Comparison: K3 and SDR 5000A

[Bill W4ZV]

KD8K:

 >The K3's close in numbers will be around the same is the 5000,
but the wider spaced measurements I expect to be much better
then the Flex-5000.

         John you are correct that the 5000 will have
constant IMD and BDR over the entire range of spacings.
But IMD does not appear to be *significantly* different
(I define significantly as >6 dB).

Quoting Eric:

Filter            20kHz  10kHz  5kHz  2kHz
200 Hz, 5 pole    100+   100+   100+   95
250 Hz, 8 pole    100+   100+   100+   95
400 Hz, 8 pole    100+   100+   100+   95
500 Hz, 5 pole    100+   100+   100+   94
1 kHz, 8 pole     100+   100+   100    94
2.7 kHz, 5 pole   100+   98      92    n/a
2.8 kHz, 8 pole   100+   100     93    n/a

...versus 96 dB for the 5000 at all spacings.

         In my opinion, BDR is a much more significant
issue.  The 5000 will be flat at 123 dB for all
spacings versus the K3 at:

"For the particular K3 I tested the phase noise on below, I measured a
typical BDR of 140-141dB at spacings 5 kHz and wider and about 138dB at
2 kHz."  (again quoting Eric)

~20 dB difference in BDR and/or Phase Noise (from my
previous post) will be very noticeable in the following
situations:

1.  During multi-transmitter operations (multiop or SO2R
in contests, Field Day, expeditions, etc).

2.  If you have a very close neighbor (either a ham or
a powerful broadcast station).

         So I agree with your comments in general, but I
feel they apply more to BDR and Phase Noise than IMD.

                                 73,  Bill  W4ZV

P.S.  And let's all not forget that there could be some
differences in what Elecraft measures versus others!





_______________________________________________
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

[John A. McCabe]

Bill Tippett wrote:
>         So I agree with your comments in general, but I
> feel they apply more to BDR and Phase Noise than IMD.
>  

Hi Bill,
Yes you are correct on that. For some reason I got the BDR and IMD
numbers mixed up. However, I do believe that one of the FAQ's mentioned
a 20KHZ dynamic range greater then 104db. That's still  pretty good :)

73,  John

_______________________________________________
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]

Hello Bill!

A few additional things to note:

WHAT THE ADC CONTENDS WITH

1a) In the K3 architecture, the 24-bit analog to digital converter (ADC)
is exposed only to signals that make it through the roofing filter. This
is typically less than a 3 kHz slice of spectrum, and may only be a
couple hundred Hz.

2a) In a quadrature sampling detector (or integrating sampling detector)
architecture, used by the SDR-1000/5000 series, the ADC is exposed to
all the signals in the detector passband, typically more than 200 kHz
wide even if the displayed passband is less.

SYNTHESIZER CONSIDERATIONS ARE ABOUT MORE THAN JUST PHASE NOISE

1b) In the K3, the synthesizer generates a single, clean signal for
mixer injection.  Subsequent oscillators/mixers (including the software
oscillators inside the DSP) are only dealing with band-limited signals
due to the roofing filter.

2b) In the typical QSD/ISD system, a quadrature signal is generated by a
DDS.  DDSes can have very low phase noise, but typically have numerous
spurs.  While mitigation techniques exist, the fact remains that there
are usually several spurs only 70 or 80 dB down inside a given 200 kHz
wide passband.  These spurs can mix with other signals.

IT TAKES TIME TO PROCESS SIGNALS

3a) Radios based on embedded DSP, like the K3, can take advantage of
processing techniques on a sample-by-sample basis.  This may translate
to reduced latency (the time it takes for a signal at the antenna to
appear in the speaker or headphones).

3b) Receivers relying on PCs for DSP deal with data in blocks, rather
than sample-by-sample.  This results in significant latency - often 100
to 200 milliseconds or even more -- regardless of the speed of the PC's CPU!

I hooked up an SDR-14 (using a dual core 3 GHz PC in block mode
processing for its DSP) to the IF output of my K3 (which uses
sample-by-sample processing in its embedded DSP) and listened to the
same signal.  The audio from the SDR-14 appeared in the speakers *much*
later than the audio from the K3.

In some cases, this may not be important.  In other cases (QSK CW comes
to mind), latency can be very important.

Enjoy!

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

[wayne burdick]

Lyle Johnson wrote:

> 2b) In the typical QSD/ISD system, a quadrature signal is generated by
> a DDS.  DDSes can have very low phase noise, but typically have
> numerous spurs.  While mitigation techniques exist, the fact remains
> that there are usually several spurs only 70 or 80 dB down inside a
> given 200 kHz wide passband.  These spurs can mix with other signals.

Just a quick follow-up to Lyle's posting.

The K3's synthesizer uses a DDS, too, but it's almost completely
isolated from the signal path, in two ways.

First, it is followed by a very narrow 4-pole crystal filter (about 2.5
kHz wide), which removes harmonics as well as both narrowband spurs
(over a carefully selected tuning range) and wideband Nyquist sampling
spurs.

Second, the DDS drives a PLL and very low-noise VCO. The VCO has a high
C-to-L ratio, and very little of the capacitance is represented by the
varactor diodes. This is accomplished by breaking the VCO tuning range
into 128 bands.

All of the above makes life difficult for your MCU firmware guy, but
the result is a *very* clean signal. And this code was finished a long
time ago    :)

73,
Wayne
N6KR


---

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

[Joe Subich, W4TV]

> 2a) In a quadrature sampling detector (or integrating sampling
> detector) architecture, used by the SDR-1000/5000 series, the
> ADC is exposed to all the signals in the detector passband,
> typically more than 200 kHz wide even if the displayed passband
> is less.

This is key in relation to IMDDR (mixing and spurious generation).
Any signal handling circuits must be able to handle the total
peak power of all signals within that 20) KHz window ... not just
the average power.  If the amplifiers/mixers cannot handle the
"sum of the peaks," there is a real chance that signals well off
frequency will generate IMD products on the operating frequency
and the IMDDR measured by a two tone process (Sherwood, ARRL, etc.)
will be significantly optimistic with a receiver design like the
SDR-1000/FLEX-5000 as it only measures third order (2F1-F2, 2F2-F1)
products and fails to consider fifth and higher order products
that result from three, four and more strong interfering signals.

This is a major issue on 160 meters where there may be several
strong AM broadcast signals just below the band or on 40 meters
(particularly in Europe) with multiple strong broadcasters IN
the band.  It's also an issue on 80/75 meters where there can
be multiple VERY STRONG "local" signals within 50 - 100 KHz
of the operating frequency.  

With a "roofing filter" design, mixers and amplifiers (other
than a preamplifier) need to contend with strong signals only
within +/- 2.5 KHz (2.7 x 1.6 / 2).  The number of potentially
interfering signals are reduced significantly with the roofing
filter design.

73,

   ... Joe, W4TV
   

_______________________________________________
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