Cape Noise Performance
Problem
Beaglebone offers noisy SYS_5V0 - TODO: add Vpp num
POE is also noisy - TODO: add Vpp num
shepherd shows residues of voltage spikes in target voltage
Solved: confirmed to be due to a “transmitting” usb-hub 50 cm away from test-setup (not powering anything)
Setup
BBGreen with Shepherd-PCB, Output VTarget set to 3V
scope: rigol ds1104z, 100 MHz, calibrated, compensated probes, up-to-date firmware
50 cm radius free of additional metal
Power-Supplies
setup1: 5V to BB via benchsupply with 1mF on input
setup2: as setup 1, but shepherd 5V and 3v3 powered by keithley sourcemeter, 1m 4-Port Cables, not buffered by Cap
setup3: BB powered by NoName POE-Splitter (EC-PD0520USB)
setup4: BB powered by TP-Link POE-Splitter (TL-POE10R), buffered by 1 mF on Input
Measurement
scope input set to AC to analyze ripple, internal 20 MHz low pass active
probe set to 1x for low noise
probe connected with short ground-loop to closest GND
timebases /div (mostly): 50 ms, 10 ms, 1 ms, 100 us, 10 us, 1 us, 100 ns
voltage range: lowest setting without clipping, mostly 1 mV/div
scope-math: Voltage-peak-peak, frequency
Scope-Screenshots (DS1Z_Quickprint#.png) are not linked or included here, only referenced for each measurement (./media_noise/)
Scope Noise with grounded signal-input
Measurements
Probe Signal shorted to ground (noise floor)
Vpp = 1.20 mV
no abnormalities or standing waves
QuickPrint 26-31
SETUP 2
A5V (setup2) - on biggest cap after regulator
Vpp = 2.10 mV
1.3 MHz Ripple (1.72mVpp)
QuickPrint 32-36
10V (setup2) - on biggest cap after regulator
Vpp = 3.40 mV
7 kHz Ripple,
QuickPrint 37-41
VTargetA = 3.3 V (setup2), 1k Load
Vpp = 1.90 mV
no abnormalities or standing waves
QuickPrint 42-46
VTargetA = 3.3 V (setup2), no load
Vpp = 1.72 mV
no abnormalities or standing waves
QuickPrint 47
VTargetB = 3.0 V (setup2), no load
Vpp = 1.65 mV
no abnormalities or standing waves
QuickPrint 48-52
SourceMeter 3V3 (setup2), ~ 4.06 mA Shepherd-Load, 4-Port-Meas
Vpp = 16.6 mV (unfiltered input)
no abnormalities, some switching noise
QuickPrint 53-57
SourceMeter 5V0 (setup2), ~ 73.9 mA Shepherd-Load, 4-Port-Meas
Vpp = 20.4 mV (unfiltered input)
no abnormalities, some switching noise
QuickPrint 63-67
BBone 3v3 (setup2) - on pinheader, shepherd not connected
Vpp = 46 mV
no abnormalities, some switching noise 20 kHz & 333 kHz
QuickPrint 58-62
BBone SYS_5V / P9-7/8 (setup2) - on pinheader, shepherd not connected
Vpp = 145 mV
all kind of fragments from msec to nsec (50 ms & 1us view shows best)
QuickPrint 68-72
BBone VDD_5V / P9-5/6 (setup2) - on pinheader, shepherd not connected
Vpp = 5.6 mV
minor abnormalities, 5Hz wave, 333 kHz switching fragments
QuickPrint 73-79
SETUP 1
VTargetA = 3.3 V (setup1), no load
Vpp = 1.8 mV
no abnormalities or standing waves
QuickPrint 80-87
A5V (setup1) - on biggest cap after regulator
Vpp = 1.9 mV
no abnormalities, some ripple ~1.3 MHz, 1.6 mVpp
QuickPrint 101-106
BBone 3v3 (setup1) - on pinheader
Vpp = 51 mV
switching noise 20 kHz, 250 kHz
QuickPrint 107-113
BBone SYS_5V / P9-7/8 (setup1) - on pinheader
Vpp = 111 mV
switching noise 10 Hz, 50 Hz, 20 kHz, 333 kHz, 2.5 MHz
QuickPrint 114-120
Bench Supply 5V (setup1) - 360 mA Load
Vpp = 37 mV
switching noise 10 Hz, 50 Hz, 2.5 kHz
QuickPrint 121-130
Bench Supply 5V (setup1) - no Load
Vpp = 20 mV
switching noise 4 kHz
QuickPrint 131-137
SETUP 3
BBone 5V Input (setup3), TP-Link Poe
Vpp = 22 mV
switching noise 10 Hz, 50 Hz, 16 kHz, 256 kHz
QuickPrint 201-207
VTargetA = 3.3 V (setup3), no load
Vpp = 2.7 mV
no abnormalities, but some oscillation around ~ 1 MHz
QuickPrint 208-215
A5V (setup1) - on biggest cap after regulator
Vpp = 3.3 mV
no abnormalities, but some oscillation around 1.3 MHz
QuickPrint 216-224
SETUP 4
BBone 5V VCC P9-5/6 (setup4), NoName Poe
Vpp = 7.6 mV
heavy switching noise ~ 75 kHz
QuickPrint 225-232
VTargetA = 3.3 V (setup4), no load
Vpp = 12.2 mV
heavy switching noise ~75 kHz
QuickPrint 233-239
A5V (setup4) - on biggest cap after regulator
Vpp = 11 mV
ramps with 50 kHz and the additional switching noise
QuickPrint 240-246
Setup1 - Tuning InAmp
VTargetA = 3.3V, 1k Load
Vpp = 1.9 mV
some oscillations at 145 kHz - 1 MHz
QuickPrint 301-307
TP2
Vpp = 1.3 mV
not enough detail, but looks like noisefloor
QuickPrint 308
TP5 - InAmp Output
Vpp = 110 mV
some oscillation just shy of 1 MHz
QuickPrint 309-318
10V - over biggest Cap
Vpp = 3.6 mV
some oscillation at 10 kHz
QuickPrint 319-324
-6V - over biggest Cap
Vpp = 3.4 mV
some oscillation at 1.3 MHz
QuickPrint 325-330
MOD: 8.2 OHm right before 1 Ohm Shunt
TP5 - InAmp Output
Vpp = 16.4 mV
no abnormalities
QuickPrint 331-336
Over 1 Ohm Shunt
Vpp = 2.24 mV
no abnormalities
QuickPrint 338
Over 1 Ohm Shunt
Vpp = 2.24 mV
no abnormalities
QuickPrint 338
MOD: 1k on InAmp-Output
TP5 - InAmp Output
Vpp = 12.6 mV
no abnormalities
QuickPrint 339-344
MOD: 100 nF over InAmp Input
Over 1 Ohm Shunt
Vpp = 2.36 mV
oscillations at 380 kHz
QuickPrint 345-351
TP5 - InAmp Output
Vpp = 11.5 mV
no abnormalities
QuickPrint 352-357
Analysis
disclaimer
some noise is very close to the scopes lower threshold (~ 1.2 mVpp)
A5V-LDO should have ~ 18 uV RMS-Noise with 60 dB Ripple Rejection (but show with higher noise)
⇾ measurements can’t be trusted when looking at individual values, but comparisons should be valid
amplifying active probe should be preferred
A5V and VTargetA are very similar ⇾ further analysis will take maximum of these two
setup comparison (worst to best)
setup4 (NoName POE) shows heavy artifacts, Noise VTarget = 12 mVpp
setup3 (TPLink POE) shows some oscillations, Noise A5V = 3.3 mVpp
setup2 (Shepherd powered by sourcemeter) shows minor ripple, noise A5V = 2.1 mVpp
setup1 (benchPowered BB) shows minor ripple, Noise A5V = 1.9 mVpp
Sys_5V-Line is surprisingly noisy, but has limited influence on VTarget
10 V / -6 V has some 7 - 20 kHz Ripple ⇾ most likely bad for InAmp
Conclusion / Mitigations
BB-Power should be avoided
switch to VDD_5V (less noisy) instead of sys-5v ⇾ NOT POSSIBLE
Sys-5V was already used in previous shepherd pcb
avoid 3v3, generate on shepherd
allow manual “switch” to only use external power (move ferrite from sys_5V input to tap into 5v_ext-rail)
InAmp - further improvement in noise
filter input, increase shunt to 10 Ohm, and 100nF parallel to the shunt brings 160 kHz Lowpass
filter output, 100k in line to ADC. ADC Pin has Capacitance of 5pF, Line ~1pF, brings ~300 kHz ⇾ option to solder a cap to Testpoint
POE-Input
should be allowed to be > 5V, and filtered by ldo
TP-Link TL-POE10R V5.0 is rated for 1A @ 9/12V, 2A @ 5V ⇾ ~ 10 W
Side Note
TP-Link POE-Splitter has heat-issues at least when powering ~300@5V, ICs are Ok, but the input Cap (47uF 100V) gets also very warm ⇾ may shorten life-expectancy
there is no cheap alternative for this unit
Stabilize Supplies
Options for Stability / less noise
avoid switching regulators (noise usually >> 1 mVpp)
use LDO, specially low noise or ultra low noise ones (low noise LDOs are somewhere around 1 mVpp)
passive LC-LowPass at the End (
6V-Switching-Regulator had big opt-potential, 6x less Vpp-Switching-Noise
+10/-6V-Regulator for InAmp
LT3487 is claiming low noise due to high frequency
regulator hysteresis should be around 6-7 mV for both rails (FBP has 7 m%/V, FBN has 1 mV/V Line Regulation)
measurements show ~ 3.4 mV after ferrite
previous shepherd pcb
LM27762 has charge pump + ldo for 22 uV_RMS and 43 dB Ripple-Rejection, 1.5 mV/V Line and 34 uV/mA Load Regulation, but only outputs +-5V
use additional very-low-noise LDOs (+10/-6V)
LP2985, 10V LDO, 30 uV_RMS, 45 dB Ripple Rejection
MIC5219 2.5 to 12 V LDO, 300 nV / sqrt(Hz), 75 dB Ripple Rejection, 9 m%/V Line Regulation
ADP7118, 2.7 to 20 V LDO, 11 uV_RMS, 50 dB Ripple Rejection, 15 m%/V Line Regulation
TPS723, -2.7 to -10 V LDO, 60 uV_RMS, 48 dB Ripple Rejection, 40 m%/V Line Regulation
Ricoh RP117x, -5.5 to -1 V, 80 dB Rejection, 16 uV_RMS,
LT3094, ultra low noise, 0 to -20V, 74 dB Rejection, 2.2 nV/sqrt(HZ), 0.8 uV_RMS
InAmp AD8421
Inputs range from 0 to 5 V,
differential Voltage of 0 to 500 mV, Gain 10 ⇾ Power supply should be at least +9 V & -5 V, Ref = GND
differential Voltage of 0 to 50 mV, Gain 100 ⇾ Power supply should be at least +9 V & -5 V, Ref = GND
Diamond Plot: https://tools.analog.com/en/diamond/#difL=0&difR=0.05&difSl=0&gain=100&l=0&pr=AD8421&r=5&sl=0&tab=1&ty=2&vn=-5&vp=9&vr=0
there are two InAmps with a budget of ~ 74 mW, ~ 4.6 mA
Positive Voltage ⇾ Voltage Doubler
LM266x not suited, no low noise
LM276x not suited, no low noise
Negative Voltage ⇾ Inverter, Charge Pump
LM27761, 20 uV_RMS, 35 dB rejection, 1.5 mV/V Line and 4.6 uV/mA Load Regulation,
LTC1550, not suitable
TPS63710
Verification of PI-Filter with Spectrum Analyzer
SA
Rigol DSA 815 (9 kHz to 1.5 GHz) with tracking generator
1 MHz Span, beginning at DC
Filter
15 uH shielded coil with 100 uF / 6.3 V on both sides
corner frequency: 4.1 kHz
TG provided 0 dBm and input was calibrated to it
open-input and filter in line showed no significant difference (no noticeable rise of the noise-floor)