Goertzel example: linear position sensor
ManAtWork
Posts: 2,080
I'm experimenting with the Goertzel mode to build a capacitive position sensor. 
I've taken Chip's example and modified it a bit.
' Goertzel input and display
' original source from Chip's "Goertzel_OUT_and_IN.spin2"
' modified by N. Benezan for capacitive position sensor or LVDT
CON
_xtlfreq = 25_000_000
_clkfreq = 180_000_000
VAR dacpins '[0] dac pins to receive output cos/sin - read once at start
adcpin '[1] adc pin to sense return signal - read once at start
adcmag '[2] adc magnification 0..4 = 1x, 3x, 10x, 31x, 100x - read once at start
frequency '[3] output frequency - read every loop
cycles '[4] number of cycles per loop - read every loop
phase '[5] 9-bit phase offset - read every loop
power '[6] measured input power - written every loop
angle '[7] 32-bit measured angle - written every loop
done '[8] done flag = 1 - written every loop
buffer[512]
PUB go() | i
dacpins := 32 addpins 3
adcpin := 38
adcmag := 2 ' gain 10x
frequency := 1000_000 ' 1MHz
cycles := 100
phase := 0
coginit(newcog, @goertzel, @dacpins)
repeat until done ' wait for cog startup
debug (UHEX_LONG_ARRAY(@buffer, 10)) ' display mirror of LUT in hub
'repeat
debug(`scope_xy s 'Goertzel' samples 10 dotsize 20 size 300 range 100_000_000 polar logscale)
repeat
waitms(10)
debug(`s `(power, angle))
' Setup
DAT org
goertzel
' Make sine and cosine tables in LUT bytes 3,2,1,0
mov ptrb, ptra ' buffer for debugging
add ptrb,#9*4
mov .z,#$1FF 'make 512-sample sin/cos table in LUT
.sincos shl .z,#32-9 'get angle into top 9 bits of z
qrotate #127,.z 'rotate (127,0) by z
shr .z,#32-9 'restore z
getqx .x 'get x
getqy .y 'get y
setbyte .x,.y,#1 'sin:cos into x
setword .x,.x,#1 'sin:cos:-sin:-cos into x
xor .x,##$FFFF
wrlut .x,.z 'write into LUT
wrlong .x,ptrb++
djnf .z,#.sincos 'loop until 512 samples
' configure dac and adc pins
setq #3-1 'read pin data
rdlong .dacpins,ptra++
mov .x,.adcpin 'get dds d operand
and .x,#%111100
shl .x,#19-2
or .dds_d,.x
mov .x,.adcpin 'get dds s operand
and .x,#%000011
decod .x,.x
shl .x,#12
or .dds_s,.x
fle .adcmag,#4 'enable adc pin
shl .adcmag,#15
add .adcmode,.adcmag
wrpin .adcmode,.adcpin
cogid .x 'enable dac pins
setnib .dacmode,.x,#2
wrpin .dacmode,.dacpins
drvl .dacpins
mov .done,#1 'set done flag
' Input Goertzel settings, start measurement, output prior results, loop
.loop setq #3-1 'read input data
rdlong .frequency,ptra++
rdlong .x,#@clkfreq 'compute NCO frequency
shl .x,#1
qfrac .frequency,.x
getqx .y
setword .dds_d,.cycles,#0 'set cycles
sets .dds_s,.phase 'set phase
setq .y 'issue new Goertzel command with new frequency
xcont .dds_d,.dds_s
getxacc .x 'get prior Goertzel reading, cos first
mov .y,0-0 '..then sin
qvector .x,.y 'convert (x,y) to (power,angle)
getqx .power 'get power
getqy .angle 'get angle
setq #3-1 'write output data
wrlong .power,ptra++
sub ptra,#6*4 'point ptra back to .frequency
jmp #.loop 'loop
'Data
.adcmode long %0000_0000_000_100011_0000000_00_00000_0 'ADC mode
.dacmode long %0000_0000_000_10111_00000000_01_00000_0 'DAC mode
.dds_d long %1111_1000_0000_0111_0000000000000000 'Goertzel mode
.dds_s long %000_000000000 ' 512 entry LUT
.dacpins res 1 'read once at start
.adcpin res 1
.adcmag res 1
.frequency res 1 'read every loop
.cycles res 1
.phase res 1
.power res 1 'output every loop
.angle res 1
.done res 1
.x res 1 'workspace
.y res 1
.z res 1
What I would have expected is to see four different waveforms with 90° phase shift at pins 32..35 (-cos, -sin, cos, sin). What I get instead is two pairs of inverted signals (-sin, +sin, -sin +sin).
The debug output shows that the LUT contents are correct.
@buffer = $FE7F_0180, $FD7F_0280, $FB7F_0480...
From the docs I read that ADC inputs can be inverted with the bits S[19..16] but DAC output cannot. What's wrong?
Comments
DDS/Goertzel section says S[19:12] for ADC polarities.
DAC polarities are set with %dddd bits in D operand. You've got them set to %1000 which is !X0 X0 !X0 X0. %1110 is probably what you want instead.
Pit: there was a bug in the engineering samples, iirc
RevA had something wrong with Goertzel hardware. A signed multiply bug or something like that. It got fixed in RevB.
EDIT: Hmm, not in the errata though. I might be imagining things:
Doh!
Yes, that's right. I knew that I've read somewhere that the DAC pins can also be configured as differential pairs. But it's not in the Goertzel chapter but in the general Streamer section.
Thanks a lot!
Ok, now it works as expected. Here is the corrected code:
' Goertzel input and display ' original source from Chip's "Goertzel_OUT_and_IN.spin2" ' modified by N. Benezan for capacitive position sensor or LVDT CON _xtlfreq = 25_000_000 _clkfreq = 180_000_000 VAR dacpins '[0] dac pins to receive output cos/sin - read once at start adcpin '[1] adc pin to sense return signal - read once at start adcmag '[2] adc magnification 0..4 = 1x, 3x, 10x, 31x, 100x - read once at start frequency '[3] output frequency - read every loop cycles '[4] number of cycles per loop - read every loop phase '[5] 9-bit phase offset - read every loop power '[6] measured input power - written every loop angle '[7] 32-bit measured angle - written every loop done '[8] done flag = 1 - written every loop buffer[512] PUB go() | i dacpins := 32 addpins 2 adcpin := 38 adcmag := 2 ' gain 10x frequency := 1000_000 ' 1MHz cycles := 100 phase := 0 coginit(newcog, @goertzel, @dacpins) debug(`scope_xy s 'Goertzel' samples 10 dotsize 20 size 300 range 1500_000 polar) repeat waitms(10) debug(`s `(power, angle)) ' Setup DAT org goertzel ' Make sine and cosine tables in LUT bytes 3,2,1,0 mov .z,#$1FF 'make 512-sample sin/cos table in LUT .sincos shl .z,#32-9 'get angle into top 9 bits of z qrotate #127,.z 'rotate (127,0) by z shr .z,#32-9 'restore z getqx .x 'get x getqy .y 'get y setbyte .x,.y,#1 'sin:cos into x setword .x,.x,#1 'sin:cos:-sin:-cos into x xor .x,##$FFFF wrlut .x,.z 'write into LUT djnf .z,#.sincos 'loop until 512 samples ' configure dac and adc pins setq #3-1 'read pin data rdlong .dacpins,ptra++ mov .x,.adcpin 'get dds d operand and .x,#%111100 shl .x,#19-2 or .dds_d,.x mov .x,.adcpin 'get dds s operand and .x,#%000011 decod .x,.x shl .x,#12 or .dds_s,.x fle .adcmag,#4 'enable adc pin shl .adcmag,#15 add .adcmode,.adcmag wrpin .adcmode,.adcpin cogid .x 'enable dac pins setnib .dacmode,.x,#2 wrpin .dacmode,.dacpins drvl .dacpins mov .done,#1 'set done flag ' Input Goertzel settings, start measurement, output prior results, loop .loop setq #3-1 'read input data rdlong .frequency,ptra++ rdlong .x,#@clkfreq 'compute NCO frequency shl .x,#1 qfrac .frequency,.x getqx .y setword .dds_d,.cycles,#0 'set cycles sets .dds_s,.phase 'set phase setq .y 'issue new Goertzel command with new frequency xcont .dds_d,.dds_s getxacc .x 'get prior Goertzel reading, cos first mov .y,0-0 '..then sin qvector .x,.y 'convert (x,y) to (power,angle) getqx .power 'get power getqy .angle 'get angle setq #3-1 'write output data wrlong .power,ptra++ sub ptra,#6*4 'point ptra back to .frequency jmp #.loop 'loop 'Data .adcmode long %0000_0000_000_100011_0000000_00_00000_0 'ADC mode .dacmode long %0000_0000_000_10111_00000000_01_00000_0 'DAC mode .dds_d long %1111_1111_0000_0111_0000000000000000 'Goertzel mode .dds_s long %000_000000000 ' 512 entry LUT .dacpins res 1 'read once at start .adcpin res 1 .adcmag res 1 .frequency res 1 'read every loop .cycles res 1 .phase res 1 .power res 1 'output every loop .angle res 1 .done res 1 .x res 1 'workspace .y res 1 .z res 1Electrodes should be connected to the pins like this:
A = 32 (dac0 = -sin out)
B = 38 (adc in)
C = 34 (dac2 = +sin out)
This results in the highest overall gain, SNR and linearity. The spot in the XY scope moves on a line through the center with around 30° angle. The phase shift comes from crosstalk between the PCB traces and RC delay, I think.
There's another option:
A = 32 (dac0 = -sin out)
B = 38 (adc in)
C = 33 (dac2 = -cos out)
Now the spot moves on a quarter circle in the first quadrant. I thought this could be used to suppress influences of gain drift (variations in capacitor plate distance, wear, play, temperature or whatever) by ignoring power alltogether and only measure phase angle. Unfortunatelly, linearity is not as good as I expected.
Hi,
when I was doing some experiments with capacitive soil moisture sensors, I learned, that the normal distance of those capacitor plates have very high impact. (I should have known) So in such setup a very precise bearing concept might be needed to keep the normal distance constant.
The coil inductance method does not need this precision there as long as the core is somewhere inside the winding diameter.
Some years ago, I made a sensor with a single cylinder coil and a moving core (varying L) together with a ne555 Multivibrator circuit. The mc only had to measure frequency.