{
  "id": 5668,
  "title": "What units are used in the HF data",
  "url": "/competitions/belkin-energy-disaggregation-competition/discussion/5668",
  "author_name": "",
  "post_date": "2013-09-05T23:16:29.457Z",
  "votes": null,
  "comment_count": 2,
  "views": 1231,
  "content": "<p>Based on the reference papers, I assume that the HF data is based on frequency analysis of the voltage transients but it is not clear what units they are in, how they were normalized or even what is being measured.&nbsp;</p>\n<p>Are we looking at:</p>\n<p><span style=\"line-height: 1.4\">Amplitude? &nbsp;</span></p>\n<p>Amplitude squared? &nbsp;</p>\n<p>log?</p>\n<p>Db?</p>\n<p>&nbsp;</p>",
  "messages": [
    {
      "id": "30254",
      "postDate": "09/05/2013 23:16:29",
      "content": "<p>Based on the reference papers, I assume that the HF data is based on frequency analysis of the voltage transients but it is not clear what units they are in, how they were normalized or even what is being measured.&nbsp;</p>\n<p>Are we looking at:</p>\n<p><span style=\"line-height: 1.4\">Amplitude? &nbsp;</span></p>\n<p>Amplitude squared? &nbsp;</p>\n<p>log?</p>\n<p>Db?</p>\n<p>&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "31391",
      "postDate": "09/23/2013 11:00:40",
      "content": "<p>hi admins, please answer the above question....</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "31422",
      "postDate": "09/23/2013 17:21:52",
      "content": "<p>Hello Noam and Amit,</p>\n<p>The units are log magnitude squared of the complex FFT. In other words, it is magnitude in dB (20*log10(sqrt(re^2 + im^2))). In the data itself, the range [0,255] maps to [0,127] dB.&nbsp;</p>\n<p>You will notice that I have explicitly left out the reference quantity for dB making the interpretation of &quot;true units&quot; arbitrary. This is because from ADC measurement to FFT, the signal goes through proprietary gain and filter functions and even I am not sure how the reference changes. At the end of the day, it is really dBmV. For all practical purposes, I have never had the need to convert it to true units other than when designing the hardware that produced these signals.</p>\n<p>Sidhant</p>\n<p>P.S.: One quick side note, the HF data is not just &quot;voltage transients&quot; but any and all signals from 0-1 Mhz on the power line. The voltage transients (if any) are subset of that data.</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 31391,
      "author_name": "stream1",
      "author_url": "",
      "post_date": "09/23/2013 11:00:40",
      "content": "<p>hi admins, please answer the above question....</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 31422,
      "author_name": "sidhantgupta",
      "author_url": "",
      "post_date": "09/23/2013 17:21:52",
      "content": "<p>Hello Noam and Amit,</p>\n<p>The units are log magnitude squared of the complex FFT. In other words, it is magnitude in dB (20*log10(sqrt(re^2 + im^2))). In the data itself, the range [0,255] maps to [0,127] dB.&nbsp;</p>\n<p>You will notice that I have explicitly left out the reference quantity for dB making the interpretation of &quot;true units&quot; arbitrary. This is because from ADC measurement to FFT, the signal goes through proprietary gain and filter functions and even I am not sure how the reference changes. At the end of the day, it is really dBmV. For all practical purposes, I have never had the need to convert it to true units other than when designing the hardware that produced these signals.</p>\n<p>Sidhant</p>\n<p>P.S.: One quick side note, the HF data is not just &quot;voltage transients&quot; but any and all signals from 0-1 Mhz on the power line. The voltage transients (if any) are subset of that data.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "30254": "",
    "31391": "",
    "31422": ""
  },
  "source": "meta"
}