Vesper VA1210 User manual

VA1210 Validation & Self-Test Guide
Rev 7

Vesper -Confidential
2
List of Contents
Page
Contents
3
Overview
4
Hardware Check
7
Wearing the TWS
8-10
Signal Quality Check
11
-
12
Setting up Hardware VAD (Voice Activity Detect)

Vesper -Confidential
3
Overview:
Validation & Self-Test
This guide will help to validate the performance of
your TWS prototype.
•
Part 1: Hardware Check.
Using BIST to check
integrity of VA1210 MEMS & wire connections
•
Part 2: Wearing the TWS
. Wear the TWS
properly to ensure the best signal.
•
Part 3: Signal Quality Check.
Measuring SNR
and Noise Floor.
•
Part 4: Setting Up Hardware VAD
. Using the
VA1210 as an ultra-low-power Hardware VAD. Validated TWS prototype
Hardware Check
Wearing the TWS
Signal Quality Check
Setting Up Hardware VAD

Vesper -Confidential
4
Hardware Check:
Overview of Built-In Self Test (BIST)
•BIST is a self-test feature which lets you
check the health of the VA1210 and its
analog connections (e.g. to the ADC).
•BIST allows you to test the VA1210 in
mass production, without using expensive
test equipment such as a shaker.
•BIST involves two tests. In the next slide,
we will explain how to perform these tests.
reserved reserved reserved reserved reserved reserved BIST_POL BIST_EN
I2C Register 0x6:
time
BIST Test Sequence
1ms

Vesper -Confidential
5
Test 1: Measure
OUT=OUTP-OUTM
Test 0: Measure
OUT=OUTP-OUTM
Hardware Check:
How to perform Built-In Self Test (BIST)
To perform BIST Test 0:
•set
MODE
pin low
•write 0x1 to Register 0x6 (BIST_POL=0, BIST_EN=1)
•Measure size of square wave at output (OUT=OUTP-OUTM)
To perform BIST Test 1:
•set
MODE
pin low
•write 0x3 to Register 0x6 (BIST_POL=1, BIST_EN=1)
•Measure size of square wave at output (OUT=OUTP-OUTM)
To exit BIST mode:
•write 0x0 to Register 0x6 (BIST_POL=0, BIST_EN=0)
Next, we will explain how to interpret the BIST result.
time
reserved reserved reserved reserved reserved reserved BIST_POL BIST_EN
BIST Test Sequence
I2C Register 0x6:
1ms

Vesper -Confidential
6
Hardware Check:
How to interpret result of BIST
•The size of the square wave measured during
BIST Test 0
and
BIST Test 1
will
determine the result of the test.
Result of BIST Test 0
Result of BIST Test 1
Interpretation
Small square wave
(
OUT < 50mVpp*)
Large square wave
(
OUT > 50mVpp*)
System is healthy.
Small square wave
(
OUT < 50mVpp*)
Small square wave
(
OUT < 50mVpp*)
VA1210, PCB wiring,
or ADC is damaged.
Large square wave
(
OUT > 50mVpp*)
Large square wave
(
OUT > 50mVpp*)
VA1210 is damaged.
Large square wave
(
OUT > 50mVpp*)
Small square wave
(
OUT < 50mVpp*)
VA1210 is damaged.
*50mVpp threshold may change in the production version of the VA1210

Vesper -Confidential
7
Wearing the TWS:
Quality of Fit and Cable Slack
•TWS prototype should be worn properly to ensure a
high-quality VA signal.
•Make sure the TWS
fits tightly
inside the ear.
•If the fit is too loose, use a
larger rubber eartip
.
•If the TWS mockup has cables, follow these steps:
1. Make sure that there is
slack
in the cables.
•Cables connected to TWS should be
loose
(not pulling tight)
•Use
tape
to secure cable and reduce tension. This also reduces cable
fluttering during wind testing.
2. Place the interface board
near
the wearer
•The interface board (connected to the TWS with cables) should not
be too far. It could be on a nearby table, or strapped to the shoulder.

Vesper -Confidential
8
Signal Quality Check:
Part 1/3 -Measuring SNR
Steps to measure SNR:
1. Wear the TWS mockup. Make
two recordings
of the VA signal:
a) Silence Recording
: 10 seconds of silence
b) Speech Recording
: 10 seconds of speech (wearer is talking, e.g. reciting Harvard Sentences)
2. Bandpass
both recordings with a bandpass filter (120Hz -2kHz, 30dB/octave)
3. Compute
the following:
a) Compute RMS of
Silence Recording
Noise,
RMSnoise
b) Compute RMS of
Speech Recording
Total Noise + Speech,
RMStotal
c) Calculate Voice RMS from previous two results
RMSvoice
= sqrt(
RMStotal2
–
RMSnoise2
)
d) Calculate SNR in dB
SNRdB
= 20·log10(
RMSvoice
/
RMSnoise
)
In a good TWS mockup with normal voice loudness, measured SNR should be
>>6dB.

Vesper -Confidential
9
Here, one audio vector (AudioRecording) includes both
Silence Recording
&
Speech Recording.
Start/stop time of each recording marked with: silenceStart, silenceEnd, speechStart, speechEnd
Code start:
audio_filtered = bandpass(AudioRecording, [120 2000], Fs, 'ImpulseResponse','iir','Steepness',[0.8 0.8]);
RMSnoise = rms(audio_filtered(silenceStart*Fs:silenceEnd*Fs));
RMStotal = rms(audio_filtered(speechStart*Fs:speechEnd*Fs));
RMSvoice = sqrt(RMStotal^2 -RMSnoise^2);
SNRva1200 = 20*log10(RMSvoice/RMSnoise);
Signal Quality Check:
Part 2/3 –Example MATLAB code
Compute RMS of
Silence Recording
Noise,
RMSnoise
Calculate Voice RMS from previous two results
RMSvoice
= sqrt(
RMStotal2
–
RMSnoise2
)
Calculate SNR in dB
SNRdB
= 20·log10(
RMSvoice
/
RMSnoise
)
Compute RMS of
Speech Recording
Total Noise + Speech,
RMStotal
Bandpass both recordings with a bandpass filter
(120Hz -2kHz, 30dB/octave)

Vesper -Confidential
10
Signal Quality Check:
Part 3/3 -Analyzing Noise Floor
•You can check for RF interference
by plotting the FFT of the
Silence
Recording
. This is the VA1210
noise floor.
•A clean noise floor (see
blue
curve)
looks like a
1/f
slope with a ~30dB
resonant peak at 4.5kHz.
•If there is RF interference, it can
look like sharp peaks on the noise
floor (see
red
curve at 5-6kHz).
•Bluetooth interference can be
identified by the noise frequency.
The peaks are usually multiples of
133Hz (i.e. 133Hz, 266Hz, 400Hz)
Clean VA1210 noise recording
VA1210 noise
recording with RF
interference