Real-Time Measurement and Control
Why does Pulse Width Modulation not work as intended with NI 9474?
Hi,
I am having some trouble implementing Pulse Width Modulation (PWM) with my cRIO. After some trouble I decided to make the following experiment today: I used the PWM example from the Example Finder to set the according FPGA VI to generate a 50% duty cycle PWM signal and did this for a set of PWM period lengths. I then measured the output using a standard handheld Voltmeter in DC mode. Hence, the voltmeter should read, what a motor that is connected to PWM port would "see". The supply voltage to the NI 9474 was 20V at all times.
For example, I set the PWM period length to 40,000 ticks for a 1kHz PWM signal with a 20,000 ticks PWM duty cycle length. However, instead of reading 10V output voltage, the Voltmeter gave me a reading of 17V. Varying the PWM period length gave the results as in the following graph:
This indicated to me that the digital output module 9474 which is supposed to have a switching speed of less than 1us, does in fact have some trouble keeping up with the speeds at much lower speeds. Hooking an oscilloscope up with the 9474 prooved, that the output of the PWM signal was every thing but a square wave (to be exact: a noisy square wave).
I attach a VI documentation of the FPGA VI I used. Any hints on why the PWM does not work as intended would be greatly appreciated.
I am having some trouble implementing Pulse Width Modulation (PWM) with my cRIO. After some trouble I decided to make the following experiment today: I used the PWM example from the Example Finder to set the according FPGA VI to generate a 50% duty cycle PWM signal and did this for a set of PWM period lengths. I then measured the output using a standard handheld Voltmeter in DC mode. Hence, the voltmeter should read, what a motor that is connected to PWM port would "see". The supply voltage to the NI 9474 was 20V at all times.
For example, I set the PWM period length to 40,000 ticks for a 1kHz PWM signal with a 20,000 ticks PWM duty cycle length. However, instead of reading 10V output voltage, the Voltmeter gave me a reading of 17V. Varying the PWM period length gave the results as in the following graph:
This indicated to me that the digital output module 9474 which is supposed to have a switching speed of less than 1us, does in fact have some trouble keeping up with the speeds at much lower speeds. Hooking an oscilloscope up with the 9474 prooved, that the output of the PWM signal was every thing but a square wave (to be exact: a noisy square wave).
I attach a VI documentation of the FPGA VI I used. Any hints on why the PWM does not work as intended would be greatly appreciated.
Ok, one of my colleagues solved the problem for me now. But I explain what was going on anyways.
This is a photo of the oscilloscope screen showing the DO running at 1kHz (Period Length: 40,000 ticks, duty length: 20,000ticks).
You can see, that the Voltage does not drop to zero immediately. this problem is getting even worse when putting the frequency up to over 10kHz:
Hence the average of the signal is nowhere near the 10V you would expect from a 50% duty cycel PWM.
The solution is to put a resistor between the channel output and COM, i sketched this below (sorry for the bad quality).
I don't really understand why it is necessary to put a resistor there, I cannot see why the Rio module should not include this extra piece of circuitry.
This is a photo of the oscilloscope screen showing the DO running at 1kHz (Period Length: 40,000 ticks, duty length: 20,000ticks).
You can see, that the Voltage does not drop to zero immediately. this problem is getting even worse when putting the frequency up to over 10kHz:
Hence the average of the signal is nowhere near the 10V you would expect from a 50% duty cycel PWM.
The solution is to put a resistor between the channel output and COM, i sketched this below (sorry for the bad quality).
I don't really understand why it is necessary to put a resistor there, I cannot see why the Rio module should not include this extra piece of circuitry.
