Whining PWM motor control

Use a clock, so that it runs constant frequency, then PWM.

The whine is probably the pole pieces flexing. I did a controller design for Bosch/Mercedes and passing the noise spec was the hardest part... The director of engineering stuck his head in the air duct, listened, and decided yes or no ;-)

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
|  Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
|  Phoenix, Arizona            Voice:(480)460-2350  |             |
|  E-mail Address at Website     Fax:(480)460-2142  |  Brass Rat  |
|       http://www.analog-innovations.com           |    1962     |
             
I love to cook with wine.      Sometimes I even put it in the food.

Hi guys,

I built a very simple PWM circuit to maintain a contant, unidirectional current through a small (24V/1A) DC motor from parts I had kicking around.

The circuit consists of a low-side MOSFET switcher with .5R current sense in the source leg. An LM393 comparator compares this voltage to an externally set reference and asynchronously resets a flip flop when the sense voltage is greater than the ref. The flip flop is clocked "on" periodically by a 20kHz oscillator; its output drives the gate driver.

Well, it works fine, actually.

My problem is that the circuit doesn't switch at 20kHz but at much lower frequencies, which are quite unstable and vary greatly depending on the preset current limit and motor speeds, which of course creates intolerable noise.

Before I start redesigning the circuit to employ a less direct form of feedback (not using the instant motor current itself to create the PWM ramp) I'd like to ask if this problem is common, and how it is usually dealt with. I've looked at data sheets of several PWM controller ICs and found pretty much the very same circuit I'm using.

What are the fine points I have missed?

I know that a certain "singing" of PWM controlled inductances such as stepper motors is common and hard to avoid. I'd just like to limit the noise to a tolerable level.

Thanks, robert

On Thu, 18 Aug 2005 11:21:11 +0200, Rene Tschaggelar wrote in Msg.

Initially of course, but once the current is up it shouldn't drop so much. The motor has 5mH, supply voltage is 27V (which means it takes 4 PWM periods to take the motor from 0 to 1 A).

I do, and as much as I hate to admit it, it's a 4001. Can this alone kill the freewheel current in 50 us? The diode is going to be replaced with an appropriate one as soon as I place my next order for parts.

Thanks, robert

Apparently the current takes longer to rise than the though 50us. When the voltage is low the risetime to the current is longer.

Do you have a diode over the motor ?

Rene

--
Ing.Buero R.Tschaggelar - http://www.ibrtses.com
& commercial newsgroups - http://www.talkto.net

On 18 Aug 2005 02:37:51 -0700, c snipped-for-privacy@hotmail.com wrote in Msg.

I'm actually using a fixed frequency; it's my circuit that made the decision to lowering the frequency at will.

Yeah, I guess that's what they mean by "voltage mode PWM".

Yes -- that it makes no sense. The time constant, given by supply voltage and motor inductance, stays the same.

robert

"It makes a noise that is not noble".

Like a fart ?:-)

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
|  Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
|  Phoenix, Arizona            Voice:(480)460-2350  |             |
|  E-mail Address at Website     Fax:(480)460-2142  |  Brass Rat  |
|       http://www.analog-innovations.com           |    1962     |
             
I love to cook with wine.      Sometimes I even put it in the food.

I hope that when faced with a spec that says "whatever he likes" you gave an estimate that said "however long I take".

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

I had this problem with a current controlled stepper motor drive, based on an integrated driver. The motor inductance was large enough that the current wouldn't fall below the current trip hysteresis in the next clock cycle. So the control skipped cycles and made noise. My solution was to add some of the clock circuit triangle wave to the current trip threshold, so that the difference of that and the current shunt voltage crossed each other (well more than the comparator hysteresis) twice each cycle, even though the current was essentially constant throughout the cycle. Instant silence.

Is your control loop hunting? Could the load of the motor be pulling the reference voltage down?

OR (a bit of a long shot) Is something oscillating above 20 Kc/s and giving an audible beat note with the clock?

--
~ Adrian Tuddenham ~
(Remove the ".invalid"s and add ".co.uk" to reply)
www.poppyrecords.co.uk

[...] Yes it is common and a side effect of this method of control. ICs suffer as well so it's not some trivial construction aspect. Annoying thing is that the stepper IC datasheets make scant mention that this will happen, showing instead textbook chopper current profiles which have bugger all relation to reality. Depending on supply voltage/motor/current, the current 'ramps' are quite happy to flip down (sometimes quite stably) into various integer, 'sub ramp' rates. An acceptable level of ramp skipping will give a 'rustling' motor noise, longer term skips can be occasionally musical. Bad cases result in harsh fluctuating wideband noise as the the ramps flip from one ramp rate to another on a random cycle basis. Plays havoc with the motor ripple currents. It's vital that the those current sensing ramps are as free as possible from hf spikes, noise and the cross coupled switching noise from the other coil. A real cure can be interestingly convoluted. regards john

Hello Robert,

As Jim already said a fixed frequency is better. What I usually do if it has to be cheap (or in your case the parts should already be flying around in your garage): Take a Schmitt trigger such as the 74HC14 or CD40106, build an oscillator and then quench that with a FET or transistor that is controlled from the feedback point. IOW if the motor current wants to exceed your limit it throttles the oscillator's "on" period but not its frequency too much. Actually the frequency changes a bit but not a lot.

If you want absolutly constant frequency you can build that oscillator and follow with a one-shot that can also be made from a Schmitt gate. Then quench that one-shot instead. There are six in a chip so you have plenty.

Regards, Joerg

Hello Jim,

They can be picky. This wasn't Daimler Benz but a big guy of a major automotive mfg was looking at a similar issue. Comment on one version: "Es macht ein unedles Geraeusch". You can't really translate that but roughly it would be "It makes a noise that is not noble".

Regards, Joerg

Hello John,

Have you ever done the opposite, like using the processor to accompany with the sound of a Harley sans muffler? That made quite an impression when doing the first demo.

Regards, Joerg

as

showing

to

result

rate

from

coil.

Belay that. There was no mention of a stepper. I must be fixated with the damned things. Sod it.

Hello Jim,

ROFL!

It was more one of those cheapy ratchet sounds like on a child's toy. Not quite in line with the price category of the vehicles.

Regards, Joerg

On Fri, 19 Aug 2005 00:24:22 GMT, Joerg wrote in Msg.

Hmm, that sound exactly like what I'm doing already: I'm using a constant-frequency oscillator to periodically set a flip flop to "on" (the oscillator clocks a D-FF with D tied to High), and the comparator resets the FF when the sense current exceeds the limit.

So I'm in fact running a constant frequency. The motor takes several cycles to get up from zero to 1A of course (f=20kHz L=5mH V=27V), but I had expected it to basically stay there via the freewheel diode and just take an additional sip of current each cycle. All I can see on my old analog scope is multiple-clock rising ramps all over the screen; I can't see if the PWM drive takes longer breaks between ramps.

It all looks as if the motor loses too much current between pulses which may be due to my KNOWINGLY poor choice of freewheel diode. I'd like to put in a BYV28-100 but my supplier is out of them. Any better ideas for a few-amps fast recovery diode?

Thanks,

robert

On Thu, 18 Aug 2005 08:00:46 -0700, Jim Thompson wrote in Msg.

That's what I'm alreay doing. See my more verbous reply to Joerg elsethread.

Whatever it is, it has to go into the inaudible range. These are small motors, it should be doable.

My sister's boyfriend's job used to be the acoustics of auto gears. He's moved on to motor acoustics now. They're really picky about what a car sound like on the inside.

robert

Of course. Also made about a dozen trips to Germany in Business Class, once in First Class ;-)

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
|  Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
|  Phoenix, Arizona            Voice:(480)460-2350  |             |
|  E-mail Address at Website     Fax:(480)460-2142  |  Brass Rat  |
|       http://www.analog-innovations.com           |    1962     |
             
I love to cook with wine.      Sometimes I even put it in the food.

Robert Latest wrote: (snip)

At 27 volts, I would use a Schottky. They have a lot of capacitance, so no larger than necessary would be good.