Single supply active LPF

[pyrohaz]

Hey,

I've got a 280KHz PWM Frequency with a sample rate of 45kHz, i'm using the MCP6001 opamp from microchip and need to make a low pass filter with the cutoff around 22.5kHz. I chose this op amp because it works at 3.3v and has rail to rail input and output. The PWM input will be between 0 and 3.3v.

I really want to minimize the amount of PWM frequency reaching the output (hence the low cutoff) along with sampling noise. Would a first order LPF suffice?

I'll include a picture of my current design. If anybody knows a better way of doing it, i'd be very appreciative!

The input to the buffer opamp is the sum of two 8 bit PWM's to produce a 16bit output. The 3k and 768k resistors are matched to 0.1%.

In my circuit, the cutoff freq is roughly 23.4kHz.

- - - Updated - - -

Hmm after simulating a standard active lowpass, a sallen key lowpass and MFB lowpass, the MFB option gives the best cut but all of these don't seem to be biased at vss/2, what is the reason for this?

Cheers,

 

[crutschow]

....................................

Hmm after simulating a standard active lowpass, a sallen key lowpass and MFB lowpass, the MFB option gives the best cut but all of these don't seem to be biased at vss/2, what is the reason for this?

A single supply op amp must be biased a vss/2 if the output signal tries to go below ground. If not, then it can be biased at ground potential.

A Sallen-Key filter uses a non-inverting amplifier so it can be biased at 0V with a positive going input.

A MFB filter uses an inverting amplifier configuration, so it must be biased at vss/2 for positive going inputs and the input capacitor coupled to provide a DC block.

 

[vikash23]

Hi I am using a PI filter 22nF-22uH-22nF for cutoff frequency of 400KHz. I calculated this from online tutorial. I used a function generator to see how it works actually. I used 1MHz frequency from function generator and feed through one end of PI FIlter (22nF-22uH). I used Oscilloscope to measure the output cutofff frequency on the other end of PI Filter (2uH-22nF) and to see the function generator output (1MHz as well).

I am not able to see the output cutoff frequency 400KHz. Where am I going wrong. Why I was not able to see the cutoff frequency ?

Help me out.

 

[pyrohaz]

Ok I understand about the Vss/2 stuff but my input is merely a square wave between 0v and 3.3v, since its a PWM waveform, the DC bias will be dependent on the pulse width surely? Will this pose a problem to any of these filters?

Cheers,

 

[crutschow]

Ok I understand about the Vss/2 stuff but my input is merely a square wave between 0v and 3.3v, since its a PWM waveform, the DC bias will be dependent on the pulse width surely? Will this pose a problem to any of these filters?

Glad you understand that "stuff". :grin:

The average of a 0-3.3v PWM waveform will certainly vary with the duty cycle. But the average is always above 0V so that's no problem for a non-inverting filter, such as a Sallen-Key.

- - - Updated - - -

Hi I am using a PI filter 22nF-22uH-22nF for cutoff frequency of 400KHz. I calculated this from online tutorial. I used a function generator to see how it works actually. I used 1MHz frequency from function generator and feed through one end of PI FIlter (22nF-22uH). I used Oscilloscope to measure the output cutofff frequency on the other end of PI Filter (2uH-22nF) and to see the function generator output (1MHz as well).

I am not able to see the output cutoff frequency 400KHz. Where am I going wrong. Why I was not able to see the cutoff frequency ?

Help me out.

It's not good manners to hijack someone else's thread to ask your own question. Please start your own thread. :-|

 

[pyrohaz]

Glad you understand that "stuff". :grin:

The average of a 0-3.3v PWM waveform will certainly vary with the duty cycle. But the average is always above 0V so that's no problem for a non-inverting filter, such as a Sallen-Key.

My apologies! I didn't mean to sound rude there.

Knowing that the slew rate of the MCP6001 is 0.6v/us and the GBW is 1MHz, will this affect the linearity of the output?

 

[crutschow]

.............................

Knowing that the slew rate of the MCP6001 is 0.6v/us and the GBW is 1MHz, will this affect the linearity of the output?

It could. To minimize this problem, you could use a single-pole passive RC low pass filter in front of the active filter. That will reduce the high frequency content of the signal to the active filter input so that any op amp frequency limitations should not significantly affect the output linearity. Here is a discussion of such a circuit to make a 3-pole filter.

 

[pyrohaz]

That website you posted is absolutely brilliant! I've designed my 3 pole LPF and the amount of ripple present at the output from the PWM is around 3mV in simulations! Obviously it won't be this good in real life conditions but thats more than good enough.

The buffer opamp that actually buffers the incoming PWM signal is the one i'm worried about for linearity though, simulations show that this is causing a low pass effect on its output. Is there anyway I can reduce this non-linearity? Could I use some form of negative feedback?

 

[crutschow]

Why do you need a buffer amp?

 

[pyrohaz]

The PWM output will have a varying but quite high impedance. The PWM output is the sum of two 8 bit outputs through a 3k and 768k resistor (the 3k resistor signifies MSB's, the 768k signifies LSB's), without buffering, surely the impedance load of the low pass filter will affect the PWM output?

 

[crutschow]

Yes, you need a buffer to avoid loading of the signal.

You could perhaps use a transistor of MOSFET for that, or else a higher frequency op amp such as an LT1498.

 

[pyrohaz]

I had originally thought of using a mosfet or transistor, the only problem being that in the case of a transistor, i'd have 3.3-0.7v range which isn't very good (equates to about a 3 bit loss). I think that moving to an op amp with a higher GBW and slew rate is the best option here, since its easily available no farnell, I think i'm going to go for the OPA2350.

Thank you for all the help!

 

[crutschow]

I had originally thought of using a mosfet or transistor, the only problem being that in the case of a transistor, i'd have 3.3-0.7v range which isn't very good (equates to about a 3 bit loss). I think that moving to an op amp with a higher GBW and slew rate is the best option here, since its easily available no farnell, I think i'm going to go for the OPA2350.

A transistor output can go between essentially between ground and a higher voltage supply, so don't understand your concern about the voltage range. Unless I misunderstand the nature of your PWM signal. It's just a duty-cycle modulated constant-amplitude square-wave, correct?

But that high frequency op amp should work fine. Just be sure you do a careful layout to keep input and output signals well separated with 0.1µF decoupling directly at the op amps power pin. Such high frequency op amps are prone to oscillations much more than you standard 1MHz GBWP op amps.

 

[pyrohaz]

I think think there is a bit of misunderstanding, because its two PWM outputs combined through a resistor network, its equivalent to a 2bit PWM output, the achievable voltage levels will be:

3.3, 3.29, 0.13 and 0v.

Power supply wise, i'm limited to the 3.3v provided by my MCU.

 

[crutschow]

OK, so the signal varies both amplitude and pulse-width?

 

[pyrohaz]

Yep thats the one!

Any advice?

 

[crutschow]

Well, that's a new one on me. :shock: What's the purpose of changing both the amplitude and the pulse width of the signal? Why not just use a 16-bit D/A converter?

 

[pyrohaz]

Well by changing the amplitude, it gives me additional bit depth, by using two higher frequency 8 bit PWM outputs, I can create a pseudo 16bit high frequency PWM output (the idea was originally proposed for some Mars exploration project: http://www.edn.com/design/analog/4329365/Combine-two-8-bit-outputs-to-make-one-16-bit-DAC), Open music labs took this idea further and applied it to music: http://www.openmusiclabs.com/learning/digital/pwm-dac/dual-pwm-circuits/

What they don't include here is how to filter the output, hence why i'm currently doing that part!

And with regards to a 16bit DAC, I don't have enough outputs for a 16bit parallel load DAC, my processor features SPI at 16bit but I was unable to find a purely 16bit SPI dac. I2C has too slow transfer time (SPI can be implemented by merely loading the data into the DR register which is lucky).

I'm using the STM32F103 on a Maple board.

Cheers,