Inverted Vee Linked Dipole vs. Endfed Antenna (Part 1)

[DK3IT]

@DD5LP

The 10 dB (i.e 10x) number has been thrown around as a guide for a while but as far as I know, it has never been accurately measured in a laboratory setup.

I think the theory for this is in the Shannon-Hartley theorem [1]: Basically, phone modes require a higher channel capacity than CW, in bits per second.

The Nyquist theorem tells us that we need at least 4000 samples per second for representing audio frequencies up to 2 kHz, which is a sampling frequency of 4 kHz. With 1 bit resolution, this means 4000 bps.

If we assume that a dit in CW lasts 40 ms (ca. 30 wpm), then a cycle of one dit and its pause lasts two dits, i.e. 80 ms.

80 ms is equal to a frequency of 12.5 Hz. Now, since a series of dits is the highest frequency we must represent/recover, we need a sampling frequency of twice that, i.e. 25 Hz, which is 25 bps at 1 bit resolution.

So CW needs a channel capacity of 25 bps, while voice audio needs 4000 bps, i.e. 1:160 - for voice, you need a channel capacity 160 times larger compared to CW.

The channel capacity, the bandwidth, and the signal-to-noise ratio are approximately linear to each other (see [1]).

If we assume a filter bandwidth of 2400 Hz for SSB and of 400 Hz for CW, the bandwidth available for SSB is six times that of CW.

However, the bit rate is 160 times as much.

Hence, the SNR for voice must be larger by the factor of 160/6 = 16.6 as compared to CW, all other things equal.

This would mean that, in an ideal world and with all my simplifications and assumptions, 5 W in CW are comparable (in terms of error rate, as a proxy for successful QSOs), ca. 80 W SSB.

Now, I am simplifying a bit regarding the bit rates, real QSOs include some degree of redundancy, and human operators can guess missing parts from character and word n-grams in natural language. Also, my reasoning might have a few flaws; I just tried to understand the theory from Wikipedia tonight (one of the fascinating side-effects of our hobby) .

But nonetheless I think that this calculated figure is a surprisingly good fit to the numbers reported by practitioners.

Any feedback and corrections are warmly welcome!

73 de Martin, DK3IT

[1] Shannon–Hartley theorem - Wikipedia

[DK3IT]

@G8CPZ

Readability is a psychophysical property, i.e. it involves the perception and interpretation of a physical stimulus, so it cannot be reduced merely to a function of the receiver bandwidth and received signal power (not alone transmit power).

Yes, but this does not conflict with the Shannon-Hartley theorem, which gives the upper limit of channel capacity for a combination of bandwidth and SNR.

The actual number of bits transmitted and received correctly cannot exceed that limit, not even for a super-human on magic mushrooms.

However, if the message contains redundancy (e.g. repeating the RST or callsign), the net amount of information that is being transmitted for a successful QSO is lower, hence the channel capacity required is lower. And slower CW requires a lower sampling frequency. This is the theory that explains why we can understand 12 WPM with RST given 3 times from much weaker signals.

Redundancy can also be in the transmitted text; if the characters are not completely random, but from a set of Q-codes, and there are some conventions in the structure of a QSO, the actual number of bits of this message is lower than that of the raw sequence of characters.

This is called entropy encoding and is the same mechanism that you are using when compressing a text file prior to sending it over the Internet.

Second, if the receiver knows part of the message (which may be just the likelihoods of some events, e.g. that a certain CEPT prefix is more likely than another, or the next word if you know the previous one), then the actual number of bits required to be transmitted and hence the net bit rate is again lower (to be precise, this is just the other side of the entropy of the transmitted information).

This is the theory behind the observation that we can read CW faster in QSOs with common Q-codes as compared to random strings of characters when training CW.

So while you are right that the actual power ratio is dependent on additional aspects, the power ratio is not outside the reach of scientific discovery, at least as an approximation.

It is not “unknowable spiritual glibbery that only gifted souls can conceive under the influence of strong homeopathic sugar drops” .

Please be assured that I am not at all implying that you were claiming or even hinting at such a perspective. I just want to carry the torch of the Age of Enlightenment as high and as far as I can.

[DD5LP]

@G8CPZ

And that’s the problem, Ed.

Readability is a psychophysical property, i.e. it involves the perception and interpretation of a physical stimulus, so it cannot be reduced merely to a function of the receiver bandwidth and received signal power (not alone transmit power).

Furthermore, despite anecdotal claims by some radio amateurs over many years regarding listening to voice vs Morse code, it’s very unlikely they are exactly equivalent for the same power-to-mode ratio, and it probably would require scientific tests involving a large number of people to determine a relationship.

That’s my tuppence worth [two old pennies].

Indeed so.
The way I understand the concept, the point put simply is that a narrower bandwidth effectively “compresses” the available power into a smaller space as well as excluding off-frequency interference, increasing the effective signal however what the “effective improvement” is - whether it is simply related to the proportion between the bandwidth values or some other scientific formulas and proven by scientific measurement, I don’t know.

(my 2 Pffenige).

73 Ed.

[G8CPZ]

@DK3IT

I think the theory for this is in the Shannon-Hartley theorem

We’re straying a long way from antennas here so I’ll make this my last point on “Voice vs Morse and power”:.

[Martin, I’m guessing you probably already know this, but for a wider audience …] The S-H theorem defines the channel capacity (in terms of maximum data bit rate) for a given bandwidth and S/N ratio. It doesn’t [attempt to] address the readability of the received data by a human operator.

For example, a machine (e.g a computer, a FT8 receiver) could correctly interpret all data bits in a noisy channel providing the data rate doesn’t exceed the capacity given by the H-R equation, but a [skilled] human operator might well not interpret (“read”) correctly voice or Morse meeting the H-R criterion.

@DK3IT

And slower CW requires a lower sampling frequency.

On a more technical point, the bandwidth occupied by a Morse code signal (transmitted in A1A mode) depends on the rise and fall times of the edges of the transmitted pulses (typ. 5-6ms) and not on the speed of the Morse. The frequency components in the pulse transitions are higher than the dot rate.

[DK3IT]

@G8CPZ

On a more technical point, the bandwidth occupied by a Morse code signal (transmitted in A1A mode) depends on the rise and fall times of the edges of the transmitted pulses (typ. 5-6ms) and not on the speed of the Morse. The frequency components in the pulse transitions are much higher than the dot rate.

Yes, but these are two different bandwidths: The bandwidth relevant for the SNR and hence the channel capacity is defined by the filter of the receiver (because this determines the power of the noise). The bandwidth occupied by the transmitted CW signal varies by rise time, but this is irrelevant in here, as long as the signal that is within the filter bandwidth is sufficient for reconstructing the original information.

And the minimal information we need is a sequence of bits, where each bit represents the signal level at one dit resolution, so

• the character S (…) would be
  101010, and
• T would be 1110 (one dash plus a dit pause).

[9A3IV]

Thank you for all explanations. It was very interested. I am simple radio amater. My experience is: I hear that is something but ssb signal is not readable. Switch to cw, no need for cw filter and cw is perfectly readable. I would say, using my earmeter, at least 10 dB above noise. That is all I am interested in.

I am very impresed by ssb operators skill and efforts

[DM1CM]

^^^^^^ What he (DK3IT) said !

That (Shannon-Hartley) is basically what I had in mind when I wrote my tongue-in-cheek comment above, but I know I could not have written it anywhere near so well or succinctly as Martin has done.

[G4IPB]

… so completely off topic how much worse is NBFM to SSB ?..( My guess is about another 6dB?)

[DM1CM]

@G4IPB

so completely off topic how much worse is NBFM to SSB ?

On 160 meters, unspeakably much worse. I think … maybe … errmmm.

[DK3IT]

“off topic” does not deserve its negative connotation. Becoming friends is getting off topic, falling in love and leaving a party early is getting off topic, searching a passage to India and discovering America is getting off topic, etc. .

Serious answer: As the bandwidth increases, theory would hint that NBFM would need less power than SSB (= can be copied at a lower SNR) assuming that the bit rate of voice information is the same for both. It just occupies a whole lot more bandwidth on the band, which is why it is mostly used on higher bands. But in practice, differences in propagation will likely be the dominanting factor when comparing NBFM with SSB - e.g. sensitivity to fading/QSB, QRM, QRN.

[G4IPB]

…I did once hear an FM QSO on 160 … which is possibly legal in the UK but is not in the band plan… … ( but I was thinking about 2m where I stick with a handheld and a J pole rather than using SSB … ).

I make a habit of being off topic… trying to explain SOTA is a bit like the classic Bob Newhart sketch…

https://www.kennedy-center.org/video/center/other/2020/bob-newhart/

( Well that cheered me up …) but the bit at the beginning about new technology and TV being transmitted across the USA for a live broadcast was interesting… I think we are at that stage with AI, its kind of fun asking it to write a poem but we (I at any rate) have not grasped the long term consequences.

Paul

[MM0FMF]

@G4IPB

I did once hear an FM QSO on 160

There are observations of well set up equipment showing occupied peak bandwidth at -30dBC for various modes as USB = 2.8KHz, AM = 5KHz, NBFM = 5KHz (12.5kHz channel spacing). You can work out from
those what the likely power needed will be based on all the simplifications etc.

@DK3IT

ca. 80 W SSB.

The figure 85W is often quoted. But 80-85W is in the right the ballpark.

[HB9PJT]

The theory does not take into account that the ear has an adaptive filter function and can still read SSB with a S/N ratio of -3 dB. According to my experiments, SSB is 7 dB worse than CW when using a good speech processor and optimal equalizer settings.

73, Peter - HB9PJT

@DK3IT

Hence, the SNR for voice must be larger by the factor of 160/6 = 16.6 as compared to CW, all other things equal.

[MM0FMF]

That is a coding function that works for sounds/speech patterns you recognise. Sending random speech sounds will remove that coding gain.

Shannon never received a Nobel prize for his work yet it is fundamental to all comms theory. If you can improve on his theory splendid rewards await you.

[DM1CM]

@MM0FMF

That is a coding function that works for sounds/speech patterns you recognise. Sending random speech sounds will remove that coding gain.

Posting random pictures will do much the same thing, while revealing a deeper truth?

[DK3IT]

@MM0FMF

That is a coding function that works for sounds/speech patterns you recognise. Sending random speech sounds will remove that coding gain.

This!

In addition, spoken language contains a lot of “empty” sections, like pauses.

An interruption of propagation for 200 mS will likely not hamper your ability to reconstruct a spoken transmission. In CW, it most likely means you are loosing three dits or one dah.

Most of the what we attribute to human ability to decipher distorted voice communications can be explained by redundancy, parts with no information entropy (like pauses, “err”, “uhm”) and knowledge about probabilities and correlation between characters and words (and context), or additional information that is not being transmitted (e.g. my operating position vs. likely CEPT prefixes on 2m).

When receiving an SSB transmission of spoken language, we do not require neither the entire spectrum nor the entire duration of the transmission (contrary to the transmission of music or dense digital modes without redundancy for error recovery).

A very obvious example is the phonetic alphabet - it is explicitly designed so that the words for each character are as distinctive as possible, and you only need part of the audio signal to reconstruct it:

Del… .ilo …ree In… tag. …oke ….able

is DK3IT/P even though about half of the transmission time is muted.

If we want a scientifically sound analysis, we would need to set up a “number stations on a summit” competition and read random sequences of numbers in SSB and in CW at controlled power levels or in combination with SNR estimates, ideally without any breaks or pauses.

As for SOTA, a likely very relevant variable is the number and geographic distribution of potential chasers wrt to the mode or modes they operate, and the equipment and antennas they have.

By the way, I do neither want to hijack this thread nor lecture anybody, I am just enjoying the inspiring interaction and food for thought!

To the moderators: I would not object if we were able to split this thread and move the discussion about modes and power ratio into a new one. It might be a useful reference, and we are close to the 100 message limit anyway.

73 de Martin, DK3IT

[G8ADD]

@DK3IT

To the moderators: I would not object if we were able to split this thread and move the discussion about modes and power ratio into a new one. It might be a useful reference, and we are close to the 100 message limit anyway.

I prefer to be hands off, let the thread wander where it will, it often roams through some fascinating scenery. Its up to the participants to start new threads - and see where they wander to!

[G4OBK]

@G8ADD

let the thread wander where it will

It’s certainly done that Brian! I’m afraid I lost interest in it many messages ago. Here is another one then to almost tip it over the 100 limit edge !

73 Phil

[G8ADD]

@DK3IT

Most of the what we attribute to human ability to decipher distorted voice communications can be explained by redundancy, parts with no information entropy (like pauses, “err”, “uhm”) and knowledge about probabilities and correlation between characters and words (and context), or additional information that is not being transmitted (e.g. my operating position vs. likely CEPT prefixes on 2m).

A lot of information in speach can be carried by things like inflection and rhythm. A couple of instances: in many western languages a question will end on a rising inflection, an imperative on a falling inflection. The rhythm of a CQ can be recognised even when no words can be made out, prompting us to turn a beam if we have one!

[DL3TU]

@G8ADD

The rhythm of a CQ can be recognised even when no words can be made out, prompting us to turn a beam if we have one!

… reminds me of this one:

CQ DX CQ DX

Sorry, now completely off topic, but couldn’t resist.