NAVTEX Survey

I was asked a question recently (without divulging the identity of the questioner) regarding the quality of reception in Shetland of 518kHz NAVTEX from Cullercoats and Malin Head.

The easy answer was to tune to 518kHz for a few hours, run YaND and see what turned up.

Indeed, there was GCC/Cullercoats, right on cue, and nice and strong, perfect decode quality. A little later came EJM/Malin Head. Not as strong, and a few errors, but it was daytime, so that's only to be expected.

Home, in the centre of a web of Navtex transmitters

The difficult question was "is Cullercoats good enough for ships at sea, and should it be publicised as having a range that extends to cover Shetland?"

The difficult answer was "I don't know, but let's see if I can make a rough estimate"

Enter the Perseus and the wonderful YaPS scheduling engine,

I made a schedule to capture 5 minutes of recorded RF, at the start of each of the published timeslots for Cullercoats and Malin Head, as well as Niton and Portpatrick for a full 24 hour period. Not satisfied with 30 recordings to pore over, I next repeated the exercise recording 5 minutes of RF, for 24 hours, on all the timeslots of Torshavn, Orlandet and Rogaland. Even this wasn't enough, and I did another 24 hours worth of recordings, 5 minutes each on the timeslots of Cullercoats, Malin Head, Portpatrick, Niton, Torshavn, Orlandet and Rogaland. I lost count of the number of recordings I now had collected, I think it's 96...

All well and good checking the "signal strength" in terms of S-Meter reading, or dBm (the Perseus has a good, linear & accurate S-meter) but what about noise level, and S/N ratio? What is the E-Field Strength? Can I calculate it from the received signal power/voltage?

The Wellbrook 1530S+, I was informed by Andy Ikin, has an Antenna Factor of 1dB/m.

As an example,  for a measured signal of -80dBm the antenna is producing a terminated voltage of +27dBuV and it generates this voltage from a field strength of +28dBuV/m. (Which is 24.1uV/m)

An OpenOfficeOrg spreadsheet was created to do the number crunching (I also had to account for the splitter loss and wanted to record noise power in a 300Hz bandwidth as well as S/N ratio)

Here's a pdf file showing the XLS sheet, split into A4 pages, making it tricky to read. I can provide the original sheet too....

http://dl.dropbox.com/u/3551430/navtex_signals.pdf

The sensitivity of Navtex receivers is generally quoted as "2uV emf". That sounds pretty good.at -107dBm. Should be okay? With an antenna factor of 10dB (small, inefficient antenna) you'd only need a field strength of +10dBuV/m. The ITU-R curves for a sea path would indicate that this field strength would be achieved at a distance of well over 1000km (from a 1kW Navtex transmitter). No problem for Cullercoats up in the waters around Shetland then? I have been calculating E-field strengths of  up to 32dBuV/m for Culleroats, so this should be plenty signal for a receiver with a sensitivity of 2uV, even fed from a small antenna.

The Path between GCC and GM4SLV = all sea apart from the final 14kM

The Path between EJM and GM4SLV = right across Caithness and Orkney, not  what  you'd call a sea path!

The Path between GPK and GM4SLV - half and half.

NOISE?

But, my QTH is "notoriously" quiet. Just because I can receive something, does it follow that a vessel at sea, with a small Navetex antenna jammed somewhere about it's superstructure, will do so? 

The ITU quote that for a representative vessel the "noise at 3MHz will be -142dBW" and this leads to an external noise factor of 83dB (...on average... in spring... at 1200UTC.... at 60 degrees North...)

The same ITU document gives a value for the noise factor of a "quiet rural" location as 62dB.

That's a 21dB difference. Will the noise floor of a ship really be 21dB above my "super low" noise? I'd imagine that's quite likely, if my previous tests looking at Perseus receivers in less than ideal locations is any hint.

It that's the case, then the "2uV" sensitivity of the receiver is a bit moot.

The limiting factor is the noise. As ever.

Anyway, here's a link to my little report, tabulating and cogitating on what I've measured.

http://dl.dropbox.com/u/3551430/Navtex%20Investigation.pdf

My guess is that Cullercoats may, indeed, be a good signal here in my low-noise backwater. But I'm uneasy at the idea that it's going to be a good signal on a ship crammed full of radars and PCs and fish finders and all manner of other electronic hash....

Wonder if I can get a grant to take a trip with my Perseus on a cruise liner around the Northern Oceans doing field strength surveys?

5MHz Sked wanted = early morning tests

I wonder if anyone reading this would be interested in a regular daily sked on 5MHz datamodes?

I'm usually in the shack between 06:30 and 07:30 local time (05:30 - 06:30 UTC) and at this time the F2 Critical Frequency is just starting to rise. It would be interesting to run a series of tests at the same time each day to observe the changes in band conditions. I'd also like to see how different data modes cope with the changing ionosphere.


I worked Keith G6NHY at 05:40UTC this morning and saw large QSB, but his signal was very good, with S/N reported between +10dB and +17dB. Keith was using 30W. Initially he reported only 75% copy of my 20W but within a few minutes, and an increase here to 40W, the signal rose to 100% copy.

The Chilton Ionogram showed the F2 Critical Freq to be changing between 3.4MHz and 3.8MHz and the MUF(1000km) to be around 5.4MHz, with MUF(800km) at 5.0MHz. The values changed each time the data refreshed, indicating a non-stable ionosphere?

Keith is approx 800km from my QTH, so we were just on the cusp of band opening.

A comparison could also be done with 80m. I don't suppose 40m would support any inter-UK communications (even longer paths of 800-1000km) at that time, but it could be checked too.

Any takers? 10 minutes for a couple of overs each and a chance to build up some experimental data?

HF Noise Floor

Prompted by discussion of non-reciprocal S/N reporting during a daytime Olivia QSO on 5MHz between myself and G4VLC - both of us in "low noise rural" locations, I thought I'd look at noise levels of a few Perseus receivers around Europe and see if it was possible to quantify what "low noise" means. I've had a few people commenting on the use of my Perseus online, as being a nice "low noise" system which helps them enjoy HF, because their own location is very noisy.

For the path between Peter and myself, I usually report a better S/N ratio of Peter's signal than he does of mine. Is the difference due to a higher noise floor at his location?

Using my Perseus, fed by a Wellbrook ALA1530S+, after a 6dB hybrid coupler that splits the signal to other receivers, I measured the noise power as reported by the Perseus S-Meter, and also as reported by the Marker function,

The S-Meter shows the amplitude (in dBm) of the signal contained in the bandwidth set for the demodulator in use. I set this to 2.5kHz.

The Marker level shows the amplitude (in dBm) of the signal contained in the Resolution Bandwidth of the FFT, which is determined by the sample rate and the Spectrum analyzer span. I used 125kS/s, for a span of 100kHz and a RBW of 122Hz.

I measured the noise floor using both methods and normalized the results to a 1Hz bandwidth, by applying a correction in dB of 10Log(RBW)

For the 2.5kHz filter this gives a correction of 34dB

For the 122Hz RBW Marker the correction is 21dB

I first measured the noise floor of the RX with a 50 ohm termination on the antenna input.

3.5MHz  = -128dBm / 122Hz and -117dBm / 2.5kHz

Normalizing these figures to 1Hz bandwidth gives:

For the 122Hz RBW Marker
-128-21 = -149dBm/Hz

For the 2.5kHz  RBW S-Meter
-117 - 34 = -151dBm/Hz

These agree well enough with each other and the average is -150dBm which I take to represent the MDS of the Perseus on 3.5MHz

I repeated the measurements at 5.3MHz, 7MHz and 10.15MHz and they all agreed within 1dB with this value.


I then repeated the measurements with the ALA1530S+ loop attached and the resulting dbm/Hz measurements are:

3.5MHz = -146dBm/Hz
5.3MHz = -147dBm/Hz
7MHz = -148dBm/Hz
10.15MHz = -148dBm/Hz

Not accounting for the loss in the 6dB splitter, the noise floor presented by the antenna is roughly 3dB above the MDS of the receiver, so the antenna is "sensitive enough".

I also repeated the measurements using my Vertical antenna that I use for TX, fed via an Auto-ATU.. I tuned the ATU to each band in turn before making the measurements on the Perseus.

The results:

3.5MHz = -144dBm/Hz
5.3MHz = -148dBm/Hz
7MHz = -146dBm/Hz
10.15MHz = -146dBm/Hz

The vertical is no more noisy than the supposedly "low noise" ALA1530 loop. My guess is that the noise floor I see from the loop is actually the internal noise generated by the loop's pre-amplifier, and the external band noise is actually below this level. Using the vertical I'm seeing the true external RF band noise.

The question is now "Is this a low-noise environment or do other people have the same levels of external band noise?"

This is where the remote Perseus Servers come in very useful.

I repeated the same measurements on a selection of Perseus servers around Europe.

The screenshots were taken several hours after making the measurements....

PI4TUE has a very low noise system too.

Here are the results, presented in the same way 

3.5MHz = -139dBm/Hz

5.3MHz = -148dBm/Hz

7MHz = -144dBm/Hz

10.15MHz = -152dBm/Hz

DJ6JZ has a higher noise level, as can be seen on the screen shot above, and the results below.

3.5MHz = -116dBm/Hz

5.3MHz = -112dBm/Hz

7MHz = -116dBm/Hz

10.15MHz = -120dBm/Hz

These are approx 30dB higher then my own or PI4TUE's figures and probably show a "normal" urban level of man-made noise, that many people have to struggle with.

OE6GFD has a good setup too

3.5MHz = -137dBm/Hz
5.3MHz = -143dBm/Hz
7MHz = -145dBm/Hz
10.15MHz = -145dBm/Hz


I also collected the same data from:

RZ3DVP

3.5MHz = -140dBm/Hz
5.3MHz = -141dBm/Hz
7MHz = -138dBm/Hz
10.15MHz = -148dBm/Hz

PA3CRP

3.5MHz = -132dBm/Hz
5.3MHz = -140dBm/Hz
7MHz = -133dBm/Hz
10.15MHz = -144dBm/Hz

G6WPJ

3.5MHz = -138dBm/Hz
5.3MHz = -139dBm/Hz
7MHz = -141dBm/Hz
10.15MHz = -144dBm/Hz

G3XVL

3.5MHz = -142dBm/Hz
5.3MHz = -143dBm/Hz
7MHz = -145dBm/Hz
10.15MHz = -148dBm/Hz

Of course just because the noise floor is low doesn't mean the S/N will be high, if the antenna is insensitive, so I don't really know the best way to interpret these results. Certainly if the noise is very high then it must mask weaker signals.

Fldigi Autoresponder in action

I thought I'd try to test a groundwave path from Lerwick town centre, in a noisy office environment, using an FT-817 with 2W RF output to an ATX Walkabout antenna, which on 5MHz is more loading coil than antenna!

The path is shown by Google Earth

There is a trailing ground/counterpoise random length of wire on the floor, and the FT817 is simply sitting on a table in the mess room at the work's QTH.

The video shows a couple of different autoresponder actions - a "ping" where the simple text :

RYRYRYRYRYRYRYRYRY
R TEST DE GM4SLV Y
RYRYRYRYRYRYRYRYRY

comes flying back from the far-end.

Then a request is sent to obtain confirmation of the power level in use at the far end and the answer comes back as "10 watts"

My camera isn't great, especially for sound, so I apologize...

The distance is approx 20km, and for a small (tiny) TX antenna, situated indoors, I think it works reasonably well!

John

5MHz Summer project

Catch Up & 5MHz Digital Mania

I've been a bit quiet on the blog for a while, but only because I've been tinkering away with a few projects and never seem to have found the time to write them up here...

It all started with working G4VLC on 5MHz Olivia. It appears that the channel centred on 3.568kHz (the UK's channel "FK") is where the digimodes are to be found on 5MHz. I've had a 5MHz NoV since they were first issued and have done some work on the band in CW and SSB, but very little using digi-modes. I never seemed to stumble across another digi-mode signal. My NoV application actually specified investigations into the use of digital modes for inter-UK communications to and from the remote Shetland Islands, so I was glad at last to find where the action was centred.

Next came a few weeks of activity based around 5MHz.

1) Re-instate the GB3RAL monitoring

This now runs on an AR7030 receiver, fed by a Wellbrook ALA1530S+ loop, in the same way is has done for several years until I stopped it last year sometime. The data will once more be sent to the RSGB's 5MHz database and also it's visible in real time graph form at G4IRX's website:

http://g4irx.nowindows.net/fivemegs/comparison.php

By carefully rotating the Wellbrook loop, while observing my TX signal on 5367kHz I was able to reduce the level of my own in-band 5MHz transmissions such that there is no measurable increase in the noise floor (or RX desensitization) on the GB3RAL beacon monitor on 5290kHz while I transmit up to 100W on 5367kHz. I can now operated on 5MHz and run beacon monitoring at the same time, with no degradation of the measurements.

The GB3RAL (and GB3WES/GB3ORK) plots give a good snapshot into current 5MHz conditions, and it's interesting to watch the beacon signal levels while also having QSOs on the same band.

2) Build a permanently available Linux machine running Fldigi connected to the IC718

I messed around for a while swapping between my Win XP PC (a mini-ITX based dual core Atom box) and an old Dell PC running Debian. I also tried running Debian on the main PC, but then found I could no longer use my Perseus RX. After much work trying to get any of the Linux ways of using a Perseus working I gave up.

I even tried running WinXP in a VMWare virtual machine, but the performance was too sluggish to use the Perseus properly.

Eventually I settled on a separate Linux Debian (Squeeze, upgraded to Wheezy) PC running "headless" and using VNC for graphical access, to run Fldigi controlling the IC718, leaving my Win XP machine to do menial duties and to give access to the Perseus receiver.

Here's a shot of a VNC session running Fldgi

I can now use HF Digimodes from any PC across my LAN, or indeed from anywhere on the internet by using OpenVPN to make a VPN connection to my LAN first, and then VNC to connect to the Linux Radio PC.

3) Fldig autoresponder


Peter G4VLC told me that Fldigi has some powerful features that can be used to detect certain text/callsigns etc. and then run external programs. He uses these "notifications" to trigger alerts via SMS message whenever such things as a CQ call etc. are received.

I decided to have a go, but instead of SMS messaging I decided to use email as the alerting method.

Simple Linux shell scripts can be run whenever certain conditions are met on decoded text. Using Postfix and the bsd-mailx command line mailer

I now receive an email if:

a)  "CQ CQ CQ" is received
b) "GM4SLV de XXXXXX" is received


The emails tell me what mode, and which frequency is being used. Fldigi's RSID is used to ensure mode-detection and automatic switching takes place. Having an Android Smartphone means I can get these alert emails in almost real-time via Gmail.

Not satisfied by this minor success....following another suggestion by G4VLC, I took it a big step further.

Firstly, for purely local, on screen, "in a real QSO" use, I came up with shell scripts that can change the audio drive level from the soundcard to the tranceiver in a controlled manner. By careful testing it was possible to create the necessary shell commands to control the "alsa mixer" and to set the audio drive so that TX power could be adjusted to any pre-set value between 100mW and 100W.

These shell scripts driving alsa mixer  became the basis for some on screen macro-buttons, making easy power level changes possible, without needing to alter the transceiver's own power controls.

Then I thought of using the "notification" system to automatically change the TX power on receipt of a "trigger command" from another station (or my own  /P station) over the air.

Much tinkering ensued....

I now have a fully fledged Auto-repsonder/Remote control facility via Fldigi's notification system

In short, by sending the correct text strings over the air my Fldigi autoresponder can:

i) Change TX power setting to any of:
100mW, 500mW, 1W, 5W, 10W, 50W and send a revertive message confirming the new power level.

ii) Report the current power level

iii) Reply to a "ping"  - with a short test message

iv) Send a Signal report of S/N etc.


v) Send a 10s plain carrier

vi) Send 10s of IDLE signal, on whichever modem is in use.

vii) Change bands between:
5MHz, 7MHz, 10.1MHz and 14MHz, auto tune the ATU and standby

viii) Reset from whichever band/mode/power combination is currently in use to the "Home" settings:

5MHz/ 10W / Olivia 16-500 modem

ix) Send a "help file" detailing the commands needed for each function.

4) Use of Autoresponder for remote testing


I've put together a portable HF station for Digimodes, consisting of an FT817, a Samsung NC10 netbook, interface, antennas (random wire, dipole, ATX Walkabout), manual ATU, SWR bridge, 8m fishing pole, SLA Gell Cells, etc. etc.

I have been able to use the home station's autoresponder to carry out multiple band groundwave/NVIS testing from /P locations around Shetland - very useful when you can't raise anyone off-island to do some tests with and you want to get some value from carting the gear around and setting up a /P station.

The system is there, and should be running 24/7, if other people wish to try some of the functionality. It's a grey area I'm afraid - remote control/operation - and despite reading the 5MHz NoV carefully I can't see the basic permissions of the main licence (Para 10) regarding remote control being proscribed on 5MHz. It isn't legal though for remote control/operation to be made generally available to anyone other than the licensee, and operation of one's station by "remote control" must be done in such a way as to prevent this - hence the use of specific (hem hem) secret "trigger commands" for my autoresponder.

Codes and ciphers aren't allowed either, so the triggering commands are in plain text.

I'll say no more..... if you want the details, you know who to ask...

Where to find me on 60m
I tend to be tuned to dial frequency 5366kHz with an audio carrier/centre frequency of 1.5kHz putting the RF signal centre at 5367.5kHz (the channel boundaries are 5366.5kHz to 5369.5kHz)

5) Antenna modifications


My 10m vertical performs quite well for inter-UK operation on 5MHz. For a typical  range of around 900km (the South Midlands/SW/SE England/ South Wales) via F2 layer the required take off angle is around 30 degrees. This isn't therefore NVIS and a vertical, with a low angle of radiation, is probably better than the usually accepted 5MHz low NVIS Dipole or Inverted V.

I got to wondering just how far the groundwave range (again requiring a vertical antenna) was on HF bands generally, and at what point NVIS became important and groundwave decayed to nill.

If I want to reliably cover the whole of Shetland out to a distance of 100km, can that be done solely on groundwave, or would I need some horizontal, high angle radiation and a critical frequency at or above 5MHz?

With a vertical TX antenna I might not be able to test the NVIS path to more distant locations in Shetland.

EZNEC modelling showed that by simply adding a 5-7m "top loading" wire to the 10m vertical to create an Inverted-L I would retain the mainly vertically polarized, low angle pattern of the original antenna, but fill in the overhead null with some horizontally polarized radiation.

The best of both worlds?

I added the top wire and so far it seems to still be working well for the longer range inter-UK traffic. I haven't had an opportunity to try any NVIS paths, I need to get 50 miles or more from home, to allow the groundwave to attenuate, and see what signals I'm able to receive from NVIS range - the autoresponder will come in handy for these tests, since there aren't any other active 5MHz operators in Shetland that I know of.
Finding a suitable /P location, however is tricky too. Shetland is long, 100 miles end to end, but I'm located in the centre/west of the mainland, to get a sufficient distance away I'm probably going to have to go up to Unst or Fetlar.

Now I need the snow/rain/gales to abate and to go /P again.

6) Portable / Man-pack operation


Thoughts turn to the summer months and chances to operate out-doors away from home. I have an FT817 - which does a good 3W on digital modes. I'd like an ex-military/commercial manpack transceiver, for the ruggedness and the option of power up to 20-30w. Finances won't stretch to such frivolous purchases.

Time to build an FT817 medium power portable station.

An HF Linear Amplifier with an output power of 20W is next on the list of hardware projects.

Phew, so much to do, so little time.

Digital Mode musings

This week I've been playing with JT65A on 160m. A new (to me) implementation of JT65A is now available that provides just the right functionality needed for HF contacts - it's called "JT65-HF" and is available here:

http://sourceforge.net/projects/jt65-hf/

I installed it, but had some difficulties with getting it to work. My PC has 2 soundcards - the onboard Realtek card and an old Soundblaster 16 PCI card.

The Soundblaster 16 PCI card is the one used for Datamodes and works perfectly with MultiPSK, WSPR, Stream etc.

 JT65-HF however would not receive anything when using this card. It's inbuilt automatic sample rate correction very quickly settled on a correction factor of around 0.3 for RX and 1.00 for TX, but never produced a decoded signal despite the same on-air signals decoding perfectly well in MultiPSK and WSJT9.

I swapped to use the RealTek card and this started to produce successful decodes.

I even managed a QSO with G3WKW on 160m.

Then I started investigating the "Reverse Beacon Network" where all your received spots are sent. This led to   http://hamspots.net/

JT65-HF seemed to work nicely, good for a simple 2-way QSO, but obviously very limited in what could be exchanged between stations, and Hamspots.net showed who & where your signal was getting to, even if no-one answered your CQ calls.

The next problem was that JT65-HF refused to "talk" to Ham Radio Deluxe, despite both programs being configured correctly. This meant that the frequency had to be entered mannually, so that any reported spots sent to the RBN and PSKReporter were labelled correctly.

Eventually I updated HRD to version 5 and they started to work together.

All good so far.


Then, while calling CQ I monitored myself on another RX and heard my tones very distorted and breaking up! I wondered it was RF feedback, so killed the TX and plugged a speaker into the soundcard output and tried again. Still distorted & breaking up. I restarted the software and normal service was resumed. For a period, but I could still hear the occasional very short break in the audio, and after a while the distortion & noise returned. No other Datamode software has this problem, with either of the 2 soundcards, so I guess it's a bug in JT65-HF.

I stopped using  JT65-HF and operated the same mode in MultiPSK very successfully, back on the Soundblaster card.

It's a shame that there's a problem with JT65-HF, because it integrates nicely with the Reverse Beacon Network run by its author, as well as sending spots to PSKReporter. I find PSK Reporter to be slow to update, and must regularly get overloaded, so the bespoke JT65 RBN is better for real time reporting.

I may not operate JT65 very often, but it's another tool in the armoury for getting a 2-way contact in difficult circumstances, but I think I'll use MultiPSK instead, as it seems to work much better with my hardware.

UPDATE : Sunday afternoon ......

I tried updating the drivers for the old SB16 card, and JT65-HF was then happy with its sample rate, and started to work okay on RX! I still wasn't happy about the odd bouts of distortion on TX that I'd seen earlier, and didn't know if this change had also cured these problems. During testing into dummy load, using the SB16 card with its new drivers, I noticed the power varying markedly on certain tones of the JT65 signal. Further investigation using Spectrum Lab showed that the SB16 card had a very odd output spectrum - with deep nulls in its response every 250Hz or so, which is why some JT65 tones were very low level. I generated an Olivia 32-1k signal and used Spectrum Lab on another PC to observe the output spectrum and there was a real problem with it, with -30dB deep nulls. This problem wasn't present using the on-board RealTek card. Something is amiss with the sound setup on this PC. I removed the old SB16 PCI card and inserted a cheap generic PCI card, and after intalling its drivers I repeated the same tests.

JT65-HF was now happy with the new card, as were MultiPSK and WSPR. What's more, after a considerable period of dummy load testing I failed to witness the JT65-HF output distortion problem using either of the soundcards.

Hopefully all is now resolved and now I can use any software, on either soundcard, for datamodes.

Digital Modes on 5MHz

I've had a few QSOs with Peter, G4VLC, over the last couple of days on the 5MHz channel  FK, centred on 5.3680. There's often a multi-tone data signal at the upper part of the channel, but the lower kHz or so has room for our signals, and the channel is largely unusable for SSB, at least in the evenings when the foreign data signal fades up.

Last night we had a QSO using Olivia in 16-500Hz mode, and although initially we were using low powers, around 5-10 Watts, as darkness fell we had to gradually increase the power until we were using 100 Watts. At this power level, using a multi-FSK signal, the duty-cycle is high, and overs are usually long, given the data-rate of Olivia. My old IC-718 coped very well, and the CG3000 Auto-ATU also showed no signs of distress after an hour or so.

Today I wondered, a) how well specified is the transceiver for this sort of high duty operation, and b) what limits should I impose on TX power for modes that require linearity (eg PSK31). On the normal "ham bands" it is usually bad to use too much power for PSK31, partly due to concerns of spectral broadening in case of poor IMD, but mainly due to concerns of receiver overload & AGC action when many stations are operating in a small slice of spectrum. These concerns are not so important on 5MHz, if the channel is mainly empty, and signals are weak - in these circumstances it is appropriate to increase power to a level that allows the QSO, until the acceptable level of IMD products is reached.

Testing into 50 ohm load I adjusted the audio drive from the PC to give 100W of single tone with the ALC bargraph just about on 1 "pixel". The IC-718 is well specified in power handling terms, the PA stage can produce up to 200W if adjusted internally, so at 100W there's a fair headroom left.

At this power level I then switched to PSK31 mode and using my Perseus SDR I observed the IPs.

The 2 tones of an idling PSK31 signal were seen to be -6dB below the level of the 100W single tone, indicating the PEP was still 100W. The 3rd order IPs were measured to be -26.1dB below either tone, or -32.1dB below PEP. Even at 100W the IC718 is "reasonably" clean, but this is probably still a little too high a level of IMD to use on 20m or 40m when the PSK31 section is very crowded. It's still much better than many of the signals that can be observed off-air though!.

100W PEP PSK31 signal

At 50W PEP the 3rd order IPs were -30.9dB below either tone, or -36.9dB below PEP. This seems perfectly acceptable.

50W PEP PSK31 signal

At 40W PEP the IPs were -32.5dB below either tone, -38.5dB below PEP

40W PEP PSK31 signal

20W PEP the IP level was -34.1dB below the tones, or -40.1dB below PEP.

20W PEP PSK31 signal

Bearing in mind that in general the S/N ratio at a distant station is unlikely to be much more than 30dB, the higher order IPs will be below the noise floor of most stations on the band. The only significant ones are the 3rd order products at higher than -30dBc.

I'd be happy to use the IC718 at 50W PEP, and on a quiet band, and away from the band edges, even 100W might be acceptable, if necessary.

For the question of duty cycle I transmitted an idling MFSK16 signal at 100W into test load and using an infrared thermometer I monitored the temperature of the switch-mode PSU and of the rear panel of the IC-718 transceiver. A sequence of 10 minute transmissions, with 5 minute receive periods, was done for an hour. The PSU case temperature rose from 23.6C to a maximum of 32.2C at which point the internal fan came on and the temperature fell to 28C over the next minute. The IC718 rear panel temperature rose from 26.9C to a maximum of 40C, and stayed reasonably steady from then onwards.

I'd be happy to run in this way for long periods, since the IC718 is really a re-badged IC-78 which is sold as a rugged HF radio to Commercial users.

I also did some plots of the spectral occupancy of various other modes, with a wider span than used for the PSK31 plots.

MFSK16 at 100W

Olivia 16-500 at 100W

DominoEX-11 at 100W