Hi everyone,
I have searched on this forum but didn’t see any topic related to this.
I noticed there is always an artefact in the PPSD results, close to the max frequency. For example, it shows around 70 Hz when the sample rate is 200Hz (Nyquist=100Hz). (e.g. screenshot below)
I have tried to change parameters like window length but it doesn’t help.
In ObsPy documentation and screenshots, this artefact appears everywhere. Is this planned to be fixed in next release ? And does someone know a way around this ?
Hey Damien,
Could that offset at 70 Hz (0.142 sec) be related to the system response? With LSBB2, I am unsure what type of sensor/digitizer is installed at that location, but could that be related to the noise floor of the sensor itself? Sensor manufactures often will supply datasheets related to the noise floor of the instrument and you may check to see if the step correlates to that instrument self noise floor. Also, digitizers also have their own self-noise which can also impact the overall sensitivity, but in most cases, the sensor is the limiting factor.
With BW.KW1 with the offset shifting downward makes me wonder if there is an issue with your (or the providers) poles-zeros response information. Earthscope only indicates that the sensor is a Trillium-C seismometer, which I am personally unfamiliar with. In my experience, a shift is not unexpected if the offset is system self-noise related, but the downward offset is curious and makes me think the pz information is possibly incorrect. I may be wrong, but just my thoughts. Just remember that the PPSD are not lines but bins and usually the noise floor goes “upward” pretty steeply at both the short/long period ends. If it were me, I would just validate that the response information you are working with is correct.
Thanks for your reply Jon.
My data are from a Trillium too… incredible coincidence ! Then I understand why I see the same problem on the ObsPy documentation. Furthermore, I see the same on data from another type of seismo I have tested, at same frequency.
And I understand I was wrong, there is probably no bug in the ObsPy PPSD tool. I think I got to look into what is common to all those sensors : digitizer (hardware and software). I am used to process seismic data, rather than seismology. I am not used to automatic pre-processing included in the digitzer; this is probably the cause of this. This really looks to me like a processing artefact, like a filter effect. I will ask to the manufacturer.
PS: I have checked my poles and zeros, everything looks good on this side.
Thanks a lot,
I’ll keep posting here when I have some more information.
I honestly might not ever have looked into that kink/step there, since it’s at the very edge of the data what usually isn’t of much interest, but maybe I get around to debug into it, see if it has something to do with the binning
obspy.imaging.cm — ObsPy 1.4.1 documentation)