[Elecraft] K3 and whats next from Elecraft

Tue Aug 20 01:28:36 EDT 2013

Warning:  rather long reply follows.
If you don't care, or have your mind made up, please skip this email.  Fred 
asked for some education, so here goes...  I'm not an expert, but these are 
my personal ramblings from what I think that I have learned.  If you don't 
agree, and the moderator doesn’t like this discussion here, maybe we can 
find another place to discuss?  Or directly?

Elecraft has done a WONDERFUL job taming the QSD/QSE "mixers" in the KX3.  I 
LOVE MINE.  I regularly tout it at meetings, including tonight.  I think 
that I "sold" at least one KX3 and one K3 or another KX3.

Having said that, the K3 and KX3 are subject to imperfections, just like any 
other analog circuitry.  Yes, the KX3 uses a QSD with "switches", but they 
have analog characteristics for an analog signal as well (on-state 
impedance, on/off switch times, off-state leakage, etc).  The op amps that 
follow them aren't perfect either.  There will be a little imbalance between 
the I and Q channels, either in amplitude or phase, due to component 
variations.  Kind of like the old phasing SSB days.  All these imperfections 
add up in a complex way, and need to be dealt with to reduce image issues, 
both on receive and transmit.  The KX3 does a remarkable job handling this. 
That's just one set of problems with "analog" radio circuitry.

The "holy grail" of SDR is to put the ADC as CLOSE to the antenna as 
possible.  The reason is simple:  the sooner you get the signal converted 
into the digital domain (represented by numbers), the fewer imperfections 
from analog world circuitry will affect that signal.  A digital mixer just 
multiplies two numbers, for example, a "LO" represented by a set of numbers 
describing a sine wave, and an input signal, possibly from the ADC.  Within 
reason (and decent design), the numbers CANNOT change due to component 
imperfections, therefore you will not see problems with a digital version of 
a mixer, where the resulting signal is sort-of, close-to, but not what you 
want.   Two times two should always equal four, not 3.9 once, 4.1 the next 
time, etc, due to component drift, or other issues.  Filters are a different 
issue, but still more consistent and reliably reproduced in DSP than in the 
analog world.

So, how can we get the HF signal right into the ADC?  That is tough, because 
the ADC itself has a limited dynamic range (and a generally poor noise 
figure).  Most SDRs use 12-bit to 16-bit ADCs, and sample from 66MHz to 
123MHz rates.  Dr. Mitola suggests in a book that good HF reception requires 
at least 130dB of dynamic range.  IIRC 130dB works out to about 22-bits of 
ADC conversion.  16-bit ADCs by themselves provide somewhat less than 96dB, 
12-bit ADCs provide slightly less than 72dB.  So, traditionally even the 
better SDRs needed some sort of filtering before the ADC - to hopefully 
reduce any strong out-of-band signals.  This may not work great on some 
bands, like 40M, where strong MW broadcast stations often appear just above 
or just below the ham band.

There is a "trick" that is real-world, and reliable, that can improve the 
dynamic range of an ADC, called oversampling, that provides "processing 
gain".  If you sample much faster than the signal of interest's frequency, 
and decimate those samples (throw out every so many samples), you can 
achieve additional dynamic range.  I believe you gain roughly 6dB (or one 
ADC bit) every time you decimate by a factor of four.

Up until recently, extremely fast and deep ADCs were too expensive for ham 
rigs, and it was hard to get more than about 20-24 dB processing gain.  The 
new Flex 6000 series has a 16-bit ADC that samples at 245Ms/s.  This allows 
them to have enough dynamic range to MAYBE not need filtering.  Notice that 
they do have filters built into the radio, that can be enabled or disabled. 
Regardless, the main analog components in the 6000 series before the ADC are 
only filters (which should contribute negligible distortion to the signal of 
interest), and a preamp, also optionally in or out-of-circuit.  I think even 
Flex has their fingers crossed regarding true dynamic range in the real 
world.

Despite the comment on the Fun Cube Dongle Pro site, I doubt they have an 
actual 32-bit ADC.  Even lower-frequency ADCs, such as in sound cards, do 
not achieve that level of success.  In fact, I doubt that power supply 
noise, PC board layout, and many other factors prevent audio-frequency sound 
cards from achieving even the 24-bits that many claim.  For RF at HF 
frequencies, 16-bit ADCs are about as good as it gets - that we can afford 
anyway.

Don't get me wrong, a DDC/DUC-based radio CAN have serious issues as well. 
Overload (poor dynamic range), bad noise figure of the ADC, not lining up 
the sample values right (value overflow), and a host of other issues. 
Filtering can be just as tricky (or more) in the digital domain as in the 
analog world.  And SDR hardware can be a pain to use and/or build.  Anytime 
a computer is involved, you are taking your chances.  Hi Hi.  Sometimes 
knobs are just plain easier to use.

Yes, there has been a lot of discussion about whether or not the K3 is an 
"SDR".  I personally believe that it is, because a significant part of the 
SIGNAL PROCESSING is done in an executing software program, and that program 
can be changed.  Not as easily as loading a new EXE in Windows, but it can 
be.  Even though the RF front-end follows a traditional analog superhet 
receiver design up to the second IF.  In my mind, there's no question that 
the KX3 is an SDR.  It is similar in very basic layout as an SDR Cube, 
Softrock/host computer, etc, in that it uses QSD/QSE "mixers", an Si570, and 
a dsPIC back-end that is fully reconfigurable.

I remember others here, including the author of much of the DSP code for the 
K3 and KX3 has previously stated that they also believe both rigs are SDRs. 
BTW, the Elecraft P3 IS A DDC RECEIVER, with a 14-bit ADC, sampling at about 
60Ms/s, and feeding an ASIC DDC chip, then a dsPIC chip for FFT and other 
DSP work.  Yes, it samples the 8.215MHz IF from the K3, but as Elecraft 
states, it can be used on other radios as well.  It does take advantage of 
the K3's RF front end, including preamps, filtering, etc.  I would NOT hook 
an antenna directly to it!

Spectrum analyzers are nice if you can afford one - many people can't.  I 
have both a homebrew W7ZOI one, and a commercial unit.  They usually don't 
have the sensitivity to look at typical weak HF signals, and they aren’t 
usually designed to demodulate anything, or at least not CW or SSB. 
Panadaptors/spectrum displays are really coming to the forefront of SDRs. 
In fact, I would argue the opposite of your comment.  Some hams are using 
SDRs, or at least hardware originally designed as an SDR, for many bench 
applications.  Spectrum analyzers, network analyzers, antenna analyzers, etc 
can be based on some of the better SDR hardware.  HPSDR Hermes, HiQSDR, even 
Softrocks, have all been used in those applications.

I'm not a "digitalist", nor am I an "analogist".  I have several radios: RTL 
dongles/upconverter, K3/P3, KX3, SDR-Cube, tons of Softrocks, SDR-IQ, Flex 
1000, SDR-Widget, SDR2GO, HiQSDR, HPSDR boardset, Flex 6500.  I know I 
forgot a couple.  I bought each for a reason, and don't disparage any of 
them.  I did finally sell my Icom IC-720a, a workhorse for over 20 years. 
Except for the clack-clacking!

I've just given talks on SDRs to two local Charleston, SC ham radio clubs 
within the last week (one tonight).  I've been asked whether someone should 
buy a K3 or a Flex 6x00.  To be honest, I told them if they wanted a radio 
to "use and rely on" in the next year or so, get a K3, or possibly a KX3. 
In about a year, the Flex 6000-series MAY be mature enough to change that 
advice.  As Jerry Pournelle used to say in Byte Magazine:  RSN (Real Soon 
Now).

I want to see more than just tiny little slices at a time.  For contests, I 
want to not only see how the present band looks, but hopefully multiple 
bands, so I can see how propagation is moving.  No, I'm not a very active 
contester, that's just one example.  I LOVE the K3/P3, and have used them on 
CW for Field day the last three years.  The Flex (any Flex) just isn't ready 
for that.

I'm not an EE, just someone who has played with leading-edge amateur radio 
for 45+ years.  I was involved in the first licensed ATV repeater (WR4AAG), 
packet radio development, spread spectrum, a little LF, and now SDR.  I know 
how to burn my fingers with an iron.

In summary: learn about Direct Digital Conversion (Digital Down Conversion) 
(DDC - whatever), look it up.  Read about it.  Don't stick you head in the 
sand, or you too may be carrying around one of those ancient Motorola 2M FM 
rigs with tubes in a Yaesu/Kenwood/Wouxun world.  There are definitely 
pitfalls to DDC/DUC radios.  But, there are several reasons why every "cell 
phone" has an SDR inside.

Sorry for the length.  To each his/her own.
73, Terry, WB4JFI

-----Original Message----- 
From: Fred Jensen
Sent: Monday, August 19, 2013 3:33 PM
To: elecraft at mailman.qth.net
Subject: Re: [Elecraft] K3 and whats next from Elecraft

I could use some education here please:

What is the advantage of sampling directly off the antenna?  I'm
surmising that's what is meant by "real SDR" vs "hybrid SDR".  I don't
know if 32-bit ADC's existed when the K3 was being designed, but
Elecraft settled on the hybrid superhet analog 1st IF/digital 2nd IF for
at least one reason and probably many.

The dynamic range of signal amplitudes at my coax connector across the 0
to 30MHz spectrum has got to be huge, even after LORAN-C shut down.  Is
there really any benefit to having that broad a spectrum with that large
an amplitude range in the number soup?

If you want to observe that spectrum, wouldn't it be better to use a
spectrum analyzer and save your communications receiver to be optimized
to deal with the tiny little slices we want 99.999% of the time?

If the K3 sampled at the 8MHz 1st IF, after the analog roofing filters
[i.e. RF amp and 1st mixer only], would it make the K3 a "real SDR" in
the eyes of those who don't see it that way now?

I'm serious about the "education" part, this "Is the K3 an SDR" thread
has been going, on and off, for more than a year. It seems that we have
two ideologies going, "Digitalists" vs "Analogists."  Unfortunately,
debating ideologies qualitatively is rarely productive.

My education is in math, my understanding of EE is the result of 60
years in ham radio colored by a math background.

73,

Fred K6DGW
- Northern California Contest Club
- CU in the 2013 Cal QSO Party 5-6 Oct 2013
- www.cqp.org

On 8/19/2013 11:38 AM, Arie Kleingeld PA3A wrote:

> We don't have to worry about processorpower. The hardware is already
> there and still developping. And Yes, sampling up to 6m is possible.
> Apart from that, the interfacing from the machine to us humans can
> certainly do better with some innovation (new ideas) as well.

> Op 19-8-2013 18:07, Lynn W. Taylor, WB6UUT schreef:
>>
>> Consider for a moment an SDR with no analog filtering at all.
>>
>> How wide would the passband be?  Are we going to cover everything from
>> MF to 6m?

______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:Elecraft at mailman.qth.net

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html 