a really bad representation of a motor!  They
> don't have the inductance, they don't generate
> back-EMF, and when their power is shut off they don't turn
> into generators and send voltage the wrong way into
> the electronics (think of Fleming's two hands...).

No, but the question here was about the switching speed of the FETs, for
which inductive or reactive load will not make that much of a
difference. And to be on the safe side I used a much heavier load than
expected. (That is: at that moment :) specifications will always
change...)

> I think you need to set up a test rig with actual motors,
> even if they don't have the actual live mechanism
> connected, but some sort of dummy load such as a flywheel.

A flywheel is not exactly a dummy load, it does not dissipate. These
motors are geared, and the gearbox does not like begin (back-) driven.

Wouter van Ooijen

-- -------------------------------------------
Van Ooijen Technische Informatica: http://www.voti.nl
consultancy, development, PICmicro products
docent Hogeschool van Utrecht: http://www.voti.nl/hvu


2006\04\22@113443 by Vasile Surducan

face picon face
On 4/22/06, Vasile Surducan <EraseMEpiclist9spam_OUTspamTakeThisOuTgmail.com> wrote:
> On 4/21/06, Alan B. Pearce <A.B.Pearcespamspam_OUTrl.ac.uk> wrote:
> > >That's part of the driver for the FETs. I can post that part of the
> > >circuit too, but I think it is unlikely to be part of the problem.
> >
> > Actually I think that is precisely where your problem is. The bottom FET
> > will switch faster than the top one in the totem pole, because you have a
> > capacitor on the gate of the top FET, but not the bottom one. Hence when the
> > top one is on, bottom off, and you swap states there will be significant
> > shoot through current while the top one turns off.
>
>  I'm afraid Allan has right. You have large timing difference in your
> FETs N and P driving for half bridges.  I didn't look at the
> datasheet, but the RC (where R is equivalent resistor in the gate
> circuit and C is the equivalent Cis + your capacitors)
> must be identical for both FETs.

Trying to be accurate indeed, this is not true if you're using one
transistor (say the PMOS) from one half bridge just for power on, and
the opposite transistor (NMOS) from the other half as PWM (as usually
used). Assuming a delay between these two ON commands, with supply ON
turned first, nothing should hapent with diodes if the other NMOS and
PMOS are indeed turned OFF.

So the problem is still the diode, the motor load or the driving algorithm.
If you'll forget to mount these diodes, the next probable step will be
drain-source short circuits via internal zenner breakdown.

> Vasile
>
>
> >
> > --

2006\04\22@114722 by Wouter van Ooijen

face picon face
>   I'm afraid Allan has right. You have large timing difference in your
> FETs N and P driving for half bridges.  I didn't look at the
> datasheet, but the RC (where R is equivalent resistor in the gate
> circuit and C is the equivalent Cis + your capacitors)
> must be identical for both FETs.

As far as I can remember now the C should not be there at all, but it
does not matter much except on direction change, beause I PWM only the N
FET.

Wouter van Ooijen

-- -------------------------------------------
Van Ooijen Technische Informatica: http://www.voti.nl
consultancy, development, PICmicro products
docent Hogeschool van Utrecht: http://www.voti.nl/hvu


2006\04\22@125939 by olin piclist

face picon face
Wouter van Ooijen wrote:
> I can't test them for long on a loaded motor of the real system (the
> mechanism would wear out quickly) so I used a bunch of lamps for a few
> hours, IIRC 2A when at 100%, running at 70% PWM.

Lamps won't simulate the inductive properties of a motor.

******************************************************************
Embed Inc, Littleton Massachusetts, (978) 742-9014.  #1 PIC
consultant in 2004 program year.  http://www.embedinc.com/products

2006\04\22@130206 by olin piclist

face picon face
Wouter van Ooijen wrote:
> Should I have a decoupling cap on the power lines near the H bridge?

That's a good idea, if for no other reason than to reduce transients to the
rest of the system.

> If so, I think I can work out a calculation of the value. But should
> it be a special type?

Make sure they are rated for the ripple current and also be conservative on
the voltage rating due to possible spikes.


******************************************************************
Embed Inc, Littleton Massachusetts, (978) 742-9014.  #1 PIC
consultant in 2004 program year.  http://www.embedinc.com/products

2006\04\22@141311 by Wouter van Ooijen

face picon face
> > I can't test them for long on a loaded motor of the real system (the
> > mechanism would wear out quickly) so I used a bunch of
> lamps for a few
> > hours, IIRC 2A when at 100%, running at 70% PWM.
>
> Lamps won't simulate the inductive properties of a motor.

correct, but any heating due to slow FET switching should be the same?

BTW what would you guys use to simulate a loaded motor? The best I can
think of is a motor driving a dynamo with a load, but can that be
transformed to for instance an unloaded motor + a parallel resistor?

Wouter van Ooijen

-- -------------------------------------------
Van Ooijen Technische Informatica: http://www.voti.nl
consultancy, development, PICmicro products
docent Hogeschool van Utrecht: http://www.voti.nl/hvu


2006\04\22@171501 by Howard Winter

face
flavicon
picon face
Wouter,

On Sat, 22 Apr 2006 16:45:38 +0200, Wouter van Ooijen wrote:

{Quote hidden}

I think there may be more than just the load - the *reaction* of the load, especially to changes of drive,
edges and so on, could be your problem, and lamps don't do that.

> > I think you need to set up a test rig with actual motors,
> > even if they don't have the actual live mechanism
> > connected, but some sort of dummy load such as a flywheel.
>
> A flywheel is not exactly a dummy load, it does not dissipate. These
> motors are geared, and the gearbox does not like begin (back-) driven.

It's a load as its speed changes, but that will fall off as it stabilises.  Perhaps if you added a fan, or
gears in an oil bath, or a water pump, or something like that, it would provide a load under stable speed
conditions.

Can you tell us what the motor will be driving in the live system, so we can see what the problem is?

Cheers,


Howard Winter
St.Albans, England


2006\04\23@035355 by Wouter van Ooijen

face picon face
> Can you tell us what the motor will be driving in the live
> system, so we can see what the problem is?

It is for a human arm supporting system. One motor (the light one)
drives a spindle that keeps the whole thing upright (a gravity sensor
privdes feedback). The other motor (heavier one) drives a spindle that
puts tension on a spring to compensate for the weight of the (human) arm
and what it holds.

Wouter van Ooijen

-- -------------------------------------------
Van Ooijen Technische Informatica: http://www.voti.nl
consultancy, development, PICmicro products
docent Hogeschool van Utrecht: http://www.voti.nl/hvu


2006\04\23@154805 by Peter

picon face


On Sat, 22 Apr 2006, Wouter van Ooijen wrote:

{Quote hidden}

A coil in series with a power resistor supplied by an auxiliary psu.

Peter

2006\04\24@042738 by Alan B. Pearce

face picon face
>GTV is 2V minimum, VCEsat (ULN2003) is 1.1 V (at 100mA,

Your problem with the ULN is that it is a darlington transistor, so the
saturation at any current will be a Vbesat + Vcesat ~= 0.6+0.3V. Personally
I would go for an array of SO-8 packaged power FET transistors.

More... (looser matching)
- Last day of these posts
- In 2006 , 2007 only
- Today
- New search...