K3 Transmit IMD

[n7ws]

Well, I've seeming stirred up a hornets' nest, or as
we have in Southern Arizona, an Africanized ("killer")
honeybees' nest.

I do stand corrected about a response from Elecraft
regarding this subject that I failed to see.

http://www.mail-archive.com/elecraft@.../msg50667.html

Furthermore, it was not my intent to necessarily
single out Elecraft regarding this issue, but
considering its fine, and well deserved, reputation
for quality products, I expected better, and I'm sure
that better will be (or is being) delivered.

The essence of what I was trying, apparently
unsuccessfully, to get across is that *all* aspects of
transceiver design are important, and the focus I've
seen while lurking in this forum for a year is on
receiver performance.  Witness the endless discussions
like, "Should I get the 5-pole or the 8-pole filter?"
and so forth.

That is certainly admirable, and I've done my share of
receiver designing and building in the last 50 years,
but no amount of bullet-proofing the receiver will
clean up your neighbor's transmitter.

And the focus of receiver performance has only
resulted from consumer demand.  We should be demanding
better transmitter performance too. That's all I was
getting at.

Respectfully,

Wes  N7WS




      ____________________________________________________________________________________
You rock. That's why Blockbuster's offering you one month of Blockbuster Total Access, No Cost.  
http://tc.deals.yahoo.com/tc/blockbuster/text5.com
_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[n7ws]

--- "Eric Swartz - WA6HHQ, Elecraft"
<[hidden email]> wrote:

> The K3 firmware used for the original QST review did
> not turn off the
> K3's RF speech compressor clipper even when the
> compressor was set to
> zero for their TX IMD two tone test. This resulted
> in a significant
> increase in measured TX IMD when using an external 2
> tone source. We
> corrected this in a firmware update last month.  All
> new K3's and anyone
> who has updated to the latest firmware have this
> fix.
>
> 73,  Eric  WA6HHQ

[snip]

Well, that sort of explains it, but raises another
question: Does the firmware change actually correct
the IMD issue or does it simply return when the
operator turns the processor on?

Regards,

Wes  N7WS


      ____________________________________________________________________________________
You rock. That's why Blockbuster's offering you one month of Blockbuster Total Access, No Cost.  
http://tc.deals.yahoo.com/tc/blockbuster/text5.com
_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[Julian, G4ILO]

Wes Stewart wrote

--- "Eric Swartz - WA6HHQ, Elecraft"
<[e-mail]> wrote:

> The K3 firmware used for the original QST review did
> not turn off the
> K3's RF speech compressor clipper even when the
> compressor was set to
> zero for their TX IMD two tone test. This resulted
> in a significant
> increase in measured TX IMD when using an external 2
> tone source. We
> corrected this in a firmware update last month.  All
> new K3's and anyone
> who has updated to the latest firmware have this
> fix.
>
> 73,  Eric  WA6HHQ

[snip]

Well, that sort of explains it, but raises another
question: Does the firmware change actually correct
the IMD issue or does it simply return when the
operator turns the processor on?

Regards,

Wes  N7WS

Nobody seems to have answered this, so I'll try. IMD is a measure of the linearity of the transmiitter. It is measured by feeding two pure sine waves into the input and looking at what unwanted mixing products come out the end.

The firmware has no effect on IMD. What it did (apparently) do is invalidate the test results by introducing distortion on the input signal so the TX did not have pure tones to start with.

It's impossible to talk about IMD when compression is used. Compression, by its very nature, distorts the signal, creating IMD. If you want to transmit the cleanest signal, leave the compressor off and make sure the ALC doesn't show either.

Most of us, though, would rather transmit a "punchier" signal in the interest of being heard at the other end, so we deliberately distort our signals by using ALC and compression. Which (it seems to me) makes the whole issue of IMD figures just another example of the pointless "numbers game" that people obsess over. What the heck does it matter whether one TX is a few dB better or worse than another when it is the component holding the microphone that really determines how clean the transmitted signal is?

[Arie Kleingeld PA3A]

If we distort our SSB audio by compression etc, then let's hope that the
sideband filter will limit the bandwidth. (I don't know where the
compression or whatever is done in a K3).
But if the PA distorts the signal even more, there is no filter to limit
this within a few kHz.
So I think the IMD measurement might be important.

Let the guru's speak.....



73 Arie

_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[Julian, G4ILO]

Arie Kleingeld PA3A wrote

If we distort our SSB audio by compression etc, then let's hope that the
sideband filter will limit the bandwidth. (I don't know where the
compression or whatever is done in a K3).
But if the PA distorts the signal even more, there is no filter to limit
this within a few kHz.
So I think the IMD measurement might be important.

Let the guru's speak.....

OK, that's true about any compression that happens before the filter. But ALC compression certainly happens after. I think that any laboratory IMD measurement will be carried out without any ALC at all. In any event, ALC should not have an effect on a laboratory IMD test since the two tones are not varying in amplitude, unlike a real SSB or PSK signal. So the laboratory IMD test bears little relationship to the real world.

I think that the characteristics of a transmitter's ALC may have more effect on the signal than any laboratory IMD figure. To give an example: in the Rev 2 K2 firmware, a modification was made to the ALC characteristic in data modes that allowed ALC to be used to control TX power without introducing unacceptable IMD products. This suggests to me that complex characteristics of the ALC will have more effect on signal width in real-world SSB operation than small differences in PA linearity that are reflected in the IMD measurements that only relate to the transmission of two constant amplitude tones. Too slow ALC and the PA could clip, causing severe splatter. Too fast and it will operate like compression creating many IMD products.

But I am not an expert and am interested in what they have to say to this hypothesis as well.

[Geoffrey Mackenzie-Kennedy]

 Julian G4ILO wrote:

> In any event, ALC
> should not have an effect on a laboratory IMD test since the two tones are
> not varying in amplitude, unlike a real SSB or PSK signal. So the
> laboratory
> IMD test bears little relationship to the real world.

Perhaps. In the case of two constant level test tones driving the
transmitter, the level of intermod products seen at the output can change
when the ALC activated. The usual cause of this problem is that the
linearity of the low level stage or stages whose gain is controlled by the
ALC gets "worse" when the ALC starts to throttle them back. How much of an
effect occurs depends on the PA's  natural IMD behaviour, the exciter's IMD
and the IMD behaviour of the controlled stage(s) vs applied ALC.

> This suggests to me that complex characteristics
> of the ALC will have more effect on signal width in real-world SSB
> operation
> than small differences in PA linearity that are reflected in the IMD
> measurements that only relate to the transmission of two constant
> amplitude
> tones. Too slow ALC and the PA could clip, causing severe splatter. Too
> fast
> and it will operate like compression creating many IMD products.

The design of ALC systems is not a casual matter especially when dealing
with human speech, and the points which you raise have to be addressed by
the designer if the ALC might be used improperly. It would appear that ALC
systems are too often used as speech processors rather than the protective
system as intended.

As a design tool and as a means for specifying transmitter odd order IMD
performance, the two tone test at least removes the variables of human
speech. When I was involved in writing Canadian Type Approval specs for
commercial SSB / ISB equipment, transmitter IMD performance was measured
with the test tones applied to the transmitter's input interface from
external generators, which was and still is the accepted practice in the
comms field. Common practice at work was to design a PA and driver stage
whose odd order IMD products relative to PEP were comfortably better than
spec at various power output levels, with both 50 ohm and "mismatched"
loads. ALC was then added later on in the design process as a protective
system for the PA.

There is an inherent problem with the type of ALC systems, when misused,
which sample the PA's output to control the gain of low level stages, which
is that splatter and other crud resulting from overdriving have been
transmitted before the ALC can reduce drive. Those ALC systems which sample
further back in the exciter chain and control the gain of a following stage
can remove these problems. The mirror image is a receiver's AGC system.

73,
Geoff
GM4ESD

_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[Julian, G4ILO]

Geoffrey Mackenzie-Kennedy wrote

 Julian G4ILO wrote:

The design of ALC systems is not a casual matter especially when dealing
with human speech, and the points which you raise have to be addressed by
the designer if the ALC might be used improperly. It would appear that ALC
systems are too often used as speech processors rather than the protective
system as intended.

Very true. However much they obsess over published performance figures, most hams when they get on the air are more interested in how far they can talk up that RF out meter so they get heard by the DX, than whether they are operating within the limits for minimum IMD products. :)

Years ago I had an outboard Datong RF Clipper and it was possible to set the output level of that (which was the mic input of the rig) so that the ALC barely moved. The K3 is the first radio I have had that has "true RF speech compression" but it doesn't seem to be possible to set it up so that the RF compressor does all the work. The method recommended by Wayne for setting the K3 up is to turn the compressor to zero, increase the mic gain until you get a few bars of ALC on peaks, and then turn up the compressor. So you are applying ALC to an already compressed signal.

[KK7P]

There is apparently some confusion about how the K3 ALC works.  Just to
help muddy the waters a bit, I'll toss in what little I know about it.

We recommend setting the Mic (or LINE IN) Gain so you get 5 bars of ALC,
with possibly the 6th bar occasionally flickering, with CMP set to 0.

No, this is not to ensure you have Tx distortion!  Just the opposite, in
fact.

In this case, the ALC meter display is acting as a "VU Meter" and is
being used to help you set the Tx audio path gain correctly.  It is not
showing you how much you are over-driving the PA.  This is why you can
set it up in Tx test mode, when you are not transmitting any RF at all.

If you have a little too much drive, the DSP will scale things back at
the output of the mic amp.  Thus at the 5th ALC bar, the DSP is at the
threshold of gain compression.  At the 6th bar, it is cutting back gain
by 6 dB or so.

Assuming SSB, after the Tx signal is created from the audio, it
optionally goes through an IF clipper. This is engaged by setting CMP to
a value other than 0.

After the optional clipper, the signal goes through a 15 kHz IF filter
followed by a gain section in the DSP, and is then output as a 15 kHz
signal to the rest of the Tx strip.

 From  this point, the MCU watches things and if it sees power levels
going awry, severe mismatch, etc., it cuts back the drive to the Tx
strip in the DSP.  Thus, the Tx strip is running at nominal gain
regardless of the power level.  OK, if you go below 12 watts you are
using the LPA and not the HPA, so the Tx strip gain is less, but you get
the idea.  What we are *not* doing is varying the bias on an IF
amplifier somewhere that might change its operating point and possibly
introduce additional distortion.  The point in the DSP path where the
change in drive is commanded is linear to the limits of 32-bit floating
point math and the resolution of the Digital to Analog converter.

Enjoy!

73,

Lyle KK7P

[ Note: For PSK and other operation, you want to set things to the level
where the 5th bar just comes on -- or just doesn't.  If you have only 4
bars in digital modes, that's OK. With less than 4, you are under-
driving things and the ALC will kick in to bring up the power level to
the amount you requested.  If you try and trick the system and run with
*zero*  ALC bars in PSK31, the ALC has to work hard to give you your
requested power - or maybe it will think there is insufficient signal to
work with - and your Tx S/N may suffer.  So, *please* set it to 4 bars
in soundcard "digital" modes. ]

_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[Bob Cunnings NW8L]

Hmmm. Your advice to:

 "...*please* set it to 4 bars in soundcard "digital" modes..."

contradicts the direction given here (in step 7):

http://www.elecraft.com/K3/k3_operating_tips.htm#psk%20audio

viz. "You should not see any ALC bars on the K3 display (when set to CMP/ALC)"

This always confused me since it didn't agree with the manual, but
your explanation helps a lot, and I'll do it that way now.

Bob NW8L

On Sun, Apr 6, 2008 at 1:06 PM, Lyle Johnson <[hidden email]> wrote:

> There is apparently some confusion about how the K3 ALC works.  Just to help
> muddy the waters a bit, I'll toss in what little I know about it.
>
>  We recommend setting the Mic (or LINE IN) Gain so you get 5 bars of ALC,
> with possibly the 6th bar occasionally flickering, with CMP set to 0.
>
>  No, this is not to ensure you have Tx distortion!  Just the opposite, in
> fact.
>
>  In this case, the ALC meter display is acting as a "VU Meter" and is being
> used to help you set the Tx audio path gain correctly.  It is not showing
> you how much you are over-driving the PA.  This is why you can set it up in
> Tx test mode, when you are not transmitting any RF at all.
>
>  If you have a little too much drive, the DSP will scale things back at the
> output of the mic amp.  Thus at the 5th ALC bar, the DSP is at the threshold
> of gain compression.  At the 6th bar, it is cutting back gain by 6 dB or so.
>
>  Assuming SSB, after the Tx signal is created from the audio, it optionally
> goes through an IF clipper. This is engaged by setting CMP to a value other
> than 0.
>
>  After the optional clipper, the signal goes through a 15 kHz IF filter
> followed by a gain section in the DSP, and is then output as a 15 kHz signal
> to the rest of the Tx strip.
>
>  From  this point, the MCU watches things and if it sees power levels going
> awry, severe mismatch, etc., it cuts back the drive to the Tx strip in the
> DSP.  Thus, the Tx strip is running at nominal gain regardless of the power
> level.  OK, if you go below 12 watts you are using the LPA and not the HPA,
> so the Tx strip gain is less, but you get the idea.  What we are *not* doing
> is varying the bias on an IF amplifier somewhere that might change its
> operating point and possibly introduce additional distortion.  The point in
> the DSP path where the change in drive is commanded is linear to the limits
> of 32-bit floating point math and the resolution of the Digital to Analog
> converter.
>
>  Enjoy!
>
>  73,
>
>  Lyle KK7P
>
>  [ Note: For PSK and other operation, you want to set things to the level
> where the 5th bar just comes on -- or just doesn't.  If you have only 4 bars
> in digital modes, that's OK. With less than 4, you are under- driving things
> and the ALC will kick in to bring up the power level to the amount you
> requested.  If you try and trick the system and run with *zero*  ALC bars in
> PSK31, the ALC has to work hard to give you your requested power - or maybe
> it will think there is insufficient signal to work with - and your Tx S/N
> may suffer.  So, *please* set it to 4 bars in soundcard "digital" modes. ]
>
>  _______________________________________________
>  Elecraft mailing list
>  Post to: [hidden email]
>  You must be a subscriber to post to the list.
>  Subscriber Info (Addr. Change, sub, unsub etc.):
>  http://mailman.qth.net/mailman/listinfo/elecraft
>  Help: http://mailman.qth.net/subscribers.htm
>  Elecraft web page: http://www.elecraft.com
>

_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[Buck - k4ia]

In a message dated 4/6/2008 3:07:22 P.M. Eastern  Daylight Time,
[hidden email] writes:
Note: For PSK and other operation,  you want to set things to the level
where the 5th bar just comes on -- or  just doesn't.  If you have only 4
bars in digital modes, that's OK.  With less than 4, you are under-
driving things and the ALC will kick in to  bring up the power level to
the amount you requested.  If you try and  trick the system and run with
*zero*  ALC bars in PSK31, the ALC has to  work hard to give you your
requested power - or maybe it will think there is  insufficient signal to
work with - and your Tx S/N may suffer.  So,  *please* set it to 4 bars
in soundcard "digital" modes. ]


Did you  mean for these instructions to apply to using the DATA A mode?  If
so, I  find this almost impossible to do - I have to crank the Line In level
all the  way up and set the MicroKeyer to almost maximum to get 4 bars on the  
ALC.   When I do, my output meter reads 50 watts even though the max  power
setting on the K3 is 20W.  That just doesn't seem right.

The  suggested "4 Bar" system does seem to work using SSB for the PSK.  


k4ia
"Buck" K3 #101
Fredericksburg, VA  




**************Planning your summer road trip? Check out AOL Travel Guides.    
  (http://travel.aol.com/travel-guide/united-states?ncid=aoltrv00030000000016)
_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[KK7P]

> The  suggested "4 Bar" system does seem to work using SSB for the PSK.  

Sorry for adding yet more confusion to the mix.  I am running firmware
that is in field test and should be released next week.  The statement I
made I believe to be true for the firmware I am running.  I haven't gone
back to the released level to see if it is true for that version.

I apologize for perhaps speaking too soon on the PSK issue.

73,

Lyle KK7P

_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[Joe Stofko-2]

Hello Lyle,

You said:  
        There is apparently some confusion about how the K3 ALC works. Just
to help muddy the waters a bit, I'll toss in what little I know about it.
<snip>

Thanks for doing so. Excellent explanation!

73,

Joe - W1AIU


_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com

[wrmoore]

In light of the Sherwood presentation referenced in the thread about K3 audio, I haven't seen much comment on the spectrum analyzer plot of the K3 transmit IMD.  Is this still an artifact of the (now fixed) firmware error mentioned by Eric earlier in this thread?

> The K3 firmware used for the original QST review did
> not turn off the
> K3's RF speech compressor clipper even when the
> compressor was set to
> zero for their TX IMD two tone test. This resulted
> in a significant
> increase in measured TX IMD when using an external 2
> tone source. We
> corrected this in a firmware update last month.

Or is this a currently existing situation with the K3?  I'm so from an expert on these things that I hesitate to even ask, but it sounds like Sherwood is saying the K3 is a "bad neighbor" in transmit IMD, and it looks like it from the plots he provides.  

73,
Randy, KS4L
K2 #337
K3 on order :-)

[Eric Swartz - WA6HHQ]

Yes - That firmware fix improved 3rd order TX IMD by 3-5 dB on the worst
case band (12M) and it is better than that on all other bands. We also
made other firmware changes in the gain of various stages to improve TX
IMD. Those changes have been in every firmware release for many months.

One note: Rob's presentation of that slide was meant to comment on and
was representative of ALL solid state 100W push-pull AB amplifiers. He
was not intending to focus on the K3. Just about every solid state ham
rig made in the past 20 years runs class AB and has a similar (or much
worse) 3rd order IMD output. We absolutely are -not- a bad neighbor :-)

73, Eric   WA6HHQ

wrmoore wrote:

> In light of the Sherwood presentation referenced in the thread about K3
> audio, I haven't seen much comment on the spectrum analyzer plot of the K3
> transmit IMD.  Is this still an artifact of the (now fixed) firmware error
> mentioned by Eric earlier in this thread?
>
>  
>> The K3 firmware used for the original QST review did
>> not turn off the
>> K3's RF speech compressor clipper even when the
>> compressor was set to
>> zero for their TX IMD two tone test. This resulted
>> in a significant
>> increase in measured TX IMD when using an external 2
>> tone source. We
>> corrected this in a firmware update last month.
>>    
>
> Or is this a currently existing situation with the K3?  I'm so from an
> expert on these things that I hesitate to even ask, but it sounds like
> Sherwood is saying the K3 is a "bad neighbor" in transmit IMD, and it looks
> like it from the plots he provides.  
>
> 73,
> Randy, KS4L
> K2 #337
> K3 on order :-)
>  

_______________________________________________
Elecraft mailing list
Post to: [hidden email]
You must be a subscriber to post to the list.
Subscriber Info (Addr. Change, sub, unsub etc.):
 http://mailman.qth.net/mailman/listinfo/elecraft   

Help: http://mailman.qth.net/subscribers.htm
Elecraft web page: http://www.elecraft.com