Periodogram Service Performance
User Guide
Using the Service:- Accessing the Service
- Accepted File Formats
- Editing an Input Table
- Plotting an Input Table
- Settings and Parameters
- Viewing Current Results
- Viewing Past Results
- Recipe: Variable Star
- Recipe: Star Spot Evolution
Performance
The Periodogram code has been implemented in ANSI-C for portability and performance.
Printer-friendly version
Performance
The following section shows the times, on a quiet machine, for calculating a periodogram with the three available algorithms.
Table 1. Comparison of Processing Times for Each of the Algorithms Using A Single Processor*
| Kepler Long Cadence Time-Series File Name | # of Points in Time-Series | Lomb- Scargle (s) | BLS (s) | Plavchan (s) |
|---|---|---|---|---|
| kplr000892010-2009350155506_llc.fits | 4,370 points | 17s | 45s | 86s |
* - On a typical desktop computer, circa 2011
Table 2. Comparison of Processing Times for Each of the Algorithms Using the Exoplanet Archive 128-core Cluster Periodogram Service
The following information corresponds to the current implementation and should be used as estimates for processing times.
| # of Stitched Together Kepler Long Cadence Time-Series | # of Points in Time-Series | Lomb- Scargle (s) | BLS (s) | Plavchan (s) |
|---|---|---|---|---|
| 1 | ~4,400 | <15s | <15s | <15s |
| 2 | ~8,700 | <15s | <15s | <15s |
| 3 | ~13,000 | <15s | <15s | <15s |
| 4 | ~18,000 | <15s | <15s | 27s |
| 5 | ~22,000 | 17s | 25s | 44s |
| 6 | ~26,000 | 25s | 31s | 65s |
| 7 | ~31,000 | 32s | 36s | 83s |
| # of Stitched Together Kepler Short Cadence Time-Series | # of Points in Time-Series | Lomb- Scargle (m) | BLS (m) | Plavchan (m) |
| 1 | ~43,000 | 1m | 1m | 3m |
| 2 | ~87,000 | 5m | 2.5m | 15m |
| 3* | ~120,000 | 1m | 0.5m | 3m |
| 4 | ~170,000 | 2m | 0.6m | 6m |
| 5 | ~210,000 | 3m | 0.8m | 9m |
| 6 | ~250,000 | 4m | 1m | 13m |
| 7 | ~290,000 | 6m | 1.5m | 19m |
| 8 | ~330,000 | 8m | 2m | 25m |
| 9 | ~380,000 | 10m | 2.5m | 33m |
| 10 | ~420,000 | 12m | 3m | 40m |
| 11 | ~470,000 | 15m | 3.5m | 48m |
| 12 | ~510,000 | 18m | 4m | 59m |
| 13 | ~560,000 | 21m | 4.5m | 70m |
| 14 | ~600,000 | 24m | 5m | 82m |
| 15 | ~640,000 | 28m | 6m | 94m |
| 16 | ~670,000 | 30m | 6.5m | 102m |
| 17 | ~700,000 | 33m | 7m | 115m |
| 18 | ~730,000 | 36m | 7.5m | 125m |
* Change in default parameters for time-series with large a number of data points (> 100,000) decreases integration time with sparser period sampling.
Notes on Performance
These computation times were calculated for Q2-Q8 time-series, including existing data gaps, for KIC/Kep ID 2571238 (= KOI 84.01 = Kepler 19b).
These computation times assume that the periodogram server load is light. Thus, these times can be longer by a factor of 2 or more under heavy server load.
These computation times assume default periodogram parameters for the indicated algorithm, including period step method and step size.
For Kepler time series, we recommend computing periodograms only for the long cadence Kepler data to obtain speedier results. Calculating a periodogram with short cadence time series is recommended if, for example, one is looking for very high-frequency periodic signals on time scales of less than 30 minutes from stellar pulsations.
There are three short cadence time-series per long cadence quarter.
Last updated: 26 April 2018