Re: [wsjtgroup] An observed Sig-Rep I made btw JT65 & JT9
Rudy, you are right, JT65 sounds way cooler and more musical than JT9.
Please read Joe Taylor's documentation for JT9 and for JT65. The occupied spectrum has
nothing to do with the noise bandwidth that affects the sensitivity. Sensitivity is determined
by tone duration, coding strength, and decoder algorithm. Joe Taylor explains it pretty well
in his documents.
I've designed systems with as much 3 GHz occupied bandwidth (UWB systems) that had
equivalent noise bandwidths of just a few kilohertz. Same is true for CDMA phone technology.
Joe Taylor's documentation states that JT9 threshold sensitivity is -27 dB SNR while
JT65 is -24 dB SNR both in a reference 2500 Hz BW. So the difference between the two
is about 3 dB, 2 dB of which is due to tone duration. Note that there is a "deep search" decoder
algorithm that can improve the SNR by another 3 or 4 dB, but that is a decoding algorithm improvement.
On 9/13/2013 12:37 PM, Rudy Benner wrote:The problem with JT9 is that it does not sound so cool as JT65, the almighty cool factor.r
While 2dB applies to the tone duration, the over all bandwidth
difference between a JT65A signal and a JT9-1 signal is more than
10 to 1 (~175 Hz vs. ~16 Hz). That would imply a greater high
bound on the difference in sensitivity - somewhere in the 11 dB
I'm sure the real difference is somewhere between 2 dB (individual
tone bandwidth) and 11 dB (occupied spectrum) ... 4 to 6 dB seems
to be consistent with observed differences on HF paths.
... Joe, W4TV
On 9/13/2013 11:39 AM, Kai wrote:
> Correction ... "JT9 tones are just 0.58 seconds in duration"
> Additionally Joe Taylor states that JT9 threshold sensitivity is -27 dB
> SNR, while JT65 is -24 dN SNR both in reference 2500 Hz BW.
> -Kai, KE4PT
> On 9/13/2013 11:04 AM, Kai wrote:
>> Hi Bob,
>> I took a peek at Joe Taylor's WSJT user's guide. JT9 tones are just
>> 0.58 Hz in duration, so the noise bandwidth is 1/0.58 = 1.7 Hz BW.
>> JT65 tones are 0.372 s in duration so the noise bandwidth is 2.7 Hz.
>> So just on the basis of noise bandwidth JT9 would operate at
>> 20log(2.7/1.7) = 2 dB better S/N ratio. All straight out of Joe
>> Taylor's documentation.
>> Improvements in the decoding software increase that even more.
>> I think that your observations are valuable confirmation of the
>> improved S/N performance of JT9 relative to JT65.
>> Kai, KE4PT
>> On 9/13/2013 8:26 AM, KD7YZ Bob wrote:
>>> Hey WSJT Group:
>>> Just an FYI (or FWIW) I just noted.
>>> On the morning 30m path to VK3AMA, I saw a report of -18 when I was
>>> on JT65.
>>> 4m later I see a report of -14 when I'd just switched to JT9.
>>> My impressions were that JT9, for me, was more sensitive. Obviously this
>>> is not even close to a scientific experiment, however.
>>> Still though, pretty neat on 1 watt with a ladder-line-fed 270-foot
>>> dipole at 80 feet up.
- On the relative sensitivities of JT9 and JT65:
JT65 and JT9 both use MFSK (multi-tone frequency shift keying).
Both modes use one tone (the lowest one) for time and frequency
synchronization: JT65 for 63/126 = 50% of the symbols (tone intervals),
JT9 for 16/85 = 18.8% of the symbols.
All remaining symbols are used to carry message information. If other
things were equal, this would give JT9 an advantage of
Other things aren't equal, though. 64-FSK modulation (as used for the
information symbols in JT65) is more efficient than the 8-FSK modulation
used for information symbols in JT9. JT65 uses a Reed Solomon code with
rate 12/63; JT9 uses a long-constraint convolutional code with K=32,
rate=1/2, and a zero tail. Both decoders use soft decisions, but the
algorithms are entirely different. Both codes are "strong", which means
that false decodes are extremely rare.
Tests with thousands of simulated transmissions over an AWGN (additive
white gaussian noise) channel show that JT9 had an advantage of 1-2 dB
over JT65. For example: on the AWGN channel JT9 decodes correctly 33%
of the time at S/N = -26 dB; JT65 decodes correctly 14% of the time at
-25 dB, 70% of the time at -24 dB.
Real-world situations involve QSB and other propagation-induced effects
that can affect the probability of correct decodes, even for signals
with the same average S/N. The low-rate JT65 code brings some advantage
in the presence of deep QSB, and the wider tone spacing of JT65 offers
slightly greater immunity to Doppler spread, for example that seen on
some trans-auroral paths.
Another thing to keep in mind, if you care about such details: every
value of S/N reported by the decoders necessarily has an associated
measurement uncertainty. In the best circumstances (well separated
signals, no QRM or QRN, flat receiver passband, etc.) the "one-sigma"
uncertainty is around 1 dB. In many other circumstances the uncertainty
is larger. JT65 was designed for EME, where true S/N values are never
greater than 0 (as usual, in a 2500 Hz reference bandwidth). At HF,
signals are sometimes much stronger. JT65 reports never get bigger than
-1 dB, no matter how strong a signal may be. JT9 is OK (and reasonably
linear, in good circumstances) up to +49 dB.
The bottom line is that the two modes have comparable sensitivity.
Under benign propagation conditions JT9 has a small (1-2 dB) advantage
over JT65, and of course it uses less than 1/10 the bandwidth.
To be sure, JT65 sounds "way cool" but JT9 does not. The reason is that
the effective Q of filters in your ear is not high enough for you to
distinguish the JT9 tones from one another. Software filters in the JT9
decoder have Q around 6000. JT9 sounds "way cool" to the computer.
-- 73, Joe, K1JT
- Reception of your 1watt sig was helped by the 30M 3 element yagi beaming
SP NA used here.
de Laurie VK3AMA
On 13/09/2013 10:26 PM, KD7YZ Bob wrote:
> Just an FYI (or FWIW) I just noted.
> On the morning 30m path to VK3AMA, I saw a report of -18 when I was on JT65.
> 4m later I see a report of -14 when I'd just switched to JT9.
> My impressions were that JT9, for me, was more sensitive. Obviously this
> is not even close to a scientific experiment, however.
> Still though, pretty neat on 1 watt with a ladder-line-fed 270-foot
> dipole at 80 feet up.
- Joe Taylor <joe@...> wrote:
> The bottom line is that the two modes have comparable sensitivity. UnderI recently changed my rig. The new one seems to drift a few cycles with heat
> benign propagation conditions JT9 has a small (1-2 dB) advantage over
> JT65, and of course it uses less than 1/10 the bandwidth.
over a 46/50 sec transmission by about 3 hz (at 10MHz, other bands pro
rata)whereas the old one didn't.
Decodes in JT65 seem unaffected but JT9 seems to suffer particularly on the
higher HF bands. Is this due to a lower immmunity to frequency drift for
- Hi Brian,
JT65A uses detection bandwidths 2.69 Hz, and JT9 uses 1.73 Hz. So yes,
JT9 will have somewhat greater difficulty with frequency drift.
The decoders for both JT65 and JT9 include "AFC" algorithms that try to
detect and remove moderate amounts of frequency drift. The AFC routine
in JT65 has been tested and optimized more thoroughly, and because of
the greater fraction of signal energy put into the Sync tone, it
probably works better. It may be possible to improve JT9's handling of
drifting signals, somewhat.
-- Joe, K1JT
On 9/13/2013 6:39 PM, Brian Duffell wrote:
> Joe Taylor<joe@...> wrote:
>> The bottom line is that the two modes have comparable sensitivity. Under
>> benign propagation conditions JT9 has a small (1-2 dB) advantage over
>> JT65, and of course it uses less than 1/10 the bandwidth.
> I recently changed my rig. The new one seems to drift a few cycles with heat
> over a 46/50 sec transmission by about 3 hz (at 10MHz, other bands pro
> rata)whereas the old one didn't.
> Decodes in JT65 seem unaffected but JT9 seems to suffer particularly on the
> higher HF bands. Is this due to a lower immmunity to frequency drift for
- I have had 5 pre-arranged DX QSO's on 15 meters using JT9 followed
immediately by JT65A. Power output was always 1 watt to a 288 ft long
horizontal loop up at 35 feet. Rig was an Icom IC-7600. At these times radio
wave propagation was pretty stable.
On every QSO JT9 beat JT65A by approximately 3 db. There were instances
where the JT9 signal got through but the JT65A signal did not.
73 & GUD DX,
Thomas F. Giella W4HM
Lakeland, FL, USA
PODXS 070 #349
FELD HELD #141
30MDG # 691
W4HM Amateur Radio & SWL Autobiography: http://www.w4hm.org