F103: quartz selection for mp3 application

[kan35]

Posted on April 24, 2010 at 13:58

F103: quartz selection for mp3 application

#stm32 #stm32-audio-usb

[kan35]

Posted on May 17, 2011 at 13:48

TLV320AIC23B is a perfect product, exept price ~

5

USD vs less then 1USD for UDA1330ATS:)

even if I will do switching betwin 12.288 and 11.2896 how can I get 48MHz for USB?...

[Tesla DeLorean]

Posted on May 17, 2011 at 13:48

Not sure you can do it all with one clock source (absent a good PLL/NCO), codecs like the TI AIC23 had a special USB mode which could use a 12 MHz clock source. While the rates weren't perfect, they might be close enough. I think I used a source capable of switching between 12.288 MHz  and 11.2896 MHz.

http://focus.ti.com/lit/ds/symlink/tlv320aic23b.pdf

8000

11025 16000 22050 32000 44100 48000

[kan35]

Posted on May 17, 2011 at 13:48

STM32F105 family does not have memory interface. I use NAND.

Another quastion: the accuracy of HSI is +-1%. Is it anougth for USB?

In my board I also have RTC quartz and can trim HSI if nesessary, should I do it?

[Nickname12657_O]

Posted on May 17, 2011 at 13:48

Hi,

To run USB on our STM32F103 devices , exact 48MHz is needed for USBCLK which is drived from SYSCLK after PLL. Thus SYSCLK should be either 72MHz or 48MHz. I would recommand you to use STM32F105 device from our connectivity line offer which is the best fit for your application : Audio + USB . A special Audio PLL is implemented and allows you a wide choice for Audio sampling rates, you can refer to our on-line presentation for more details :

http://www.st.com/stonline/domains/support/epresentations/stm32connectivity/stm32connectivity.htm

Cheers, STOne-32.

[Nickname12657_O]

Posted on May 17, 2011 at 13:48

Hi,

Regarding HSI, it is not enough to cover all cases, USB Clock jitter at 48MHz should not exceed  absolute ''0.25%'', But it can work in your application for enumeration. However If you try to perform electrical testing , it may violate the USB specs. Cheers,      STOne-32.

[kan35]

Posted on May 17, 2011 at 13:48

As far as I see trimming may be used. The step of trim is 40kHz. 0.25% means window 120kHz. So I think it is possible, I'll make experiment.

thanks

[Tesla DeLorean]

Posted on May 17, 2011 at 13:48

You'd probably have to actively discipline it rather than just trim an forget. It is likely to change with voltage supply and temperature (not stable). The 40KHz LSI is certainly all over the place compared to an 8MHz TCXO

Given how easy a digital NCO (DDS) could have been implemented, It's kind of a pity that the STM32 doesn't have a handful.