Minimum Discernible Difference

Amateur radio, as with most things in life,  is always filled with tradeoffs and compromises, particularly when it comes to matters of signal strength.  Is that bigger antenna worth the extra cost?  What would a few more watts buy me?  Should I replace my feedline with something  that has lower loss?  How much more readable would I be if I added a few feet to my tower, or added a second antenna at some lower height, in order to gain a few db at the takeoff angle of interest?

In my case, it was the antenna question that first led me to try to quantify the difference in readability between two signals of comparable, but unequal strength.  I was trying to decide between two antennas that differed by roughly 2 db in forward gain and by several hundred dollars in cost.  My gut reaction said that 2 db wasn't worth the extra cost, but as a DXer and contester I changed my mind after I generated some audio files similar to the ones I've included here ... especially when I considered the probable reduction in background noise that the tighter pattern of the larger antenna would provide.  No doubt others with different motivations or constraints would have reached a much different conclusion, but hopefully these examples may offer some small assistance to anyone faced with similar choices.

Any comments (favorable or otherwise) would be warmly received via email sent to ... dave (at)    I am not an audio professional so be gentle with me.


All of the files on this page use background noise recorded from a moderately noisy and otherwise dead 20m band within approximately a 500 Hz bandwidth (roughly 350 Hz to 850 Hz with some rolloff in the 70 Hz segments at each end).  The CW Signals were generated and recorded to .wav files using CW Player, a nice freeware application from F6DQM.  Various tones and speeds were used depending upon what I was trying to accomplish, but in general I tried to choose those that I thought were mid range for the majority of people (roughly 550 Hz and 26 wpm).  All audio fles were converted to .mp3 format and mixed with appropriate level shifts using Goldwave, a reasonably competent audio editor.

These files were generated with decent fidelity (32 kHz and 64 kbs), but they may "play" differently depending upon individual situations such as headphone or speaker response, software equalization, user hearing response, user tone or speed preference, etc.  Your mileage may vary ... a lot.

Each of the files must be downloaded from the links and played locally .... they won't play directly from this website.  Embedding a player here would have been possible but wouldn't have given the user much control over playback, at least not with my limited knowledge of web crafting.  If you fear that these may not be legitimate .mp3 files, my recommendation is that you ignore them.

Stepped Signal Strength

As a first analysis, and as could be inferred by the title of this page, I was curious what the minimum discernible difference was between two weak signals of approximately  equal strength.  Clearly the answer to that is a function of how close they are to the background  noise level.  A difference of a few db might be unnoticeable for two relatively strong signals, but if one of the signals is beneath the noise threshold and the other is above it the comparison would be more striking.  I was interested in the weak signal case, so I generated this file with CW text stepped in six 1 db increments beginning more or less beneath the noise and rising above it.


In my case, I could reasonably discern a difference in volume for each one db step, and my threshold of easy readability was at the third level ("two db" sent text).  Your results might be considerably different.

Here is the same set of signals, except in reverse order from loudest to weakest.


Dueling Callsigns

The file above tries to display readability purely with respect to background noise.  A  somewhat different consideration is how two concurrent signals of unequal strength compare with each other ... how much difference in volume does it take to be heard first when both are near the noise level.  To explore that I created four files with different tones and different volumes, 
, each with two callsigns sent at 26 wpm.  The multiple files were an attempt to provide enough tone difference between the callsigns to distinguish them from each other, but with mirror image tone assignments to compensate for possible personal preference.   Two of the files  have a one decibel difference in volume between the two callsigns and the other two files have a two decibel difference.  Each file starts out with Callsign A louder than Callsign B and then changes midstream to make Callsign B louder than Callsign A.  Version 1 files use 550 Hz for one callsign and 520 Hz for the other, while version 2 files use the opposite assignment.  Here are the four files:





In my opinion, one decibel difference is at least noticeable and two decibels represents a competitive edge.

Effect of Speed

In the presence of spikey background noise (my best attempt at a suitable scientific term) it is sometimes helpful to use lower CW speeds so that our brain can integrate each element over a longer period and thus more effectively sort out the noise.  To investigate that effect I created the following file.  It contains 20 different callsigns ... five each sent at 20, 25, 30, and 35 wpm ... all of them using a 550 Hz tone.  The volume of each callsign is the same as the louder of the two callsigns in the Dueling Callsigns files.


I wouldn't try to take this illustration too literally, though.  Slower character speeds may help the dits bridge noise spikes but faster speeds will slip in the gaps between static crashes more effectively.  Faster speeds may also facilitate copy during conditions of rapid QSB.  Slower speeds may get you ignored by impatient operators or by those who naturally copy better at faster speeds.  Still ... if you're trying to make contacts with an area of the world when the band is opening or closing, or if you know the other end of the path is in an area of high QRN, cranking down the speed might make a difference.  If the other operator keeps asking for repeats it might not be because his/her code speed is necessarily poor ... it might be because the QRN spikes on his end are obliterating your dits.

Dave   AB7E