Did any of you try to resample the signal as in \nhttps://nl.mathworks.com/help/signal/ref/resample.html\nDid that work?
", "messages": [ { "id": "515316", "postDate": "04/12/2019 12:26:29", "content": "Did any of you try to resample the signal as in \nhttps://nl.mathworks.com/help/signal/ref/resample.html\nDid that work?
", "rawMarkdown": "Did any of you try to resample the signal as in \nhttps://nl.mathworks.com/help/signal/ref/resample.html\nDid that work?", "votes": null }, { "id": "515719", "postDate": "04/13/2019 02:38:13", "content": "I like this idea - but at what frequency should you consider resampling? maybe the machine used to make those experiments uses variable frequencies?
", "rawMarkdown": "I like this idea - but at what frequency should you consider resampling? maybe the machine used to make those experiments uses variable frequencies?", "votes": null }, { "id": "516330", "postDate": "04/14/2019 02:28:54", "content": "I tried it out. Most of the energy is centered around 250 kHz and below (assuming a 4 MHz sample rate). I isolated a small band from 240 kHz to 255 kHz, modulated it to baseband, and decimated the signal by a factor of 150. It seemed to work pretty well, and helps remove a lot of the quantization noise. If you're curious check out the beginning of my kernel.
", "rawMarkdown": "I tried it out. Most of the energy is centered around 250 kHz and below (assuming a 4 MHz sample rate). I isolated a small band from 240 kHz to 255 kHz, modulated it to baseband, and decimated the signal by a factor of 150. It seemed to work pretty well, and helps remove a lot of the quantization noise. If you're curious check out the beginning of my [kernel](https://www.kaggle.com/mrganger/lanl-finding-l-estimators-via-pca).", "votes": null }, { "id": "517651", "postDate": "04/16/2019 10:13:31", "content": "The sample rate is assumed to be 4Mhz, however as is also shown in a few kernels the time in between two succesive measurements is not a constant. Perhaps resampling towards 4Mhz is better than just assuming 4Mhz.
", "rawMarkdown": "The sample rate is assumed to be 4Mhz, however as is also shown in a few kernels the time in between two succesive measurements is not a constant. Perhaps resampling towards 4Mhz is better than just assuming 4Mhz.", "votes": null }, { "id": "518315", "postDate": "04/17/2019 04:44:54", "content": "Seems dangerous. If there are small higher frequency signals, they will be removed in the AA filter. These signals could perhaps help the model.
", "rawMarkdown": "Seems dangerous. If there are small higher frequency signals, they will be removed in the AA filter. These signals could perhaps help the model.", "votes": null } ], "comments": [ { "id": 515719, "author_name": "etaifour", "author_url": "", "post_date": "04/13/2019 02:38:13", "content": "I like this idea - but at what frequency should you consider resampling? maybe the machine used to make those experiments uses variable frequencies?
", "votes": null, "replies": [ { "id": 518315, "author_name": "petewills", "author_url": "", "post_date": "04/17/2019 04:44:54", "content": "Seems dangerous. If there are small higher frequency signals, they will be removed in the AA filter. These signals could perhaps help the model.
", "votes": null, "replies": [] } ] }, { "id": 516330, "author_name": "mrganger", "author_url": "", "post_date": "04/14/2019 02:28:54", "content": "I tried it out. Most of the energy is centered around 250 kHz and below (assuming a 4 MHz sample rate). I isolated a small band from 240 kHz to 255 kHz, modulated it to baseband, and decimated the signal by a factor of 150. It seemed to work pretty well, and helps remove a lot of the quantization noise. If you're curious check out the beginning of my kernel.
", "votes": null, "replies": [] }, { "id": 517651, "author_name": "hmcranbercourt", "author_url": "", "post_date": "04/16/2019 10:13:31", "content": "The sample rate is assumed to be 4Mhz, however as is also shown in a few kernels the time in between two succesive measurements is not a constant. Perhaps resampling towards 4Mhz is better than just assuming 4Mhz.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "515316": "Did any of you try to resample the signal as in \nhttps://nl.mathworks.com/help/signal/ref/resample.html\nDid that work?", "515719": "I like this idea - but at what frequency should you consider resampling? maybe the machine used to make those experiments uses variable frequencies?", "516330": "I tried it out. Most of the energy is centered around 250 kHz and below (assuming a 4 MHz sample rate). I isolated a small band from 240 kHz to 255 kHz, modulated it to baseband, and decimated the signal by a factor of 150. It seemed to work pretty well, and helps remove a lot of the quantization noise. If you're curious check out the beginning of my [kernel](https://www.kaggle.com/mrganger/lanl-finding-l-estimators-via-pca).", "517651": "The sample rate is assumed to be 4Mhz, however as is also shown in a few kernels the time in between two succesive measurements is not a constant. Perhaps resampling towards 4Mhz is better than just assuming 4Mhz.", "518315": "Seems dangerous. If there are small higher frequency signals, they will be removed in the AA filter. These signals could perhaps help the model." }, "source": "meta" }