Switched Cap Low Pass Filter in audio circuit

[Jester]

I'm working on an audio circuit with a frequency response of 600 - 3000Hz. The input signal has significant high frequency noise (sounds like 10kHz white noise). An 8th order SCF (MAX7400) with a cutoff of 3kHz has little effect audio wise, however if I inject a signal and measure the attenuation of the filter the signal is down by 50dB before 4kHz at the output of the filter. So the filter is working, however the background noise (when I listen with an audio amplifier) seems about the same at the input and output of the filter, the tone of the nose changes slightly but that's about it.

Looking at the signal with a scope there is significant noise at the SCF oscillator frequency of 300kHz (10mVp-p), and some even higher frequency, lower level noise ~0.5mV at perhaps 1MHz but nothing obvious at lower frequencies. So I can't see what I hear.

One more thing if I listen with my amplifier to the Vcc line feeding the SCF and opamps the noise sounds almost identical the the filter output. If I listed to ground, it's silent.

Any suggestions?

Looking at the data sheet, does this graph imply that with really low signal levels (my problem case) that the filter does little to nothing (see red arrow)

 

[KlausST]

Hi,

sounds like 10kHz white noise

these are two totally different signals.

An 8th order SCF (MAX7400) with a cutoff of 3kHz has little effect

The SCF should attenuate the 10kHz signal. But it won´t change SNR within your frequency range of 600 - 3000Hz.

the tone of the nose changes slightly but that's about it.

Then i guess the problem is noise wihtin your frequency range of 600 - 3000Hz.

One more thing if I listen with my amplifier to the Vcc line feeding the SCF and opamps the noise sounds almost identical the the filter output. If I listed to ground, it's silent.

I think here is the problem. Try to clean your VCC (and other voltages like reference). Use tantalum and/or film capacitors (instead of electrolytic and ceramic capacitors).
Bypass every IC Vcc connection with a capacitor to your ground plane.

Looking at the data sheet, does this graph imply that with really low signal levels (my problem case) that the filter does little to nothing (see red arrow)

I did no calculation on the cahrt data.. but what you see is typical:
From left to right you see a falling line.
On the left side it starts with a signal amplitude of "0". No signal. But noise is still there. So the signal_to_noise ratio is a bad as it can be.
Now traveling to the right the signal is increasing. If you imagine that the noise level is constant (i don´t know) but the signal is increasing this means that the signal_to_noise ratio is improving.
At almost the end you see the best SNR. But then - caused by clipping distortions - the THD increases. And the chart shows SNR + THD, therefore the signal quality becomes worse....

Please tell us more about your application. Show us your schematic and PCB layout and we can better assist you.

Hope this helps.

Klaus

 

[Jester]

Klaus,

Every IC has decoupling, however the decouple capacitors are 100n ceramic

I can't share the schematic (NDA).

Board is 4 layer with solid ground plane.

This is how the ground, Vcc and output of the SCF look on the scope.





If I connect an audio amp to the green signal , it's silent, however the pink and yellow signals both sound about the same (significant background hiss).

 

[crutschow]

So the Vcc line sounds the same as the signal?
What is generating all that noise on Vcc?
Do you have a digital circuit operating nearby?

 

[Jester]

The yellow line is Vcc, looks to be about 200uVp-p, to my eye it looks identical to the green (silent) ground line.

The purple trace is the output of the filter and what you see is 10mV of clock feed through, however that's at 300kHz so I don't expect I hear any of that even with my bat ears on.

 

[KlausST]

Hi,

however that's at 300kHz so I don't expect I hear any of that

true, but you may hear every any deviation from that 300kHz. like change in frequency, change in duty cycle and so on.

from datasheet:

Anti-Aliasing and Post-DAC Filtering
When using the MAX7400/MAX7403/MAX7404/
MAX7407 for anti-aliasing or post-DAC filtering, synchronize
the DAC and the filter clocks. If the clocks are
not synchronized, beat frequencies may alias into the
passband.

What is your clock frequency and where does it come from?

Just to post filter the 300kHz you could use a simple 1k / 33nF RC low pass filter.

The signal should look better on scope, but i don´t expect audible changes.

Klaus

- - - Updated - - -

Hi,

some adds:

* did you read datasheet about pre- and post-SCF filtering?

May i ask why you need that SCF filter.
I personally don´t like them and instead try to use analog filters, ADCs with built in filters (Delta sigma.....) or (the most flexible one) use digital filters...
... but maybe you don´t have ADC and digital processing...

All that should give better SNR+THD, especially with low input signal. The MAX7400 isn´t cheap either..

***
Did you ever think of using a dedicated signal processor like the ADAU1700? Just out of curiosity i bought an eval board and installed the software... amazing simple..
There are movies on youtube...


Klaus

 

[Jester]

Hi,



true, but you may hear every any deviation from that 300kHz. like change in frequency, change in duty cycle and so on.

from datasheet:

What is your clock frequency and where does it come from? The MAX7400 has an internal oscillator, just add a capacitor at pin 8

Just to post filter the 300kHz you could use a simple 1k / 33nF RC low pass filter.

The signal should look better on scope, but i don´t expect audible changes.

Klaus

- - - Updated - - -

Hi,

some adds:

* did you read datasheet about pre- and post-SCF filtering?

May i ask why you need that SCF filter.
I personally don´t like them and instead try to use analog filters, ADCs with built in filters (Delta sigma.....) or (the most flexible one) use digital filters...
... but maybe you don´t have ADC and digital processing... Correct, in this application no ADC, straight to the amplifier, I inherited the design, I'm thinking an analog filter would have been a better approach.

All that should give better SNR+THD, especially with low input signal. The MAX7400 isn´t cheap either..

***
Did you ever think of using a dedicated signal processor like the ADAU1700? Just out of curiosity i bought an eval board and installed the software... amazing simple..
There are movies on youtube... Link?, I did a quick Google search and did not find much


Klaus

See response in BOLD above.

 

[D.A.(Tony)Stewart]

SCF has a high input THD threshold so input signal must be boosted to >1V with prefilter and high SNR to work effectively. It cannot used with mV level signals.

Rather it is intended as a telephone SLIC filter for phone BW and line levels to prevent alias distortion at >=3.5khz with 8kHz sample rate for 64kbps u or A law 8 bit ADC with high dynamic range.

Output switch rate is easily filtered with 12db/octave 2nd order filter over 7 octaves for max SNR.

it wont work well for low level signals.

if this is for piezo Rx, you may consider better low noise preamps and active filters that match requirements for rejection, interpolation and decimation such as integrating sigma-delta ADC

 

[mar0791]

Instead of IC MAX7400 , maybe use RC active filter , or several RC active filter in various combinations . I am thinking about RC filter low-pass , high-pass filter , or other . If we use low-noise op-amp , they will be really selective










I am learning English . If I am writing bad ,please correct me ..
Man is learning all time, and stupid die.

 

[KlausST]

Hi,

Then correct SigmaDSP name is ADAU1701, the software is SigmaStudio

https://m.youtube.com/watch?v=PPmkCVz5LAs

Then there will be automatically related videos

Klaus

 

[Audioguru]

Switched capacitor filters are used to convert a square wave or stepped-wave into a low distortion sinewave. I made some that have distortion and noise at only 0.002%.

 

[Jester]

Switched capacitor filters are used to convert a square wave or stepped-wave into a low distortion sinewave. I made some that have distortion and noise at only 0.002%.

What are your thoughts for use of SCF in an non-sampled audio application (music not voice), where the audio will have low levels (for example the pause between songs)

 

[Audioguru]

What are your thoughts for use of SCF in an non-sampled audio application (music not voice), where the audio will have low levels (for example the pause between songs)

A SCF works perfectly for wideband high fidelity audio and its very low level noise cannot be heard when the music is muted. But your application is narrowband (no bass and no treble) which is AM radio or telephone quality, so noise and distortion do not matter.

 

[Jester]

A SCF works perfectly for wideband high fidelity audio and its very low level noise cannot be heard when the music is muted. But your application is narrowband (no bass and no treble) which is AM radio or telephone quality, so noise and distortion do not matter.

Well it may be narrowband, however I can certainly hear background hiss between songs or in low volume portions of the song.

 

[Audioguru]

What is playing the songs? A noisy tape player? In your first post you said, "the background noise (when I listen with an audio amplifier) seems about the same at the input and output of the filter" so the signal source is producing the wideband noise and this lowpass filter has "little effect" as you said.

Why do you play songs with no bass and no treble?
I do not know why your power supply has noise. Add a 100uF capacitor parallel to the supply then try using a battery to power the filter circuit and nothing else.