Re: LF: RFspace SDR-IQ

[Andy Talbot <andy.g4jnt@googlemail.com>]

AH,  forgot that note only gave the results, and didn't include any explanation of the findings...

Perhaps the contents of this email should be added to it

 

The complete unpredictability - far more than just the IMP-3 asymmetry - was the whole point of making the measurements.   It was actually G3PLX who asked me to do them as he didn't have the test equipment.   What they  show is that the 'classic' third order linearity model is not applicable to  direct sampling SDRs and high speed A/D converters and a completely new approach is needed.

 

The fact shows up dramtically when you observe that the level of the third order products remains reasonably constant with varying two-tone input amplitude whereas conventionally you should see a 3dB/dB variation.

 

But then it changes dramatically when a third tone is introduced that cannot itself contribute to the IP3 tone being measured.   Peter spotted something like this and asked me to confirm with the more controlled measurements

 

One explanation we can think of is that there is no "real" third order product being generated at all, certainly not one above the A/D quantisation noise, but there is leakage from the digital lines.      A single tone into the A/D will give signal components on the digital A/D ouptuts that contain components at the input frequency and  its harmonics, which can leak into the RF path.

 

Two tone signals will include I/M sidebands as well within this spectrum and those on the Lowest Significant Bits will probably remain pretty constant whatever the input amplitude,  provided it is above the minimum quantising level.  When a third non related tone in intoduced, whatever its level, the LSBs will be jittered around a lot more, so reducing the level and changing the spectrum of teh leakage.   This effect is observed and can be seen in the measurements

 

All a bit empirical but if you web-search on "SDR Linearity"  you'll find a large number of papers and observations now; many showing similar results and offering similar conclusions.

 

So, in conclusion :-

We cannot use conventional analogue-receiver linearity definitions or measurement techniques on direct samplibg SDRs.

 

What we can safely say, is that direct sampling receivers will work best in the presence of multiple signals with a spread of amplitudes.  That will ensure the spectrum of RF leakage from the digital outputs will be noiselike with no discrete components and hence allow higher dynamic range.  In other words, just what you see by connecting to an antenna.

 

In fact the very highest specification top end A/D converters do optionally deliberately jitter the clock to spread out the leakage spectrum.  The jitter is taken out digitally by DSP within the A/D chip itself so the user sees a tranparant conversion, or this can be done subsequently by the user if preferred.   Probably a perusal of the Analog Devices web site www.analog.com  will reveal a plethora of papers on linearity specifications and results.

 

I believe a formalised route to DD Receiver specification is being developed and prbably has been by now, , but I have had no inclination to follow the story these days - that's all a bit too much like the work I used to do and couldn't wait to retire from.

Andy
www.g4jnt.com

On 30 October 2010 20:07, Clemens Paul <cpaul@gmx.net> wrote:

Hello Andy,

 

your test results show quite strong differencies between the

upper and lower IM3 product,up to 18dB.

Unsymmetrical IM3 products *always* imply  that there is more than one

IM3 producing source.

Maybe it's a good idea to check the inherent IM3 behaviour of your test

setup itself.

3dB-combiners have only 30-35dB port isolation,true hybrid combiners (6dB)

are better by at least 20dB,if they are made tunable up to 80dB isolation

is achievable.

Also the 70dB resistive isolation between the two crystal oscillators seems to

be a bit on the short side. 

 

73
Clemens
DL4RAJ

 

 

 

---- Original Message -----

From: Andy Talbot

To: rsgb_lf_group@blacksheep.org

Sent: Saturday, October 30, 2010 7:47 PM

Subject: Re: LF: RFspace SDR-IQ

http://www.g4jnt.com/SDRIQ_Linearity.pdf

 

Showing some of the minor peculiarities of DS SDRs

Andy
www.g4jnt.com

On 30 October 2010 18:04, Alan Melia <alan.melia@btinternet.com> wrote:

Hi Clemens, you seem to impute that there are no shortcommings in direct
sampling!
 :-))    I suspect there are, but they are just "different".  You may be
right about the units mentioned, but it is just as easy to make a poorly
preforming DSP radio as to make a poor analogue one. We are in danger of
being conditioned to accept that "digital" is the magic dust that solves all
problems......it is definitely not so!

Having said that, the performance of some units is quite remarkable and I do
own a couple very good conventional receivers.

Alan G3NYK

----- Original Message -----
From: "Clemens Paul" <cpaul@gmx.net>
To: <rsgb_lf_group@blacksheep.org>

Sent: Saturday, October 30, 2010 4:11 PM
Subject: Re: LF: RFspace SDR-IQ


> Tony,
>
> the most important difference between SDR-IQ/-IP(from RF Space)
> or Perseus and the $200 SDR on
> http://www.lazydogengineering.com/LD-1A_SDR.html
> is that  the latter is no direct sampling receiver.
> It uses a downconverting technique with a number of shortcomings,e.g.
> phase noise of the LO and others.
> I would go for a direct sampling SDR like those e.g. from RF space or
> Perseus.
>
> 73
> Clemens
> DL4RAJ
>
> ----- Original Message -----
> From: "Tony" <ei8jk@o2.ie>
> To: <rsgb_lf_group@blacksheep.org>
> Sent: Saturday, October 30, 2010 3:49 PM
> Subject: Re: LF: RFspace SDR-IQ
>
>
> > Thanks guys.
> >
> > You have just tipped the balance for me and I shall be placing an
> > order
> > for one on Monday morning.
> > It's about time I joined the 21st century, but it will be
> > interesting to
> > compare it to my beloved FT 102 rx.
> >
> > Tony, EI8JK.
> >
> >
> > On 29/10/2010 16:50, Tony wrote:
> >> Hello group.
> >>
> >> Has anyone any experience with this SDR ?
> >> It looks interesting and I'm a bit tempted, especially as it works
> >> full spec down to 500 Hz and "usable" to 100 Hz.
> >> But as I have never used an SDR, I haven't got a clue if it's any
> >> good
> >> or not.
> >>
> >>
> >>
> >
> >
>
>
> --------------------------------------------------------------------------
------
>
>
>
> Eingehende eMail ist virenfrei.
> Von AVG überprüft - www.avg.de
> Version: 9.0.864 / Virendatenbank: 271.1.1/3226 - Ausgabedatum:
> 10/29/10 20:34:00
>
>

---

Eingehende eMail ist virenfrei.
Von AVG überprüft - www.avg.de
Version: 9.0.864 / Virendatenbank: 271.1.1/3227 - Ausgabedatum: 10/30/10 08:34:00