Triplen harmonics in star-connected 3-phase system?

Thread Starter

powersys2005

Joined Jan 3, 2010
29
Hello,

I've a 3-phase PM brushless motor. It's operated as a BLDC drive. The output of one of the phase current sensors is connected to Agilent Dynamic Signal Analyzer 35670A to obtain the FFT of the phase current. I find out that there are triplen harmonics (their amplitudes are not very high compared to other non-triplen odd harmonics, e.g. 5th, 7th, 11th, 13th, etc) in the FFT spectrum. If I'm not mistaken, for a star-connected 3-phase system, triplen harmonics should not appear because neutral (star point) is left open (i.e. not connected to ground). So, any idea why triplen harmonics are observed in the phase current in my system?

Thanks.
 

3ldon

Joined Jan 9, 2010
82
"(their amplitudes are not very high compared to other non-triplen odd harmonics, e.g. 5th, 7th, 11th, 13th, etc)"

This makes sense if you are driving it w/i a six step drive, its not going to filter out 100%, its being driven with a square wave.
 

Thread Starter

powersys2005

Joined Jan 3, 2010
29
Hi 3ldon, thanks for your reply. Yes, the PM brushless motor is driven by a six step drive (i.e. pseudo-squarewave current waveform). What do you mean by "not going to filter out 100%"?
 

3ldon

Joined Jan 9, 2010
82
Its not going to filter out the 3rd harmonic with 90db attenuation because your motor isn't ideal. it is a complex network of inductors, stray capacitance and iron loss.

This should not surprise you.

Also, "not really high" is really vague.
 

lmartinez

Joined Mar 8, 2009
224
The following attachment will assist your original question. As part of the spectrum of signals within a delta system, there can be third harmonic voltages observed within the phases of such said system. Take a look at page 8 of the following attachment. Good observation :)
 

Attachments

Thread Starter

powersys2005

Joined Jan 3, 2010
29
Hi Imartinez, thanks for your document. However, I still could not relate the information in page 8 to my question because in my case, the motor has star-connected or Y-connected windings.

By the way, I'm very interested in the question posted in page 11. Would you share the answer here?

Thanks.
 

3ldon

Joined Jan 9, 2010
82
Hi Imartinez, thanks for your document. However, I still could not relate the information in page 8 to my question because in my case, the motor has star-connected or Y-connected windings.

Thanks.
Don't connect the motor in delta, the whole point of using a 6 step drive is to boost torque at the expense of higher iron and copper losses.
and you still haven't said how much third harmonic current is flowing into the motor, so there is no determining whether or not this is an issue.
.
A few percent is normal, the 5th is going to be higher than 20% anyway.
I stated the reason is in my previous post.

.
 

lmartinez

Joined Mar 8, 2009
224
What you are observing is an additional benefit (annoyance in this case) from the inverter(not a pure positive sequence sinusoidal voltage generator :eek:). It does not apply to a three phase electrical power system as described by your post above. The electrical system you have described is composed of many harmonics generated by the inverter !!!!


"240VAC -> DC Power Supply -> 3ph Inverter -> 3ph Permanent Magnet Brushless DC Machine (with star-connected windings). The motor is operated as BLDC drive (i.e. six-step PWM current control)"

The question brought up in page 11 might be answered by the math shown in page 7 of the such document.
 

Thread Starter

powersys2005

Joined Jan 3, 2010
29
Don't connect the motor in delta, the whole point of using a 6 step drive is to boost torque at the expense of higher iron and copper losses.

And you still haven't said how much third harmonic current is flowing into the motor, so there is no determining whether or not this is an issue.
.
A few percent is normal, the 5th is going to be higher than 20% anyway.
I stated the reason is in my previous post.
.
Kindly find the harmonic plots below. Even harmonics as well as triplen harmonics (e.g. 3,9,15, etc.) are quite significant in the frequency harmonic plotted in logarithmic scale. Kindly advise.


Figure 1.


Figure 2.
 

Attachments

3ldon

Joined Jan 9, 2010
82
1% != significant

This is not an issue.

Also, keep in mind that the higher harmonics may be caused by the inherently non linear feed back loop that invariably exists in there somewhere, as your FFT is averaging the data over several cycles..
Or is this open loop?

Looking at a square wave from a 311 op amp on a cheap electronic oscope/FFT, the even harmonics are at the same Dbm as the 23rd harmonic.

EDIT: square wave wasn't too square... rise and fall time was 2%
the oscope compensation 1Khz square wave is showing no discernible even harmonics.

What do the voltage waveforms look like going into the motor?
 
Last edited:

Thread Starter

powersys2005

Joined Jan 3, 2010
29
Also, keep in mind that the higher harmonics may be caused by the inherently non linear feed back loop that invariably exists in there somewhere, as your FFT is averaging the data over several cycles..
Or is this open loop?
It's a closed loop speed control. Would you please give some examples of non-linearity in a feedback loop?
The frequency spectrum is measured using Agilent 35670A and the number of averaging is 20. May I know how "FFT is averaging the data over several cycles" will cause higher harmonics?

What do the voltage waveforms look like going into the motor?
The motor is operated as BLDC drive, similar control approach described in the attached AN885 PDF document (see Figure 1). I usually call this control approach as conventional six-step PWM current control. The terminal voltage and current waveform look similar to that shown in Figure 2.


Figure 1.
http://forum.allaboutcircuits.com/attachment.php?attachmentid=15354&stc=1&d=1263708958

Figure 2.
http://forum.allaboutcircuits.com/attachment.php?attachmentid=15356&stc=1&d=1263711325
 

Attachments

Last edited:

3ldon

Joined Jan 9, 2010
82
I'm not sure what your question is.

while the feedback loop may or may not be linear, the fact is the load is not. when the motor is free wheeling, your system still has a certain amount of jitter. Take a look at the duty cycle of a cheap 12 to 5 volt converter in a portable hard drive under no load for an example, and I guarantee you that your FFT will show something crazy compared to a standard square wave.
Also try backman or hamming or hann instead of the default window, or tell it to do a vector average instead of an RMS maximum, it may give you a better picture of what is going on.
 

Thread Starter

powersys2005

Joined Jan 3, 2010
29
while the feedback loop may or may not be linear, the fact is the load is not. when the motor is free wheeling, your system still has a certain amount of jitter. Take a look at the duty cycle of a cheap 12 to 5 volt converter in a portable hard drive under no load for an example, and I guarantee you that your FFT will show something crazy compared to a standard square wave.
Also try backman or hamming or hann instead of the default window, or tell it to do a vector average instead of an RMS maximum, it may give you a better picture of what is going on.
I use Agilent 35670A Dynamic Signal Analyzer in my experiment. At the moment, I use Hanning window and RMS averaging. I'm not sure if Agilent 35670A can perform "vector averaging" or not. I need to check it out. Or, do you mean "time-domain averaging".
Thank you very much
 
Top