PWM chip that can drive an H bridge?

[Pooh Bear]

Do you have any suggestions, such as specific chips.

A UC3525 appropriately configured will do that.

Graham

 

[Ignoramus30105]

EXAR !

You're missing the point.

These chips aren't designed as PWM controllers.

No dead time for starters.

No, their output will be the logic input for the gate driver
chips. That's my plan. In light of this, do you have any specific
suggestions, thanks.

i

 

[Spehro Pefhany]

Ignoramus, this is about the time you'd better start telling us
what it is that you're trying to accomplish, because as the
thread wears on you're getting more and more off track.

I am trying to build a tig inverter. To that end, I want to have gate
driver controlling a full H bridge. To send signals to gate
drivers, I want to find an appropriate chip that would send square
wave signal to gate drivers and have its frequency and duty cycle
separately settable. Frequency should vary between 30 and 1000 Hz, at
least, and duty cycle between 15% and 85%, or better.

Wtf do you mean by "duty cycle"? Looks to me like the job of the
output inverter is just to smartly reverse current to the arc at a
certain (low) frequency, and the primary inverter is tasked with
actually controlling the arc current. The IRF app note even suggests a
slight negative "duty cycle"-- forcing a small amount ( a few hundred
ns) of cross-conduction in the bridge.


Best regards,
Spehro Pefhany

 

[Ignoramus30105]

Pooh, I found a very simple chip XR2206, it is not perfect because
duty cycle and frequency are not independently selected, but it is
simple.

Ignoramus, this is about the time you'd better start telling us
what it is that you're trying to accomplish, because as the
thread wears on you're getting more and more off track.

I am trying to build a tig inverter. To that end, I want to have gate
driver controlling a full H bridge. To send signals to gate
drivers, I want to find an appropriate chip that would send square
wave signal to gate drivers and have its frequency and duty cycle
separately settable. Frequency should vary between 30 and 1000 Hz, at
least, and duty cycle between 15% and 85%, or better.

Wtf do you mean by "duty cycle"?

Duty cycle is a term used in PWM chips. It means, in the context of
square wave signal, the time that the signal is ON, as percentage of
the total oscillation period.

For example, if the period is 0.1 seconds (10 Hz), and the signal is
ON for 0.07 seconds (and therefore OFF for 0.03 seconds), that this is
a 70% duty cycle signal.

Our normal symmetrical square waves have, by definition, 50% duty
cycle.

In tig welding, especially aluminum, it helps to vary the duty cycle
and vary the time of electrode negative and electrode positive
time. That lets the welder balance the cleaning action of electrode
positive with deeper penetration of electrode negative.

Looks to me like the job of the output inverter is just to smartly
reverse current to the arc at a certain (low) frequency, and the
primary inverter is tasked with actually controlling the arc
current. The IRF app note even suggests a slight negative "duty
cycle"-- forcing a small amount ( a few hundred ns) of
cross-conduction in the bridge.

Well, yes, you are correct, but it greatly helps to also vary the
EN and EP times as percentage of every cycle.

i

 

[Ignoramus30105]

A UC3525 appropriately configured will do that.

Thanks Graham... They have only less than 50% duty cycle, whereas I
would like to see a greater variation. Nevertheless, it is a
possibility. I will check out tig welding books, maybe I can get away
with less tan 50% duty cycle. I appreciate your input.

i

 

[Pat Ford]

Thanks Graham... They have only less than 50% duty cycle, whereas I
would like to see a greater variation. Nevertheless, it is a
possibility. I will check out tig welding books, maybe I can get away
with less tan 50% duty cycle. I appreciate your input.

i

I think I've seen a 555 set up ( or was it a 565) I think it was the daul
timer chip, the first timer was a VCO and the other did duty cycle, I think
it was in a radio shack book on powersupplies.
Pat

 

[Ignoramus30105]

I think I've seen a 555 set up ( or was it a 565) I think it was the daul
timer chip, the first timer was a VCO and the other did duty cycle, I think
it was in a radio shack book on powersupplies.
Pat

Yes. The problem is that regulation of frequency vs. duty cycle is
very screwy. You have two pots, R1 and R2, and frequency and duty
cycle are functions of R1 and R2. You basically have to solve a linear
equation system every time you want to set frequency and duty cycle,
to find out what pot settings to use. Very counterintuitive.

i

 

[Robert Latest]

On Fri, 07 Oct 2005 11:13:30 GMT,

I am trying to build a tig inverter. To that end, I want to have gate
driver controlling a full H bridge. To send signals to gate
drivers, I want to find an appropriate chip that would send square
wave signal to gate drivers and have its frequency and duty cycle
separately settable. Frequency should vary between 30 and 1000 Hz, at
least, and duty cycle between 15% and 85%, or better.

Do you have any suggestions, such as specific chips.

LM393 (dual comparator). Get the data sheet from National.com. The first
example they show is the "square wave generator". Replace the top 100k
resistor with a trimpot and resistor in series and get an appropriate
capacitor to cover the desired frequency range.

This circuit will produce a somewhat triangularish waveform at the
capacitor terminal whose frequency is settable by the trimpot.

Now connect this node to the + input of the other comparator. Wire
up another trimpot as adjustable voltage divider and connect the wiper
to the - input. Now the output of the other comparator will go high
whenever the sawtooth voltage exceeds the pot voltage. High pot voltage
-> only the tips of the sawtooth icebergs will stick out -> low duty
cycle. Low pot voltage -> sawtooth voltage exceeds pot voltage most of
the time -> high duty cycle. Limit adjustment range of duty cycle pot by
appropriate resistors in the top and bottom leg of pot.

Not that since the sawtooth isn't a real triangle wave (it doesn't have
straight edges) the duty cycle percentage isn't exactly linear with pot
setting.

If you have an oscilloscope you won't even have to do any math to tailor
this circuit to your needs.

robert

 

[Robert Latest]

On Fri, 07 Oct 2005 12:29:37 +0100,

You're missing the point.

These chips aren't designed as PWM controllers.

It took me a long time to realize this, but his application isn't
exactly the target of PWM controllers. He's really looking for a
variable frequency, variable duty cycle oscillator.

Dead time is another matter -- he'll have to ensure that there's no
shoot-through, but I think the HIP4081A will do that for him.

robert

 

[Ignoramus30105]

On Fri, 07 Oct 2005 12:29:37 +0100,


It took me a long time to realize this, but his application isn't
exactly the target of PWM controllers. He's really looking for a
variable frequency, variable duty cycle oscillator.

that's correct.

Dead time is another matter -- he'll have to ensure that there's no
shoot-through, but I think the HIP4081A will do that for him.

If you are referring to controlling the H bridge, then the gate
drivers will take care of shoot through. All I want from the
oscillator chip is a square wave 5V signal that I can properly control
and that can assume proper frequency and duty cycle range.

i

 

[Ignoramus30105]

Okay, I'd call that symmetry, since the load is 100% on, just the
polarity swaps around. In any case.. .

I kind of agree, but duty cycle is an accepted term for pwm chips.

You can probably do this acceptably well with a CMOS 555 and a CMOS
comparator and a couple of pots.

See my another answer, I could, but control is cumbersome. You
basically have to solve a 2x2 linear system to figure out proper pot
positions for any given frequency/duty cycle combination.

Maybe I am missing something.

i

 

[Spehro Pefhany]

Well, yes, you are correct, but it greatly helps to also vary the
EN and EP times as percentage of every cycle.

i

Okay, I'd call that symmetry, since the load is 100% on, just the
polarity swaps around. In any case.. .

You can probably do this acceptably well with a CMOS 555 and a CMOS
comparator and a couple of pots.

Alternatively, if you can live with switch selected frequencies (like
a rotary switch with a number of set approximate frequencies) and a
pot for duty cycle you could just use a single TLC555. The switch
would select capacitors.


Best regards,
Spehro Pefhany

 

[Tim Wescott]

Yes. The problem is that regulation of frequency vs. duty cycle is
very screwy. You have two pots, R1 and R2, and frequency and duty
cycle are functions of R1 and R2. You basically have to solve a linear
equation system every time you want to set frequency and duty cycle,
to find out what pot settings to use. Very counterintuitive.

i

If you wire a 555 as an astable multivibrator and put a comparator on
the trigger input, the duty cycle out of the comparator will vary as you
change it's threshold voltage. You can adjust the frequency of the 555
with a single pot, and the threshold voltage with a separate pot. At
those frequencies things will even be somewhat consistent.

 

[Spehro Pefhany]

I kind of agree, but duty cycle is an accepted term for pwm chips.


See my another answer, I could, but control is cumbersome. You
basically have to solve a 2x2 linear system to figure out proper pot
positions for any given frequency/duty cycle combination.

Using the two-chip method,. the TLC555 would set the timebase and
generate a somewhat nonlinear triangular waveform. The pot setting
would be linear with period (1/frequency) and completely independent
of symmetry.

The symmetry would be set by the comparator threshold (completely
independently of frequency) and would be somewhat nonlinear.

A single TLC555 could do this if you switch-select the capacitor and
adjust the symmetry with a pot and a couple of switching diodes.
That's the really easy way to go if you don't need a lot of resolution
on the frequency, or would like to increase it in a series like
50Hz/100Hz/200Hz/500Hz/1000Hz or whatever.


Best regards,
Spehro Pefhany

 

[Tim Wescott]

Thanks Graham... They have only less than 50% duty cycle, whereas I
would like to see a greater variation. Nevertheless, it is a
possibility. I will check out tig welding books, maybe I can get away
with less tan 50% duty cycle. I appreciate your input.

i

IIRC the 3525 has two open-collector outputs which can be paralleled to
get nearly 100% duty cycle (I know the LM3524 can, because I have the
data book open to that page).

If you're driving an H bridge for electronically switched polarity, and
if you're clever, you could get square-wave AC out of the thing by
clever application of your '3524.

 

[Tim Wescott]

Okay, I'd call that symmetry, since the load is 100% on, just the
polarity swaps around. In any case.. .

You can probably do this acceptably well with a CMOS 555 and a CMOS
comparator and a couple of pots.

Alternatively, if you can live with switch selected frequencies (like
a rotary switch with a number of set approximate frequencies) and a
pot for duty cycle you could just use a single TLC555. The switch
would select capacitors.


Best regards,
Spehro Pefhany

Or a PIC with a couple of buttons, a beeper and an LCD. But then he'll
have so much fun with the operator's interface that he'll never get to
the power electronics.

I'd go with the 555 and a 393, myself, but then I've been reading too
many of Jeorg's posts.

 

[Robert Latest]

The IRF app note even suggests a
slight negative "duty cycle"-- forcing a small amount ( a few hundred
ns) of cross-conduction in the bridge.

Yes, because the supply is inductive. If the IGBT bridge were to
switch off altogether (as in any "normal" PWM design), there
would be IGBT bits all over the place.

robert

 

[Robert Latest]

If you are referring to controlling the H bridge, then the gate
drivers will take care of shoot through. All I want from the
oscillator chip is a square wave 5V signal that I can properly control
and that can assume proper frequency and duty cycle range.

Look at my other post -- you can do that with a single LM393.

As for dead time -- in "normal" PWM applications, the dead time
is a life saver for the IGBT bridge. In your application, as the
IRF app note mentions, you need some shoot-through due to the
inductive nature of your supply (I just realized this). In fact
if you had dead time in your application, your bridge would
indeed be dead right away.

So you CANNOT use a gate driver that "takes care" of this,
because they're all designed to avoid cross-conduction.

What is the max current/max voltage of your TIG supply?

robert

 

[Tim Wescott]

Look at my other post -- you can do that with a single LM393.

As for dead time -- in "normal" PWM applications, the dead time
is a life saver for the IGBT bridge. In your application, as the
IRF app note mentions, you need some shoot-through due to the
inductive nature of your supply (I just realized this). In fact
if you had dead time in your application, your bridge would
indeed be dead right away.

So you CANNOT use a gate driver that "takes care" of this,
because they're all designed to avoid cross-conduction.

What is the max current/max voltage of your TIG supply?

robert

Shoot-through connects the two power supply rails together, leading to
high currents in the switches and some rather dramatic behavior (wear
safety goggles).

Catch diodes would probably be more appropriate, and are routine in this
sort of application.

 

[Ignoramus30105]

Look at my other post -- you can do that with a single LM393.

As for dead time -- in "normal" PWM applications, the dead time
is a life saver for the IGBT bridge. In your application, as the
IRF app note mentions, you need some shoot-through due to the
inductive nature of your supply (I just realized this). In fact
if you had dead time in your application, your bridge would
indeed be dead right away.

Here's where I am a little confused. I thought that snubber capacitors
placed across the IGBT would protect them during "dead time".

So you CANNOT use a gate driver that "takes care" of this,
because they're all designed to avoid cross-conduction.

What is the max current/max voltage of your TIG supply?

It is a digitally controlled Hobart CyberTig.

You can see it here:

http://igor.chudov.com/projects/Welding/00-Hobart-CyberTig-Welder/


It is rated for 200A maximum. (and it does put out 200A).

If I set it to operate at, for instance, 137A, that would be very
close to its actual output.

Open Circuit voltage 80V, typical arc voltage 30V.

i