Espritmodel.com Telemetry Radio
Reply
Thread Tools
Old Oct 25, 2002, 12:46 PM
Señor Meember
Sloper Mario's Avatar
Joined Sep 2001
973 Posts
Switching Frequency Discussion

Hey guys,
I've seen a couple discussions that mentioned speed controller switching frequency along the way but they never went into any kind of detail. I've also read the FAQ on the subject and I understand what it is and how it works but I still have questions.

In an old thread a person showed surprise that a Jeti 110 speed controller had a switching frequency of 1000hz and that they didn't think any controllers were that low anymore. The E-zone FAQ it seems to suggest that anywhere between 1000hz and 5000hz was about the same but near 3000hz was optimum for the motors we run. Can anyone explain this further? How was 3000hz determined to be optimum?

I've also read a couple threads indicating that some motor failures might have been caused by too low a frequency. Again this was suggesting that 1000hz was too low. The only thing I could get from the FAQ on this was that really low frequencies like 50hz would heat the motor more implying more wear. Does each motor have a specific frequency at which it will last the longest? Is this frequency also the best for performance/efficiency?

In a real world example I have a Jeti 050 (1000hz) and 2 Castle Creations Pixie-7P (2800hz) controllers. With the CC controller I've burned up 2 GWS EDF-50 motors reverse wired as pushers inside of 5 flights each. Yet with the Jeti the motor has yet to go. When I switched to the EPD-50XCs on the CC controllers I didn't have any problems. All three planes are like this one http://www.datascape.net/plane/fwing-e/index.html . In all cases I was running 2/3 AAA 7 cell batteries so that's not a variable. I don't know if this is a frequency related problem but I thought I'd throw it into the discussion.

Your thoughts?
Mario
Sloper Mario is offline Find More Posts by Sloper Mario
Reply With Quote
Sign up now
to remove ads between posts
Old Oct 25, 2002, 03:48 PM
S55
Registered User
S55's Avatar
Beaverton, OR, USA
Joined Feb 2001
772 Posts
An ideal speed controller would provide a DC voltage to the motor. This could be done by using a switching DC/DC converter and adjusting its output voltage. These DC/DC converters use an inductor (plus other things) and the higher the switching frequency, the smaller (and lighter) the inductor. To put some numbers here, switching at hundreds of KHz is very common today.

Unfortunately you cannot use these high switching frequencies in RC, because of the even higher frequency harmonics that will be generated and which will interfere with the receiver. This high frequency noise cannot be removed completely, it can only be attenuated. An easy way to do this is to slow down the turn on and off of the mosfets. By doing this the instantaneous power dissipation during the transition between on and off and viceversa gets higher and your efficiency is in jeopardy. Things get back under control if the switching frequency is considerably reduced. The kilohertz region is good for that. But this hits you because now you need a big and heavy inductor.

Fortunately the motor is an inductive component, so it can play the role of the inductor. And this is how you get to the currently available light and cheap ESC’s. They are basically low frequency DC/DC buck converters where the inductor is removed and replaced by the motor. Things are not nice and clean yet though. An inductor needs to be reset. This means that if you apply V1volts across it for S1 seconds, then you need to allow it to develop the opposite voltage V2 for S2 seconds. (To avoid confusion about the numbers, we talk about micro and milliseconds here). V1 x S1 = V2 x S2, where V1 and V2 have opposite signs. The product volts x seconds has to be constant in steady state, otherwise the inductor will saturate.

Now let us take an example: 20000 RPM. This is 333 Rpsec, so one rotation lasts 1/333 = 3 milliseconds. Each rotor pole changes polarity over one rotation. Half of the time (1.5 millisec in our case) it is biased as a North pole and the other half it is a South. To make the volts x seconds product happy you do not have to interfere with the voltage change that is required by the North / South polarity change. If the ESC switching frequency is too low, you may run into the situation when the inductor is in need of a negative voltage and right then its voltage is switched to a positive one because of the North / South switching. I do not know whether saturating it will damage the motor, but it is not something to look for anyway. To minimize the effect of this kind of events the ESC switching frequency needs to be much higher than the switching dictated by the motor rotation. In our case the motor frequency is 333 Hz per pole, so the ESC frequency should be 10 to 20 times higher. This is how you get to 3 to 6 KHz.

For a low RPM motor a lower switching frequency ESC is OK. For a high speed one, a higher frequency ESC is desirable.

Be aware that even at these low switching frequencies the interference problem is still present and you should avoid long leads between battery and ESC and also between ESC and motor. Twisting the leads helps, but may not solve all interference problems. The true DC/DC converter I was talking about in the beginning (the one with the inductor) has less noise on the motor leads even at high switching frequencies, so these ones can be long. As far as I know there is no commercially available ESC of this type.

I hope this was not too boring.

S55
S55 is offline Find More Posts by S55
Reply With Quote
Old Oct 25, 2002, 04:52 PM
Registered User
Gordon's Avatar
Joined Aug 2000
3,918 Posts
Quote:
Originally posted by S55
I hope this was not too boring.
Absolutely not. This also explains eloquently why a 10-pole motor (eg Pletti washing machine) has a much lower max rpm than a 2-pole one (eg Hacker) on the same esc. Schulze state the figures on their website, but your post explains why.

Many thanks :-)

Gordon
Gordon is offline Find More Posts by Gordon
Reply With Quote
Old Oct 25, 2002, 06:47 PM
Señor Meember
Sloper Mario's Avatar
Joined Sep 2001
973 Posts
Good stuff S55. Not boring at all. I'll admit that I'm going to have to reread your post a few times before I get it all but I'll enjoy the lesson. Thanks!

Mario
Sloper Mario is offline Find More Posts by Sloper Mario
Reply With Quote
Old Oct 25, 2002, 11:08 PM
We want... Information!
Bruce Abbott's Avatar
Hastings, New Zealand
Joined Jan 2001
5,169 Posts
Using a low frequency increases motor heating because resistive power loss is proportional to current squared. For example if the duty cycle is 50%, on-time current is double the average, but heating is four times greater. Over the whole cycle the heating effect is twice as much as what would be created by the same amount of continuous current. The situation gets worse as the duty cycle is reduced (ie. lower throttle setting). Efficiency loss shortens flight time, and heats the motor more than expected at lower revs (when there is usually less cooling). Another factor often overlooked is that the brushes and commutator have to handle the higher peak current, causing more arcing and brush wear.

The minimum frequency required depends on how much loss is acceptable, and the inductance and resistance of the motor windings. Iron-cored motors have a large inductance (the iron concentrates the magnetic field), whereas coreless motors have low inductance and so need a much higher switching frequency. The differences caused by varying wire size and number of turns tend to cancel out, because both resistance and inductance increase with more turns.

At the high end, increasing frequncy causes greater losses in the motor due to eddy currents and skin effect. More importantly, switching losses in the ESC increase, and there is often a tradeoff between having short switching times (to increase efficiency) and reducing harmonics that cause RF interference. In cheaper units software execution speed may set the upper frequency limit.
Bruce Abbott is offline Find More Posts by Bruce Abbott
Reply With Quote
Old Oct 26, 2002, 06:17 PM
S55
Registered User
S55's Avatar
Beaverton, OR, USA
Joined Feb 2001
772 Posts
In fact the current through the motor has a triangular shape. It linearly goes up when the ESC mosfet is ON and then linearly goes down when the ESC mosfet is turned OFF. During the OFF time the current continues to flow through the motor and the Schottky diode. You need an oscilloscope and current probes to see this. A multimeter set on DC current will measure the average current and this will be higher or lower depending on the duty cycle, but the current will keep the triangular shape.

S55
S55 is offline Find More Posts by S55
Reply With Quote
Old Oct 26, 2002, 09:58 PM
We want... Information!
Bruce Abbott's Avatar
Hastings, New Zealand
Joined Jan 2001
5,169 Posts
These oscillographs show the effect. In each graph the upper trace is the drive voltage to the FET, and the lower trace is current through the motor. At 50Hz current is basically a sqaure wave, and as the frequency is increased it gets smoother. At 1KHz the waveform has become triangular, and current never quite gets down to zero. The peak current at 50Hz is almost twice as large as it is at 3KHz.

You can see the rise and rise and fall times of the drive waveform at 3KHz. As the frequency increases this becomes more apparent. Also there are spikes in the current waveform, caused by stray circuit inductance (some of it is due to using a wirewound resistor to sense the current). These spikes could cause RF interference.
Bruce Abbott is offline Find More Posts by Bruce Abbott
Reply With Quote
Old Oct 26, 2002, 10:16 PM
We want... Information!
Bruce Abbott's Avatar
Hastings, New Zealand
Joined Jan 2001
5,169 Posts
And here's what the voltage across the FET looks like (motor voltage is similar but upside down). During the low part of the waveform the motor is getting full voltage, and current ramps up. During the high part the motor gets no voltage except for about 0.5V across the shottkey diode, while current ramps back down. You can see a sharp drop towards the end as the diode turns off (compare this to the 500Hz current waveform - it reaches zero at the same point in time).

Below that is another graph showing what happens when you don't wire in the shottkey diode. With nowhere for the current to go at switch-off, the collapsing magnetic field creates a huge voltage spike that threatens to blow the FET (in this case it got up to about 45V ).
Bruce Abbott is offline Find More Posts by Bruce Abbott
Reply With Quote
Old Oct 27, 2002, 02:48 AM
Registered User
Gordon's Avatar
Joined Aug 2000
3,918 Posts
Great stuff, Bruce!

Will your setup show what's happening to current and voltage at the battery side of the esc?

There have been many queries on these forums regarding esc lead lengths for brushless motors (though Astrobob did comment that the results apply also to brushed motors), specifically whether one should lengthen motor or battery leads when the distance between motor and battery "as the wire goes and not as the crow flies" is long, as in a twin with batts in the fuz, motors in wing-mounted nacelles.

Astrobob, Tom Cimato (Maxcim) and Ulf Herder (Schulze esc designer) have explained why one keeps the battery leads short (long leads promote voltage spikes, caused by the inductance of the leads, which can blow the FETs) and it'd be great to see the current and voltage waveforms the battery sees, plus, the spikes on your 'scope pics if you can do it!

Using the foregoing advice, in a brushless-powered twin, one would mount the escs close to the batteries in the fuz, and take the motor leads out though the wing roots to the motors. Guys seem worried about having to feed 3 wires through for the brushless, but twisting them together is supposed to reduce potential interference and inductance problems. However, looking at the voltage spikes on the motor leads, which you illustrated above, it makes one wonder if lengthening the motor leads is the correct option for a brushed motor-powered twin. Or do all modern escs incorporate the Schottky diode, thus removing the voltage spike problem in the motor leads?

Many thanks for your very illuminating posts. Maybe I'm more than averagely interested in such things as I used to lecture RAF trainee technicians in radar principles oodles of years ago. I've forgotten it all now, but I still find the waveform diagrams (there were plenty of those in a radar timing diagram) and the principles really fascinating.

Gordon
Gordon is offline Find More Posts by Gordon
Last edited by Gordon; Oct 27, 2002 at 02:59 AM.
Reply With Quote
Old Oct 27, 2002, 02:07 PM
S55
Registered User
S55's Avatar
Beaverton, OR, USA
Joined Feb 2001
772 Posts
Great waveforms, Bruce!

The low frequency test clearly shows the inductor core is saturating. At least one effect would be higher core losses. When the current does not go up and down linearly you have core saturation. At 3 kHz everything is fine, 1 kHz is acceptable, while 50 Hz shows deep saturation.

S55
S55 is offline Find More Posts by S55
Reply With Quote
Old Oct 28, 2002, 04:43 AM
rebmeM roineS
MrBungle's Avatar
Rockingham, Perth - Western Australia
Joined Apr 2001
615 Posts
Quote:
Originally posted by Bruce Abbott
With nowhere for the current to go at switch-off, the collapsing magnetic field creates a huge voltage spike that threatens to blow the FET (in this case it got up to about 45V ).
.... and a huge amount of RF noise.
MrBungle is offline Find More Posts by MrBungle
Reply With Quote
Old Oct 28, 2002, 06:52 AM
Speed Demon
GregG's Avatar
Antioch,CA,USA
Joined Dec 1999
12,296 Posts
Bruce, great photo!! Thanks for the enlightenment.
GregG is offline Find More Posts by GregG
Reply With Quote
Old Oct 28, 2002, 10:01 AM
Registered User
morristown,nj,usa
Joined Sep 2002
83 Posts
Bruce - good snaps.

Have some questions.
Pardon my ignorance on real world parameters.

Are the motor & battery & ESC real & typical or is this a "simulation"? (S400,9.6v,...)
How did you manage to change the switching periodicity so much, vary some r/c somewhere?

For the display @3KHz, the current ramps up & never reaches saturation - saturation being sort of "full" current. Which suggests that, say, a 9.6v (perfect) battery can & should be used with a "lower voltage" motor, or alternatively, the frequency may be tweaked to maximize/optimize current for a "higher voltage" motor?
Is that realistic?
It seems intuitive that running the motor @ just saturation is optimal. Is that reasonable?

Since many ESCs seem to run ~1-3KHz, how far from
max current are the motors really running @? Is a typical 9.6v battery with a 7.2v motor combination optimal?
escapee is offline Find More Posts by escapee
Last edited by escapee; Oct 29, 2002 at 11:04 AM.
Reply With Quote
Old Nov 01, 2002, 05:58 AM
now that's a wattmeter...
simingx's Avatar
Singapore
Joined May 2002
1,136 Posts
Just my $0.02...

I monitored the current into my (brushless) Schulze 12.46e driving a Lehner Basic 4200 at full speed/no load, and this is what it looks like:
(sorry I can't remember what the scope settings were)



The output voltage (not current) waveform at half speed is:



The current waveform was taken from a short bit of shunt wire into a MAX4372H current sense amplifier. The voltage waveform was obtained by connecting the scope leads between two phases of the output.
simingx is offline Find More Posts by simingx
Reply With Quote
Old Nov 01, 2002, 08:16 AM
We want... Information!
Bruce Abbott's Avatar
Hastings, New Zealand
Joined Jan 2001
5,169 Posts
Quote:
Originally posted by escapee
Are the motor & battery & ESC real & typical or is this a "simulation"? (S400,9.6v,...)
How did you manage to change the switching periodicity so much, vary some r/c somewhere?
The motor is real (a Mabuchi 050) but the ESC is simulated. It's just a FET and diode driven from a signal generator. I held the motor stalled, to eliminate ripple caused by the motor generating voltage during the 'off' periods.
Quote:

For the display @3KHz, the current ramps up & never reaches saturation - saturation being sort of "full" current. Which suggests that, say, a 9.6v (perfect) battery can & should be used with a "lower voltage" motor
Remember these waveforms show a 50% duty cycle, producing about 25% of full power. The interesting thing is that average battery current is even lower, as it only flows during the 'on' period. Thus it is possible to use a low voltage, high current motor on a high voltage, low current battery. This is sometimes done in practice - when a motor is loaded for maximum power (way beyond the level of best efficency) but you only use full throttle sparingly!
Quote:
running the motor @ just saturation is optimal. Is that reasonable?
The waveforms do not show saturation, but simply the amount of current smoothing created by winding inductance. Even at full current the motor is probably nowhere near saturated (if it was the waveform would be asymmetric, rising sharper and flattening sooner at the top than at the bottom - due to inductance decreasing as current increases).

Greatest efficiency with a brushed ESC occurs at full power, as there are no switching losses and current is smooth. However, we often want to cruise at less than full throttle. A slight reduction in efficiency is acceptable as you are still saving lots of energy compared to simply trimming for level flight at full throttle!
Quote:
Since many ESCs seem to run ~1-3KHz, how far from
max current are the motors really running @? Is a typical 9.6v battery with a 7.2v motor combination optimal?
It depends on loading. If the motor is allowed to rev high enough (and doesn't fly to bits) it can handle a lot more than its rated voltage. A 7.2V Speed 400 can comfortably handle 12V. The trick is to keep current down by increasing the gearbox ratio (or using a smaller prop - but then you lose aerodynamic efficiency).
Bruce Abbott is offline Find More Posts by Bruce Abbott
Reply With Quote
Reply


Thread Tools

Similar Threads
Category Thread Thread Starter Forum Replies Last Post
Will changeing the switching frequency help BAR-STOOL Castle Creations 7 Nov 21, 2003 10:55 AM
Switching frequency module on MPX 3000 series MaxFx Radios 1 Aug 04, 2003 06:38 AM
Best switching frequency (FUN 600-18 + Futura 18.48K) ThomasJoehnk Power Systems 6 Jul 12, 2003 03:59 PM
ESC switching frequency... high or low? paulaac Power Systems 22 Feb 27, 2002 11:14 AM