Denso Window Motors, Warning - Don't use with Jaguars !

Tom,
It is common for a motor to be commanded to not move and the simple way to do this is alternate FWD/REV at 50% duty cycle each. Although the average current is zero, the instantaneous (and absolute) is stall current. The net effect is the motor does not move in either direction and the motor sings at the switching frequency. You are correct that at 50% duty cycle the inductance plays into current waveform distortion. However, the Victor is capable of the same motion control you describe, as we have done it in the past. Why it would switch one way with Victors and the other way with Jaguars is puzzling. Are you making a change in software when you change controllers? I know from experience using this method with high current motors eats the battery. We used this when a stable platform was needed for shooting.

yes, if the PWM frequency is high enough for the given motor inductance.

When the input to the motor is 50% fwd, 50% reverse, the average voltage is zero,

correct

therefore there is no average current,

the algebraic average will be zero. the rms average will be near zero only if the PWM switching frequency is high enough for the given motor inductance.

and hence it does not draw any current or heat up.

algebraic current can be zero and yet rms non-zero, in which case the motor will heat up. see above.

The impedance (inductance) of the windings will not allow the current to change at anywhere near the rate of the 15 kHz switching.

correct, depending on the motor inductance. if the inductance is very very low, it will start to draw current even at high switching speeds. most DC motors today are designed with sufficient inductance that 15KHz is plenty fast.

The motor is not “stalled”, but rather it has no net current flow even though the voltage is toggling wildly.

correct. the algebraic average current is zero, and the rms is near zero. the instantaneous current is also near zero.

Well, there may be some dynamic power dissipation 1/2 CV^2 or some such, but certainly nowhere near what you are thinking as “stall current”???

correct

Do I have this all messed up?

No.

Ether,
There is a thread from last year where inductance vs. input waveform for CIM motors was discussed and plotted. As I remember, the CIM motor inductance limits the rise time of the motor controller output below about 60% duty cycle at 15kHz. It is this that adds to the Jaguar’s “linearity” at low throttle values. (long rise time contributes to better response by lowering the average current at low throttle levels) With the suspected lower inductance of the Window motor, I would expect at 50% duty cycle, there would be little effect on the rise time. It is hard to use the term RMS in this context since this is not really an AC signal. It is more of a pulsed bi-polar DC signal and is easier to discuss in those terms when using DC motors. Following the definition of stall current as the current flow when a motor is energized but not moving, then in the example, stall current would indeed flow. All of the KOP motors draw stall current in the instant they are commanded to move and continue to do so until moving. What is particularly intriguing is the fact that motors do not heat up.
Tom, when you mentioned 50% duty cycle were you saying the positive pusle is only at 50% and then it remains off for 50% before changing polarity? I expected you meant that 50% of the time it was positive and 50% was reversed. The former would give an average current at half of the latter and may account for different motor heating.

With the suspected lower inductance of the Window motor, I would expect at 50% duty cycle, there would be little effect on the rise time.

I would expect that the Window motor has sufficient inductance so that at 50% -12/+12 duty cycle at 15KHz, the RMS current would be rather small and thus very little heating. If I get a chance later tonight, I will try to find the datasheet for the window motor and see if it lists an inductance… and I’ll crunch the numbers.

It is hard to use the term RMS in this context since this is not really an AC signal.

It is indeed an AC signal, just not perfectly sinusoidal. I agree that the calculation is more difficult, but RMS can be calculated for any periodic waveform, sinusoidal or not. The point I was making is that the RMS is correlated with I^2R heating, whereas the algebraic average is not.

It is more of a pulsed bi-polar DC signal

I could be wrong, but I’d bet a buck that the current in the motor windings probably looks more like a triangular waveform than a pulsed bi-polar square wave.

Following the definition of stall current as the current flow when a motor is energized but not moving, then in the example, stall current would indeed flow.

Ah, there’s the confusion. We have different definitions of stall current. “Stall current” in the motor datasheets refers to the current flowing in the motor windings with the motor stalled against an external torque with a constant specified DC voltage applied. A motor which is receiving 50% -12/+12 15KHz excitation is not “stalled” in this sense; if the motor inductance is high enough and the PWM frequency is high enough, there will be little or no current flow, and thus little or no I2R heating of the motor windings under these conditions.

All of the KOP motors draw stall current in the instant they are commanded to move and continue to do so until moving.

Yes they draw current, but not “stall current” (as used defined above) unless the command is +12V DC.

What is particularly intriguing is the fact that motors do not heat up

They don’t heat up because there is little or no RMS current, because the inductance and PWM switching frequency are sufficiently high that the current never has a chance to rise above a very small value.

~

Ether,
I don’t see any inductance specs on the window motor data sheet.
At full voltage with a 50% FWD and 50% REV, the motor should be drawing stall current except during the period of charge pump refresh in the Jaguar. The RMS voltage would be near stall under these conditions. You are correct, the inductance will affect the full stall current during rise time. With the Jaguar, dependent on the length of the pulse this average current could be less than stall. I have not scoped the window motor connected to a Jaguar. I have looked at window motors connected to Victors. The only limiting factors are the series resistance of the wiring and speed controllers at 120Hz.

Al / Ether:

The “idle” waveform I observed from the Jaguar (joystick centered) was a square wave with a ~ 15 KHz frequency, toggling between +12V and -12V. There was no time where it was zero volts. When the joystick was pushed forward, it continued to toggle between +12V and -12V, while the duty increased until it was a constant +12V. When the joystick was pulled back, it continued to toggle between +12V and -12V, while the duty decreased until it was a constant -12V.

With the Victor, it was was literally zero volts when the stick was centered. When the stick was pushed forward, it began to toggle between zero volts and +12 volts at it’s 117 Hz rate with varying duty until it was a constant +12V. When the stick was pulled back, it began to toggle between zero volts and -12 volts at it’s 117 Hz rate with varying duty until it was a constant -12V.

BTW – I think the window motor would have substantially higher inductance than a CIM… It is a much lower power motor (finer gauge windings), higher impedances across the board, --> higher inductance.

-Tom

Tom,
Where was your scope probe connected? Where was the ground lead? I am leaning towards a smaller motor, and less power leads to smaller windings although higher series resistance (due to smaller gauge wire). Add to that the smaller field structure (smaller core) would also lead to lower inductance. Of course since none of us has measured the inductance or looked at rise times, all of this is just a guess. We do not use Jaguars, so I can’t make any measurements on a robot.
I am guessing that the -12 volts is a ghost due to the reference lead of the scope. Are you viewing in AC mode or DC mode?

This is what I was about to throw in to this conversation as well, although, I think I would have used the term “coupling mode”.
I must admit though, Tom did say it was a square wave and ac coupling will not display a square wave well. There will always be considerable rise and fall spikes that hint at the coupling error. Depending on the scope, this should be true at 15KHz as well.

For this set up, the scope probe should be connected directly to the +M of the Jag with the reference on the -M. The scope should be in DC coupling mode.

Can you put a small resistor (tenth of an ohm maybe?) in series with the motor and measure the voltage drop across it on your scope please, and tell us what you find?

I’m guessing you’re gonna see a very low amplitude triangle wave when running with the Jags at “zero” command.

~

Robot completely isolated from earth ground. Scope was most definitely DC coupled. Bilbo911 has it exactly right. “ground” lead of scope on the M- and probe tip on M+. Measuring the differential voltage across the motor winding. If the scope reference was tied to the robot battery (either + or -), it would not be possible to observe a bipolar signal.

Nope, definitely not a “ghost”.

Smaller wires, higher impedances, --> higher inductance to me.

I don’t have a CIM in my hands right now, but a quick check of the Denso with a scope and a pulse generator indicates an inductance on the order of 200 uH.

-Tom

Aren’t we all pathetic !!!

Any normal human being with half a life would have better things to do than debate over the particulars of PWM motor control.

I don’t have this stuff in front of me, will have to go to the robot shop. Just to keep the geek-nerds amused, I will obtain a tenth ohm resistor at work tomorrow and run the current test.

If you guys are really nice, I may capture some scope plots of voltage and current and post them.

After Wednesday, we will be preoccupied at the Chesapeake regional, so additional nerd experiments will go on hold.

-Tom

You mean, better things as in browsing the web to see what the hell is Paris Hilton doing today? :stuck_out_tongue:

I’ll take particulars of PWM motor control over that, any day…

Wow. Reading that last post really robbed me of any desire to respond. I had to re-read Ether, Jared and Al s’ posts a few times first. I hope that anyone who reads the dire warning on page 1 also reads page 3, so they may understand the source of the misinformation.

There are two major questions.

  1. Is the Jaguar switching when given a command of “0”?
    I just tested a grey jag on PWM, a grey jag on CAN, a black jag on PWM, and a black jag on CAN. All signs point to no.

  2. Why do I see +/-12V on Jaguars, but not on Victors?
    The answer has actually been posted several times before: Inductance vs Frequency. Assuming 12V, 200uH and no resistance, a victor changes current by 500 amps during its switch cycle, whereas a Jaguar changes by only 4 amps. This means that, just as Al said, a Victor will ramp all the way to stall or 0 during its switching period. However, a Jaguar will maintain a relatively steady current.

    If the motor has a load that relates to a current greater than its delta current ((Vin-Vemf)/Ht), it will always be conducting current. While any h-bridge is conducting current, its motor will report a voltage equal to the input voltage (minus losses). So, a Jaguar driving a stalled* denso with more than 4Amps will show a square wave that looks a lot like +12 / -12. It’ll actually be more like +12 / ~-11.

A victor will never show this wave form, unless you can stall your motors at above 500A without lighting them on fire.

This is actually a GoodThing™, and for many reasons. For a given torque, reducing the ripple current reduces the power dissipation in the motor (rms vs mean, per Ether).

*Remember that back emf is subtracted from this.

Al, could you post a link to this documentation please. I can see how 50% FWD/REV would slow down a moving motor, but not hold a motor stationary against a sustained external torque.

Thanks.

~

This so called Locked Anti phase drive is much nicer when you are trying to do speed control for reasons I won’t go into here but you are all right, in noting that the frequencies have to be high when compared to the time constant of the motor circuit or bad things can happen.

Joe J.

The Denso motors have a rubber flex element in them that absorbs shock at the end of the window travel. It also makes them very sloppy if you are trying to use them for precise positioning.

Tom,
When you connect the scope in this fashion the reference lead is switching between +12 and common. So yes, you will see a signal go from +12 to -12 volts. In reverse the scope common is now at +12 volts and the probe is 12 volts below it. The Victor should show the same waveform. As to why the Jaguar exhibits this in zero throttle, I can’t answer.
At 200uH and about 18 ohms of series resistance the time constant turns out to be 1.1uSec. Giving it 6 time constants to fully charge that is 6.6uSec. A 15kHz square wave has a period of 66uSec or 33uSec high or low for a 50% duty cycle. So yes, there will be a little rounding of the waveform but displaying several (say 5 periods) waveforms horizontally will not really show much effect of the inductance on the waveform rise time.
Ether, I am quoting the Jaguar manual from last year. It lists analog position control as only available under CAN control in Table 2-1. If the motor is given a FWD command and then a reverse command of equal amplitude, the armature will stay in one place given that the switching frequency equals or exceeds the ability of the motor shaft to turn in between pulses. The trade off is excessive current demand. We used this technique in the game of shooting balls into the overhead goal. By locking our wheels in place while shooting, other robots were not capable of moving the robot to throw off our shooter. Using crab, we simply rotated the high friction wheels 90 degrees to our shooting direction and engaged the lock. We could not be pushed far enough to prevent our balls launching into the goal. After ten balls or whatever we were carrying, the lock went off and we drove normally. Remember that the game left us unmolested during parts of the game so it was not needed more than a few seconds each match. Still it was enough to deplete the battery more than I would have liked.

Originally Posted by Mr.G:
Tom, what you are saying is that power is flowing at a 50% duty cycle throught the motor when it is at a zero speed command?*

No, he is not saying that “power is flowing”. He is saying that a 50% duty cycle -12/+12 voltage is present at the motor coils. He is saying that the Jag, at zero command, apparently is providing a 50% voltage duty cycle between +12 and -12 volts (see attached GIF) at the motor. Whether or not this produces any significant current flow in the motor windings is what we are discussing.

Why would they have power flowing at zero speed?

There’s not supposed to be. Locked antiphase PWM is supposed to be used with motors that have sufficient inductance so that negligible current flows under these conditions. Someone mentioned that the Denso Window motor inductance is 200uH. If this is true, I would expect to see less than 500ma* RMS flowing through the motor under the voltage conditions mentioned above. Compare this to the Denso’s stall current of 18.6 amps.

*Have you measured the current? *

I have asked Tom to measure the current using a small series resistor. He said he would do it, but hasn’t had a chance to yet. Tom: since we are dealing with such a small sense resistor, please be sure to measure the voltage drop across the actual resistor leads, so you don’t pick up extra contact resistance from the alligator clips or whatever you use to connect it.

Wouldn’t this cause the motor to heat up over time?

Yes, it would, IF there were sufficient current flowing. But apparently there’s not, because the the high PWM frequency and the motor’s inductance.

~

  • the Denso window motors are fairly low-tech motors not specifically designed to be used with PWM. Even though the low inductance draws more current at the zero-command 50% duty cycle condition than would normally be expected, it is still a small enough current not to cause motor heating. The Denso’s stall current is listed at 18.6 amps at 12VDC, so its resistance is about 0.65 ohms. 500ma flowing through 0.65 ohms produces only about 0.16 watts of I2R heating, which is why the motor does not heat up.

~





Al Skierkiewicz wrote:
The Victor should show the same waveform.

Not necessarily. The Victor may not be operating in locked antiphase. In fact, Tom’s scope measurements show that the Victor is not.

…about 18 ohms of series resistance

The data I have for the Denson window motor is 18.6 amps stall current at 12VDC. This would give a DC resistance of 0.65 ohms.

Ether, I am quoting the Jaguar manual from last year. It lists analog position control as only available under CAN control in Table 2-1. If the motor is given a FWD command and then a reverse command of equal amplitude, the armature will stay in one place given that the switching frequency equals or exceeds the ability of the motor shaft to turn in between pulses.

It is not the locked antiphase per se which holds the motor in position, it is the closed-loop servo control.

If an external load causes the motor to move from the commanded position, the closed-loop servo control adjusts the PWM duty cycle away from 50% in order to produce average motor current (and thus motor torque) to counter the load and hold the motor position. It is this non-50% duty cycle which draws current.

A locked antiphase motor controller not operating in a closed-loop position mode and providing a 50% -12/+12V signal at zero command would NOT hold the motor against an external load, because the 50% -12/+12V creates no net torque in the motor.

~

18 ohms is the impedance of a 200 uH inductance at 15kHz. I did not consider that this test is running in locked anti phase since no mention of feedback was listed. There does not appear to be any closed loop control. I am just speculating that the Jaguar can perform position control only under CAN. As strange as it seems, the Jaguar does appear to be very confused compared with the Victor in the same application commanded to zero throttle. I know that the Victor has a deadband at zero throttle, I am not aware of the modes of the Jaguar at zero throttle. This very well may be a jitter phenomena either supplied by the Jaguar itself or the code in use.