1
0
mirror of git://projects.qi-hardware.com/ben-wpan.git synced 2024-11-05 04:38:26 +02:00

More RF research for the ECNs (balun and layout)

- atrf/ecn/ecn0001.txt, atrf/ecn/ecn0002.txt: changed ECN format to
  separate title from body by two blank lines and, where a conclusions has
  been reached, put it into a section separated from body also by two blank
  lines
- atrf/ecn/ecn0003.txt: researched balun design and availability
- atrf/ecn/INDEX, atrf/ecn/ecn0004.txt: new ECN: Take into account layout
  considerations for RF
- TODO: removed item covered by ECN0002
- TODO: removed bug fixed in f32xbase commit
  79396b17772639cea615d1c87870d55f08c11850
- TODO: added reference to ECN0003
This commit is contained in:
Werner Almesberger 2010-08-26 15:41:14 -03:00
parent fe8bdac37f
commit 2406f674f5
6 changed files with 137 additions and 14 deletions

11
TODO
View File

@ -37,11 +37,6 @@ inside a Ben.
Things not done yet
-------------------
- define values for crystal load capacitors. Measurements with instruments not
quite precise enough (+/- 100 ppm) for the task (+/- 40 ppm, preferably
< +/- 10 ppm) suggest that we're about 300 ppm off with no capacitors at all
and can only correct about 150 ppm with the internal trim capacitors.
- examine spectrum around carrier frequency and first harmonic to look for
obvious distortions. Vary transmit power.
@ -60,7 +55,7 @@ Things not done yet
- verify that the Ben can output an a) 16 MHz clock, and b) with +/- 40 ppm
- replace discrete balun and filter with integrated solution, to reduce BOM
size, maybe cost, insertion loss, and PCB space
size, maybe cost, insertion loss, and PCB space (see ECN0003)
- check if we really need three DC blocking caps in the RF path
@ -80,10 +75,6 @@ Things not done yet
Bugs to fix
-----------
- two of my systems (tv and u1010) flat out refuse to talk to the board's USB
application, but have no problem talking to its DFU boot loader. Very
strange.
- atrf vs. atspi naming is a bit confusing

View File

@ -3,3 +3,4 @@ Number Status Description
0001 Edit Adjust balun component values
0002 Done Add load capacitors to 16 MHz crystal
0003 Edit Replace balun and filter with integrated balun
0004 Edit Take into account layout considerations for RF

View File

@ -1,5 +1,6 @@
Adjust balun component values
Some of the components specified in the schematics were not at hand
in my lab and were thus replaced with similar parts. Furthermore, the
LED color was changed to ease visual identification:

View File

@ -1,5 +1,6 @@
Add load capacitors to 16 MHz crystal (C14, C15)
The crystal has a specified load capacitance of 8 pF. The AT86RF230 has
an internal capacitor array that can be trimmed in 16 steps from 0 pF to
4.5 pF.
@ -41,6 +42,7 @@ Board Error Meas. accuracy C14/C15 Trim
CNTR measurements are repeatable within 1 ppm, so the nominal accuracy
appears to be far too pessimistic.)
These measurements suggest that, combined with parasitic capacitance,
load capacitors of 12 pF make the crystal roughly half the trim range
faster than 16 MHz.
Conclusion: these measurements suggest that, combined with parasitic
capacitance, load capacitors of 12 pF make the crystal roughly half the
trim range faster than 16 MHz.

View File

@ -1,3 +1,58 @@
Replace balun and filter with integrated balun
(pick suitable balun)
We consider the balun needs for the AT86RF230 and the TI/Chipcon
CC2520 we may try as a design alternative.
For the AT86RF230, Atmel recommend baluns with integrated filter,
namely the Wuerth 748421245 and the Johanson 2450FB15L0001, both in the
AR86RF230 data sheet.
For the CC2520, TI recommend a microstrip-based design for the balun,
both in the CC2420/CC243x/CC2480 application note [1] and the reference
design [2], without fully characterizing neither the transceiver's RF
output nor all the components in the balun.
Johanson lists the 2450FB15L0001 [3] and even a 2450BM15B0002 [3] for
the TI/Chipcon CC2520, but not even Octoparts is able to find a
distributor for these parts.
Digging deeper, TI reveal more information about the balun in [5].
Finally, TI somewhat hesitatingly acknowledge that the Murata balun
LDB182G4510C-110 can be used for the CC2520 [6], with a 3.9 nH inductor
connecting RF_P and RF_N, the balun coupled to GND not directly but via
10 nF, and finally an LC low-pass filter with 1.5 nH and 2.2 pF at the
output, for EMI compliance.
Note that this also means that CC2520 and AT86RF230 both have an
impedance of 100 Ohm.
A balun without filter similar to the Murata part would be the Johanson
2450BL15K100.
Manufacturer Part number Package Digi-Key Price/Qty
--------------- ----------------------- ------- --------------- ---------------
Wuerth 748421245 0805-6 732-2230-1-ND 2.15/25
732-2230-2-ND 0.753/4000
Johanson 2450FB15L0001 0805-6 - -
Johanson 2450BM15B0002 0805-6 - -
Johanson 2450BL15K100 0805-6 712-1045-1-ND 0.488/10
712-1045-2-ND 0.225/4000
Murata LDB182G4510C-110 0603-6 490-5023-1-ND 0.325/10
490-5023-2-ND 0.114/4000
[1] http://www.ti.com/litv/pdf/swra098d
[2] http://focus.ti.com/docs/toolsw/folders/print/cc2520em_refdes.html
[3] http://www.johansontechnology.com/images/stories/ip/baluns/Balun_Filter_Combo_Matched_2450FB15L0001_v11.pdf
[4] http://www.johansontechnology.com/images/stories/ip/baluns/balun_filter_combo_matched_2450bm15b0002_v2.pdf
[5] http://www.ti.com/litv/pdf/swra236a
[6] http://e2e.ti.com/support/low_power_rf/f/155/t/15910.aspx
Conclusion: the Wurth balun appears to be the safest choice for
prototyping the AT86RF230. Due to its high cost, a circuit with a
discrete filter may be considered for larger quantities.
For the CC2520, it's probably safest to directly try the Muarta balun
with the recommended discrete filtering circuit.

73
atrf/ecn/ecn0004.txt Normal file
View File

@ -0,0 +1,73 @@
Take into account layout considerations for RF
There are a number of layout considerationg when designing RF systems
that were not taken into account or not quantified when making the
first design.
- transmission line width
The microstrip [1] transmission line connecting the balun and filter
circuit with the antenna must be impedance-matched with the antenna.
The rule of thumb according to [2] is to make its width twice the
board thickness, in this case 0.8 mm or 31.5 mil.
The microstrip calculator at [3] also takes into account the
thickness of the copper, 1 oz, and yields a slightly narrower 57.5
mil or 1.46 mm.
A more elaborate calculator can be found at [4].
- via spacing
Section 4.2 of [5] recommends a via spacing of no more than
Lvia = C/sqrt(Er)/Fres
where
C = the speed of light, 3*10^8 m/s
Er = the board's dielectric constant, 4.5 for FR-4
Fres = the resonance frequency, at least 24.5 GHz
We thus obtain Lvia = 5 mm.
- component placing
[5] places DC blocking, balun, and filter close to the transceiver,
with only the feed line between the RF circuit and the antenna. Thus,
no changes are needed.
- feed line termination
Point 12 of [6] warns us that we may need to terminate the
transmission line if it is longer than 20% of the signal's rise time.
Point 1 of [6] gives the rise time as 1/(10*Fclk), which looks as if
it's meant for digital signals. But we'll use it anyway.
[2] gives us the typical propagation delay for a microstrip as
150 pS/in.
This means that Lmax = 0.2*tr*v
with
tr = 1/24.5 GHz
v = 1 in/150 pS
We thus obtain Lmax = 1.4 mm
[2] suggests that the maximum unterminated stub is L(in) = tr(nS).
With tr = 1/(10*Fclk), we thus obtain Lmax = 1.04 mm.
Not sure if all this even applies to antennas. This needs looking to by
someone who understands about RF.
[1] http://en.wikipedia.org/wiki/Microstrip
[2] http://www.hottconsultants.com/techtips/rulesofthumb.html
[3] http://www.cepdinc.com/calculators/microstrip.htm
[4] http://mcalc.sourceforge.net/
[5] http://www.ti.com/litv/pdf/swra236a
[6] http://www.pcbmotif.com/home/index.php?option=com_content&view=article&id=104&Itemid=137
Conclusion: the antenna feed line needs to be revised. The via spacing
of the RF area needs to be examined. The recommended spacing may be
beyond the capabilities of a DIY process, though.