anelok: revisiting Bluetooth, chips beginning with A

Werner Almesberger werner at almesberger.net
Fri Nov 22 21:37:34 EST 2013


I already mentioned Dmitry Grinberg's creative abuse of chips that
are supposed to be BTLE-incapable. He eventually found a chip that
can handle large enough packets, the Amiccom A7125. Unfortunately,
that one only does FSK but not GFSK, which means that it would
be able to communicate with a BTLE device, but it would have a
very "dirty" spectrum.

Fortunately, Amiccom have a whole family of chips and the A7105
looks much better. Coming from Taiwan, the data sheet is
"China-open", i.e., it's probably considered to be "confidential",
but you still find it all over the Web:

http://www.avantcom.com.tw/AVANTCOM/TC/DATA/PRODUCT/SOLVE/18_3.pdf

Of course, none of those Western racist distributors have the
chips, but Chinese pragmatism prevails here as well:

http://www.aliexpress.com/popular/amiccom-a7105.html

Picking the seller with more medals it would cost USD 1.49 apiece
(at 10 units), it comes in a 20-QFN 4x4 mm package, the frequency
deviation is an okay-ish +/- 186 kHz, and the other parameters
look sane as well.

It also supports clock output so we could use it to provide the
MCU with a crystal-based clock, like we currently do with the
AT86RF232. This means that the MCU doesn't need a crystal,
reducing BOM cost and making the PCB less crowded. The downside is
that the RF chip has to run (and consume power) when the MCU needs
that clock, but since we need the clock only for USB, power is not
an issue.

There's one more chip that came up in various discussions, and
that's the Analog Devices ADF7242 [1]. It comes in a 32-QFN 5x5 mm
package and is the priciest of the bunch, at USD 3.05 @ 1000. It
passes all the RF requirements with flying colors and even
supports O-QPSK. And yes, the data sheet even states that it can
do 802.15.4, and there should even be a Linux driver for it.

Inside it is a little 8 bit CPU core of unknown type with a ROM
and a program RAM (!) for downloading code snippets for certain
low-level functions (CSMA and such, not whole protocol stacks as
in the CC254x), but it seems that all this is non-open.

It is spoken of favourably on #ubertooth.

[1] http://www.digikey.com/product-detail/en/ADF7242BCPZ/ADF7242BCPZ-ND/2606616

Conclusions: two more candidates, the Amiccom A7125 and the Analog
Devices ADF7242.

Now, if you expected 25 more posts for the rest of the alphabet,
I must disappoint you since these are already all the chips I
found this far.

Next: integration issues and interim summary.

- Werner



More information about the discussion mailing list


interactive