AND9035/D
BELASIGNA[ 250 and 300
for Low-Bandwidth
Applications
http://onsemi.com
APPLICATION NOTE
Introduction
BELASIGNA 300 there are multiple options available.
Such decimation can be efficiently performed by the
HEAR Accelerator signal processing engine. While the
standard library of HEAR functions does not include a
specific down−sampling module, there are several filtering
modules available (both FIR and IIR) that can be used to
low−pass filter the incoming data, as well as vector copy
modules that can be used to extract to the down−sampled
stream from the filtered data. Alternatively, a custom
decimation filter function can be written for execution on the
CFX core. Finally, ON Semiconductor provides an
optimized HEAR module that combines FIR filtering and
deimation to achieve greater efficiency.
In the example described below, the system is configured
for 16 kHz sampling. A custom HEAR module performs
multi−stage decimation using low−pass FIR filters. The first
stage decimates by a factor of 10 and the second by a factor
of 4 to achieve an overall decimation factor of 40. The signal
bandwidth is effectively reduced from 8 kHz to 200 Hz.
This application note describes the use of
BELASIGNA 250
and
BELASIGNA 300 in
low−bandwidth applications. The intended audience is
portable electronics equipment designers who are working
with low−bandwidth signals, typically under 200 Hz. An
example application is the analysis of biological signals for
monitoring and/or diagnosis. The goal of this document is to
provide customers with information about the performance
of these products in low−bandwidth situations, as well as
information on how best to configure the devices for use in
such environments.
This application note also describes the expected
performance of BELASIGNA 250 and BELASIGNA 300
in terms of analog−to−digital conversion noise and
accuracy, system current consumption and MIPS required in
a typical application, and to suggest configuration and
implementation strategies to ensure the best possible
performance in these categories.
ANALOG TO DIGITAL CONVERSION
ADC Calibration on BELASIGNA 300
The operational amplifiers (op−amps) used in the input
stage of BELASIGNA 300 produce DC offsets that are
accumulated in the closed loop of the sigma−delta ADC.
This results in undesirable spikes and/or bumps in the noise
floor. The value of the DC offset is configurable through a
dedicated register. To prepare BELASIGNA 300 for
near−DC sampling the ADC must be properly calibrated.
Typically this would be performed at the production line,
with calibration values stored in an SPI EEPROM attached
to BELASIGNA 300. These values are then loaded at boot
time. Each ADC can be calibrated for each of the 8 available
pre−amplifier gain settings. However, it is more efficient to
calibrate the ADCs for only those pre−amplifier gain
settings used in the application. For more information on
ADC calibration for BELASIGNA 300, see Reference 1.
Analog to Digital Converters
BELASIGNA 250 and BELASIGNA 300 provide
16−bit, 8x over−sampled analog−to−digital converters with
configurable sampling frequencies from 1.27 kHz to
60 kHz. BELASIGNA 250 offers two ADCs, while
BELASIGNA 300 offers four.
The above sampling frequency range does not extend low
enough for the low−bandwidth applications discussed in this
application note. However, down−sampling can be
performed in software on the DSP itself. Down−sampling by
a factor of N typically consists of two steps:
• Low−pass filtering: limits the bandwidth of the
incoming signal to prevent aliasing in the
down-sampled signal
• Decimation: reduces the data rate by selecting every
Nth sample and discarding the remaining samples
Input Stage Noise
In many low−bandwidth applications, the signals
involved are of a very low level – often less than 10 mV.
Input stage noise is thus an important consideration for the
system designer. Noise characteristics of the input stages of
BELASIGNA 250 and BELASIGNA 300 are given below.
This process can be repeated (multi−stage decimation) to
achieve a higher overall decimation rate.
On BELASIGNA 250 a custom down−sampling routine
can be written for the RCore DSP, while on
© Semiconductor Components Industries, LLC, 2011
September, 2011 − Rev. 0
1
Publication Order Number:
AND9035/D
AND9035/D
noise is calculated as the maximum positive sample minus
the minimum negative sample, and the RMS noise value is
calculated as:
Of interest are zero−input peak−to−peak and RMS noise in
two bandwidth regions: 0.5 to 40 Hz and 0.5 to 150 Hz.
Measurements were made by acquiring a buffer
containing 30 seconds of digital audio samples through the
PCM interface at a sampling frequency of 16 kHz. The
analog input pin was grounded through a 0.1 nF capacitor.
The data was band−pass filtered offline (using MATLAB®)
to the bandwidths mentioned above. The peak−to−peak
X RMS +
Ǹ
x 0 2 ) x 1 2 )AAA ) x 2N*1
N
All values are in mV and are input−referred.
Table 1. RMS NOISE (mV, input referred) IN THE 0.5 − 150 Hz REGION
BELASIGNA 300
BELASIGNA 250
Chopper Clock Divisor
Chopper Clock Divisor
Off
0x1
0x5
0x9
0xF
Off
0x1
0x5
0x9
0 dB
31.7
31.1
30.1
31.3
30.7
8.1
10.0
10.1
10.1
12 dB
8.1
7.8
7.8
7.9
8.0
3.1
2.8
3.1
3.1
18 dB
4.5
4.0
4.0
4.1
3.9
1.7
1.5
1.5
1.6
24 dB
3.0
2.3
2.2
2.0
2.1
1.0
0.9
0.8
0.9
30 dB
2.1
1.2
1.2
1.2
1.3
0.7
0.8
0.5
0.5
Table 2. PEAK−PEAK NOISE (mV, input referred) IN THE 0.5 − 150 Hz REGION
BELASIGNA 300
BELASIGNA 250
Chopper Clock Divisor
Chopper Clock Divisor
Off
0x1
0x5
0x9
0xF
Off
0x1
0x5
0x9
0 dB
198.8
203.8
200.4
200.0
198.4
99.5
97.5
90.1
87.2
12 dB
54.1
53.4
50.0
51.3
50.9
27.2
23.6
29.4
28.5
18 dB
28.7
27.4
26.3
26.4
26.0
16.8
14.7
12.2
14.3
24 dB
19.0
15.1
13.9
13.6
13.7
9.0
8.0
7.2
7.2
30 dB
14.8
8.0
8.3
8.0
8.6
5.5
6.4
4.9
7.0
Table 3. RMS NOISE (mV, input referred) IN THE 0.5 − 40 Hz REGION
BELASIGNA 300
BELASIGNA 250
Chopper Clock Divisor
Chopper Clock Divisor
Off
0x1
0x5
0x9
0xF
Off
0x1
0x5
0x9
0 dB
25.1
25.2
24.0
25.3
24.9
6.0
6.3
6.5
6.4
12 dB
6.6
6.2
6.3
6.3
6.4
2.2
1.7
2.1
2.1
18 dB
3.6
3.3
3.2
3.3
3.3
1.2
0.9
0.9
1.0
24 dB
2.6
2.0
1.8
1.7
1.7
0.7
0.5
0.5
0.6
30 dB
1.7
1.0
1.1
1.1
1.2
0.5
0.5
0.3
0.3
Table 4. PEAK−PEAK NOISE (mV, input referred) IN THE 0.5 − 40 Hz REGION
BELASIGNA 300
BELASIGNA 250
Chopper Clock Divisor
Chopper Clock Divisor
Off
0x1
0x5
0x9
0xF
Off
0x1
0x5
0x9
0 dB
142.3
150.0
145.2
140.7
141.7
69.3
51.4
59.3
60.0
12 dB
37.8
36.4
35.0
37.4
37.4
19.2
15.3
16.4
19.2
18 dB
20.9
20.3
18.3
19.2
18.5
12.1
6.7
7.7
10.3
http://onsemi.com
2
AND9035/D
Table 4. PEAK−PEAK NOISE (mV, input referred) IN THE 0.5 − 40 Hz REGION (continued)
BELASIGNA 300
BELASIGNA 250
Chopper Clock Divisor
Chopper Clock Divisor
Off
0x1
0x5
0x9
0xF
Off
0x1
0x5
0x9
24 dB
14.5
11.2
10.4
9.7
9.5
6.3
4.5
3.6
4.0
30 dB
10.7
6.3
6.1
6.4
7.0
3.7
4.1
2.7
2.7
Chopper Stabilization
Conversion Accuracy
When examining low−bandwidth signals close to DC, 1/f
noise typically dominate the noise floor. In op−amps a
modulation scheme called chopper stabilization can be used
to reduce the effects of 1/f noise. This technique involves
modulating the op−amp’s input signal to a higher frequency
where 1/f noise is minimal, and then demodulating
following amplification. In BELASIGNA 250 chopper
stabilization is implemented in the pre−amplifiers and the
ADCs, and in BELASIGNA 300 chopper stabilization is
implemented only in the pre−amplifiers. The chopper carrier
signal is a square wave with a configurable frequency. The
recommended chopper frequency is an integer multiple of
the sampling frequency, typically 8x the sampling
frequency.
On BELASIGNA 250, the chopper clock is derived from
the Master Clock signal (MCLK) with the frequency defined
by a 4−bit integer divisor. To enable the chopper, set bit 6 of
analog control register A_ADC_CUR_CTRL (0x1C). The
configuration value (CHOPPER_CTRL) used to define the
frequency is stored in bits (7:4) of analog control register
A_CLK_CTRL (0x11). The default CHOPPER_CTRL
value of 0 disables chopper stabilization. For non−zero
values, the chopper clock frequency is defined as:
Another important factor in ADC performance is
conversion accuracy, often specified by differential
nonlinearity (DNL). Nominal static DNL values for
BELASIGNA 250 and BELASIGNA 300 ADCs with a
16 kHz sampling frequency are given in Table 5 below.
Chopper Frequency +
Table 5. STATIC DIFFERENTIAL NONLINEARITY
DNL
NOTE:
BELASIGNA 300
±0.4 LSBs
±0.4 LSBs
Typical, not qualified.
Conversion accuracy measurements were made using the
histogram method with a sine wave input signal [2]. The sine
wave signal was generated with a function generator (Audio
Precision) and the digital converted values were transmitted
to a PC through the Serial Peripheral Interface (SPI) port,
collected with an SPI analyzer device. MATLAB was used
to perform the final calculation to obtain DNL values. Since
the ADCs provide 16−bit samples for both systems, the
number of samples collected was in excess of 70 million to
ensure DNL error of less than 0.1 LSB and a 99% level of
confidence.
MCLK
(CHOPPER_CTRL ) 1)
Example Low−Bandwidth Application
On BELASIGNA 300, the chopper clock is derived from
SLOW_CLK with the frequency defined by a 4−bit integer
divisor. The configuration value (CHOPPER_CLK) used to
define the frequency is stored in bits (19:16) of analog
control register A_INPUT_CTRL (0xE141). The default
CHOPPER_CLK value of 0 disables chopper stabilization.
For non−zero values, the chopper clock frequency is defined
as:
Chopper Frequency +
BELASIGNA 250
BELASIGNA 300 in a Cardiac Monitoring Application
To demonstrate the capabilities of BELASIGNA 300 in a
low−bandwidth application, a simple cardiac monitoring
application is presented. In this application, ECG lead
signals (right arm (RA), left arm (LA), left leg (LL) and
chest (V2)) are input to BELASIGNA 300 and the DSP is
responsible for down−sampling (by a factor of 40),
calculation of differential signals I (LA – RA), II (LL – RA)
and III (LL − LA), heart rate detection, and for the
transmission of this data back to a host PC through the I2C
interface.
SLOW_CLK
(CHOPPER_CTRL ) 1)
Example: On BELASIGNA 300, if the SLOW_CLK
frequency is set to 1.28 MHz and the sampling frequency is
16 kHz, the recommended chopper frequency is 8 * 16 kHz
= 128 kHz. To obtain this frequency, SLOW_CLK must be
divided by 10. Thus, the appropriate value for
CHOPPER_CLK is 9.
System Configuration
In this application, the four input signals are capacitively
coupled to the BELASIGNA 300 analog input pins. A buffer
is used to match impedances between the sensor output and
http://onsemi.com
3
AND9035/D
determined by the down−sampling factors. With a block size
of 160 samples per channel, the first stage of
down−sampling (by a factor of 10) results in 16 new samples
at an intermediate sampling rate of 1.6 kHz. A further stage
of down−sampling by a factor of 4 results in four new
samples at the target sampling rate of 400 Hz. All
down−sampling is performed by the HEAR Accelerator.
The HEAR uses a custom microcode module to efficiently
perform stereo FIR low−pass filtering and decimation. The
first stage uses a 16−tap filter while the second uses a
128−tap filter.
Upon completion of the decimation, the CFX begins
feature computation using the V2 lead’s decimated signal. A
simple derivative−based method is used for peak detection
[3]. The steps performed in feature computation include:
• Calculation of first−order derivative:
the BELASIGNA 300 input stage. The front end used for
this application is shown in Figure 1.
Optional Filter
+
100 nF
×1
+
+
Sensor
Output
To
BELASIGNA
Input
Figure 1. Front End for Cardiac Monitor
Application
In this application, the op−amps are configured for unity
gain and are used for buffering the input signals. The 100 nF
coupling capacitor gives a cut−off frequency of ~3 Hz. The
beat detection in this algorithm is based on the R peak in the
ECG waveform. For a more detailed view of the ECG
waveform a larger capacitor (and thus a lower cut−off
frequency) is recommended.
While the ADCs provide effective anti−alias filtering
through an 8x oversampled architecture, additional
low−pass filtering can be performed in the analog domain to
further improve performance. A sample low−pass filter is
shown in the diagram above.
The BELASIGNA 300 system is configured by the
application firmware as described in Table 6.
y o[n] + x[n] * x[n * 1]
• Calculation of second−order derivative:
y 1[n] + y 0[n] * y 0[n * 1]
• Linear combination of the magnitude of derivatives:
y 2[n] + 2
2.56 MHz (internal oscillator, calibrated)
SLOW_CLK
1.28 MHz
Sampling Rate
16 kHz
ADC Chopper Rate
128 kHz
Preamp gain
30 dB
|y 0[n]| ) w 1
Ťy 1[n]ŤǓ
Where w0 = 0.65 and w1 = 0.55.
• Moving average filter:
y 3[n] + sumǒy 2[n], AAA , y 2[n * M ) 1] ǓńM
Table 6. SYSTEM CONFIGURATION
System Clock
ǒw 0
Where M = 8.
• Finally, set a binary variable to indicate peaks:
y 4[n] +
{
1, if y 3[n] u threshold
0, otherwise
The remaining processing consists of examining the y4
signal to determine the likelihood of a peak. If a peak is
detected, the peak−to−peak time interval is estimated as the
number of blocks that have elapsed since the previously
detected peak. If this block count falls within a valid range,
corresponding to 40 to 200 beats per minute (BPM), then the
block count is added to a circular buffer. The average block
count in this buffer is calculated, and converted to a BPM
value.
For this application a low−level I2C slave handler was
included to facilitate communications with a host PC. The
PC communicates with BELASIGNA 300 through the
ON Semiconductor Communications Accelerator Adaptor
(CAA) to request BPM information, as well as up to three
sets of decimated signal data (any three of the RA, LA, LL,
V2, I, II and III signals).
Using
an
ECG
patient
simulator
device,
BELASIGNA 300 is able to perform heart rate detection
with input signals down to 0.5 mV.
Four FIFOs are used in this application. The four input
signals are combined into two interleaved signals and routed
to two stereo FIFOs. The remaining two FIFOs are used to
store the down−sampled versions of these signals.
Signal Processing
Processing in this application includes the following
steps:
• Down−sampling by a factor of 10
• Down−sampling by a further factor of 4
• Differential signal calculation
• Feature computation
• Beat detection
The signal processing is split between the CFX and the
HEAR, with the HEAR performing the filtering and
decimation and the CFX performing feature computation
and beat detection. The processing in this application is
block−based, meaning the DSP waits for one full “block” of
ADC samples and then processes all data samples in the
block. This process is then repeated. The block size is
Resource Requirements
BELASIGNA 300 is capable of running the sample
application described above with a current consumption of
511 mA at 1.8 V. In terms of CPU usage, BELASIGNA 300
http://onsemi.com
4
AND9035/D
Hardware Considerations
requires roughly 3000 cycles to perform down−sampling
and beat detection per block. At 16 kHz sampling and a
block size of 160 samples, there are 0.01 seconds per block.
With a 2.56 MHz system clock, this equates to 25600
available cycles per block. Thus, in this application
BELASIGNA 300 is utilizing 12% of the CPU for the
detection algorithm. Some additional CPU usage is required
for communications with the host PC.
• System current consumption can be reduced by using
•
Design Considerations
When designing BELASIGNA 250 or BELASIGNA 300
into a low−bandwidth application, there are several
considerations, both hardware and software, that must be
taken into account.
•
Software Considerations
• Current consumption can be minimized by reducing the
overall system clock. When designing the firmware for
a given application, the best approach is to split
processing between the CFX and HEAR as evenly as
possible, maximizing parallel processing. By doing so,
the system clock can be drastically reduced compared
to a serial processing approach.
• When dividing processing between the CFX and the
HEAR, it is advisable to allocate typical signal
processing functionality to the HEAR while the CFX
handles application−level processing and signal
processing functions which are not available in the
standard HEAR library.
• When performing down−sampling with a high
decimation rate (e.g. 40, as seen in the above example),
it is more efficient to split the decimation into multiple
stages. While this approach may make the resulting
firmware somewhat more complicated, it provides
significant savings of both processing time and memory
by replacing one large monolithic filter with multiple
smaller filters with a lower overall number of taps.
•
•
the internal oscillator rather than an external oscillator.
A properly calibrated internal oscillator offers accuracy
to within ± 1% of the calibrated frequency. For higher
accuracy, an external oscillator should be used.
BELASIGNA 250 and BELASIGNA 300 can operate
with a supply voltage down to 1 V, further reducing
system current consumption.
Input impedance for analog inputs is nominally 500 kW
for both BELASIGNA 250 and BELASIGNA 300.
When selecting decoupling capacitors, a value should
be chosen such that the corner frequency is low enough
to pass all desired frequencies.
High impedances are frequently encountered when
working with biological sensors. To deal with
impedance mismatches between sensors and
BELASIGNA’s input stage, it is recommended that
front end buffers be used.
The ADCs in BELASIGNA 300 should always be
calibrated to ensure best performance.
References
1. “ADC Offset Calibration of the BELASIGNA[
300 Series” (AND8341/D), ON Semiconductor,
2008.
2. W. Kester, “The Data Conversion Handbook”,
Newnes, 2005.
3. B−U Köhler, C. Hennig and R. Orglmeister, “The
Principles of Software QRS Detection”, IEEE
Engineering in Medicine and Biology, 2002.
Company or Product Inquiries
For more information about ON Semiconductor products
or services visit our Web site at http://onsemi.com.
Technical Contact Information
[email protected]
BELASIGNA is a registered trademark of Semiconductor Components Industries, LLC (SCILLC).
MATLAB is a registered trademark of The MathWorks, Inc.
ON Semiconductor and
are registered trademarks of Semiconductor Components Industries, LLC (SCILLC). SCILLC reserves the right to make changes without further notice
to any products herein. SCILLC makes no warranty, representation or guarantee regarding the suitability of its products for any particular purpose, nor does SCILLC assume any liability
arising out of the application or use of any product or circuit, and specifically disclaims any and all liability, including without limitation special, consequential or incidental damages.
“Typical” parameters which may be provided in SCILLC data sheets and/or specifications can and do vary in different applications and actual performance may vary over time. All
operating parameters, including “Typicals” must be validated for each customer application by customer’s technical experts. SCILLC does not convey any license under its patent rights
nor the rights of others. SCILLC products are not designed, intended, or authorized for use as components in systems intended for surgical implant into the body, or other applications
intended to support or sustain life, or for any other application in which the failure of the SCILLC product could create a situation where personal injury or death may occur. Should
Buyer purchase or use SCILLC products for any such unintended or unauthorized application, Buyer shall indemnify and hold SCILLC and its officers, employees, subsidiaries, affiliates,
and distributors harmless against all claims, costs, damages, and expenses, and reasonable attorney fees arising out of, directly or indirectly, any claim of personal injury or death
associated with such unintended or unauthorized use, even if such claim alleges that SCILLC was negligent regarding the design or manufacture of the part. SCILLC is an Equal
Opportunity/Affirmative Action Employer. This literature is subject to all applicable copyright laws and is not for resale in any manner.
PUBLICATION ORDERING INFORMATION
LITERATURE FULFILLMENT:
Literature Distribution Center for ON Semiconductor
P.O. Box 5163, Denver, Colorado 80217 USA
Phone: 303−675−2175 or 800−344−3860 Toll Free USA/Canada
Fax: 303−675−2176 or 800−344−3867 Toll Free USA/Canada
Email: [email protected]
N. American Technical Support: 800−282−9855 Toll Free
USA/Canada
Europe, Middle East and Africa Technical Support:
Phone: 421 33 790 2910
Japan Customer Focus Center
Phone: 81−3−5773−3850
http://onsemi.com
5
ON Semiconductor Website: www.onsemi.com
Order Literature: http://www.onsemi.com/orderlit
For additional information, please contact your local
Sales Representative
AND9035/D