[SOLVED] pulse width modulation linearity

[kishore3677]

Hello,

need some help. I was trying linear PWM with a comparator(AD790) using a triangle wave and DC voltage. To my surprise, PWM output duty cycle is not linear when DC voltage changes. Any suggestions..

gvk

 

[ZekeR]

Biggest suggestion is to probe the waveforms and see if you're getting what you expect. I would think that, even with non-idealities such as comparator delay and noise injection into the DC voltage, it would still be a linear relationship between DC voltage and duty cycle. At the moment, my thought is that your triangle wave isn't very linear. (Perhaps the triangle is curved, or has ringing superposed atop it).

 

[kishore3677]

Hi,

Thanks a lot. But I have checked everything, I am using a agilent wave generator. DC precision source and comparators bypassed with decoups. But there is still a problem. Have you done anything like this before or would you give any suggestions to improve. Also are you sure you would get 100% linear output?

regards
gvk

 

[fareen]

why dont you post your circuit, gives the better idea. to get linear PWM triangular wave should be linear
try changing decopling capacitor

 

[FvM]

Besides analysing the circuit, the first point would be to check the results against expectable non-linearity according to AD790 specifications, e.g. "propagation delay versus overdrive".

Generally, no analog PWM modulator can be perfect. Please tell your detail results to allow a meaningful discussion. It would be also interesting how you actually measure PWM linearity.

 

[ZekeR]

But I have checked everything

If you checked everything, then you would've found the answer =)

Please tell your detail results to allow a meaningful discussion. It would be also interesting how you actually measure PWM linearity.

I concur. I still don't see anything that would be causing nonlinearity, just from the basic description you've provided.

Another possibility, how high does your ramp go to? The comparator is only specified to work properly when the input voltages are below Vin-2V... Perhaps you're exceeding this.

 

[kishore3677]

Hi,
To measure PWM linearity, I was just measuring the width in oscilloscope. I was expecting a proportional change in width with respect to a DC voltage. My VIN goes upto 8V. My triangular peak is at 9V. By seeing all your answers, I am still unclear as how to make Pulse width linear to DC voltage.

gvk

 

[fareen]

why dont you post your circuit ???
that will help solving your problem

 

[kishore3677]

Please find the attached circuit for your reference

gvk

 

[FvM]

According to AD790 data, the input voltage range is -15 to +13V with +/-15V supply voltage. So there's no obvious reason why your "PWM-modulator" shouldn't work. At which point do you see an unexpected behaviour?

 

[godfreyl]

why dont you post your circuit ???
that will help solving your problem

A photo may help too. Never underestimate the power of bad layout and wiring.:wink:

Please find the attached circuit for your reference

Thanks, but what's wanted is your circuit, not a copy of a circuit from the datasheet.

 

[crutschow]

........................
By seeing all your answers, I am still unclear as how to make Pulse width linear to DC voltage.

One way is to use negative feedback. You filter the PWM output to generate the average DC voltage and feed that back as negative feedback to an op amp that also has the DC control voltage as an input. With proper scaling of the voltage, the op amp output voltage to the PWM modulator will be such as to correct for any non-linearity in the modulator. Some form of that feedback circuit is typically used in PWM switching voltage regulators to regulate the output voltage.

 

[kishore3677]

Would you suggest some ICs which gives highly linear PWM output for input DC Range of 0- 10V?

 

[ZekeR]

To measure PWM linearity, I was just measuring the width in oscilloscope. I was expecting a proportional change in width with respect to a DC voltage. My VIN goes upto 8V. My triangular peak is at 9V. By seeing all your answers, I am still unclear as how to make Pulse width linear to DC voltage.

Sorry, when I said that "the comparator is only specified to work properly when the input voltages are below Vin-2V," what I meant by "Vin" was the supply voltage. What's your power supply's positive voltage?

A big thing that you still haven't mentioned: Instead of observing a linear relationship, what sort of relationship are you seeing? Showing us a graph of V_dc vs. Dcycle could be helpful.

 

[kishore3677]

My supply voltages are +/- 12V. I did plot the graph between Input DC voltage and output pulse width measured value and the equation of the graph is order 2. Same is the case with Duty cycle. Now can I use a DDS chip ? will it give me high accuracy triangle wave?

 

[crutschow]

If you want a high accuracy PWM output then perhaps a digital audio IC such as this would work.

 

[FvM]

I did plot the graph between Input DC voltage and output pulse width measured value and the equation of the graph is order 2.

That doesn't say much without a linearity number. You also didn't mention the PWM frequency.

Same is the case with Duty cycle. Now can I use a DDS chip ? will it give me high accuracy triangle wave?

What do you want to achieve? Highest pwm linearity can be achieved with a pure digital pwm controller, you don't need to take a DDS triangle generator detour. The said "agilent wave generator" is a DDS design, I assume. Depending on the triangle frequency, limited output filter bandwidth affects triangle linearity however, something that can be easily calculated.

 

[kishore3677]

Hi FvM,

What I am looking for is that I want to prove that my pulse width changes proportionally to my input DC voltage and that should be linear. for example , 1 V DC input results in 200 us pulse width(say),then 2V input should result in 400us and so-on. My tri wave frequency is 1KHz.

regards
gvk

 

[FvM]

1 kHz is in fact a moderate frequency. Generator bandwidth or comparator propagation delay shouldn't matter in this range. How much is the linearity devation?

 

[kishore3677]

Hi FvM,

Here are the data points captured:
VIN(V) VOUT(width in us)
0 500 ( for 1 KHz)
1 450
2 400
3 345
4 293
5 243
6 193
7 143
8 92
9 41

any clues?

regards
gvk