Power measurement theory and practice, and S-meter operation

K1LSB
Posts: 638
Joined: Wed Feb 05, 2020 5:25 pm

Re: Thetis 2_8_11 changes/ideas/wants/wishes

Postby K1LSB » Tue Jul 27, 2021 5:52 pm

Richie,

I'm risking getting kicked off this forum for even venturing into the S-meter discussion, but here goes...

I want to set aside for the moment your very good point that S9 is defined as 50uV below 30MHz and 5uV above 30MHz, given that you are perfectly correct in that statement, and if Simon Brown doesn't recognize that shift then he's doing something wrong.

However (confining ourselves for the sake of argument to 30MHZ and below), what Simon is doing right is displaying the correct reading in his S-meter.

The IARU spec (officially only a "recommendation" but it's the de facto standard) states that the S-meter reading shall (not may) be based on the peak (actually "quasi-peak") voltage level detected anywhere in the passband. NOTE: The term "quasi-peak" is defined in the spec simply as a signal that is sustained for a minimum specified length of time (10 msecs +/- 2 msecs) before it is reported as an actual signal (otherwise it is simply ignored as noise).

That's exactly what Simon's S-meter does, and that's why his meter reading consistently agrees very closely with the dB scale on the edge of the panadapter scope.

There are many (and I mean many) people who claims that the S-meter should "integrate the total power in the passband". However, such a notion is mentioned nowhere in the spec! Nor anywhere in the spec is there any mention of any notion of "passband width", or "summing of power" or anything else except the quasi-peak voltage detected anywhere in the passband. Not a single person who claims the S-meter should vary its reading based on "the width of the passband" or "the total power in the passband" has been able to show me where in the spec it says that!

The function of an S-meter is not to report the "total power" or "average power" or "integrated power" in the passband, it is to report the highest voltage detected anywhere in the passband, irrespective of any nebulous "total power" contained within the filter. That's what Simon's S-meter does.

Anyone can easily see this for themselves by simply opening a session of SDR Console and clicking on any broadcast AM station. If the carrier is steady then the S-meter reading will also be steady, regardless of how much the carrier is being modulated and regardless of the passband width.

Simon gets it right, and everyone else gets it wrong, including Flex and Thetis. As soon as someone mentions "integrating the power in the passband" they lose my attention because the spec mentions no such silly concept. The spec recognizes only the peak voltage detected anywhere in the passband, nothing more and nothing less.

Let me repeat, no one who claims the S-meter should vary its reading based on the width of the passband or "the total integrated power in the passband" can show me in the spec where it says any such thing.

So anyone who wants the S-meter to report anything other than what the spec says it should report wants something besides an S-meter.

Either the S-meter does what the spec says it should do, or we're talking about something else that's not an S-meter.

Mark
User avatar
ramdor
Posts: 1468
Joined: Wed Jul 03, 2019 3:07 pm
Contact:

Re: Thetis 2_8_11 changes/ideas/wants/wishes

Postby ramdor » Tue Jul 27, 2021 9:24 pm

K1LSB wrote:Richie,

I'm risking getting kicked off this forum for even venturing into the S-meter discussion, but here goes...

I want to set aside for the moment your very good point that S9 is defined as 50uV below 30MHz and 5uV above 30MHz, given that you are perfectly correct in that statement, and if Simon Brown doesn't recognize that shift then he's doing something wrong.

However (confining ourselves for the sake of argument to 30MHZ and below), what Simon is doing right is displaying the correct reading in his S-meter.

The IARU spec (officially only a "recommendation" but it's the de facto standard) states that the S-meter reading shall (not may) be based on the peak (actually "quasi-peak") voltage level detected anywhere in the passband. NOTE: The term "quasi-peak" is defined in the spec simply as a signal that is sustained for a minimum specified length of time (10 msecs +/- 2 msecs) before it is reported as an actual signal (otherwise it is simply ignored as noise).

That's exactly what Simon's S-meter does, and that's why his meter reading consistently agrees very closely with the dB scale on the edge of the panadapter scope.

There are many (and I mean many) people who claims that the S-meter should "integrate the total power in the passband". However, such a notion is mentioned nowhere in the spec! Nor anywhere in the spec is there any mention of any notion of "passband width", or "summing of power" or anything else except the quasi-peak voltage detected anywhere in the passband. Not a single person who claims the S-meter should vary its reading based on "the width of the passband" or "the total power in the passband" has been able to show me where in the spec it says that!

The function of an S-meter is not to report the "total power" or "average power" or "integrated power" in the passband, it is to report the highest voltage detected anywhere in the passband, irrespective of any nebulous "total power" contained within the filter. That's what Simon's S-meter does.

Anyone can easily see this for themselves by simply opening a session of SDR Console and clicking on any broadcast AM station. If the carrier is steady then the S-meter reading will also be steady, regardless of how much the carrier is being modulated and regardless of the passband width.

Simon gets it right, and everyone else gets it wrong, including Flex and Thetis. As soon as someone mentions "integrating the power in the passband" they lose my attention because the spec mentions no such silly concept. The spec recognizes only the peak voltage detected anywhere in the passband, nothing more and nothing less.

Let me repeat, no one who claims the S-meter should vary its reading based on the width of the passband or "the total integrated power in the passband" can show me in the spec where it says any such thing.

So anyone who wants the S-meter to report anything other than what the spec says it should report wants something besides an S-meter.

Either the S-meter does what the spec says it should do, or we're talking about something else that's not an S-meter.

Mark


All above my pay grade Mark to decide what should or shouldn't be done. I haven't looked for (or at) the IARU spec on it so I can't say one way or another if you are right, but I will take your word for it.

My main worry is that if you were to makes the changes, and then put an anan up against a flex for example you'll have a case of S9 on the flex is S7 on the anan and people will incorrectly assume it is deaf or the Smeter is out. I can guarantee that it would result in loss of sales in the long term. Simon doesn't care in that regard, he just wants donations for dog food.

R.
User avatar
w-u-2-o
Posts: 5525
Joined: Fri Mar 10, 2017 1:47 pm

Re: Thetis 2_8_11 changes/ideas/wants/wishes

Postby w-u-2-o » Tue Jul 27, 2021 10:40 pm

The discussion of how S-meters should work is extraordinarily important in order to maintain commonality with other radio equipment and worthy of further discussion.

Bottom line up front: if Simon is doing as Mark says, then he is doing it differently than every other radio on the planet. It is therefore not a good choice, and I strongly recommend it be rejected as an approach. It represents a completely fundamental change to a standard, de facto or de jure.

The de jure (based on some written or legal standing) standard comes from the IARU technical recommendations agreed to by IARU signatories at the Brighton conference of 1981 and as further refined at the Torremolinos conference of 1990 (paper attached below):

Code: Select all

IARU Region 1 Technical Recommendation R.1
BRIGHTON 1981, TORREMOLINOS 1990

STANDARDISATION OF S-METER READINGS

1. One S-unit corresponds to a signal level difference of 6 dB,
2. On the bands below 30 MHz a meter deviation of S-9 corresponds to an
available power of -73 dBm from a continuous wave signal generator
connected to the receiver input terminals,
3. On the bands above 144 MHz this available power shall be -93 dBm,
4. The metering system shall be based on quasi-peak detection with an
attack time of 10 msec ± 2 msec and a decay time constant of at least
500 msec.

All RF power measurements MUST be referenced to some bandwidth. This is physics and not me just saying this. If you look at only the voltage of an RF waveform what frequency components does it represent? The answer is "All of them". So that bandwidth must be specified, and controlled, or it can't be said what part of the spectrum has been measured. Further, power is energy/time. The crudest way to measure this is apply RF energy for some time and measure the heating of the sensor (and many old time RF power meters did exactly that). Do not confuse item (4) of the recommendation to mean anything but over what time periods this heating, or power, effect is measured. The "quasi-peak detection" refers to the output of the power detector and NOT the peak voltage of the RF signal.

With the engineering and science out of the way, there ARE two problems with the standard as written.

A. As Mark quite rightly points out, the IARU neglected to specify an RF power measurement bandwidth. Nevertheless, the de facto (the accepted way even though it is not written or law) standard is to use the bandwidth of the final passband. This allows for both good signal power measurements and noise power measurements, and to be able to easily compare the two.

B. The IARU did not define S-9 between 30 and 144MHz. This has become a bit of a dog's breakfast in the real world, with no de facto standard appearing. However, in the IARU HF Manager's Handbook V9, dated June 2018, on page 38 the IARU now writes that -93dBm starts immediately above 30MHz. The fix for this is conceptually easy for Richie: make an option for -93dBm to start at 30 or 144MHz. It may be harder in practice (in coding).

In summary: I have not looked at or analyzed Simon's SDR Console software. But based on Mark's description, it seems Simon is looking at the peak value in a very small bandwidth (1 or 2Hz) within the larger bandwidth of the final passband and reporting that as the S reading. This is incorrect and it should not be done that way.

IARU Technical_Recommendation.pdf
(58.4 KiB) Downloaded 164 times
K1LSB
Posts: 638
Joined: Wed Feb 05, 2020 5:25 pm

Re: Thetis 2_8_11 changes/ideas/wants/wishes

Postby K1LSB » Tue Jul 27, 2021 11:49 pm

Scott,

Per the IARU spec, for the HF bands, "a meter deviation of S-9 corresponds to an available power of -73 dBm from a continuous wave signal [CW] generator connected to the receiver input terminals" [which directly correlates to a signal strength of 50 microvolts assuming a standard input impedance of 50 ohms] (wording in brackets is mine).

Note the test condition is a CW signal. Nowhere in the spec is "total passband power" mentioned at all, nor is there any "waveform" mentioned. There is only a reference to the greatest (literally "quasi-peak") signal whose strength is to be measured, with no concern for the strength of any other signal that may be found anywhere in the passband.

Passband width has no relation to the maximum voltage measured in the passband. A passband of 20 Hertz (assuming it captures the carrier of an example AM transmission) will read the same peak voltage as a 20 kilohertz passband that captures the carrier and both sidebands of that transmission, because a (proper) AM transmission will always have the greatest voltage on the carrier, as any spectrum display will clearly show. If all you're trying to measure is the maximum voltage detected anywhere in that passband (which is all we should be doing, per the spec) then the carrier level is all you will be reporting, because (in the case of the AM signal) everything else (i.e., the greatest voltage in either sideband) is below the carrier level, and so is of no concern to a meter whose sole function is to report the highest voltage (not the "total power") detected.

You say "it seems Simon is looking at the peak value in a very small bandwidth (1 or 2Hz) within the larger bandwidth of the final passband and reporting that as the S reading", but that is not correct. Simon is looking at the entire passband and reporting the peak voltage found anywhere in that band. And that's exactly what the spec says to do. No regard is given to any other signal anywhere in the passband, only to the peak signal, regardless of the width of the passband - the spec clearly and specifically references the "peak" signal and nothing else. Not "average", not "integrated", nor any other consideration (including passband width).

With all of that said, I do want to go on record as saying that I've always held your understanding of our radios in the very highest regard, Scott - none of us here would be anywhere near the level of operator we are today if it weren't for your tireless support!

Best to all,

Mark
User avatar
W2PA
Posts: 165
Joined: Sun Apr 09, 2017 6:34 pm
Location: LaGrangeville, NY
Contact:

Re: Thetis 2_8_11 changes/ideas/wants/wishes

Postby W2PA » Wed Jul 28, 2021 12:20 am

This has been a very interesting discussion! And my compliments to all for the cordial tone. I think I see the issue. The language in the spec is just a tad imprecise and lacks an explicit definition of "quasi-peak detector".

First, spec #2 says "On the bands below 30 MHz a meter deviation of S-9 corresponds to an available power of -73 dBm from a continuous wave signal generator connected to the receiver input terminals,"

Note there is no mention of any other signals. We must assume that whatever the passband is, the CW signal (which, we presume is constant, unkeyed, therefore as close to zero bandwidth as has any meaning) is the only signal being measured.

Secondly, spec #4 says "The metering system shall be based on quasi-peak detection with an attack time of 10 msec ± 2 msec and a decay time constant of at least 500 msec."

According to the definition on Wikipedia, "Quasi-peak detectors for specific purposes have usually been standardized with mathematically precisely defined dynamic characteristics of attack time, integration time, and decay time or fall-back time." (Anyone here is welcome to find another definition.)

Note this definition includes integration time.

The article goes on to state that "The CISPR quasi-peak detector applied to most conducted emissions measurements (0.15 - 30 MHz) is a detector with an attack time of 1 ms, a decay time of 160 ms and an IF filter setting of 9 kHz. The quasi-peak detector applied to most radiated emissions measurements (30 - 1000 MHz) has an attack time of 1 ms, a decay time 550 ms and an IF filter bandwidth of 120 kHz." CISPR is the International Special Committee on Radio Interference.

Note that bandwidth is in fact stated as a parameter. So why does the IARU fail to mention it?
73,
Chris, W2PA
K1LSB
Posts: 638
Joined: Wed Feb 05, 2020 5:25 pm

Re: Thetis 2_8_11 changes/ideas/wants/wishes

Postby K1LSB » Wed Jul 28, 2021 1:42 am

Chris,

1) The IARU spec is explicit in its definition of "quasi-peak detector". The fact that it doesn't include a specification for integration time or filter bandwidth does not make it "imprecise". There is a fundamental maxim of law which states, "That which is not included is excluded". That means that anything that is not specifically included in a document was by intent left out. The fact that there is no "integration time" (whatever that even is) specified in the IARU spec was intentional. The Wiki article you mention states that quasi-peak detectors are typically intended "for specific purposes", and "usually" specify attack time, integration time, and decay time. In short, that article is nothing more than a layman's explanation of what a quasi-peak detector is, not a "definition" as recognized by any academic discipline or industry.

2) The two examples of quasi-peak detectors given in the Wiki article are specifically intended for measuring RF interference, and are not intended for use as an S-meter. Further, each of those detectors specify wildly different IF filter bandwidths, both of which are much greater than any passband you or I would typically use in HF operations. The attack time in both examples is only 1 msec, much smaller than what is specified in the IARU spec. In short, any comparison of the IARU spec to any particular "specific purpose quasi-peak detector" described in a Wiki article is meaningless. That article has no nexus to the operation of an S-meter.

3) Notably, that same Wiki article further states that integrated power detection is used by both the CISPR and the FCC for RFI measurements above 1GHz, while quasi-peak detection is used for measurements below 1GHz! So measuring integrated power is clearly a distinct and separate function than measuring quasi-peak power. And the IARU spec clearly states that a quasi-peak detection method shall be used, with no mention at all made of "integrated power" detection.

Edit:

4) The Wiki article also states that "integration time" (among other things) is typically included in the definition of a quasi-peak detector, yet neither of the examples given in the article include an integration time characteristic. So clearly the fact that the article says any particular characteristic is "usually" included in a typical spec doesn't at all mean that that characteristic should be included in any particular spec. The fact that the IARU spec doesn't include a bandwidth characteristic is as irrelevant as the fact that it also doesn't include an "integration time" characteristic -- neither of those characteristics are relevant to the IARU spec.

Good night, all!

Mark
User avatar
W2PA
Posts: 165
Joined: Sun Apr 09, 2017 6:34 pm
Location: LaGrangeville, NY
Contact:

Re: Thetis 2_8_11 changes/ideas/wants/wishes

Postby W2PA » Wed Jul 28, 2021 12:18 pm

Well, Mark, I take issue with your assertion that technical specs follow the maxims of law. Technical specs are normally explicit in the extreme. But, no matter.

We're still left with questions:
Why, as Scott asserts, do none of the equipment (hardware) manufacturers adhere to the "peak" interpretation of the spec?
In an era of SDR and precision measurements, has the S-unit has outlived its usefulness except as a relative measure?

I guess I'll still give out reports that way - and a 5NN (or 59) to everyone in a contest even if I'm straining to hear them at ESP signal level. ;-)
73,
Chris, W2PA
w9mdb
Posts: 446
Joined: Sun Apr 09, 2017 5:53 pm

Re: Power measurement theory and practice, and S-meter operation

Postby w9mdb » Wed Jul 28, 2021 2:22 pm

If you don't measure over some bandwidth you'll get a different answer depending on the FFT size used.

In analog space it's easy to measure peak voltage. In digital space no-can-do-so-easily as the FFT is just an approximation of the real signal. As the FFT size decreases the peak value decreases and spreads out into other bins.

So in digital space you HAVE to measure against some bandwidth.

Mike W9MDB
Mike W9MDB
K1LSB
Posts: 638
Joined: Wed Feb 05, 2020 5:25 pm

Re: Power measurement theory and practice, and S-meter operation

Postby K1LSB » Wed Jul 28, 2021 2:46 pm

Chris,

Given that the IARU spec is on its 2nd revision, I can only surmise that any characteristics that are omitted (such as bandwidth and "integration time") were intentionally omitted. For my own part, I see no need for either of those functions to be included in the spec, given that 1) the bandwidth is whatever the user has selected, and 2) there is no "integration" process or algorithm defined in the spec to be performed so there is no need for an "integration time" function.

As to why no manufacturer's S-meter operates per the spec, I can think of several possible explanations:

1) It may be difficult to design (and/or expensive to make) an analog meter that reports only the greatest voltage detected anywhere in the passband while completely ignoring any other signal that may be present (even though that's exactly what the spec says to do).

2) Per my understanding, in an analog radio the traditional S-meter gets its input from the AGC circuit. However, the correct sample point should be at the antenna input to the receiver, which is where Simon Brown captures his sample (in fact, Simon says the AGC isn't involved in his S-meter function at all).

3) The manufacturers simply ignore what the spec says, given that virtually everyone has become so used to the traditional S-meters reporting the total power captured in the passband that that's what they've come to expect the meter to report. Everyone has come to think of the S-meter as a power meter, which it is not. It's a peak voltage meter. To expect the S-meter reading to vary according to passband width even though the total power captured may be nothing but noise floor is patently absurd!

Mike,

Per a post on Simon's site, his smallest sample size is whatever the FFT bin size is (you'll have to ask him how he determines that).

Mark
User avatar
w-u-2-o
Posts: 5525
Joined: Fri Mar 10, 2017 1:47 pm

Re: Thetis 2_8_11 changes/ideas/wants/wishes

Postby w-u-2-o » Wed Jul 28, 2021 2:49 pm

K1LSB wrote:Scott,

Per the IARU spec, for the HF bands, "a meter deviation of S-9 corresponds to an available power of -73 dBm from a continuous wave signal [CW] generator connected to the receiver input terminals" [which directly correlates to a signal strength of 50 microvolts assuming a standard input impedance of 50 ohms] (wording in brackets is mine).

Note the test condition is a CW signal. Nowhere in the spec is "total passband power" mentioned at all, nor is there any "waveform" mentioned. There is only a reference to the greatest (literally "quasi-peak") signal whose strength is to be measured, with no concern for the strength of any other signal that may be found anywhere in the passband.

No argument from me on any of those points. It is a poorly written specification or standard that may assume the reader is familiar with the fundamentals of RF power measurement.

Passband width has no relation to the maximum voltage measured in the passband.

Actually it does. More specifically, the voltage and current conditions impinging on the impedance of the measurement device will differ depending upon what signals and what noise exists in the measurement bandwidth. If you change the bandwidth you change what signals and what noise impinge on the detector.

A passband of 20 Hertz (assuming it captures the carrier of an example AM transmission) will read the same peak voltage as a 20 kilohertz passband that captures the carrier and both sidebands of that transmission, because a (proper) AM transmission will always have the greatest voltage on the carrier, as any spectrum display will clearly show. If all you're trying to measure is the maximum voltage detected anywhere in that passband (which is all we should be doing, per the spec) then the carrier level is all you will be reporting, because (in the case of the AM signal) everything else (i.e., the greatest voltage in either sideband) is below the carrier level, and so is of no concern to a meter whose sole function is to report the highest voltage (not the "total power") detected.

This is where I'm afraid you've gone wrong. A power measurement is NOT equivalent to a voltage measurement. One cannot measure power by putting a 50 ohm impedance voltmeter on an antenna and declare that X volts equals Y power except in very simple cases. When there are multiple signals (sinusoids) summing together with noise this cannot be done.

There must also be a time component of the measurement because power = energy/time. A million volts for zero time equals zero power. This is why the IARU language includes the attack and decay times for the measurement, although even that is very poorly defined. It would have been better if they had simply specified an integration time period for the measurement, e.g. 250ms or something like that.

This brief might provide some better insight into this: https://www.boonton.com/Portals/0/RF_Power_Measurement_Basics_1.pdf

Where it comes to RF power measurement people make two fundamental mistakes:

1. They think that it involves a voltage measurement of the RF waveform. It does not. It involves a voltage measurement of the output of an RF power detector. There is a huge difference! That power detector might be an old time microcalorimeter, or a crystal detector diode, or mathematical function applied to digitized RF data, but the power detector comes first, and it measures power, NOT voltage by itself.

2. They think that peak RF power readings are accomplished instantaneously. That is impossible. So called "peak reading" power measurements always involve an integration time. Peak reading meters simply make that integration time very short, typically 10 to 100ms. Average reading meters will typically use integration times in the 100-500mS range.

You say "it seems Simon is looking at the peak value in a very small bandwidth (1 or 2Hz) within the larger bandwidth of the final passband and reporting that as the S reading", but that is not correct. Simon is looking at the entire passband and reporting the peak voltage found anywhere in that band. And that's exactly what the spec says to do. No regard is given to any other signal anywhere in the passband, only to the peak signal, regardless of the width of the passband - the spec clearly and specifically references the "peak" signal and nothing else. Not "average", not "integrated", nor any other consideration (including passband width).

I'll have to take your word for what Simon is doing, I certainly don't know, I was only guessing based on your previous posts. But if Simon is only looking at the peak RF voltage after filtering to a given bandwidth, he cannot possibly assign that voltage to any specific sinusoid in that bandwidth unless there is only one. That might provide sufficient accuracy for a single CW signal with good SNR, but it won't work for an SSB, AM or any other signal environment that involves a multiplicity of sinusoids.

With all of that said, I do want to go on record as saying that I've always held your understanding of our radios in the very highest regard, Scott - none of us here would be anywhere near the level of operator we are today if it weren't for your tireless support!

That is very kind of you Mark, thank you :)
w9mdb
Posts: 446
Joined: Sun Apr 09, 2017 5:53 pm

Re: Power measurement theory and practice, and S-meter operation

Postby w9mdb » Wed Jul 28, 2021 2:53 pm

It doesn't matter what bin size you use....an FFT does not measure peak voltage correctly unless the frequency being measured is an exact multiple of the bin size which never happens in the real world. You get a gaussian peak which, when integrated, reflects the peak power...ergo integration over some bandwidth.
Mike W9MDB
K1LSB
Posts: 638
Joined: Wed Feb 05, 2020 5:25 pm

Re: Power measurement theory and practice, and S-meter operation

Postby K1LSB » Wed Jul 28, 2021 4:27 pm

Gents,

Per the info on Simon's website explaining the operation of his S-meter, "[The signal level] is the peak spectrum FFT bin (output value) which is within the current filter."

That page is found here: https://www.sdr-radio.com/s-meter

Here's another post where he states, "SDR Console’s S-meter value is simply the maximum FFT bin value within the filter range, the only way the IQ data from the radio is modified is via the Windowing function (Hann, Hamming, etc.). There’s a peak hold and decay, otherwise it’s a faithful value.":

https://sdr-radio.groups.io/g/main/mess ... 0,20440519

He is 100% faithful to the spec and only reports the highest FFT bin value in the passband. His meter doesn't care one whit about anything else. It can't get any simpler than that.

Edit:

Btw, his meter always agrees with the edge scale in the panadapter! :)

With that said, I have to let this discussion go, gentlemen. I simply don't have the free time or energy to devote to this anymore (I still work full-time).

Best regards to all,
Mark
User avatar
w-u-2-o
Posts: 5525
Joined: Fri Mar 10, 2017 1:47 pm

Re: Power measurement theory and practice, and S-meter operation

Postby w-u-2-o » Wed Jul 28, 2021 6:19 pm

K1LSB wrote:Gents,

Per the info on Simon's website explaining the operation of his S-meter, "[The signal level] is the peak spectrum FFT bin (output value) which is within the current filter."

That page is found here: https://www.sdr-radio.com/s-meter

So I was correct when I wrote "But based on Mark's description, it seems Simon is looking at the peak value in a very small bandwidth (1 or 2Hz) within the larger bandwidth of the final passband and reporting that as the S reading." Because that's really the only way you can get that behavior. Too bad it's flat wrong. And, if you read that page, he's calculating noise power in the passband wrong, too.

He is 100% faithful to the spec and only reports the highest FFT bin value in the passband. His meter doesn't care one whit about anything else. It can't get any simpler than that.

It is possible to meet a specification and still be wrong. This is a classic tenet of system engineering.

Regardless of what anyone may think of the IARU language, the de facto standard is:

- For panadapter/spectrum analyzers: use relatively small channel (bin) bandwidths in order to provide a detailed representation of RF power vs. frequency.

- For S-meters: perform a channel power measurement where the channel bandwidth is equal to the final passband of the radio.

For what it's worth at this point, these are intended to be different measurements, different views of the same data. They are NOT intended to be the same. Each measurement provides important and useful data. If you somehow force the issue with the S-meter, then you lose access to valuable data, which is channel power. Hell, if the S-meter measurement is changed to match the panadapter, why even have an S-meter? Just turn on peak blobs and be done with it.

Btw, his meter always agrees with the edge scale in the panadapter!

Clearly that is his design intent, and he has chosen a way to make that happen. His methodology has consequences. I may be overly harsh in my use of the word "wrong". So let's see if I can write something a bit more measured (Ha, a pun! :D )

Bottom line: Simon has chosen to make the metering on his radio work differently than every other radio on the planet. As a result it becomes impossible to make accurate channel (passband) power measurements of either signal and noise on his radio and, more importantly, it means that any signal reports from his radio will not be comparable to any other radio. He has done this because he wants his metering to match signal peak values on the panadapter.
User avatar
ramdor
Posts: 1468
Joined: Wed Jul 03, 2019 3:07 pm
Contact:

Re: Power measurement theory and practice, and S-meter operation

Postby ramdor » Wed Jul 28, 2021 7:23 pm

K1LSB wrote:Btw, his meter always agrees with the edge scale in the panadapter! :)


Not for me it doesn't, it is all over the place. The following image is a 4k (0-4000) lsb filter, peak hold on (the yellow line) and there is nothing in the panadapter filter area that is anywhere near -129.5dBm

not4me.png
not4me.png (38.17 KiB) Viewed 7425 times


It is all over the place. I thought you said it was the same on the left?

wtf.png
wtf.png (22.91 KiB) Viewed 7425 times


.... and nothing at all being done at 30MHz

pinch_of_salt.png
pinch_of_salt.png (837.24 KiB) Viewed 7425 times


EDIT: ok i could get everything to align if I turned off smoothing to NONE (which is not default btw). That has a massive impact, but still no changes around 30MHz.

R.
K1LSB
Posts: 638
Joined: Wed Feb 05, 2020 5:25 pm

Re: Power measurement theory and practice, and S-meter operation

Postby K1LSB » Wed Jul 28, 2021 7:54 pm

Richie,

As you may have seen, Simon's S-meter is a multi-meter in the sense that it can display several things at once: It can show the noise level (that's the user-selectable gray needle in the meter), the instant peak level (that's a user-selectable yellow needle that may not have been selected in your session), a red user-selectable peak-hold needle, and a separate peak-readout numeric window (the small yellow window in the meter) that may have a longer decay threshold than the red peak-hold needle. Due to the different decay values, at any given instant the value showing in the yellow window may or may not agree with the red needle.

Further, the user can also apply varying degrees of smoothing to the display, along with user selectable Windowing functions. Any smoothing applied to the displayed signal has the potential to cause a disparity between the displayed panascope curve (including the peak-hold panascope curve) and the S-meter needles. It would quite possibly be more revealing if you had enabled the yellow instant-peak needle in the S-meter panel.

Given that I have no idea what smoothing or Windowing options you have selected (or what options are set as default in SDRC), I should probably revise my previous post to say "Simon's S-meter usually closely agrees with the edge scales".

Mark
User avatar
ramdor
Posts: 1468
Joined: Wed Jul 03, 2019 3:07 pm
Contact:

Re: Power measurement theory and practice, and S-meter operation

Postby ramdor » Wed Jul 28, 2021 8:37 pm

K1LSB wrote:Given that I have no idea what smoothing or Windowing options you have selected (or what options are set as default in SDRC), I should probably revise my previous post to say "Simon's S-meter usually closely agrees with the edge scales".


Yes I know about the three needles, but if you look at the images the panadapter does not match the Smeter. I had a fresh install, in fact I had to use the reset tool to get it running.

The only way I could get the scales to match was to turn smoothing off which is not default. Actually you can move that panadapter data all over the place with the smoothing stuff. He even mentions it not lining up in his info you linked :

Signal Level
This is the peak spectrum FFT bin (output value) which is within the current filter. Note: This value will probably differ from the actual spectrum trace as the spectrum trace uses smoothing to make the display more aesthetically pleasing and less tiring on the eyes.


So unless he actively moves the scale to match the trace (which it doesn't look like he is) then it will be out of line if using any form of smoothing, he even acknowledges that.

R.
User avatar
DL2XY
Posts: 116
Joined: Sun Jul 30, 2017 9:47 am

Re: Power measurement theory and practice, and S-meter operation

Postby DL2XY » Thu Jul 29, 2021 11:04 am

Please don't touch the S-Meter in Thetis.
It's perfectly precise und correct in the sense of metrology.

We have the choice between voltage peak (Signal) and spectral density(Sig Avg), the latter is perfekt for SSB use.
You can give reports like "You are S9+10 over S7 noise" stating the SNR directly. The link quality is purely definded by ratio of channel powers .

If someone urgently needs a quasipeak display, please add it as a third choice leaving the other as is.

73 Walter

ps: Using the hightest bin of FFT makes no sense. You have to use an appropiate interpolation of at least 3 bins. Still it gives no usefull information.
User avatar
KA5KKT
Posts: 129
Joined: Thu Aug 06, 2020 6:51 pm

Re: Power measurement theory and practice, and S-meter operation

Postby KA5KKT » Fri Jul 30, 2021 7:44 pm

Early versions or PowerSDR or variants thereof offered the choice of either the dB scale or and S-Unit scale on the left (...center or right) hand side of the pan display. I'd like to see that option available in Thetis.
User avatar
w-u-2-o
Posts: 5525
Joined: Fri Mar 10, 2017 1:47 pm

Re: Power measurement theory and practice, and S-meter operation

Postby w-u-2-o » Sat Jul 31, 2021 1:31 am

KA5KKT wrote:Early versions or PowerSDR or variants thereof offered the choice of either the dB scale or and S-Unit scale on the left (...center or right) hand side of the pan display. I'd like to see that option available in Thetis.

This won't work. What is shown on the panadapter is not measureable in S-units because of the difference in measurement bandwidth between the two methods. The panadapter shows power in small bandwidths on the order of 2Hz or so. The S-meter shows channel power where the channel bandwidth is equal to the final passband width.

Again, these are intended to be different measurements, different views of the same data. They are NOT intended to be the same. Each measurement provides important and useful data.
User avatar
KA5KKT
Posts: 129
Joined: Thu Aug 06, 2020 6:51 pm

Re: Power measurement theory and practice, and S-meter operation

Postby KA5KKT » Sun Aug 01, 2021 1:02 pm

w-u-2-o wrote:
KA5KKT wrote:Early versions or PowerSDR or variants thereof offered the choice of either the dB scale or and S-Unit scale on the left (...center or right) hand side of the pan display. I'd like to see that option available in Thetis.

This won't work. What is shown on the panadapter is not measureable in S-units because of the difference in measurement bandwidth between the two methods. The panadapter shows power in small bandwidths on the order of 2Hz or so. The S-meter shows channel power where the channel bandwidth is equal to the final passband width.

Again, these are intended to be different measurements, different views of the same data. They are NOT intended to be the same. Each measurement provides important and useful data.


Thanks, Scott.

I can understand that there are different ways of measuring dB's. The full implications of such are over the head of this non EE / E-Tech. Apparently the dB scale now imployed in Thetis cannot translate into an S-Meter type scale.

Walter, DL2XY suggests that spectral density (Sig Avg) is appropriate for SSB use.

If that is the case, is it possible, either in theory or practice to have a pan display scaled on notion of spectral density?
User avatar
w-u-2-o
Posts: 5525
Joined: Fri Mar 10, 2017 1:47 pm

Re: Power measurement theory and practice, and S-meter operation

Postby w-u-2-o » Sun Aug 01, 2021 2:55 pm

KA5KKT wrote:
w-u-2-o wrote:
KA5KKT wrote:Early versions or PowerSDR or variants thereof offered the choice of either the dB scale or and S-Unit scale on the left (...center or right) hand side of the pan display. I'd like to see that option available in Thetis.

This won't work. What is shown on the panadapter is not measureable in S-units because of the difference in measurement bandwidth between the two methods. The panadapter shows power in small bandwidths on the order of 2Hz or so. The S-meter shows channel power where the channel bandwidth is equal to the final passband width.

Again, these are intended to be different measurements, different views of the same data. They are NOT intended to be the same. Each measurement provides important and useful data.


Thanks, Scott.

I can understand that there are different ways of measuring dB's. The full implications of such are over the head of this non EE / E-Tech. Apparently the dB scale now imployed in Thetis cannot translate into an S-Meter type scale.

Walter, DL2XY suggests that spectral density (Sig Avg) is appropriate for SSB use.

I agree. Sig Avg provides a pure channel power measurement. This is not spectral density, however. Spectral density is measured in power per unit bandwidth, most typically in milliwatts/Hz, also in dBm/Hz (decibels relative to one milliwatt per 1Hz of bandwidth). Sig Avg measures the total RF power, shown in either in S-units, dBm, or dBuV (decibels relative to one microvolt referenced to a 50 ohm impedance).

There is a big difference between channel power and spectral density. Spectral density can only be applied to signals that have a relatively wide bandwidth and are generally the same power level at every frequency. Atmospheric noise is a perfect example of this. A CW signal is not. And an SSB signal is a big mess. But to use noise as an example, let's say you have a noise level in a 2700Hz passband of S5, which is also -97dBm (S9 = -73dBm, and every S-unit is equal to 6dBm). So that's the channel power of the noise, and that's what is displayed on the S-meter when in Sig Avg mode. But the spectral density of that is -131dBm/Hz (without getting into the math, but you can find the math at this link here).

If that is the case, is it possible, either in theory or practice to have a pan display scaled on notion of spectral density?

Again, the answer is "No". The pan display shows energy in very tiny bandwidths, typical 2 to 4Hz wide. The S-meter shows channel power in bandwidths of typically 500 to 3000Hz, more or less.

Here's a rough analogy: think of a shower head. The shower head has lots of little holes in it, each creating a separate stream. The flow rate of that stream is equivalent to RF power. But the shower head is old, and some of the holes flow more water than others. Let's pretend it's a really, really big shower head and there are 100,000 holes, and each one corresponds to a small piece of the spectrum, say 1Hz, so you've got a 100KHz shower head :D . Each pixel on the panadapter corresponds to the amount of flow of just one hole in that shower head. It gives you a great picture of how each hole in that shower head is performing all at the same time. A map of that shower head, if you will, in terms of flow rate for each hole.

Now think about your receive passband. It's going to be the equivalent of say 3,000 of those holes. The passband filter is like a big funnel placed under the 3,000 adjacent holes that you want. And the S-meter measures the flow rate of the water coming out of that funnel. It gives you a measurement of the energy in the entire signal, not the energy coming out of each tiny little shower head hole.

They are both flow rate measurements. The panadapter measures each shower head hole individually. The S-meter measures the flow rate of a large number of shower head holes added together. So they can't EVER be the same. And they are not supposed to be the same. They are measuring different things.

One other aspect of the shower head analogy: you can't measure flow rate instantaneously. You can't count water going by molecule by molecule. You have to wait until enough water flows by to get a good, reliable, smooth reading. So maybe it takes a few tens of milliseconds for each reading. This is no different for our panadapter and S-meter. So when you have them set to measure "peak" power, it's just power averaged over a very short time, it's not an instantaneous reading.
User avatar
KA5KKT
Posts: 129
Joined: Thu Aug 06, 2020 6:51 pm

Re: Power measurement theory and practice, and S-meter operation

Postby KA5KKT » Sun Aug 01, 2021 6:35 pm

Thanks again, Scott.

Is it suffice to say that that the S-Meter scale along side of the Pan display or PowerSDR offered little value to the user?
User avatar
w-u-2-o
Posts: 5525
Joined: Fri Mar 10, 2017 1:47 pm

Re: Power measurement theory and practice, and S-meter operation

Postby w-u-2-o » Sun Aug 01, 2021 8:17 pm

KA5KKT wrote:Thanks again, Scott.

Is it suffice to say that that the S-Meter scale along side of the Pan display or PowerSDR offered little value to the user?


Again, that is not an S-meter scale. That scale shows the power level of the energy in whatever FFT bin size you have programmed.

It has great value to the user. How else would you be able to compare relative signal strengths on the panadapter? Or look at how much difference there might be in different portions of someone's SSB signal. Or measure the SNR of a large signal visually as you adjust attenuation?

Remember that everything is relative and everything has value. You just need to remember that when you are describing something to another ham to tell them where the information came from so that they can understand how the measurement was made.

When panadapters were just a bunch of squiggles with no scale life was easy. When panadapters became legitimate spectrum analyzers with the advent of SDR technology it really caused some people to get very confused, and the learning curve definitely went up.
User avatar
KA5KKT
Posts: 129
Joined: Thu Aug 06, 2020 6:51 pm

Re: Power measurement theory and practice, and S-meter operation

Postby KA5KKT » Sun Aug 01, 2021 11:14 pm

To clarify my previous post...

"Thanks again, Scott.

Is it suffice to say that that the S-Meter scale along side of the Pan display o<f> PowerSDR offered little value to the user?"

My question might have been better understood had I referenced the proper context. My question had to do with previous versions of PowerSDR and / or variants there of that included an option(?) of an S-Meter as an alternative to the dB scale.
User avatar
w-u-2-o
Posts: 5525
Joined: Fri Mar 10, 2017 1:47 pm

Re: Power measurement theory and practice, and S-meter operation

Postby w-u-2-o » Mon Aug 02, 2021 1:09 am

KA5KKT wrote:To clarify my previous post...

"Thanks again, Scott.

Is it suffice to say that that the S-Meter scale along side of the Pan display o<f> PowerSDR offered little value to the user?"

My question might have been better understood had I referenced the proper context. My question had to do with previous versions of PowerSDR and / or variants there of that included an option(?) of an S-Meter as an alternative to the dB scale.

Sorry, I misinterpreted your question!!! I can't remember any version of PowerSDR that showed an S-meter scale on the panadapter. That includes the legacy Flex version that I used with my (now long gone) Flex-3000, nor PowerSDR mRX PS. However, if there was such a version, then the answer to your question as I now, finally :) , understand it, is definitely "Yes!". Since the panadapter does not offer a channel power measurement (well, they are teensy, weensy little channels ;) ) it really has no meaning in that context.
User avatar
KA5KKT
Posts: 129
Joined: Thu Aug 06, 2020 6:51 pm

Re: Power measurement theory and practice, and S-meter operation

Postby KA5KKT » Mon Aug 02, 2021 2:08 am

w-u-2-o wrote:
KA5KKT wrote:To clarify my previous post...

"Thanks again, Scott.

Is it suffice to say that that the S-Meter scale along side of the Pan display o<f> PowerSDR offered little value to the user?"

My question might have been better understood had I referenced the proper context. My question had to do with previous versions of PowerSDR and / or variants there of that included an option(?) of an S-Meter as an alternative to the dB scale.

Sorry, I misinterpreted your question!!! I can't remember any version of PowerSDR that showed an S-meter scale on the panadapter. That includes the legacy Flex version that I used with my (now long gone) Flex-3000, nor PowerSDR mRX PS. However, if there was such a version, then the answer to your question as I now, finally :) , understand it, is definitely "Yes!". Since the panadapter does not offer a channel power measurement (well, they are teensy, weensy little channels ;) ) it really has no meaning in that context.


Some versions of NaP3 included the option of displaying an S-Meter scale where we are used to seeing the left-hand dB scale. I don't know if that was in other variants of PowerSDR, but susepect some have have had it. It has been out of NaP3 for several versions.

Return to “Thetis”