[hpsdr] Odyessey-Siren Rev B - 180 vs 90 degree conversion loss

[Tayloe Dan-P26412]

I think the only difference between the two may be the input impedance
seen into the detector.  However both approach sum a large number of
average pulses.  From a detection point of view, both are the same.  A
series of half cycle pulses results in lower conversion gain than a
series of quarter cycle pulses as the average peak contribution is
different due to the sampling period.

Just perform a mental experiment.  Picture in you mind a train of half
wave pulses center on the peak of the sine wave, and a train of 1/4
cycle pluses centered on the peak of the sine wave.  Which is going to
give a higher detected voltage?  Since the voltage on the detection caps
only change when a pulse is presented to it and does not discharge
between pulses, the higher average signal shape will produce a higher
detected voltage.  Conceptually this is very simple.  You can simulate
the circuit in LTSpice and see that this is indeed what happens.

Also, I don't think we are discussing matched losses here.  At least in
the original implementation, the detector is not at all matched with the
source.  The source provides the "R" factor in what appears to be an R/C
low pass filter.  That is not a matched situation.  What this is doing
is providing a detected voltage or a detected current depending on the
op-amp connection to the detector.

The lowest noise configuration is a very low noise op-amp as a pre-amp.
Because of the effect of the inverting and non-inverting terminals, the
+ sees a detected open circuit voltage, and the - side sees a detected
short circuit current.  Open circuit voltage and short circuit current
are duals of each other, but, as you can see, the situation can hardly
be seen as matched.  The disadvantage to the single op-amp approach is
that it provides a flatter frequency response.  This is good for SDR
applications, but bad if you are using it for a traditional receiver and
are trying to get as much frequency discrimination as early in the
signal chain as possible.

I am uncertain of the advantage of the active integrator.  It may be
that this provides a flatter frequency response.  Has anyone built and
compared both approaches?  Four detection caps with four separate
integrators is inherently more noisy than two differential op-amps
because the differential op-amp is proving a 1x op-amp noise
contribution while combining two differential input signals. Thus, a 2x
combined signal is produced containing only a 1x op-amp noise
contribution. If one integrating op-amp is required for each of the four
detection outputs, each op-amp provides a 1x op-amp noise contribution
for a 1x signal output.  Thus when differentially combined in the
following stage, a 2x combined signal is produced containing 1.414x
op-amp noise contribution (noise combines with the square root of the
sum of the squares). 

In addition, if the integration approach does not have gain, it will
introduce additional excess noise, as the noise contribution of the
following combining op-amp will need to be added as well, further
degrading the sensitivity of the detector.

This is why I use a single normal ultra-low noise op-amp in my designs
rather than an instrumentation op-amp, which is internally three op-amps
configured much the same as the proposed integrator scheme.  The
instrumentation approach provides for better frequency roll off, but is
more noisy.

- Dan, N7VE


-----Original Message-----
From: Ahti Aintila [mailto:oh2rz.sdr at gmail.com] 
Sent: Wednesday, December 20, 2006 6:06 AM
To: Tayloe Dan-P26412; Robert McGwier
Cc: hpsdr at hpsdr.org
Subject: Re: [hpsdr] Odyessey-Siren Rev B - 180 vs 90 degree conversion
loss

Hi Dan,

Finally, I think that I understood the essential difference between the
passive integrator and the active integrator. In the passive integrator
as your original Tayloe Detector is, you really get the average voltage
during the sampling times. In the active integrator the resistor and
capacitor are connected to the virtual ground potential of the opamp.
That way you get the integral of the sampled charges in the capacitor.
Over the big number of successive charge pulses the voltage over the
capacitor follows closely the modulated envelope shape and amplitude.

Would that be an acceptable explanation?

73, Ahti OH2RZ

On 20/12/06, Ahti Aintila <oh2rz.sdr at gmail.com> wrote:
> Dan,
>
> Sorry the low clocking speed of my brains, I need more time to
understand.
>
> You mention "each of the two detection caps" in the case of the 180 
> degree switching. Actually I'm using four capacitors like the 90 
> degree circuits. That should increase the detected voltage.
>
> Then another point. We cannot study the case during one RF cycle only.
> Our integrators have high time constants, so within one RF cycle the 
> voltage of the integrating capacitor cannot change very much, but 
> during the thousands of cycles of the modulated envelope the voltage 
> can reach close to the peak value of the input signal, taking into 
> account the loss in the source resistance and assuming high load 
> impedance of the capacitors.
>
> To my layman's understanding, in the matched case the losses should be

> same for the 90 degree and 180 degree circuits. The main difference is

> that in the 90 degree case you have 4 switches charging the capacitors

> 90 degrees minus the rise and fall time of the switches, and in the 
> 180 degree case the 4 switches carry the current 180 degrees minus the

> same rise and fall times.
>
> 73, Ahti OH2RZ
>
>
> On 20/12/06, Tayloe Dan-P26412 <Dan.Tayloe at motorola.com> wrote:
> > ***** High Performance Software Defined Radio Discussion List *****
> >
> > The detector integrates the fraction of the RF pulse onto the
detection cap.
> >  Physically, this integration is done using the R/C low pass filter 
> > effect of R, the system impedance, and C which is the detection cap.

> > The loss appears across the system impedance and is not visible at 
> > the input to the detector since the voltage drop across the system 
> > impedance has already happened.  By selecting C we affect the roll 
> > off of the R/C low pass filter and thus C is relative small for an 
> > SDR front end (0.01 uf - small RC time constant = wider bandwidth, 
> > 10's of KHz wide), and relatively large (0.82 uf - larger R/C time 
> > constant = narrower bandwidth; ~1 KHz wide) for a traditional analog
DC phasing receiver such as the NC2030.
> >
> >
> >
> > Integrating 180 degrees of the RF waveform onto each of two 
> > detection caps mathematically gives square root of 2 times the peak 
> > or 0.707x the voltage, a 3 db detection loss.  Evaluating the 
> > integration of a sine wave from 0 to 180 degrees is a 
> > straightforward mathematical exercise.  On the other hand, 
> > integrating over 90 degrees of the RF waveform on to each of four 
> > detection caps gives 0.9x the peak RF voltage, a 1 db loss.  This is

> > the same as evaluating the integration of a sine wave from 45 to 135
degrees, integration over the 90 degree section of a sine wave that
covers the peak.
> > Thus, these losses do have a mathematical basis.
> >
> >
> >
> > The fact that folks look at the input to the detector and see that 
> > 1v pk-pk RF waveform into the detector gives 1v pk-pk of detected 
> > base band audio are missing the fact that voltage drop (loss) that 
> > has already happen across the 50 ohm system impedance before the 
> > detector input. Thus it is easy to come to the conclusion that the 
> > detector has basically no conversion loss and is thus more of a 
> > "sample and hold" type process rather than an integrating process.
> >
> >
> >
> > - Dan, N7VE
> > _______________________________________________
> > HPSDR Discussion List
> > To post msg: hpsdr at hpsdr.org
> > Subscription help:
> > http://lists.hpsdr.org/listinfo.cgi/hpsdr-hpsdr.org
> > HPSDR web page: http://hpsdr.org
> > Archives: http://lists.hpsdr.org/pipermail/hpsdr-hpsdr.org/
> >
> >
>

 1166626528.0