Maker Pro
Maker Pro

Duty cycle and current capacity relation

J

Jon Slaughter

Jan 1, 1970
0
How is the duty cycle and current capacity generally related? linear,
exponential, logarithmic? (approximate ofcourse)


Maybe if they specified an exponential approximation Ae^(-Bx) + C then only
3 parameters would need to be given(which is just one more) for a
description of an arbitrary(within reason) of any duty cycle.

This assumes that a linear approximation is not very good.

Right now I'm trying to figure out if a linear interpolation is good enough.
1/10 duty has a current of 100mA and duty of 1(average) is 25mA.

This gives the linear equation -83.3x + 108.3 where x is the duty cycle.

So at 50% duty I would expect that I could have a current capacity of 66mA.
This seems aweful high though? (for some reason I guess I'm expecting to
double it from 25mA to 50mA since I cut the duty in half)

It also seems that there would then be an optimal duty cycle for a given
device. If f(x) is the current capacity given duty cycle x then x*f(x) is
the average current. If f(x) is approximately linear then this has a
maximum.

For my example above it occurs at a duty cycle of .65 and gives a maximum
current capacity of 54mA and average current of 35mA. An increase of 15mA(a
40% increase)!!!

Also, since at the maximum it is somewhat "flat" the duty doesn't have to be
perfect. A variation of 10%, i.e. 55% to 75% duty, gives a variation of <
1mA in the average current capacity.

I expect though that the curve is not linear but exponential... can anyone
verify this?
 
J

Jon Slaughter

Jan 1, 1970
0
I expect though that the curve is not linear but exponential... can anyone
verify this? By this I mean that a linear approximation is a bad one.


Also, if the component is an LED, then one can bring in the brightness and
maximize brightness vs duty cycle. (i.e., the would maximize efficiency)
 
B

Bill Sloman

Jan 1, 1970
0
Jon Slaughter said:
How is the duty cycle and current capacity generally related? linear,
exponential, logarithmic? (approximate ofcourse)

Mostly it is about heat. The average heat output at a 10% duty cycle is 10%
of
that which you'd get at a 100% duty cycle, and that's a linear relationship.

If you want to try for more current when the device is on, you have to work
out
how much the voltage drop across the device will rise when you raise the
current, and that is a lot trickier.

The only sensible answer is that it depends on the device, and you need to
experiment to find out what actually happens.

Buy quire a few devices before you start experimenting - they tend to blow
up.

Imagining possible mathematical relationships isn't a particularly useful
way
of spending your time.
 
J

Jon Slaughter

Jan 1, 1970
0
Bill Sloman said:
Mostly it is about heat. The average heat output at a 10% duty cycle is
10% of
that which you'd get at a 100% duty cycle, and that's a linear
relationship.

If you want to try for more current when the device is on, you have to
work out
how much the voltage drop across the device will rise when you raise the
current, and that is a lot trickier.

The only sensible answer is that it depends on the device, and you need to
experiment to find out what actually happens.

Buy quire a few devices before you start experimenting - they tend to blow
up.

Imagining possible mathematical relationships isn't a particularly useful
way
of spending your time.

what the heck is that suppose to mean? There are only so many possible
relationships... (just a handleful... like 3... It's not going to be some
random function and we don't need a perfect model)

I say blowing up devices just to test them is much worse.

if the device dissipates the duty cycle less heat and it is linear and the
heat dissipation is the main factor then my analysis is correct...

And hence an optimal duty cycle can be found for giving the most current
capacity. That is pretty damn significant IMO. You may not care too much
about mathematics but history has proven it to be much more useful than
trial an error.

I suppose you didn't understand what I was doing because I definitely wasn't
"imagining" possibilities....
 
what the heck is that suppose to mean? There are only so many possible
relationships... (just a handleful... like 3... It's not going to be some
random function and we don't need a perfect model)

You still haven't told us what kind of device you plan on misusing.

If the critical element was a bipolar transistor acting as a
saturating switch, with just enough base drive to keep it in
saturation with 100mA going through it, none of your "three possible
relationships" would be useful in modelling the relationship between
heat dissipation and peak current.

Unfortunately, one needs an educated imagination to create useful
computer models, and experimenting is pretty much the only way of
educating the imagination.
I say blowing up devices just to test them is much worse.

If you understand what you are doing, you won't blow them up.
if the device dissipates the duty cycle less heat and it is linear and the
heat dissipation is the main factor then my analysis is correct...
Unlikely.

And hence an optimal duty cycle can be found for giving the most current
capacity. That is pretty damn significant IMO. You may not care too much
about mathematics but history has proven it to be much more useful than
trial and error.

That is a thoroughly wrong-headed proposition. In fact, history proves
that
computer modelling is a very useful technique for making sense of
experimental results. Unvalidated computer models are an invitation to
disaster.
I suppose you didn't understand what I was doing because I definitely wasn't
"imagining" possibilities....

I understand all too well what you are trying to do. Too many of my
junior engineers had
tried to substitute guesswork for systematic measurement, and spent a
couple of days
devising what they think is a plausible computer model, and rather
longer on doing an
elaborate design that could never work before I got around to getting
them to take some
measurments.
 
J

Jon Slaughter

Jan 1, 1970
0
You still haven't told us what kind of device you plan on misusing.

If the critical element was a bipolar transistor acting as a
saturating switch, with just enough base drive to keep it in
saturation with 100mA going through it, none of your "three possible
relationships" would be useful in modelling the relationship between
heat dissipation and peak current.

Unfortunately, one needs an educated imagination to create useful
computer models, and experimenting is pretty much the only way of
educating the imagination.


If you understand what you are doing, you won't blow them up.


That is a thoroughly wrong-headed proposition. In fact, history proves
that
computer modelling is a very useful technique for making sense of
experimental results. Unvalidated computer models are an invitation to
disaster.


I understand all too well what you are trying to do. Too many of my
junior engineers had
tried to substitute guesswork for systematic measurement, and spent a
couple of days
devising what they think is a plausible computer model, and rather
longer on doing an
elaborate design that could never work before I got around to getting
them to take some
measurments.

Fortunately I don't like your condescending attitude...
 
J

Jon Slaughter

Jan 1, 1970
0
Anthony Fremont said:
I wondered how long it would take before the pent up belligerence started
seeping out. Stop arguing and experiment, that's how it's done. You
can't change reality thru argument, but the reciprocal is entirely
possible.

I knew you would reply... didn't think it was going to be so quick. You and
Bill must have came from the same bad apple.

Anyways... I guess you will go back on the ignore list... That won't stop
you from replying though.
 
J

Jon Slaughter

Jan 1, 1970
0
Jon Slaughter said:
How is the duty cycle and current capacity generally related? linear,
exponential, logarithmic? (approximate ofcourse)


Maybe if they specified an exponential approximation Ae^(-Bx) + C then
only 3 parameters would need to be given(which is just one more) for a
description of an arbitrary(within reason) of any duty cycle.

This assumes that a linear approximation is not very good.

Right now I'm trying to figure out if a linear interpolation is good
enough. 1/10 duty has a current of 100mA and duty of 1(average) is 25mA.

This gives the linear equation -83.3x + 108.3 where x is the duty cycle.

So at 50% duty I would expect that I could have a current capacity of
66mA. This seems aweful high though? (for some reason I guess I'm
expecting to double it from 25mA to 50mA since I cut the duty in half)

It also seems that there would then be an optimal duty cycle for a given
device. If f(x) is the current capacity given duty cycle x then x*f(x) is
the average current. If f(x) is approximately linear then this has a
maximum.

For my example above it occurs at a duty cycle of .65 and gives a maximum
current capacity of 54mA and average current of 35mA. An increase of
15mA(a 40% increase)!!!

Also, since at the maximum it is somewhat "flat" the duty doesn't have to
be perfect. A variation of 10%, i.e. 55% to 75% duty, gives a variation of
< 1mA in the average current capacity.

I expect though that the curve is not linear but exponential... can anyone
verify this?

I believe that in general the total thermal resistance would just be a sum
of individual thermal resistances which would be constant excluding
temperature. The capacity would be dependent on the power dissipation which
is a direct consequence of the thermal resistances.

Hence the relation I am looking for should be linear(approximately at
least). Hene the analysis I gave for optimality would hold and there is an
optimal duty cycle for maximizing current capacity.
 
How is the duty cycle and current capacity generally related? linear,
exponential, logarithmic? (approximate ofcourse)

Maybe if they specified an exponential approximation Ae^(-Bx) + C then only
3 parameters would need to be given(which is just one more) for a
description of an arbitrary(within reason) of any duty cycle.

This assumes that a linear approximation is not very good.

Right now I'm trying to figure out if a linear interpolation is good enough.
1/10 duty has a current of 100mA and duty of 1(average) is 25mA.

This gives the linear equation -83.3x + 108.3 where x is the duty cycle.

So at 50% duty I would expect that I could have a current capacity of 66mA.
This seems aweful high though? (for some reason I guess I'm expecting to
double it from 25mA to 50mA since I cut the duty in half)

It also seems that there would then be an optimal duty cycle for a given
device. If f(x) is the current capacity given duty cycle x then x*f(x) is
the average current. If f(x) is approximately linear then this has a
maximum.

For my example above it occurs at a duty cycle of .65 and gives a maximum
current capacity of 54mA and average current of 35mA. An increase of 15mA(a
40% increase)!!!

Also, since at the maximum it is somewhat "flat" the duty doesn't have tobe
perfect. A variation of 10%, i.e. 55% to 75% duty, gives a variation of <
1mA in the average current capacity.

I expect though that the curve is not linear but exponential... can anyone
verify this?

Duty cycle should be based on the circuit requirements. Once you know
the operation of the circuit, you size up the components accordingly.
At that point, you let things like average current due to duty cycle
come into play.

Basically, I think your approach is backwards.

While I mentioned the 1ms rule for pulsed electromigration, it
occurred to me that diodes used in power line rectification run a lot
longer than 1ms. Granted, the current flow is not a pulse but a piece
of a cycle, but still that makes me think the 1ms rule is really guard
banded.
 
x-no-archive:
what the heck is that suppose to mean? There are only so many possible
relationships... (just a handleful... like 3... It's not going to be some
random function and we don't need a perfect model)

I say blowing up devices just to test them is much worse.

if the device dissipates the duty cycle less heat and it is linear and the
heat dissipation is the main factor then my analysis is correct...

the point you miss is that the heat produced in the device may not be
(probably isn't) linearly related to the current through the device

Mark
 
J

Jon Slaughter

Jan 1, 1970
0
x-no-archive:

the point you miss is that the heat produced in the device may not be
(probably isn't) linearly related to the current through the device

Mark

The point I missed? That was my question.. duh... maybe you should read the
OP?

Now you say "probably isn't" but no proof? The thermal resistance is not
constant with respect to temperature or linearly related although it is
linear with small changes... see the temperature coefficients.

It seems you guys like to do a lot of guessing instead of relying on
fundamental physical prinicples. You claim I'm "guessing" when in fact I was
asking then turn around and guess that I'm wrong.

In fact it seems you are the one who is missing the point.
 
L

legg

Jan 1, 1970
0
How is the duty cycle and current capacity generally related? linear,
exponential, logarithmic? (approximate ofcourse)


Maybe if they specified an exponential approximation Ae^(-Bx) + C then only
3 parameters would need to be given(which is just one more) for a
description of an arbitrary(within reason) of any duty cycle.

This assumes that a linear approximation is not very good.

Right now I'm trying to figure out if a linear interpolation is good enough.
1/10 duty has a current of 100mA and duty of 1(average) is 25mA.

This gives the linear equation -83.3x + 108.3 where x is the duty cycle.

So at 50% duty I would expect that I could have a current capacity of 66mA.
This seems aweful high though? (for some reason I guess I'm expecting to
double it from 25mA to 50mA since I cut the duty in half)

It also seems that there would then be an optimal duty cycle for a given
device. If f(x) is the current capacity given duty cycle x then x*f(x) is
the average current. If f(x) is approximately linear then this has a
maximum.

For my example above it occurs at a duty cycle of .65 and gives a maximum
current capacity of 54mA and average current of 35mA. An increase of 15mA(a
40% increase)!!!

Also, since at the maximum it is somewhat "flat" the duty doesn't have to be
perfect. A variation of 10%, i.e. 55% to 75% duty, gives a variation of <
1mA in the average current capacity.

I expect though that the curve is not linear but exponential... can anyone
verify this?

Depends on the current vs voltage relationship of the device, and
which specified limit you hit first.

Parts have continuous and peak current limits, maximum junction
temperatures, thermal impedances and time constants.

A pure resistance (like a fet) will exhibit a peak to average penalty
that a fixed forward voltage drop device (like a thyristor or diode)
will not. This is one reason why cuurrent ratings for fets are often a
waste of print and why rds dominates their spec.

I believe you're dealing with leds. These respond to average current
like most diodes, though some have more resistive tendencies than
others. Thyristors, diodes and bipolar devices are typically supplied
with pulsed duty nomograms, in their specification.

You neglect the first problem - actual required luminous intensity in
the application. I understand that this intensity is maintained in
leds at lower average current, when pulsed duty is employed. This is
useful.

RL
 
Top