Cape v2.1 - PCB Changes
Implemented Changes
level translators - implement single voltage supply
smooth 6v-boost ⇾ larger capacitors 58uF to 158uF, ferrite, lower boost-voltage 5.38 V instead of 6.766
typical on A5V are 38mA for Quiet/On, 150mA max, mp3221 gives formular for upper thresholds of Vpp
old config: 164 uVpp (38mA) to 645 uVpp (150mA)
new config: 27 uVpp (38mA) to 104 uVpp (150mA) ⇾ effect of ferrite not included
LDO is rated for 60 dB dampening, but measurements show a factor of ~ 500x of ripple rejection
6V Regulator overdrive is edgy, sw-pin is good for +8V, currently there are ~ 7.14 V
Target-IO-PUs - get powered by unmonitored voltage buffer
add 3v3 converter (linear), because BB provides very noisy power (both 5V & 3V)
make sure shepherd works (theoretically) with BB Black (not only Green) ⇾ Boot-CFG is identical
recheck which IC gets which voltage (5V Sys is exceptional noisy)
allow to run completely from external power ⇾ move sys-5v ferrite
main power switch not needed
mark ext input
sensitive lines should get more distance from gnd-polygon
consider ultra low noise ldo for -6 and 10V
No suitable 5 to 10V voltage doubler found (fast switching, low noise)
there are quiet inverters (LM27761) though
see “noise_performance.rst” for more info
adjusted Pads of some footprints - reduce width a bit when solder resist was too narrow on ICs
improve GPIO signal integrity (1k from target can be removed), 100k PU on Sys-Side replace by 10 k
better buffer the io voltage, needs single OPA, but gets rid of one NLAS-Switch
PI-Filter is bad for varying loads, so only use with dac / adc
keep 100 uF constant, but vary inductance to optimize voltage drop ⇾ << 1 mV
external cables, filter and buffer (100nF)
(NOT DONE) if there is enough space ⇾ switch out 47uH Coil of LC-LP to 150uF (larger)
gpio - RC lowpass ⇾ Line-Capacity vs. current-limiting resistor
L1L2 distance 2x 180 um, e_r ~4.2
trace width 250 um
length BB-Side [50; 87], LVL ~ 17, Target-Side [13; 27]
C = e*A/d = 3.4 pF ⇾ max capacitance of 1 gpio-trace on shepherd-pcb
C_lvl = 12.5 pF
R = 1k
fc = 10 MHz (neglecting target and BB capacitances)
re-evaluate spi-lines ⇾ fine
via-fence near the lan-port
QR-Code is readable, but still a bit messed up ⇾ negative
target-connector-redefinition, bring GND to first and last position (EMV)
(target) reset (P0.18) with Resistor bridge to gpio-header
(target) remove more metal around antenna
compare LSF-Versions of TI with nexperia ones ⇾ seems to be exact copy
would 1000 Hz on-off-pattern be possible? YES
(if there is time) - implement fixed recorder design
Additional Parts
lvl trans: 2x 240k, 2x 1M, 1x NLAS4684
emu vSwing: 10R 0.1%, 1.1k 1%
emu vDrop: NLAS4684, 100nF?
A3v3: lp2989-3.3, 10nF X7R,
VSenseStabilize: 1k
6V Stabilize: 576k, 100uF, Ferrite
InAmp Stabilize: 100nF, 100k
DNP: Ferrite 5V_SYS (for pwr-rerouting)
16VStability: 4x 33uH, 4x 10uF
20x 10k, Opa388, 100nF, 1uF
removed: 1x NLAS, 2x Ferrite
removed: 20x 100k, 10x 1k
new: opa189 for recorder
removed / rec: mosfet, 1M, 10k
Power-Budget
see separate spreadsheet (PowerConsumption.ods)
BB takes <= 2W
Shepherd MAX ~ 1 W (4mW @ 3V, 743mW @ 5V, 36mW @ 6V, 74mW @ 16V) (with 2x 50mA Target)
Shepherd ON ~ 340 mW (4mW @ 3V, 190mW @ 5V, 36mW @ 6V, 74mW @ 16V)
ON Quiet Current matches with reality (66mA measured, 68mA calculated)
Open Issues
additional PCB for POE
poe is noisy (up to 25 mVpp for TL-POE10R, 300 mVpp for NoNameThing)
switching regulator for 12/9 to 6V
linear regulator with proper noise resistance for 6 to 5V
OR just 2-Stage LC-LowPass (15uA / 770mA from previous Order), additional 100 uF
fix layerwindows
optimize position of current limiting resistors
power-supply-pins? upgrade path to stacked pcb, because current space is already maxed out
check / measure real reverse current of diode
evaluate higher driving strength for Target-Supply TODO:
new coil is now 963-NR3015T3R3M