[hpsdr] SDR Dynamiic Range Considerations

[Rob Frohne]

Hi Bob, et. al., 

When you state below that the SDR-1000 does have a roofing filter, are
you are referring to the first order RC filter from charging the
capacitors for the sample and hold, and/or the low pass filter in the
sound card?  It seems from Tom Rauch's tests (see the cross post on this
reflector entitled, "I wonder how HPSDR will affect this?"), that the
SDR-1000 needs a better roofing filter in order to be competitive to
those receivers using crystal roofing filters.  Perhaps Tom's
observations and Phil's below both indicate that the SDR-1000 has some
of the same problems you suggest Mercury and Quicksilver may have
because the roofing filter in the SDR-1000 is not as good as it could be
and/or because of the A/D converter voltage sensing input. 

I think you do have a good point that the input to an A/D converter is
not the same as what we have been used to in a classical receiver.
While we may get wonderful 3rd or

I'm looking forward to reading more of Phil's tests.

73,

Rob, KL7NA


On Tue, 2007-02-20 at 14:36 -0500, Robert McGwier wrote:
> p.covington at gmail.com wrote:
> >
> > On Feb 18, 1:24 pm, Robert McGwier <rwmcgw... at gmail.com> wrote:
> >   
> >> I have a few comments, etc. If this falls far of the mark, please forgive.
> >>
> >> I am glad the DDC is being pushed back into the Cyclone II but I have
> >> not yet understood the need for large FIR filter. Have your tried to
> >> replace the CIC with Halfband filters? I am certain this will work
> >> better. But I suspect that the desire is to have multiple 192 ksps
> >> channels since one or more want to aim Mercury and/or Quicksilver at HF.
> >> I never once considered these as high performance HF engines and feel
> >> that any desire to do this has misunderstood some truly serious problems
> >> with doing that. I always envisioned it as a VHF, UHF, Microwave engine
> >> that could be used on HF but at some severe cost/penalty. If the
> >> problems I am about to describe are being witnessed, then it is clear
> >> there is a lack of understanding that needs to be overcome to proceed to
> >> realistic designs at minimal cost that achieve a reasonable objective.
> >>
> >> Please allow me to explain.
> >>
> >> Blocking dynamic range is an often touted figure of merit and it is an
> >> indicator of something and in my argument, it really does matter. IP3 or
> >> DR3 are both good figures of merit to describe the analog front-end of a
> >> receiver because they are both related to the fundamental transfer
> >> equation of the receive system. Sensitivity is another that is of great
> >> interest to all of us. Two tone tests for the SDR-1000 show it performs
> >> better with near in signals than anything I have ever seen SO LONG AS
> >> THE A/D is not saturated. However, A/D converters used in digital radios
> >> are not well described by the familiar formulas for their nonlinear
> >> transfer function. If multiple strong signals are allowed to enter the
> >> A/D converter, the relevant figure of merit is the A/D saturation limit.
> >> Below this limit, A/D converters are typically very linear, extremely
> >> so, because they are not power amplification things, they are voltage
> >> sensing things with large impedances - they produce practically no IM3
> >> all the way up to the saturation limit where they fail completely. A
> >> digital receiver will have IP3 and DR3 values that look very good, but
> >> in real life they can not be compared to "good old analog" receivers
> >> that will continue to function with input voltages high above the range
> >> indicated by DR3. For example, an analog receiver that is subjected to
> >> 10 signals at the DR3 level will produce IM3 for all combinations of
> >> these signals, but these spurious responses will be near the noise
> >> floor. A digital receiver will see the 10 signals sometimes add in
> >> amplitude to give a peak level that is 20 dB or more higher, so it may
> >> become heavily saturated and useless. At the present state of the art,
> >> digital receivers need roofing filters to limit the number of signals
> >> entering the A/D converter, but inside the roofing filter bandwidth they
> >> remain vulnerable to saturation effects when the band is crowded with
> >> very strong signals. Test procedures must take realistic account of this
> >> and they typically do not. The SDR-1000 DOES have a roofing filter. The
> >> sampling detector is a mixer and roofing filter in one circuit. The
> >> wider that roofing filter is, the greater the likelihood that you will
> >> have N signals add up and saturate the A/D.
> >>
> >> Mercury and Quicksilver will have NO Roofing filter at the A/D. You will
> >> be adding hundreds of signals up on HF and with probability one, they
> >> will sometimes add up super constructively. This will cause the envelope
> >> of your signal to exceed the dynamic range of ANY A/D you can put in
> >> front. This will be visible as a rise in the noise floor of the system.
> >> To limit this at all, bandpass filters for bands of interest will be
> >> needed and even there on a crowded open band, signals will add up and
> >> saturate the A/D and act like a wideband impulse that raises the noise
> >> floor of the system. However, the construction of these analog band pass
> >> filters is nontrivial and will have insertion loss and ripple. The
> >> absolutely beauty of the QSD or ISD system is that it has very low
> >> insertion loss, high IP3/DR3, high IP2, etc. what is critical is to get
> >> the operating point just right to get the noise floor of the mixer
> >> matched to the noise floor of the A/D system so you maximize the
> >> blocking dynamic range so you minimize these constructive adds of
> >> multiple signals that are so destructive to performance.
> >>
> >> I never thought the designers were targeting HF so I never brought all
> >> of this up before. I do believe the Mercury and QS1R will be useful as a
> >> wideband "surveillance" system on HF, probably better than ANY currently
> >> in operation. But as a high performance receiver in comparison to the
> >> analog ones or the SDR-1000 or any SDR offerings from commercial
> >> vendors, they will never be competitive. Any attempt to try to backload
> >> more selectivity after the problem has occurred will be costly of time
> >> and parts and will ultimately fail. The problem, the sum of myriad
> >> signals in the A/D, will have already occurred and any remedy is too late.
> >>
> >> Again, if I have completely missed the boat here, I apologize.
> >>
> >> Bob
> >> N4HY
> >>     
> >
> > Yesterday I tried to cause a rise in noise floor in the QS1R prototype
> > by hitting it hard out of band with a HP8640B signal generator.  I can
> > see a rise in noise floor when the signal generator is cranked up to
> > about +9 dBm.  That appears to be where the ADC saturates.  Of course,
> > if I set the QS1R attenuator at 12 dB then I cannot overload the ADC
> > with the 20 dBm max output of the HP840B.  Since I don't have multiple
> > signal generators to throw at QS1R all at the same time, I cannot make
> > any further determination.
> >
> > Do you have any method that you can suggest to test this?
> >
> > I disagree about Mercury or QS1R not being able to compete with the
> > SDR-1000.  Side by side using PowerSDR, I see very little difference
> > in performance (either with the D44 or JANUS connected to the
> > SDR-1000) other than the SDR-1000 gives more spurs than the QS1R
> > especially above 20 MHz.
> >   
> Oh my goodness the SDR-1000's AD9858 is pretty bad and the LO leakage 
> and subharmonics are objectionable.  Sounds like QS1R and Mercury are 
> going to be great tools to have.   I am just as happy to be completely 
> wrong here but I will cogitate on ways to test.
> 
> Since you are using libusb,  I see no reason this should prevent us from 
> having code running for both of them on Linux for Ozy/Janus/Mercury/QS1R 
> in a short period of time.
> > Phil N8VB
> >
> >
> > >
> >
> >   
> 
> 
-- 
Rob Frohne, Ph.D., P.E.
E.F. Cross School of Engineering
Walla Walla College
100 SW 4th Street
College Place, WA 99324
(509) 529-3585
http://www.wwc.edu/~frohro/