21 October, 2013

Studies on Morse code recognition

Nice drawing of Morse key by Dutch
radio amateur 
Dick Kraayveld, PA3ALM
(http://www.morsecode.nl/pa3alm.html)
In the early 90's there were some interesting studies performed on Morse code recognition and the effects of pitch frequency, signal to noise ratio, and code speed. They were part of the PhD work of Peter Montnémery who is a medical doctor and also radio amateur SM7CMY. I reread these papers now since there was a discussion on the pitch frequency (CW pitch resolution) just recently on the Elecraft mailing list. This discussion comes up from time to time, as it did for instance in 2006 also (Sidetone questions -- copy speed vs sidetone pitch).

The two first papers are probably the most interesting ones for radio amateurs, so therefore I have posted their abstracts and a key figure from each of them also.
Press figure for larger size

Paper 1: Montnemery, P., Almqvist, B., & Harris, S. (1991). Recognition of Telegraphy Disturbed by Noise at Different S/N-Ratios and Different Telegraphy Speeds A Comparison to Short-Tone Audiometry. Scandinavian Audiology, 20(1), 33-39.

Abstract: "The ability to recognize telegraphy at different S/N ratios and telegraphy speeds was compared for 10 normal-hearing men trained in detecting telegraphy. The ability to recognize telegraphy disturbed by noise decreased with increasing telegraphy speed and reduced S/N ratio. The recognition of telegraphy disturbed by noise seems to follow the relations known from experiments with short-tone audiometry."

The figure above (Fig. 5) shows recognition in % vs signal to noise ratio (SNR) for speeds 13, 25, 40, 63, 82, and 103 characters per minute (CPM) which when divided by five correspond to 2.6, 5, 8, 12.6, 16.4, and 20.6 words per minute (WPM). Note how recognition at the 90 % rate is easiest (works under lowest SNR) at 5 WPM, then 8, then the exception for the slow speed 2.6 WPM, and then gets even harder at speeds 12.6, 16.4, and 20.6 WPM.
Press figure for larger size

Paper 2: Montnemery, P., Almqvist, B., & Harris, S. (1992). Recognition of telegraphy signs at different listening levels and frequencies. Scandinavian Audiology, 21(4), 255-260.

Abstract: "The recognition of telegraphy masked by noise at 40 and 80 signs/min telegraphy speed was studied in 10 normal-hearing subjects at different sound pressure levels (25–85 dB SPL in steps of 5 dB) as well as at different test frequencies (2000, 1000, 800, 630, 500 and 250 Hz). The ability to recognize the signs varied with varying SPL. Recognition for most of the subjects was best at an SPL close to 70 dB. All subjects improved their recognition as the frequency was lowered to 500 Hz, some even at 250 Hz. These facts should be taken into consideration when training telegraphy operators as well as in the construction of radio receivers to permit listening at low frequencies. Furthermore, the critical ratio was calculated at the different test frequencies."

The figure above (Fig. 6) shows recognition rate at different pitch frequencies at a speed of 80 CPM = 16 WPM. There is an improvement as the pitch is lowered and it more or less seems to level off at 500 Hz, except for the lowest SNRs (-13 and -14 dB) which improve at a pitch of 250 Hz. As a side note, the SNR scale must be different from that in the former figure since the values are considerably lower here.

The other papers in the thesis "Signal Detection in Noise, with special reference to telegraphy" are:
The first paper shows how lowering the speed enhances recognition when conditions are poor. Unfortunately only 250 and 500 Hz were tested in the second paper, but it should be clear that a low tone is preferable. If one were to do a follow up study, it would be interesting to zoom in on the frequencies between 250 and 500 Hz for further testing. Since 40 and 80 CPM is only 8 and 16 WPM, it would also be interesting to test at higher speeds.

5 comments:

  1. Hei Sverre,

    A very interesting posting. I will have to make some time to read some more of the subject in your references.

    I noted long ago my preference for a lower pitch when copying higher speed CW. Right around 500 to 600 hz is my sweet spot.

    I like CW contests and operate QRP almost exclusively. Some tricks of the trade I use are to slow down when conditions are noisy (QRN or QRM) and to slightly offset tune the station I am calling so as to change the pitch of signal. I learned many years ago that both techniques can be quite effective under the right condtions.

    cheers es 73, Graham ve3gtc

    ReplyDelete
    Replies
    1. Hei Graham,

      Thanks for your comment. What you are saying fits exactly in with Fig. 6 above. For copying above the 75% level, 500 and 630 Hz come out more or less the same in that sample of 10 operators. At lower recognition levels (lower SNR) a lower pitch seems to be more important..

      And Fig. 5 indicates that one can gain ~2 dB SNR by slowing down from 20 to 12 WPM, just like your experience also says.

      Delete
  2. Nice article.

    A while ago I did some research of my own (some findings at http://fkurz.net/ham/stuff.html?noise) and later also found Peter's dissertation. It is my plan (I have said this for years, but I finally need to do it!) to include a "SNR challenge" into LCWO.net (Learn CW Online), where it would be possible to grade hundreds of users in terms of their ability to copy at low SNRs, with different tone frequencies, etc.

    By the way, additionally to Peter's papers, F5ULS hinted me to another thesis about this subject, "Improving the Morse Intercept Operator's Audio Display" by Jesse M. Washburn (1995), which is freely available on the net.

    ReplyDelete
    Replies
    1. Hi Fabian

      Thanks for the comments. I look forward to the SNR challenge in LCWO! Your links are here:
      DJ1YFK's Ham Radio Stuff
      Improving the Morse Intercept Operator's Audio Display

      Delete
  3. Hi Fabian

    Copying weak CW signals and psychoacoustics has been a passion of mine since I was first licensed (1957). More recently it has taken the form of DXing on the low bands (especially Topband) and SOTA operating (QRP from summits). I've signed up on LCWO and hope you will implement your SNR challenge. BTW here is a related website by AB7E about weak signals:

    http://www.ab7e.com/weak_signal/mdd.html

    Many years ago I discovered 270 Hz was my personal optimum pitch for copying very weak signals on the low bands. This was using a TS-930S which had continuously adjustable pitch. I even found 240 Hz was sometimes useful. I was disappointed that the Elecraft K3's lowest pitch setting is 300 Hz but have learned to "live with it".

    73, Bill W4ZV

    ReplyDelete