K3 FSK power transients

[Joe Subich, W4TV-4]

I don't know what a *monitor* output proves since it is *not* an RF
sample.  However, I have attached a Spectrum Lab Time Domain capture
as requested.  Notice the crest factor is well less than 1 dB.

73,

     ... Joe, W4TV


On 2015-05-29 7:45 AM, Ken_ke2n via Elecraft wrote:

> Joe -  I am using data mode with "DATA A" selected.  (This mode uses the SSB
> filter and turns off equalization and compression. But of course the ALC is
> "on").
>
>
>
> What would be really helpful is if you would download a copy of Spectrum Lab
> and run the monitor output of your K3 through it while transmitting FSK441.
> Then we would both have a common point of comparison (eliminating the
> vagaries of power meter PEP algorithms).
>
>
>
> http://www.qsl.net/dl4yhf/spectra1.html
>
>
>
> The time domain scope is under the "components" tab.  The default settings
> with the initial install will work fine, although you will have to select
> your input device (audio card) if it is not the system default. That is
> under the Options/Audio settings tab.
>
>
>
> Regards
>
> Ken
>

______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html
Message delivered to [hidden email]

[Joe Subich, W4TV-4]

 > But I stand corrected on the average. I was computing an average of
 > the crest, not the effective value of the wave.  Since these are sine
 > waves, the effective value is 0.707 of the crest. Of course, when the
 > crest value is varying over the period under consideration, you might
 > want to average the crests and then multiply by 0.707, which is kind
 > of what I was doing.

That is not the definition of peak to average in the RF domain.

For a CW (sine wave modulation) signal, the peak to average ratio is
1:1 or 0 dB.  Since FSK (specifically, in this case FSK-441 - which is
4FSK) is a sequence of sine waves, the theoretical peak to average
ratio is 0 dB.  Now, if you introduce AM into the FSK modulation, e.g.,
some tone or tones have different levels, the crest factor departs from
0 dB.

In this case, Peak Power (or PEP) is the maximum average peak (CW) case
and "average power" is the longer term average of the peak power.  In a
linear system, PEP is the CW (single tone) power while average power -
per FCC definition is the value to which a capacitor would charge based
on multiple peaks (or the average of the peaks of the modulation).  For
4FSK where three of the 4 tones were 1.4 V Peak (1.0 V RMS) while the
fourth tone was 1.12V (0.8 V RMS), the average voltage would be:
(3 * 1.4 + 1.12)/4 or 1.33 V Peak.  The Peak voltage is the highest of
the four tones or 1.4 V.  Thus the Peak to average ration for a 4FSK
signal with one tone at 80% of the other three would 1.4/1.33 - or
0.45 dB.

I'd strongly suggest you get an audio oscillator with a known, stable
output level, pick one tone - either 1323 or 1764 Hz - and run it into
you K3, set the mic gain to "4 bars of ALC" and measure the CW output
power at that frequency.  Then, without changing the audio level or
mic gain, measure the CW level of your K3 at 882, 1323, 1764 and 2205
Hz.  My guess is that you will find that one tone or another produces
significantly less power than the others - probably because of a filter
alignment or ripple issue.

73,

    ... Joe, W4TV


On 2015-05-29 11:46 AM, Ken wrote:

> I was going by this definition:
>
> http://en.wikipedia.org/wiki/Crest_factor
>
> But I stand corrected on the average. I was computing an average of
> the crest, not the effective value of the wave.  Since these are sine
> waves, the effective value is 0.707 of the crest. Of course, when the
> crest value is varying over the period under consideration, you might
> want to average the crests and then multiply by 0.707, which is kind
> of what I was doing.
>
> That means the PAPR (peak to average power ratio) for a sine wave is,
> by definition,  3 dB.  If you have more than one tone present at any
> moment the PAPR must be higher than 3 dB (see the URL above).
>
> But that is not what our fancy power meters measure.
>
> These power meters are supposed to measure Peak Envelope Power and
> average power. With complex waveforms like speech there is no
> analytical relationship between PEP and average. It depends on the
> user's voice and the amount of speech processing used. The peaks can
> easily be 10-20 times the average in an uncompressed voice signal.
>
> For FSK441, the modulating wave is mostly pure tones (of a constant
> amplitude if all goes well) and some minimal "keying sidebands" due
> to the 441 baud modulation. So I was thinking there must be some way
> to calculate PEP/average.
>
> The result of a pure tone, in SSB, is a steady carrier and the
> filtered envelope of that RF is the same as the average power - so RF
> PEP/average is 1.0. Mind you, the *instantaneous* value of the RF
> voltage wave will be 1.414 times the RMS value, but the usual
> definition of PEP says that a steady carrier has PEP equal to average
> power. And, when I send one of the steady tones, "73" for example,
> the Wave Node meter reads PEP equal to average  (within some
> tolerance).  What happens is that I see the average power reading
> rise to the PEP.  The PEP reading stays the same. I imagine the K3 is
> riding herd on the peak and keeping it constant (even with no bars
> showing).
>
> Time for lunch here ;-)  But I will send a note to the Wave Node guy
> and see what he says about those peaks and dips.
>
> Ken
>
> -----Original Message----- From: Joe Subich, W4TV
> [mailto:[hidden email]] Sent: Friday, May 29, 2015 10:50 AM To:
> Ken; [hidden email] Subject: Re: [Elecraft] K3 FSK power
> transients
>
>
> A quick reply since I'm on my way out ... I'll look more in depth
> later.
>
> I am concerned about the audio waveform generated by your WaveNode
> device.  I see peaks to 128 +/- and dips to 64 +/- with a significant
> amount of asymmetry.  The asymmetry would make me wonder if there is
> not some clipping (bad transistor) in the LPA, KPA3 or in the
> WaveNode device itself.
>
> Perhaps the only way to know for sure would be to use a second
> receiver and look at the *recovered* audio.
>
>> Mine is using the more steep-sided 8 pole filter and has the new  >
>> synthesizer board and the latest firmware updates.
>
> I'm also using the 2.8 KHz filter, new synthesizer and most recent
> production firmware for the KSYN3A.  Note there was no change in DSP
> code to support the KSYN3A so I would not expect the firmware to be
> an issue.
>
>> PEP/average (power) ratio is the (voltage) crest factor squared.
>> It  > is hard to get the effective(average) value of the wave by
>> eyeball. But you could take the min/max and divide by 2.
>
> No - Crest factor is 10*log(Peak/Average) in terms of power and
> 20*log(Peak/Average) in terms of voltage.  Since the modulation is
> single sine waves (not a complex waveform) the average (RMS) is 0.7 x
> peak.  The lowest voltage I see is 0.8V, the highest is slightly less
> than 1.0 (call it 0.98).  Using the average of 0.98 and 0.80 (there
> should be about as many of each tone) we get 20Log(.98/.89) or 0.84
> dB which is a reasonable match to the Peak to Average ratio reported
> by the LP-100.
>
>> I tried backing down the line level to where no ALC action was
>> visible but there are still "dips".
>
> Can't do that.  With less than 4 bars of ALC the DSP code will try to
> "follow level" and may make things worse depending on the time
> constant.
>
> 73,
>
> ... Joe, W4TV
>
>

______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html
Message delivered to [hidden email]

[Elecraft mailing list]

Today I put the old Bird wattmeter in line and found that the average power out of the K3 does *indeed* drop down and become somewhat erratic in fsk441 mode. As noted previously the PEP output stays the same although my watt meter shows some erratic behavior.

So my wattmeter is telling the truth. And my K3 has a problem. (I tried different computers and sound cards - all the same). The power fold back is  in FSK441 and a similar mode called JTMS. Modes that vary the tone more slowly - JT65 and JT6M - do not cause the same large power drop (minor fluctuations in power only).

To use an separate receiver I had to switch from 50 MHz to 28 MHz and there I eventually found that if I run <12 watts (final amp not switched in) then the PEP/Average is quite near 1:1 (or 10:10).  Its only when I go to a power level where the final is required, that I get this strange effect.  

I get quite a high PEP/Average ratio on voice too, but it does not sound bad ...

The final still produces plenty of power, so it is not fried. I went through and repeated the transmitter gain calibration - to no effect.

So I guess I am going to contact Elecraft on Monday.

73

Ken



-----Original Message-----
From: Joe Subich, W4TV [mailto:[hidden email]]
Sent: Friday, May 29, 2015 3:12 PM
To: Ken; [hidden email]
Subject: Re: [Elecraft] K3 FSK power transients


 > But I stand corrected on the average. I was computing an average of  > the crest, not the effective value of the wave.  Since these are sine  > waves, the effective value is 0.707 of the crest. Of course, when the  > crest value is varying over the period under consideration, you might  > want to average the crests and then multiply by 0.707, which is kind  > of what I was doing.

That is not the definition of peak to average in the RF domain.

For a CW (sine wave modulation) signal, the peak to average ratio is
1:1 or 0 dB.  Since FSK (specifically, in this case FSK-441 - which is
4FSK) is a sequence of sine waves, the theoretical peak to average ratio is 0 dB.  Now, if you introduce AM into the FSK modulation, e.g., some tone or tones have different levels, the crest factor departs from
0 dB.

In this case, Peak Power (or PEP) is the maximum average peak (CW) case and "average power" is the longer term average of the peak power.  In a linear system, PEP is the CW (single tone) power while average power - per FCC definition is the value to which a capacitor would charge based on multiple peaks (or the average of the peaks of the modulation).  For 4FSK where three of the 4 tones were 1.4 V Peak (1.0 V RMS) while the fourth tone was 1.12V (0.8 V RMS), the average voltage would be:
(3 * 1.4 + 1.12)/4 or 1.33 V Peak.  The Peak voltage is the highest of the four tones or 1.4 V.  Thus the Peak to average ration for a 4FSK signal with one tone at 80% of the other three would 1.4/1.33 - or
0.45 dB.

I'd strongly suggest you get an audio oscillator with a known, stable output level, pick one tone - either 1323 or 1764 Hz - and run it into you K3, set the mic gain to "4 bars of ALC" and measure the CW output power at that frequency.  Then, without changing the audio level or mic gain, measure the CW level of your K3 at 882, 1323, 1764 and 2205 Hz.  My guess is that you will find that one tone or another produces significantly less power than the others - probably because of a filter alignment or ripple issue.

73,

    ... Joe, W4TV

______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html
Message delivered to [hidden email]