After a few years of Olivia MFSK regular use, I observed a fair number of stations who are unable to transmit good and clean signals. I have also seen users struggling to squeeze performance out of Olivia or receive signals in the most challenging conditions.
However, seasoned Olivia users know that it is possible to extract legible text in almost impossible conditions, when the other station’s signal is buried under broadband radar interference or amid strong RTTY signals during contests. Olivia users can also be heard talking (or chatting) about “ghost signals”, so what is this all about and how can it be achieved?
For those who are not familiar with Olivia MFSK, commonly referred to as Olivia, this digital mode comes in various combinations of bandwidth and number of tones. The most common combinations you find these days (October 2020) are 8/250, 16/500, 32/1000. Broadly speaking, narrowband combinations such as 8/250 are the ones that can be decoded when the signal is very low, say down to about -13dB s/n, at the cost of speed; wideband combinations such as 32/1000 are somewhat less resilient in weak signal conditions but are faster and can be incredibly robust in the presence of narrowband interference.
Below is a surely incomplete list of issues and tips that will help Olivia users to increase their chances of success.
SOFTWARE OR INTERFACE SETTINGS
It all starts from here. Your choice of software and the way you adjust it can make the difference in your ability to put out a good signal that can still be decoded in challenging conditions, and in your ability to receive signals deep in the noise or amid interference.
When transmitting, you need to ensure that you are not sending a distorted signal from your computer to the radio and that you are not overloading your transmitter with your audio signal. Although Olivia isn’t too badly affected by less than optimal linearity, I have seen signals clearly overwhelming the transmitter. Unfortunately those signals were fairly strong and I had no chance of testing how they could be decoded at -12dB s/n but, by judging from the way they looked in the waterfall and by how they sounded in the earphone, I suspect that decoding those signal successfully in challenging conditions would be, to some extent, not optimal. How to achieve optimal settings depends on what radio, interface (if any), PC operating system and digital modes software, you are using. My strong advice here is to read carefully the software manual to learn how the audio interface (regardless of whether this is embedded in your radio or external) has to be set to minimise distortion and to avoid overloading the transmitter. If you use a Linux operating system, you will be able to see when you start adding some pre-amplification to the PC audio output (note that this output is the audio input to your radio, which is then transmitted as a modulated RF signal); if you use Windows, this is somewhat more complicated and the rule of thumb is not to set the PC soundcard to a too high level. Setting these parameters correctly will deliver a clean audio signal to your radio.
For receiving, the same principles apply but things can be tricky. You could, for example, have a weak output from the radio and manipulate your audio signal in the digital modes software to have a waterfall that looks good, yet still not be able to use the maximum dynamic range achievable with your combination or software and hardware. Similarly, you could have an exceedingly strong output from the radio and get a good looking waterfall but still fail to extract the best possible performance. So it is important to read your software calibration instructions, following them to ensure you are sending the best and cleanest audio signal from the radio to your PC in first place.
Having an interface, in simple terms, means adding a sound card to the system. Although this sounds like introducing an unnecessary element in the chain, a well adjusted hardware interface will provide better RX performance and results.
In summary you need a good I/O setting to a) receive clean signals and use the full possible dynamic range and b) transmit the loudest possible signal without distortion.
RSID
Experienced users can easily recognise their favourite data modes when they hear them in the headphone or when they see them in the waterfall, though there are chances that the rare, QRP, or distant signal you are looking for, is deep in the noise and difficult to pinpoint. This is where RSID comes to the rescue.
You have probably heard of the Olivia calling frequencies, these are “channels” where most stations will place calls and where traffic can be found but, as we are sharing the spectrum with many other users who use different modes around the same frequency slots, you may still be unable to immediately recognise the signals you are looking for. RSID, usually available in mainstream multi-mode digital modes applications, is a burst of tones that can be transmitted and received to identify a digital mode. I can’t remember for sure, but I think I read somewhere that it can be decoded down to s/n levels around -18dB. That’s impressive and way deeper in the noise than any signals that you will be able to decode in Olivia. A station with the RSID enabled on RX will see a pop-up message showing mode and frequency of the transmitter and will be able to just click on the pop-up button to switch mode at once and precisely tune on the frequency of the transmitter. With the receiver set correctly, the receiving station will have maximised their chances of decoding the weak signal they just detected.
For the transmitter, transmitting the RSID burst maximises the chance of being heard and decoded. The RSID burst can be set at the beginning and at the end of the transmission. I have seen plenty of stations transmitting it at the end and after thinking and rethinking why they do this, I failed to find an answer. However I know that whenever I click on an RSID burst that I am interested and that was transmitted at the beginning of a message, I can decode the station (if the band conditions and signal strength allow). You do your maths and decide when you want to transmit the RSID tones burst.
Then there is another type of ID, called Video ID in Fldigi, which prints the selected mode in the waterfall – this is, in my opinion, a far less effective mode identification method as it will most likely be not visible at the receiver end when the signal is weak, thus defeating the purpose of launching a mode identification burst. However it can be very useful when an RSID burst is not available, for example when the user sets an unusual combination of tones and bandwidth, e.g. Olivia 32/500 or Olivia 16/250. In this case, the Video mode identification may be the only possible option.
RECEIVER ADJUSTMENTS
A good receiver is a life changing experience and it goes without saying that if your receiver belongs to the digital era you have a distinctive advantage. Relatively new receivers capable of adjusting the receiving bandwidth, offering the option of setting narrow and possibly multiple notch filters anywhere in the audio spectrum, can greatly enhance the receiving performance of the radio as well as that of the software.
In my experience the preamp settings commonly found on Icom radios bring no benefits when the signal is very weak as they amplify both the noise and the signal. Probably they amplify the noise better than the signal as in the cases I refer to the signal is usually near or below the noise floor. I also found that the noise reduction function doesn’t help either and the best possible action when you face difficult waterfall conditions is to narrow the passband and leave a narrow slot open and centred on the signal that you intend to decode. That, with some RF gain adjustment and the best possible combination of radio-interface-software settings (see above), is the most effective way I found to decode challenging signals. On occasions I found that narrowing the bandwidth on receive to 400 Hz and even less I managed to pull signals out of the noise despite the presence of either wideband radar signals or other strong signals such as multiple contesting RTTY stations. EDIT (January 2020): In recent times I started experimenting some new settings, where the receiver’s AGC is turned off and most of the DSP is left to the software. At the time of writing I haven’t yet come to a conclusion on whether this option can extract more signal from the noise, I will update this paragraph as soon as I can reach a sensible conclusion.
FORWARD ERROR CORRECTION (FEC)
In addition to a good control of your receiver, you can adjust Olivia’s FEC to the needs and conditions of the band at any given time. In Fldigi (this is the software I use and that I’d recommend to anyone) the default setting is 4 and this is good in most circumstances, however, I found that increasing this value to 8 can enhance Fldigi’s decoding ability. This comes at the cost of some delay, as it can take a few more seconds to start decoding however most experienced stations are aware of this and when they call CQ they will leave sufficient time for their system to detect a distant station’s response before launching the next call. The rule of thumb is to start low, say level 4, then if you hear an incoming RSID and see a weak signal that does not decoed well, increase this level to the required level.
SETTING UP YOUR CQ CALLING PROCESS CORRECTLY
This is another area often overlooked. Some time ago I found a weak signal from Central Europe on the 80 metres. Not a distant station but one of those located too far for the ground wave but too close for the skywave. The s/n was somewhere around -12dB but I managed to recognise Olivia 8/250 by the sound of the start/stop two tones and to decode the CQ call. I never managed to make contact because, every time I responded, the other station started a new CQ call; I even tried my best to launch my call without the RSID (the other station wasn’t using it at all on TX so I assumed it was off for RX too) immediately after the end of their transmission but probably their CQ call started before their software had a chance of decoding the first characters in my response. I finally gave up.
Calling in Olivia mode requires some method. When a call starts, it normally takes a few seconds, depending on their FEC settings, to start decoding at the other end. The same delay is carried through the whole transmission until the end and, if our signal is very weak at the other end, the distant station will most likely start transmitting after the whole call is decoded (though experienced users may be able to shorten the delay of their response). When the response is sent back, a similar delay occurs and the cycle goes on. The take away message here is that leaving 20-30 seconds space between the end of one CQ call and the beginning of the next helps weak and distant signals to catch up and begin the two-way QSO process.
TRANSMITTER’S AUDIO BANDWIDTH
This is an issue that is mostly encountered by stations operating in PSK mode who hear an Olivia signal some 2.9 kHz above the frequency displayed on their VFO operating frequency. For example a station operating in PSK mode with the radio tuned on 14070 kHz hears an Olivia signal at 14072.5 or 14072.9 kHz. The Olivia station’s RSID gives them the hint on mode and audio centre, they click and then respond but their signal comes across with the higher tones suppressed or strongly attenuated. The root cause of the problem is that the station tuned on 14070 kHz is pushing their transmitter beyond the 2.7 kHz or 3 kHz audio TX bandwidth.
Being one of those operator that doesn’t give automatic 599 signal reports, I tend to care about the quality of the signals that I receive. Unfortunately I often find that my explanation and suggestion to tune the VFO up by 1 kHz or so and then centre the audio again is not followed and the stations move on. Well, if this happens to you, the solution is simple and the fix takes less than five seconds.
CONCLUSION
I hope and trust that the above paragraphs will help most people to enjoy digital modes such as Olivia, which allow free communication beyond the usual signal report.
M0ZZM 
Thank you for taking the time to write this out. It has been enlightening! 🙂