[gta02-core] Audio ECN update

[Rene Harder]

Werner Almesberger wrote:
> Rene Harder wrote:
>   
>> Do you mean the EMI protection of the headset jack!  Sure using an
>> integrated filter works with me, I just thought we do not want too many
>> new components.
>>     
>
> Yes. Ideally, I would use a "general-purpose" EMI filter, such as the
> Murata NFL18ST series.
> http://search.murata.co.jp/Ceramy/image/img/PDF/ENG/L0112S0119NFL18ST.pdf
>   

This one has a really good performance over a large frequency range.
But as you can see in the data sheet it's a T filter which is best
suited for low source and load impedances like power lines and doesn't
work well for data lines, for audio this might work though.

> They're 0603, and the data sheet promises truly magic characteristics,
> including zero insertion loss at DC. The downside is that it seems only
> Murata makes such things.
>
> Also the Rohm MCF18 series, particularly the MCF182CN102 may be suitable
> (and is a lot cheaper):
> http://www.rohm.com/products/databook/emi/pdf/mcf18.pdf
>
> Fun things I found along the way: a one-stop shopping EMI and ESD
> solution for USB:
> http://www.st.com/stonline/products/literature/ds/10950/emif02-usb01f2.pdf
>
> The same for SDIO:
> http://www.st.com/stonline/books/pdf/docs/13195.pdf
>   

Here you can exactly see what in meant in my previous email. Those
integrated filters are mostly optimized for one specific purpose and
they do work very well there but. However general purpose filter are
rarely to find and even then you might have to cut back on something.

>   
>> However, I think discrete filter are more flexible regarding vendor or
>> parameter changes.
>>     
>
> What I'm not sure about is what parameters we really need. Most of
> the signals we want to filter are slow: < 20 kHz for audio, < 1 MHz
> for UART, < 4 MHz for SIM, <= 25 MHz for SDIO. The big interferences
> are all in the GHz range. So a filter with a cut-off frequency around
> 100-300 MHz ought to be okay, right ?
>   

If those filter have an "ideal" low-pass characteristic this cut-off
frequency is fine for most cases, but in reality they are more like a
notch or bandstop filter where the attenuation is decreasing with higher
frequencies through internal resonances.
I think we need to block those frequencies where the FR is exposed to
high power RF, (self generated or external). The highest field density
we'll probably get from our own RF which is generated by BT, WLAN, GSM,
internal clocks etc., so i guess everything between a couple hundred MHz
to at least 3GHz.