| Shop |
| Products |
| Downloads |
| Services |
| Research |
| Nature |
| Press |
| Tech Notes |
| Contact |
FAQ Topic
What is "burst sampling" and why is it used?
The LogDator and SediMeter can be equipped with 2 optional sensors that are sampled up to 84 times during each measurement. This is burst sampling. The minimum rate of burst sampling is 0.5 Hz (1 sample per 2 seconds), and the maximum rate is 50 Hz (in the present firmware). The user can set how many samples to take and with what frequency. If no optional sensor is present always set to 0 samples, to save power.
 This timeline illustrates 2 measurements at an interval of 5 seconds, with a burst sampling setting of 3 samples at a rate of 1 Hz. Legend RED = sedimeter sensor sampled (it takes about 0.5 s) YELLOW = the white indicator LED on the sedimeter sensor blinks BLUE = optional sensors are sampled—this is the burst sampling GREEN = data is written to flash memory |
The reason for burst sampling is to provide a better temporal resolution for optional sensors without decreasing the measurement interval of the sedimeter sensor. The goal is to optimize data quality without sacrificing battery and memory. Some parameters change slowly and do not need such frequent sampling, e.g. bottom level and temperature. Others vary rapidly, and for them burst sampling can be quite useful. Burst sampling benefits include:
When using a light sensor in a shallow environment on a sunny day, the light level fluctuates by an order of magnitude on a time scale of fractions of a second. One way to deal with that would be to create a low-pass filter in hardware so that only the average was measured. However, perhaps the value of interest for one study is the RMS value, for another the peak value, and for the third the standard deviation. By taking a burst sample all of those parameters can be calculated in the post-processing software. Incidentally, the standard deviation can be used to determine if it is cloudy or sunny on a coral reef, for instance.
When using a pressure sensor the average can be calculated to estimate the depth, tide, or water level, without the instantaneous errors created by waves. By performing a frequency analysis the dominant wave period can be calculated. One can also calculate the energy content over the entire wave spectrum, so that both swell and wind wave components can be identified. Having access to the raw pressure data is of course always preferable to only having some statistical measures. Potential areas of use include wave energy conversion, wave damping, and measuring the swell from vessels.
Tip for SediMeter deployment
An instrument accuracy only says how accurate the measurement value is. It does not say if it is representatve for the population. To determine that, one must use statistical methods, and to use them, one must have many measurement values. It can be shown that the cost-optimal sampling strategy is not an even rate, but burst sampling. The bursts serve to detect short-scale fluctuations, which is essential to know in order to estimate the uncertainty. This is equally true in space, i.e., when considering spatial variability in geostatistics.
Suppose that you want to measure the sedimentation pattern away from a point source. If you just place one SediMeter on each station, you will not know how much of the measured variability is due to gradual spatial variation, and how much is due to the local variation caused by chance on each station. But if you deploy 2 to 5 instruments at each location, close to each other, you will get an idea of the local variation. This local over-sampling enables you to determine the so-called nugget effect. For more details look up kriging in a textbook on geostatistics, such as Myers, J.C., Geostatistical Error Management, Van Nostrand Reinhold, New York, 1997, ISBN 9780471285564.
The advice is thus, use burst sampling for temporal over-sampling, and use more instruments than there are measurement stations for spatial over-sampling.
Footnote: The main measurement interval can be set between 1 second and 24 hours. However, a shorter interval than 5 s is not recommended for a standard SediMeter, due to power issues. Contact us if this is a concern, since there is a workaround.