JT65-HF - How is "dB" Calculated?
I realize this is an indication of relative signal strength - but relative to what? At first I thought it might be signal to noise ratio. But I never see a positive value (> 0 db), even for local stations. I realize that JT65-HF can successfully decode signals with a negative S/N ratio. However, I don't believe every signal it decodes has a negative S/N ratio. Listening by ear, I know some the tone sequences sound well above 0 dB S/N ratio, and yet they show readings like -6dB. Can someone explain to me what this is really measuring?
Wed Dec 15, 2010 12:36 am
--- In 30MDG@yahoogroups.com, "Don" <dhobson123@...> wrote:
> One more note on the JT65a-HF software and S/N reports. I would like to thank Phil KA1GMN for taking the time to do some tests on S/N on JT65a-HF vs FLDIGI S/N reporting.
> I think Phil and I spent well over an hour of testing using FLDIGI Contestia mode vs JT65a-HF and I turned the antenna until I was receiving a pretty constant -11 S/N on FLDIGI just on the edge of 100% decode to 90% or so and then we switched back and forth to JT65a-HF and noticed JT65a-HF at around -18 S/N for its report. So this was around a 7 difference between the two programs.
> I just noticed that no matter what it seemed that JT65a-HF S/N reports seemed off or too far below S/N on its reporting (I do keep my sound level as close to 0 as possible as the program/software suggests). I have never received anything but a minus -S/N report..even my Ham digital friend about a mile or so away from me a report of -2 S/N was produced on JT65a when he was at 40 over s9. I also noticed that 75% or so of all the JT65a spots/reports on the JT65a-HF log of mine were from -18 to -6 which would indicate to me that if from -14 to -6 those are easily heard and signals seen thus maybe PSK more QSO friendly digital contacts could be made and from -14 to -18 better weak signal digital QSO modes like Olivia or Contestia could be made but that 25% that was from -18 to -28 S/N on JT65a-HF that I noticed for sure JT65a-HF wins out and most likely the only weak signal mode that would produce a contact.
> I'm not technical and not sure I'm expressing my point but it seems to me that a few things are happening and number one that JT65a-HF S/N reporting seems to be reporting much lower S/N numbers than the FLDIGI program but then again I'm not sure how those numbers are produced/calculated (I think Phil had a theory and I will let him state that if he wishes). I also think that those wanting more of a QSO that if signals are heard and seen and at -18 or better that using more QSO friendly modes could be had if that is desired (example might be the South African station that I worked that was -10 to -8 S/N on JT65a-HF, clearly heard signal and seen, strong well above the noise but had never heard him on PSK or other digital modes..then again maybe JT65a is his mode of choice and that is fine too).
> Anyway, just thought I would note this and thanks again to Phil for taking the time to experiment.
> Don kb9umt 30MDG#0001
Wed Dec 15, 2010 5:37 pm
--- In 30MDG@yahoogroups.com, Phil Williams <ka1gmn@...> wrote:
> Sorry about being slow on my response about my thoughts on differences in
> SNR measurement between JT65-HF and Fldigi. My ham operations have been
> curtailed as of late.
> I am going off cuff here, but I think it’s in the right direction.
> First, when an SNR measurement is taken, you also have to take in to account
> the bandwidth at the SNR is made. Signal-to-noise ratio is expressed in dB
> and is ratio of power of the received signal power to the power of the
> noise. Noise is dependent upon bandwidth or really the amount of noise power
> is dependent on bandwidth.
> The next paragraph is based the assumption that the bandpass is perfect w/
> steep skirts and no insertion loss along with the other assumption that the
> distribution of noise power is even.
> If we measure noise power with a bandwidth of 2000 hz and then decease the
> bandwidth to 1000 Hz the change would be 10 * log(1000/2000) or -3 dB. The
> amount of noise power is now 50 percent less @ 1000 Hz then when the measure
> was taken with a bandwidth of 2000Hz. This is the reason why it is
> important to know at what bandwidth the SNR measurement. If the the signal
> level has remained the same then the SNR measured would have improved by 3
> dB. If the SNR @ 2000Hz was -18 dB then the SNR with a 1000 Hz bandwidth
> would be -15 dB.
> I do believe that JT65-HF performs SNR measurements based using a 2000 Hz
> audio bandpass. Also, I think I recall that the SNR is based on the average
> SNR of a JT65A transmission. The bandwidth of a JT65A transmission is
> approximately 175 Hz and this is contrasted against an audio bandpass of
> 2000Hz. Let’s say that we have a JT65A station that we are hearing with an
> SNR of -15 dB based on an audio bandpass of 2000 Hz. Now, say we were able
> to make the same measurement again with a 200 Hz bandwidth - 10
> log(200/2000) = -10 dB. The amount of noise power has decreased by -10 dB or
> a 10 dB improvement in SNR or -5 dB. Note that the value of SNR now is more
> representative of what we see in the more QSO oriented modes like PSK,
> OLIVIA, MFSK, ect.
> A good way to prove may be to have a two stations exchange JT65A
> transmissions, not the SNR, and then adjust add in audio bandpass filtering
> that is 200 Hz wide - just enough to allow the JT65A signal to pass - and
> repeat the SNR measure. I would think that the SNR values would improve.
> As for Fidigi and how it measure SNR for a given mode, I am not sure. I did
> a quick experiment on a OLIVIA 16/500 signal using the methodology described
> in the previous paragraph and did not notice difference in SNR value when I
> had the audio bandpass set to 600 Hz. So maybe the SNR measurement in this
> case is based on a bandwidth of around 600 Hz. Dave could answer this for
> Those are my thoughts about SNR measurements and that one should always take
> in to account the bandwidth at which the SNR measurement was made.
> I could be wrong about all of this. Oh heck, who wants cake?
> philw de ka1gmn
I had the same thoughts and questions thus my first post so maybe this too might help....response from Joe W6CQZ creator of JT65-HF (or join the Google JT65-HF group and ask away!)
Positive sig strength edited as negative
Bert - W0RSB
More options Sep 22 2011, 4:49 pm
Finally encountered a signal with a positive signal strength, an LU station
on 10 meters reading +1db.
Always ready to exercise the infrequently used paths in software, I just had
to double-click it to see what happened.
The Rpt field was edited as -01.
How's it supposed to appear, +01 or just 01?
Or, was the error in recording the +1dB in the received signal strength?
Sep 22 2011, 4:57 pm
On Sep 22, 2011, at 5:49 PM, Bert - W0RSB wrote:
> Or, was the error in recording the +1dB in the received signal strength?
That's probably it.
In the digital world, full scale is generally regarded as one or more successive samples at maximum value, which is 0dBfs.
More options Sep 24 2011, 9:59 am
In the JT65 protocol there is no definition for signals > -1 dB.
Considering it is an EME mode this makes sense.
The primary difficulty is that the protocol itself does not "understand" the
concept of a signal greater than -1 dB and, as such, produces unexpected
result when faced with that situation.
In JT65-HF I explicitly limit the signal report to -1 dB. Even if the
decoder indicates a value > -1 I hard limit it to -1 dB. Why? Beyond what
I said above, the code that calculates the signal strength "breaks" above -1
dB and the resulting value has no meaning as it's essentially random in
nature. In other words, if the decoder returns a value of (+) 3 dB it's
just as likely that the real value is +10 dB or +1 or + whatever dB as it
has overflowed it's computational range maximal value of -1 dB.
73 - W6CQZ
On Thu, Sep 22, 2011 at 6:57 PM,
So then it sounds like I need to do the following to get a ballpark estimate of signal to noise ratio for an individual signal:
1. Add 10dB to the reported value to account for the fact that noise is being measured over the entire 2kHz band whereas a single JT65 signal is only 200Hz wide.
2. Disregard as unreliable any values reported as > -2dB.
I just log whatever signal report I get from JT65 for my sent report and whatever the other station sends me. Never gave it any thought, just in the log, contact was made. the lower the # the stronger the signal, good enough for me
I wondered about the dB ratings too. Amazed that on occasion I pick up a -24 signal and cannot even see it on the waterfall yet get full copy of the message. Seems like -1 is similar to S9 in SSB and -24 is S0?