Jump to content


Photo

noise reduction in sensor output


  • Please log in to reply
9 replies to this topic

#1 taan_in

taan_in

  • Members
  • Pip
  • 1 posts

Posted 14 September 2005 - 07:09 PM

Hi
I am trying to implement on DSP based DTC control for Induction Motor.....so I thought as a step of progress to run my motor through pwm inverter by generating one pwm through dsp..........and simultaneously estimate flux ,speed...etc. from the voltage and current sensor outputs........when i am running the motor through inverter my sensor outputs are getting distorted......can anybody suggest anything so that I can reduce the noise.....
thanks

#2 marke

marke

    Posting Freak

  • Moderator
  • PipPipPipPipPipPip
  • 2,603 posts
  • Gender:Male
  • Location:Christchurch, New Zealand

Posted 17 September 2005 - 10:29 PM

Hello taan_in

Welcome to the forum.

One of the major issues is pcb layout and earthing. If you have long wires and/or tracks, they must be screened. It is important that the screen does not add extra noise to the signal, so the screen should be connected to the earth at the measuring device. Sometimes it is best to leave the other end of the screen open circuit.
The noise you are seeing is most probably the result of noise generated by the output switching waveform. Perhaps ypu can take measures to reduce this noise by applying a filter in the AC input to the drive, connecting a small capacitor from the DC bus +ve leg to earth, and adding a ferrite filter to the output. Layout and earthing is very important to reduce generated noise.

Best regards,

#3 Carl

Carl

    Member

  • Full Member
  • PipPipPipPip
  • 67 posts

Posted 18 September 2005 - 08:22 AM

Hello tann_in,
Circuit layout is crucial and the selection of you electronic components is important.
I shall assume that you have a shunt resistor in the arm of the motor ..usually of a value 0.01ohms. This resistor should be a non wire wound resistor, as short as possible….basically it should ideally have 0 inductance. If you have such a resistor tell I and I’ll make you a millionaire.
Basically your PWM introduces a ripple current …the idea is to read the motor current and not the ripple.
If your PWM is greater than 16kHz then you could filter it. If it is less then the measured value is something less than real time ..because of the filter delays.
If you look at one pwm pulse you could sample the shunt in the middle of the pulse thus eliminating the rising and falling edge and all the harmonics that come with it.
If you take the above point a little further you would want to sample the current when none of the IGBT’s are switching……………….and tat point would be in a centre aligned PWM switching strategy.
If you are using a differential op amp, to read the shunt, the layout of this is component is crucial. I suggest you find a practical book on EMC that gets down to the nitty gritty of EMC.

When you set up the op amp be aware that the current can flow in both directions.

Hope this helps you in the right direction..
The selection of the op amp is important basically one with a good CMRR and slew rate

Good luck
carl

#4 marke

marke

    Posting Freak

  • Moderator
  • PipPipPipPipPipPip
  • 2,603 posts
  • Gender:Male
  • Location:Christchurch, New Zealand

Posted 18 September 2005 - 09:16 AM

Hello Carl

I would not use a resistor in series with the motor, this will certainly create noise problems due to the switching of the output circuit plus the isolation will be a major problem. Modern Hall effect CTs are useful for measuring motor currents on inverters. This is the technology that is usually used for measuring inverter signals..
Some inverters rely on the measurment of the DC bus current only.
Best regards,

#5 Carl

Carl

    Member

  • Full Member
  • PipPipPipPip
  • 67 posts

Posted 18 September 2005 - 09:41 AM

Hello Mark
I don't see the problem with a 0.01 ohm shunt; this will have negligible impact on the motor circuit. If you consider the typical inductance of a component lead is 1nH/mm
And the switching frequency is +-15kHz. I don’t see the impact.
I would agree if it is an inverter greater than 2.2kW then isolation vs. cost…isolation would win. If it is <2.2Kw, then isolation to the user would be the only isolation that is of importance. Certainly international rectifiers advocate this in some of there designs.

#6 marke

marke

    Posting Freak

  • Moderator
  • PipPipPipPipPipPip
  • 2,603 posts
  • Gender:Male
  • Location:Christchurch, New Zealand

Posted 18 September 2005 - 10:25 AM

Hello Carl

It is not the resistor in series with the motor that is the problem, it is how you will couple this measured voltage back into the monitor circuit/micro without introducing a lot of noise beariung mind that you will probably want a broad bandwidth.
The motor terminal will be leaping up and down at your switching frequency (perhaps 5 - 15 kHz) between the +ve DC bus and the -ve DC bus at a rate of higher than 1000V/uS and a voltage swing of perhaps greater than 600V pk-pk.
Using a high speed hall effect CT eliminates these problems.

Best regards,

#7 Carl

Carl

    Member

  • Full Member
  • PipPipPipPip
  • 67 posts

Posted 18 September 2005 - 10:52 AM

Hello Mark,

That is the point of sampling between the pulses and filtering the signal.

#8 Carl

Carl

    Member

  • Full Member
  • PipPipPipPip
  • 67 posts

Posted 18 September 2005 - 11:10 AM

hello Mark
The op amp should have a GBW > 1Mhz, slew rate > 0.5v/uS and a low input offset voltage. It shoul also be arail to rail output/input. The TLV2460 would do.

|
----[ ]-----|

#9 marke

marke

    Posting Freak

  • Moderator
  • PipPipPipPipPipPip
  • 2,603 posts
  • Gender:Male
  • Location:Christchurch, New Zealand

Posted 18 September 2005 - 06:47 PM

Hello Carl

While I see it as being accedemicaly possible to monitor the current by using a shunt resister in series with each phase to the motor, I consider that there are many potential problems that make it a less desirable approach.
You have a high common mode signal with a small signal to be measured. Your common mode voltage is in the order of 600vpk-pk and you are looking at voltages in the order of fractions of a volt. This is a tall order! Yes you can sample between the pulses, but some of the pulses are very narrow which would not allow for any settling time.
You could use an amplifier at each shunt to increase and preprocess the signal before translating it to the micro for measurement, but this is going to require additional circuitry and power supplies not to mention some means of isolation. - more complicated than a high speed Hall effect CT!
If you sample between the pulses, you will need to allow sufficient setling time after the rising or falling edge which means that you will not see the effects of a short circuit on the output for some time. This will compromise the protection of the inverter.
If you really want to use a shunt resister, use a shunt in series with the -ve DC Bus. That way, there is only one sense circuit and there is a much smaller common mode problem.
- why make it hard when you can chase the impossible??

Best regards,

#10 Carl

Carl

    Member

  • Full Member
  • PipPipPipPip
  • 67 posts

Posted 18 September 2005 - 08:05 PM

Hello Mark
You might find the ic IR2171 or IR2271 of interest. This is a dedicated current feedback ic in the motor leg. I think your CT has some competition!
Regards
Carl




0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users