ADAU1761 Adaptive filter

Hi John:

Thank you for your post, we'll be happy to make the "FIR default to zero" modifications for your application. That's not a big deal but can you elaborate on" "You cannot use this block for noise cancellation as that requires an additional FIR filter in the error update path which is internal to the block which is a pity. " perhaps by showing us a block diagram.

Regards

Miguel

Hi Miguel

Thank you for your prompt reply. The "coefficient start at 0" NLMS will really help me - thanks.

I should explain my remarks about noise cancellation - I meant active active noise control - creating a zone of quiet at a point. For echo cancellation where the filter acts directly on the signal to be removed your NLMS block as it stands is fine.

For the filtered-X noise control algorithm used in a good deal of noise control algorithms, an extra FIR filter is required to filter the input signal before it is used to update the coefficients. I will knock something together in power point tomorrow to illustrate.

Regards

John

Thanks for adding this post, as it has been very helpful to understand what the NLMS block is doing in lieu of docs from ADI.

Has there been any progress on this topic in terms of further examples etc posted by ADI that I'm not aware of? I have found a few things on the forum, but there doesn't appear to be much new posted on this topic.

I have a new project in which I need to have echo cancellation, so I'm playing with the ADAU1761 evaluation board. It works great at removing white noise, but when I feed the input with a streaming audio source (radio station coming from my PC) it doesn't seem to have much of an effect. My 'real' audio source that goes to pin0 is a pure sinewave source.

I see from a July 9th 2012 post by Brett (entitled 'Echo Cancellation') that the use of a IIR filter to roll off more of the low end. Since I don't have a means of generating the necessary coefficients for that type, I've simply added a cascaded high pass filter. Sof far, I'm still having minimal luck.

Any suggestions as to what else I'm doing wrong?

If the circuit is working OK with white noise (I assume you mean the circuit in my first post) then have you tried feeding a white noise into the analogue input of this circuit? In theory, white noise is white noise whether it is calculated or digitised from an analogue input. This should tell you whether it is the content you are trying to cancel that is wrong or there is a fault in your program (unlikely). If you can cancel an analogue noise input, then you are in business. It then is a case of changing the coefficients of the filter to optimise the filter for the inputs you are using. I suspect for a radio station you will need to increase the value of Alpha (unless you are listening to heavy rock and then we are back to white noise). Increasing it too high will cause it to go unstable and it will be obvious if it does.

Good Luck

John

Hi John, I appreciate the reply. I should correct my statement from the 10th. It works great on eliminating a pure sinewave, but is really in-effective on white noise or 'real' audio. As you know, the upper mixer input is simulating the microphone audio I'm trying to keep. I took the analog input and fed it with PC radio source in place of white noise (using both music as well as talk radio). I've played around with the values (various combinations of Alpha and TC) for several days and had absolutely no luck in eliminating the radio source at all and minimal luck with the white noise.

My apologies, but I'm very much a novice when it comes to acoustic noise cancellation, so I'm still trying to get up to speed. After your original project post, where you able to get something working?

Is there another part within the ADAU series of parts that is more appropriate than the 1761?

I'm after whatever feedback I can get on this one before I completely abandon the 1761..

Hi Byron

I think I can see where you might have made a mistake. The Sigma Studio adaptive filter has two inputs. The top one needs to be fed with a delayed version of your noise to be removed AND the signal you want to keep. The second input is fed with an un-delayed version of the noise. This configuration will let the filter converge.

I suggest you start with the circuit in my first post and confirm that the noise gets removed and the tone is left. Then, leaving everything else as it is, replace the W noise generator with the analogue input with the radio signal and see what happens. The important thing is to ensure that the mixer remains.

Good luck and let me know how you get on.

John

Hi John,

I'm also trying to wrap my brain around this. Do you think it's at all possible to use the NLMS block for feedback cancellation in a PA-system type environment? (live amplification which feeds back into the microphone and then generates howling)

Best regards, Roman

Hi Roman

Sorry this is a bit late! I don't think an adaptive filter is the correct architecture for acoustic howl reduction. Adaptive filters are good when you know the noise you want to remove (such as an echo or engine noise). I think the FFT based howl reducers are best when you want to eliminate a peak in the frequency response like a howl.

Regards

John

This question has been assumed as answered either offline via email or with a multi-part answer. This question has now been closed out. If you have an inquiry related to this topic please post a new question in the applicable product forum.

Thank you,
EZ Admin