All modern real-time digitizers and oscilloscopes incorporate a wide band front-end amplifier and an Analog-to-Digital Converter (ADC). The inherent noise of the front-end amplifier tends to increase to the square root of the bandwidth.
A solution to the problem of characterizing signals in the presence or below the noise level is to use signal averaging accumulation. It is a well known technique used to increase the vertical resolution. The noisy signal is acquired many times. A memory buffer is used to add up N waveforms and then divide by N.
There are a few restrictions: First, the signal must be repetitive. Second, the signal waveform must be stable in time with respect to the ADC sampling clock when in synchronous averaging mode or to the trigger event in asynchronous averaging mode. Thirdly, it should be noted that the averaging process removes only the random/uncorrelated noise on the signal.
It is important to note that in synchronous averaging mode the signal repetition may be more stable. The synchronous sampling guarantees that each period of the periodic signal is sampled with the same sampling phase. This allows averaging a periodic signal with a known period. The ADP7104 must be properly synchronized with an external clock source via 1 GHz external clock or 500 MHz, 250 MHz, 200 MHz, 100 MHz reference clock input/output.
In time–frequency analysis, waveform signals are analyzed in both time and frequency domains to capture the progressive change of spectrum components with time (e.g., a transient or nonstationary signal such as music). A waveform is a graphic representation of a wave. Physical phenomena such as vibrations, temperature, or electrical phenomena such as current or power can be converted to a voltage by a sensor. One cycle of a wave is the portion of the wave that repeats. A voltage waveform shows time on the horizontal axis and voltage on the vertical axis.
In asynchronous averaging mode the noise in the trigger signal may cause the trigger time to shift from acquisition to acquisition and therefore can distort the averaged waveform. Asynchronous sampling can also be affected by the drift in the sampling time base.
Knowing the difference between both modes described above, allows to choose the most suited option for required signal averaging.
Guzik ADP7104 16/32 Gsa/s 10-bit digitizer features both synchronous and asynchronous averaging capabilities. The averaging is carried out in the FPGA-s in real-time with 100% trigger utilization and up to 4B (4,000,000,000) total waveform accumulations. The real-time averaging functionality is available as a command line utility and as software SDK library API for system integration purposes.
Example:
ADC’s LSB RMS | uV, RMS | SNR, dB | ||
| 2 microsecond 1 GHz LFM chirp | Signal | 9.3 | 144.2 | |
Before Averaging | Noise | 26.9 | 416.9 | -9.3 |
After 1M (1,000,000) Averaging | Noise | 0.0382 | 0.592 | 47.7 |
Total SNR improvement | 57 | |||
Operation throughput from command, arming to final averaged .csv file stored to SSD:
2.1 microsecond Pulse Repetition Interval (PRI) and 100% trigger utilization. Resulting .csv file 1.37 MB (1,437,932 bytes):
1 PRI elapsed time: 0.293267 seconds
1,000 PRI elapsed time: 1.27325 seconds
100,000 PRI elapsed time: 1.32839 seconds
1,000,000 PRI elapsed time: 3.28767 seconds
10,000,000 PRI elapsed time: 21.308 seconds
100,000,000 PRI elapsed time: 210.367 seconds
1,000,000,000 PRI elapsed time: 2100.55 seconds
ADP7104 per channel Synchronous Averaging Mode (Option ADC_AVG) | |||
Sampling rate | Minimum accumulator length | Maximum accumulator length | Bandwidth |
32 GSa/s | 30 ns | 40 uS | 10 GHz |
16 GSa/s | 30 ns | 40 uS | 6.5 GHz |
8 GSa/s | 60 ns | 80 uS | 3 GHz |
4 GSa/s | 120 ns | 160 uS | 1.6 GHz |
2 GSa/s | 240 ns | 320 uS | 800 MHz |
1 GSa/s | 480 ns | 640 uS | 400 MHz |
ADP7084 per channel Synchronous Averaging Mode (Option ADC_AVG) | |||
Sampling rate | Minimum accumulator length | Maximum accumulator length | Bandwidth |
20 GSa/s | 48 ns | 64 uS | 8 GHz |
10 GSa/s | 48 ns | 64 uS | 4 GHz |
5 GSa/s | 96 ns | 128 uS | 2 GHz |
2.5 GSa/s | 192 ns | 256 uS | 1 GHz |
1.25 GSa/s | 384 ns | 512 uS | 500 MHz |
0.625 GSa/s | 768 ns | 1024 uS | 250 MHz |
ADP7104 per channel Asynchronous Averaging Mode (Option ADC_AVGA) | |||
Sampling rate | Minimum accumulator length | Maximum accumulator length | Bandwidth |
32 GSa/s | 80 ns | 10 uS | 10 GHz |
16 GSa/s | 80 ns | 10 uS | 6.5 GHz |
8 GSa/s | 160 ns | 20 uS | 3 GHz |
4 GSa/s | 320 ns | 40 uS | 1.6 GHz |
2 GSa/s | 240 ns | 80 uS | 800 MHz |
1 GSa/s | 480 ns | 160 uS | 400 MHz |
ADP7084 per channel Asynchronous Averaging Mode (Option ADC_AVGA) | |||
Sampling rate | Minimum accumulator length | Maximum accumulator length | Bandwidth |
20 GSa/s | 128 ns | 16 uS | 8 GHz |
10 GSa/s | 128 ns | 16 uS | 4 GHz |
5 GSa/s | 256 ns | 32 uS | 2 GHz |
2.5 GSa/s | 512 ns | 64 uS | 1 GHz |
1.25 GSa/s | 1024 ns | 128 uS | 500 MHz |
0.625 GSa/s | 2048 ns | 256 uS | 250 MHz |
The Asynchronous Averaging using segmented memory approach uses the ADP7000 digitizer to acquire two signals: the synchronization signal and the signal-to-be-averaged.
The digitizer acquisition is triggered by the synchronization signal in multi-segment acquisition mode, where the signal-to-be-averaged is divided by individual segments each starting with its own synchronization event. The multi-segment acquisition greatly reduces the acquisition memory consumption, when the rate of synchronization events is low.
The acquired data is downloaded to the host computer for software post processing. Software performs precise detection of synchronization event locations with sub-sample accuracy. Then it performs sub-sample averaging of multiple segments of the signal-to-be-averaged. The sub-sample accuracy of both trigger detection and averaging improves the averaging accuracy when compared with the traditional approach. Efficient multicore algorithms are used for software post processing, so the Asynchronous Averaging performance is limited only by the data download speed from the digitizer to the host computer.
The software post processing implementation imposes some limits on the Asynchronous Averaging parameters. Particularly, only one signal channel is averaged at a time, the maximum acquisition size is limited by the digitizer memory, and the performance is limited by the data download speed.
Signal Waveform
The Asynchronous Averaging algorithm is implemented in GSA SDK library and is available as the command line utility and as software API.
Key specifications for the software implementation of Asynchronous Averaging with (Option ADC_SM):
Ppg Signal Waveform
| Specification | Value |
| Sampling rate | Up to 32 Gsa/s for ADP7104 |
| Up to 20 Gsa/s for ADP7084 | |
| Number of ADC inputs used | Two: one for trigger and one for signal to average |
| Maximum acquisition size | 52.5 GSamples |
| Maximum segment size for 1M (1,000,000) averages | 1.6 us |
| Maximum averages for 10 us segment | 160,000 |
| Time to process max. acquisition | 135-140 sec |
| Inter-segment dead time | < 200 ns |
Signal Waveform Amplitude
For more information contact us at sales@www.guzik.com