What do we know that works? (2024)

Ford Peterson

I want to pose a series of questions in hopes that somebody has measured and analyzed the issues. My goal is to build a better mouse-trap (read: radio). The notion of using baseband audio to form the basis of detection and GUI has been documented to work and work very well. My ponderings relate to the use of the IQ method of mixing as compared to traditional superhets.

It appears to me that IQ method sensitivity (MDS) drops off with increasing frequency, with the current state-of-the-art maximum usable frequency in the neighborhood of 25MHz. Traditional mixing methods are virtually unrestricted in this regard. Since most of the amateur spectrum allocations lie above 25MHz, is using the IQ method for mixing a benefit or a distraction?

Traditional mixers have a difficult time with distortion free dynamic range. However, with the availability of substantially improved mixers (e.g. AD831 and others), the IP3 in excess of 100mW is well documented. In addition, the LO energy needed to achieve this has dropped to about -10dBm. This is a $9 part at Digikey. Historically, high power LO chains have been required to achieve >20dBm IP3, causing all sorts of problems with tweets and pops all over the place as the LO becomes a transmitter in its own right! Does the IQ method offer any hope of improvement in dynamic range?

Phase noise and spurious responses have been troublesome to control in oscillators. The more sophisticated the VFO (DDS and the like) seemingly the more difficult it is to make a nice clean LO chain. Since both traditional superhets and the IQ method of mixing both require the use of high quality oscillators, traditional mixers have the benefit of needing to tune relatively narrow frequency ranges by comparison to the IQ, which need to tune a 4:1 increase in range in order to work effectvely.

For many of the ham bands, noise is the road block to higher performance. For example, on 160M, noise will significantly limit the MDS in a traditional radio. By its very nature, the IQ method should be able to filter out noise better than a traditional mixing method. Is this true? Is the IQ mixer able to filter out noise better than a traditional mixer? Has anybody measured this?

Spurious responses can be somewhat controlled with the use of multiple IF stages. Hence we find a number of quad conversion radios on the market. To make things equal, the IQ method would also require multiple IFs in order to control the images and other unwanted characteristics of the single conversion radio. Is this not easier to perform using traditional mixers? Or does the IQ method offer some hope of simplification or improvements?

What are the benefits of using the IQ method? It seems to this tinkerer that building a high performance radio using baseband audio as the interface is the way to go. But are Tayloe detectors the path to nirvana? I am almost to the point of thinking the IQ method is a distraction from the path to higher performance radios. So I would like to identify the benefits before abandoning this methodology.

I eagerly await some educated comments on the matter.

Ford-N0FP
ford@...

KD5NWA

At 09:34 AM 3/15/2006, you wrote:

I want to pose a series ofquestions in hopes that somebody has measured and analyzed theissues. My goal is to build a better mouse-trap (read:radio). The notion of using baseband audio to form the basis ofdetection and GUI has been documented to work and work very well.My ponderings relate to the use of the IQ method of mixing as compared totraditional superhets.

It appears to me that IQ method sensitivity (MDS) drops off withincreasing frequency, with the current state-of-the-art maximum usablefrequency in the neighborhood of 25MHz. Traditional mixing methodsare virtually unrestricted in this regard. Since most of theamateur spectrum allocations lie above 25MHz, is using the IQ method formixing a benefit or a distraction?

It is capable of higher frequency with fast switches and careful design,it's benefit is higher efficiency, and very high IP3 capability.

Traditional mixers have adifficult time with distortion free dynamic range. However, withthe availability of substantially improved mixers (e.g. AD831 andothers), the IP3 in excess of 100mW is well documented. Inaddition, the LO energy needed to achieve this has dropped to about-10dBm. This is a $9 part at Digikey. Historically, highpower LO chains have been required to achieve >20dBm IP3, causing allsorts of problems with tweets and pops all over the place as the LObecomes a transmitter in its own right! Does the IQ method offerany hope of improvement in dynamic range?

The IQ method in itself doesn't nothing about that, using FETs asswitching devices instead of diodes increases the dynamic range of themixer.

Phase noise and spuriousresponses have been troublesome to control in oscillators. The moresophisticated the VFO (DDS and the like) seemingly the more difficult itis to make a nice clean LO chain. Since both traditional superhetsand the IQ method of mixing both require the use of high qualityoscillators, traditional mixers have the benefit of needing to tunerelatively narrow frequency ranges by comparison to the IQ, which need totune a 4:1 increase in range in order to workeffectvely.

For many of the ham bands, noiseis the road block to higher performance. For example, on 160M,noise will significantly limit the MDS in a traditional radio. Byits very nature, the IQ method should be able to filter out noise betterthan a traditional mixing method. Is this true? Is the IQmixer able to filter out noise better than a traditional mixer? Hasanybody measured this?

IQ detector is nothing but a mixer, it does not filter outnoise, using direct conversion with a clean oscillator however results insuperior performance, since it tends to have higher gain that diodemixers and that improves the S+N/N ratio a little bit.

Spurious responses can besomewhat controlled with the use of multiple IF stages. Hence wefind a number of quad conversion radios on the market. To makethings equal, the IQ method would also require multiple IFs in order tocontrol the images and other unwanted characteristics of the singleconversion radio. Is this not easier to perform using traditionalmixers? Or does the IQ method offer some hope of simplification orimprovements?

That is actually a curse, the more conversion stagesyou have the more opportunities for birdies, spurs and signal overload. Asingle conversion super-heterodyne can be more free of spurious signalsthan a equally designed quad conversion heterodyne, They get into fourstages and such to avoid image problems and in order to accomplishsome things like variable bandpass filters. I think they now stilldo it out of habits. Direct conversion receivers that are well designedtend to be very free of spurious signals, but like the heterodyne hasimage problems, that's where a Quadrature detector can come in, with Iand Q signals you can use hardware ala NC2030 or software ala SDR-1000,and SoftRocks to eliminate the image issue.

What are the benefits of usingthe IQ method? It seems to this tinkerer that building a highperformance radio using baseband audio as the interface is the way togo. But are Tayloe detectors the path to nirvana? I am almostto the point of thinking the IQ method is a distraction from the path tohigher performance radios. So I would like to identify the benefitsbefore abandoning this methodology.

By allowing a receiver to be single conversion, RF ---> Audio aQuadrature based receiver avoids the problems of multiple IF's and dealswith images as well. The result is that you have a simple receiver thatis free of spurious signals, sounds terrific, the best of allcombinations.

Hope it helps.

I eagerly await some educatedcomments on the matter.

Ford-N0FP
ford@...

Yahoo! Groups Links

<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/softrock40/

<*> To unsubscribe from this group, send an email to:
softrock40-unsubscribe@...

<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/

Cecil Bayona
KD5NWA
www.qrpradio.comI fail to see why doing the same thing over and over and getting thesame results every time is insanity: I've almost proved it isn't; only afew more tests now and I'm sure results will differ this time ...

--- In softrock40@..., "Ford Peterson" <ford@...> wrote:

I want to pose a series of questions in hopes that somebody has

measured and analyzed the issues. My goal is to build a better
mouse-trap (read: radio). The notion of using baseband audio to form
the basis of detection and GUI has been documented to work and work
very well. My ponderings relate to the use of the IQ method of mixing
as compared to traditional superhets.

It appears to me that IQ method sensitivity (MDS) drops off with

increasing frequency, with the current state-of-the-art maximum usable
frequency in the neighborhood of 25MHz. Traditional mixing methods
are virtually unrestricted in this regard. Since most of the amateur
spectrum allocations lie above 25MHz, is using the IQ method for
mixing a benefit or a distraction?

Traditional mixers have a difficult time with distortion free

dynamic range. However, with the availability of substantially
improved mixers (e.g. AD831 and others), the IP3 in excess of 100mW is
well documented. In addition, the LO energy needed to achieve this
has dropped to about -10dBm. This is a $9 part at Digikey.
Historically, high power LO chains have been required to achieve

20dBm IP3, causing all sorts of problems with tweets and pops all

over the place as the LO becomes a transmitter in its own right! Does
the IQ method offer any hope of improvement in dynamic range?

Phase noise and spurious responses have been troublesome to control

in oscillators. The more sophisticated the VFO (DDS and the like)
seemingly the more difficult it is to make a nice clean LO chain.
Since both traditional superhets and the IQ method of mixing both
require the use of high quality oscillators, traditional mixers have
the benefit of needing to tune relatively narrow frequency ranges by
comparison to the IQ, which need to tune a 4:1 increase in range in
order to work effectvely.

For many of the ham bands, noise is the road block to higher

performance. For example, on 160M, noise will significantly limit the
MDS in a traditional radio. By its very nature, the IQ method should
be able to filter out noise better than a traditional mixing method.
Is this true? Is the IQ mixer able to filter out noise better than a
traditional mixer? Has anybody measured this?

Spurious responses can be somewhat controlled with the use of

multiple IF stages. Hence we find a number of quad conversion radios
on the market. To make things equal, the IQ method would also require
multiple IFs in order to control the images and other unwanted
characteristics of the single conversion radio. Is this not easier to
perform using traditional mixers? Or does the IQ method offer some
hope of simplification or improvements?

What are the benefits of using the IQ method? It seems to this

tinkerer that building a high performance radio using baseband audio
as the interface is the way to go. But are Tayloe detectors the path
to nirvana? I am almost to the point of thinking the IQ method is a
distraction from the path to higher performance radios. So I would
like to identify the benefits before abandoning this methodology.

I eagerly await some educated comments on the matter.

Ford-N0FP
ford@...

What exactly do you mean by the "IQ Method"? Are you talking about
the QSD (Tayloe) based samplers/mixers? You can't get away from I and
Q no matter how hard you try (even in the analog world) - all signals
have a magnitude and phase representation.

73 de Phil N8VB

Ford Peterson

What exactly do you mean by the "IQ Method"? Are you talking about
the QSD (Tayloe) based samplers/mixers? You can't get away from I and
Q no matter how hard you try (even in the analog world) - all signals
have a magnitude and phase representation.

73 de Phil N8VB

My line of query relates to the Tayloe based mixer...

Ford-N0FP
ford@...

Ford Peterson

Cecil-KD5NWA wrote about the Tayloemixer:

> It is capable of higher frequency withfast switches and careful design, it's benefit is higher efficiency, and veryhigh IP3 capability.

FP: Cecil, has anybody actually measured thisand posted it someplace where a fella can look and compare? I would beinterested to know how much better and how much higher the efficiency. Inmy illustration I talked about the AD831 ($9 at Digikey) and find that usingonly -10dBm of LO drive provides +24dBm IP3 and +10dBm 1dB Compressionpoint. Is it better than that?

> IQ detector is nothing but a mixer, itdoes not filter out noise, using direct

> conversion with a clean oscillator however results in superiorperformance,

>since it tends to have higher gain that diode mixers and thatimproves the S+N/N ratio a little bit.

FP: I find this curious. The Tayloedetector/mixer appears quite similar to a switched capacitor filter. Thesampling rate will dictate accuracy of the filter. But you say there is nofiltering benefit. This begs the question, why not? Is it that weare fixated on the Nyquist limit? (e.g. 2x frequency) Does it seemplausable that sampling at substantially higher rates would result in filteringaction? For those not familiar with the switched capacitor filter, atypical application would be the Tayloe approachapplied to audio.E.g. if you want a 700Hz tone, you sample a set of capacitors at 700Hz using acommutator of sorts. Tones appearing at 700Hz will basically charge eachof the capacitors to a certain level and remain there. Each successivecyle will result in the same phase on the capacitor and thereforetransformation. Random noise, and undesired tones,will fallout-of-sync with the commutator sequence and disappear. In this approach, having 10 capacitors separated in the commutator by 36 degrees has the effect of10x oversampling of the waveform. It has been some time since I playedwith one, but it seems to me that the higher the oversampling rate, the greaterthe immunity to noise. Is this the case or am I missingsomething?

FP continued: "...we find a number of quad conversion radios on themarket. "

> That is actually a curse, themore conversion stages you have the more opportunities

>for birdies, spurs and signal overload. A single conversionsuper-heterodyne can be more

>free of spurious signals than a equally designed quad conversionheterodyne, They get

>into four stages and such to avoid image problems and in order toaccomplish some things

>like variable bandpass filters. I think they now still do it outof habits. Direct conversion

>receivers that are well designed tend to be very free of spurioussignals, but like the heterodyne

>has image problems, that's where a Quadrature detector can comein, with I and Q signals

Cecil continued:

> By allowing a receiver to be single conversion, RF ---> Audio aQuadrature based receiver

>avoids the problems of multiple IF's and deals with images aswell. The result is that you have a

>simple receiver that is free of spurious signals, soundsterrific, the best of all combinations.

FP: Agreed... A very "simple"receiver. My dictionary says simple (def 4) "without embellishment" and(def 5) "not elaborate, elegant, or luxurious." It seems to me that thesupport hardware needed to make the Tayloe work well quickly becomes VERYcomplex when you attempt to apply all the traditional notions of what makes agood radio--like accurate frequency agility over many octaves--like good dynamicrange--like noise fighting ability--like spurious free passbands.

I don't want to sound like the devil's advocatehere. I am clearly impressed with the performance of a simple radio likethe softrock operating in any 48KHz of any band. I want to understand thelimits of this design method. To date, I see and hear a lot of chestbeating about the greatness of it all. I'm from Missouri on this--"Showme."

Thanks for the comments & 73

Ford-N0FP

I think you equate I/Q with QSD, which is not the case. For example,
the R2 and the binaural receivers use 2 DBM and produce I/Q output.
Also, IF DSP radios use the I/Q method: the first thing done by the
DSP is to apply a transform on the input signal to extract I & Q. For
example the Pic-A-Star uses the same chip as the QSD, but in a
conventional mixer configuration and the DSP extract I & Q.

The problem is the QSD which runs out of gas as the frequency
increases, but it is just because the parts used are bus switches and
have never been designed to be used in a Softrock. It should be
possible to design a better part for SDRs but I guess the market is
too small.

Jean-Claude, PJ2BVU

--- In softrock40@..., "Ford Peterson" <ford@...> wrote:

Cecil-KD5NWA wrote about the Tayloe mixer:

................

IQ detector is nothing but a mixer, it does not filter out

noise, using direct

conversion with a clean oscillator however results in superior

performance,

since it tends to have higher gain that diode mixers and that

improves the S+N/N ratio a little bit.

FP: I find this curious. The Tayloe detector/mixer appears quite

similar to a switched capacitor filter. The sampling rate will
dictate accuracy of the filter. But you say there is no filtering
benefit. ............
Ford, Cecil said "it does not filter out noise" (I am not sure why
Cecil mentions it). The Tayloe detector as a filtering action and if
you read Dan's original article he was not trying to invent a
detector but to design a filter.

FP: I think the primary benefit behind the superhet is the

ability to better manage the image frequency, and provide frequency
accuracy.

As the QSD permits the detection of amplitude and phase, the
software knows that the image is an image and can ignore it. As it
is also a digital filter, aliases will appear and the software will
have no way of knowing it. A bandpass filter ahead of the detector
should take care of it.
The frequency has nothing to do with the type of receiver. It all
depends on the oscillator(s).

Obviously, a dirty LO in any receiver will result in unpredicible

spurs and tweets. Frequency agility is an obvious problem with the
Tayloe, as is the powerful fundamental and image frequency.

What do you mean by "frequency agility is a problem with the QSD"?
Regarding the powerful fundamental, I suppose you refer to the 0Hz
spike. There is not much we can do about it but ignore it. The
SDR1000 tune signals around 11.5KHz in order to be far enough from
the 0Hz junk.

Of the two methods used in the Tayloe, either method creates a

serious headache when you attempt to apply the methods into upper HF
or low VHF. Using the phase shift approach, you have an endless
number of optimum delay-line components (one set for every frequency
on the dial). Using the Flip-Flop approach, the delays are managed
to perfection, but this necessitates the use of very high speed ECL
as clock rates become impossible to produce. Where is the
simplicity in all of this?

It sure looks like the operation on VHF is a problem - by the lack
of components, not because the principle is flawed. I have been
thinking of a 50MHz SDR for a while and it seems like one way to go
is using a softrock as an IF and a downconverter. The other would be
to use DBMs a la R2 and a DDS prividing I & Q and 50MHz would be the
upper limit with DDS available currently.

Jean-Claude, PJ2BVU

Ford Peterson

Jean-Claude, PJ2BVU wrote:

Ford, Cecil said "it does not filter out noise" (I am not sure why
Cecil mentions it). The Tayloe detector as a filtering action and if
you read Dan's original article he was not trying to invent a
detector but to design a filter.

So there is some noise filtering. My keen interest is for the low bands, since atmospheric noise seems to dominate. But with only a Quad sampling, I'm not sure much benefit results. That was why I was inquiring. So my line of inquiry moves towards a higher sampling rate. Has anyone done work on this? Is this new territory? I certainly don't want to re-invent a broken wheel.

FP: I think the primary benefit behind the superhet is the

ability to better manage the image frequency, and provide frequency
accuracy.

JC:

As the QSD permits the detection of amplitude and phase, the
software knows that the image is an image and can ignore it. As it
is also a digital filter, aliases will appear and the software will
have no way of knowing it. A bandpass filter ahead of the detector
should take care of it.
The frequency has nothing to do with the type of receiver. It all
depends on the oscillator(s).

Obviously, a dirty LO in any receiver will result in unpredicible

spurs and tweets. Frequency agility is an obvious problem with the
Tayloe, as is the powerful fundamental and image frequency.

What do you mean by "frequency agility is a problem with the QSD"?
Regarding the powerful fundamental, I suppose you refer to the 0Hz
spike. There is not much we can do about it but ignore it. The
SDR1000 tune signals around 11.5KHz in order to be far enough from
the 0Hz junk.

FP: I have seen report after report where the detector seems to fall apart above high HF. Output becomes low, the need for preamps, etc. Also, clocking these little devils is a pain when you need to QSY great distances. The phase shifting network is very frequency dependant. Let's assume 1MHz to 50MHz is wanted. Using a clock at the frequency of interest and a phase shift network would take very awkward variable components to adjust. Using the 4X method with a pair of flip-flops suggests a 200MHz clock is needed at 6M and a 4MHz clock at the low end. That's 50 octaves! Ouch! Alternatively, a pair of UHF/SHF oscillators could be swept, like in an analyzer, and extract the difference to use as a clock that would remain in proper quadrature over frequency. Suddenly, the notion of 'simple' is taking on an all new meaning. That is what I meant by frequency agility.

JC:

It sure looks like the operation on VHF is a problem - by the lack
of components, not because the principle is flawed. I have been
thinking of a 50MHz SDR for a while and it seems like one way to go
is using a softrock as an IF and a downconverter. The other would be
to use DBMs a la R2 and a DDS prividing I & Q and 50MHz would be the
upper limit with DDS available currently.

Jean-Claude, PJ2BVU

FP: My ponderings have been along those lines, but lately I'm thinking about picking an IF frequency where a guy can locate some saw filters in the 200KHz - 400KHz bandwidth. Maybe at 45 MHz? 70MHz? I am still pondering. Then use DBM convert the 1st IF down to a 2nd IF that is very simple for the Tayloe--perhaps 2.5MHz? 10.7? Use a very high quality clock run at 4x the 2nd IF and a pair of flip-flops to generate the IQ. Suddenly, you have an all band receiver. A series of converters to the first IF and you have a complete receiver. A hybrid of conventional clocks and mixers coupled with the tayloe at the last IF conversion down to baseband.

The 2005 ARRL handbook goes through the theory behind the famous Omni VI low noise oscillators. A very interesting read. The VFO runs at just under a GHz and gets divided down to 5 to 5.5MHz. In the process, the phase noise gets divided by the same factor. Very interesting read.

But then I get back to the notion of why? I can buy a brand new $450 Icom that does all the nifty functions of a radio, and includes a 100w transmitter as well. So what does the topology buy me? Is it better? Where is it better? Can I make it filter out my noise limited bands?

Help me define "better."

Thanks & 73

Ford-N0FP
ford@...

The intent of this thread is to identify what I

KY1K

Ford,

I think you left out a very important question:

Does the Tayloe detector offer the potential for a more sensitive receiver??

A conventional superhet has at least 6 db insertion loss for every single mixer. Transformers, of which many are required...also means a lower gain by nature of the losses.

In a Tayloe, the 'conversion loss' is about 1 db, which is basically the insertion loss of the analog switche(s). In theory, the Tayloe detector should be able to provide around 6 db (more sensitive) NEP. Crystal filters needed for superhet selectivity have 6 to 10 db loss. Each of the additional amplifier stages needed to overcome losses in the superhet adds noise. And the non liner mixing process introduces scads of miscellaneous spurs. It is tough to actually quantify this type of degradation, mainly because you can't attach a number to how the receiver 'sounds', whether it's clean and pure or muddy sounding as some have called superhets. An NC2030 user can attest to the 'character' and 'feel' of the radio.

The enhanced NEP values of the QSD might not be 'usable', due to (atmospheric, terrestrial and noise from space) noise on HF frequencies. Modern analog switches do not operate at VHF frequencies, where the enhanced NEP would provide the ability to copy weaker signals, but this should change as semiconductor manufacturers create new products for the QSD type receivers.

Regards,

Art

Every high performance Mil RX schematicI have seen has at least 1 conversion stage to an IF containing a roofing filter. Then there is the conversion to base band. Some have high performance preselectors. The most high performance preselector is in the Cubic R3150 but the newer models (3250)have 1/3 octive filters. I have a double balancedtayloe hanging off the secondmixer of my Racal RA6830. The spectrum display and variable bandwidth features of I2PHD software are pretty cool. Performance wise the SDR appears to have a veryvery slight advantage under poor conditions.

The SDR runs in parallel with the stock demodulation. Still learning and forming an opinion. I too have the same performance questions. I do know the SDR takes the place of 3 modules in the 6830 and gives similar performance at the expense of being connected to a computer. I hape these sharp FPGA guys can get rid of the computer.That would be a cool module to drop into the chassis. My next quest will be to go directly to the 40.455 MHz first IF. The Racal has a fractional n synthesizer

Ford Peterson

Art, KY1K wrote:

Ford,

I think you left out a very important question:

Does the Tayloe detector offer the potential for a more sensitive receiver??

FP: Sensitivity is a function of bandwidth. The narrower the bandwidth, the more sensitive the MDS can be. At 500Hz it is impossible to achieve better than about -135dBm to -138dBm or so. Why? I think it could be described as that is the amount of noise power in 500Hz. That's the noise that comes from outer space--the big bang. You can add amplification all you want and you just amplify noise. Most radios available today are actually at the best MDS the laws of physics will allow.

A conventional superhet has at least 6 db insertion loss for every
single mixer. Transformers, of which many are required...also means a
lower gain by nature of the losses.

FP: The noise figure of a receiver is determined by the components in front of the first stage of amplification. If the first stages are all passive and a lossy mixer, you have shot yourself in the foot in a big way to begin with. If the conversion loss of the Tayloe is only 1dB, that is pretty dang good without a question. But even if it was -6dB, you can counter act the negative effects with an amplifier, which is what most radios do--they add an RF Amp in front of the first mixer. It is this first stage of amplification that largely determines the MDS of the radio.

In a Tayloe, the 'conversion loss' is about 1 db, which is basically
the insertion loss of the analog switche(s). In theory, the Tayloe
detector should be able to provide around 6 db (more sensitive) NEP.

FP: That is true if your first active stage is the mixer.

Crystal filters needed for superhet selectivity have 6 to 10 db loss.

FP: Which is why most radios have amplifiers in front of the crystal filters and generally use L/C type filters in the RF section with lower insertion loss.

Each of the additional amplifier stages needed to overcome losses in
the superhet adds noise. And the non liner mixing process introduces
scads of miscellaneous spurs. It is tough to actually quantify this
type of degradation, mainly because you can't attach a number to how
the receiver 'sounds', whether it's clean and pure or muddy sounding
as some have called superhets. An NC2030 user can attest to the
'character' and 'feel' of the radio.

FP: I was always under the impression that phase noise and other spurious responses in the oscillators had a lot to do with the non linear mixing you are talking about. If the LO chain is grubby to begin with, all sorts of mixing products will result no matter what you do in or after the mixer.

The enhanced NEP values of the QSD might not be 'usable', due to
(atmospheric, terrestrial and noise from space) noise on HF
frequencies. Modern analog switches do not operate at VHF
frequencies, where the enhanced NEP would provide the ability to copy
weaker signals, but this should change as semiconductor manufacturers
create new products for the QSD type receivers.

Regards,

Art

FP: Perhaps you are correct. Since none of us are crazy enough to invest the money to build actual chips, we do have to wait for the good stuff to become available in nature. Meanwhile, the quest for a better radio can continue... Great fun actually!

Thanks & 73

Ford-N0FP
ford@...

Ford Peterson

Francis wrote:

> Every high performance Mil RXschematicI have seen has at least 1 conversion stage

>to an IF containing a roofing filter.Then there is the conversion to base band. Some have

>high performance preselectors. The mosthigh performance preselector is in the Cubic R3150

>but the newer models (3250)have1/3 octive filters. I have a double balancedtayloe hanging

>off the secondmixer of my RacalRA6830. The spectrum display and variable bandwidth features

>of I2PHD software are pretty cool.Performance wise the SDR appears to have a veryvery slight

>advantage under poor conditions.

>

> The SDR runs in parallel with the stock demodulation. Still learningand forming an

>opinion. I too have the same performance questions. I do know theSDR takes the

>place of 3 modules in the 6830 and gives similar performance atthe expense of being

>connected to a computer. I hape these sharp FPGA guys can get ridof the computer.

>That would be a cool module to drop into the chassis. My nextquest will be to go

>directly to the 40.455 MHz first IF. The Racal has a fractional nsynthesizer

Thanks & 73

Ford-N0FP

1/3 octive filters are bandbass filters on the input of the receiver. My Racal preselector has 13 filters. There are some nice tunable ones around made by Cubic and Harris 10% bandwidth. They are usually in a 1U rack.

The dsp seems to pull signals out of the noise slightly better than the stock radio demodulation. Yup a saw would be nice but they are fairly lossy. I would think a pair of them would be required for good ultimate rejection. Nice preselector, H Mode mixer and High IF with a pair of Saws. Then the SDR. Hmmmm where are my 70 MHz saw filters.

My SDR at 455 KHz has a 6 dB pad at the input is double balanced tayloe with fst3253 and a pair of INA163s off plus and minus 15v. It functions without spurs -135 dBM to -40 dBM. with 2 signals spaced 20 KHz. At 2.5 KHz spacingmid 80s. I'm impressed with the close in performance but wide band not that great. fc

Ford Peterson wrote:

KY1K

Does the Tayloe detector offer the potential for a more sensitive

receiver??

FP: Sensitivity is a function of bandwidth. The narrower the bandwidth, the more sensitive the MDS can be. At 500Hz it is impossible to achieve better than about -135dBm to -138dBm or so. Why? I think it could be described as that is the amount of noise power in 500Hz. That's the noise that comes from outer space--the big bang. You can add amplification all you want and you just amplify noise. Most radios available today are actually at the best MDS the laws of physics will allow.

Actually, it's not space noise that fundamentally limits the MDS at HF. It's shot noise, which is associated with the temperature of the surroundings. Because the world is above absolute zero, there is always shot noise...which is why we often define the fundamental limit in terms of noise temperature, commonly in degrees Kelvin.

A conventional superhet has at least 6 db insertion loss for every
single mixer. Transformers, of which many are required...also means a
lower gain by nature of the losses.

FP: The noise figure of a receiver is determined by the components in front of the first stage of amplification. If the first stages are all passive and a lossy mixer, you have shot yourself in the foot in a big way to begin with. If the conversion loss of the Tayloe is only 1dB, that is pretty dang good without a question. But even if it was -6dB, you can counter act the negative effects with an amplifier, which is what most radios do--they add an RF Amp in front of the first mixer. It is this first stage of amplification that largely determines the MDS of the radio.

But, the extra amplifier reduces the dynamic range, adds noise and introduces spurs and extraneous outputs. The softrocks do as well as the best receivers without any rf amp, up to 14 MHz or so. Even there, it's debatable whether the gap between the softrock and a conventional receiver is due to the lack of an rf amp or whether it's caused by the slow acting analog switch. I believe the lack of a proper analog switch and the usable resolution of the sound card limits the softrock, not the detector itself. Adding a wonkin' big rf amp in the front end when it isn't needed only degrades the dynamic range-the world is full of real life high buck over amplified/poor dynamic range superhet receivers::> They should be avoided unless absolutely necessary.

In a Tayloe, the 'conversion loss' is about 1 db, which is basically
the insertion loss of the analog switche(s). In theory, the Tayloe
detector should be able to provide around 6 db (more sensitive) NEP.

FP: That is true if your first active stage is the mixer.

Crystal filters needed for superhet selectivity have 6 to 10 db loss.

FP: Which is why most radios have amplifiers in front of the crystal filters and generally use L/C type filters in the RF section with lower insertion loss.

Yes, an amplifier overcomes the loss, but it introduces noise-no matter how perfect it is. Every amplifier in an system introduces noise. Every additional amplifier also introduces spurs and parasitic oscillations. The point I was trying to make is that the crud from the (additional) amplifier stages needed in a superhet adds up, the Tayloe detector based receivers do not get selectivity from crystal filters.

Each of the additional amplifier stages needed to overcome losses in
the superhet adds noise. And the non liner mixing process introduces
scads of miscellaneous spurs. It is tough to actually quantify this
type of degradation, mainly because you can't attach a number to how
the receiver 'sounds', whether it's clean and pure or muddy sounding
as some have called superhets. An NC2030 user can attest to the
'character' and 'feel' of the radio.

FP: I was always under the impression that phase noise and other spurious responses in the oscillators had a lot to do with the non linear mixing you are talking about. If the LO chain is grubby to begin with, all sorts of mixing products will result no matter what you do in or after the mixer.

No argument there! A sloppy 'anything' degrades receiver performance in one way or another....which is why good receivers aren't cheap! Every resistor and cap and amp that goes in them needs to be properly selected and designed-or else you get crap.

Regards,

Art

windy10605@juno.com

Can anyone with access to an SDR-1000 tell us how well it works (with
some data) on 15m and 10m ....and 6m. With/without the added RF amplifier
(I believe it's plug selectable).

73 Kees K5BCQ

windy10605@juno.com

>FP: I was always under the impression that phase noise and other
>spurious responses in the oscillators had a lot to do with the non
>linear mixing you are talking about.

I often hear about the critical need for reducing phase jitter and OK, Ican see where a rapid change in frequency/phase would result in unwanted mixerproducts. However please explain how the +/-24Khz or +/-48Khz signal to thesound card is affected by a Tayloe mixer whose differential I or Q signals areeach the result of apositive "audio" and negative "audio" which is shifted180 RF degrees. Since the Audio is still several orders of magnitude less thanthe RF, that apparently makes no difference ...and buys you 6dBm of voltage gainin the process. I would still prefer chargingand reading diferentially,two capacitors (one positive, one negative) simultaneously for each oftheI and Qsignals. I believe that's also 6dBm.

Another question is that Phil Covingtonhas some very informativeswitch data on his web site relative to the FST switches we're using. The datawas gathered at 10Mhz andI wonder what those plots look like at 30Mhz ?....because I agree with Art, I think the "higher frequency" weak point is theswitch in this particular design. But, as Bill reminds us, ....how does SDR-1000do 54Mhz ?I'm sure there arenewer/faster switches available by now....or maybe ways of maximizing the speed/use of the existing FSTswitches.

73 Kees K5BCQ

Ford Peterson

Kees wrote:

> I often hear about the critical need for reducing phase jitter andOK,

>I can see where a rapid change in frequency/phase would resultin

>unwanted mixer products. However please explain how the+/-24Khz

>or +/-48Khz signal to the sound card is affected by a Tayloemixer

>whose differential I or Q signals are each the result ofapositive "audio"

>and negative "audio" which is shifted 180 RF degrees. Since theAudio

>is still several orders of magnitude less than the RF, thatapparently makes

>no difference ...and buys you 6dBm of voltage gain in theprocess. I

>would still prefer chargingand reading diferentially, twocapacitors (one

>positive, one negative) simultaneously for each ofthe I andQsignals. I believe that's also 6dBm.

>

My guess is that there is a nifty trick going onwith the Q and notQ outputs of the flip flops that flip the signals 180 degreesand allow you to just look at one side or the other. I'm in over my headon that one.

Regarding noisy oscillators, phase noise is a BIGculprit when it comes to passband noise. Just a few Hz away from the LO,there is random noise. In an earlier email I referred to atmospherics andinter-stellar noise. The shot noise that is inherent in all oscillators isthe cause of that rushing sound when you hook to the dummy load and crank up thevolume on the radio. Hssssssss...

Earlier I spoke of the OMNI 6 VFO as beingfamous. It really is nifty. The VFO runs at just under a GHz as Irecall. The thing gets divided down to 5 to 5.5MHz. The phase noiseat 900MHz is not unlike the phase noise at any other frequency and extends outsome 100KHz or more. The synthesizer steps by the same divide-by ratio sothat each step size at 900 is a 1Hz step at 5. Larger step sizes serve toincrease the comparator frequency and pushes the spurs out and away from theLO. Because only the main LO tweet will trigger the counter, much if notall the spurious garbage just goes away since they are unable to trigger thecounter! Nifty huh? And the phase noise gets divided down too!That's the magic. A VFO running at 5MHz would have phase noise extendingout 100KHz or more. But when you start at 900, divide it down, and run itthrough a LP filter, the phase noise has beendivided too. Thearticle suggests a 35dB improvement in noise! This is why the OMNI 6 andnow the ORION have such low noise receivers!(and partly why thelatter has a big price tag) There is a big write-up in the 2005 ARRLHandbook on the topic, complete with schematics and other diagrams.

These softrocks use a fairly clean crystaloscillator, which is partly why the performance is as good as it is. Ihave not measured the phase noise but it is likely to be fairly good.Making a frequency agile softrock would mandate the use of pretty dang goodoscillators if you are to expect similar performance. Making it frequencyagile (while salvaging performance)would make the "Dick Tracy wrist watchradio" impossible.

The bottom line is that the big money in radios isnever spent on mixers. The big money is spent on DSP, processors, filters,and high power, and/or highly complex components.

Which brings me back to my original questionsregarding the use of the Tayloe. If a radio's "goodness" is in partmeasured by its ability to QSY, is the Tayloe a good choice for the mixer?Does the improvement in efficiency (e.g. -1dB instead of -6dB conversion gain)and its inherent close-inBDR ability (close-in is superior but far distantis poor), justify the complexity of the support electronics to make the radiointo a serious contender when it comes to performance? The out-of-band BDRissues can be cleaned up with traditional 'hard' filters. But then the"simplicity" feature is gone for good.

I have appreciated all the responses in thisthread. I thinkthe threadhas allowed me to reduce the issue tothe previous paragraph.

Thanks & 73

Ford-N0FP

Another question is that Phil Covingtonhas some very informativeswitch data on his web site relative to the FST switches we're using. The datawas gathered at 10Mhz andI wonder what those plots look like at 30Mhz ?....because I agree with Art, I think the "higher frequency" weak point is theswitch in this particular design. But, as Bill reminds us, ....how does SDR-1000do 54Mhz ?I'm sure there arenewer/faster switches available by now....or maybe ways of maximizing the speed/use of the existing FSTswitches.

73 Kees K5BCQ

KY1K

But, as Bill reminds us, ....how does SDR-1000 do 54Mhz ?

Hi Kees and all,

I don't have any SDR-1000 data at 6 meters, but I was told 'the performance on 6 meters is well down'. I'm not sure if the ARRL measured the receiver performance on 6 meters or not, they did a second product review fairly recently with the Delta 44 and the Dell supplied desktop.

I'm sure there are newer/faster switches available by now ....or maybe ways of maximizing the speed/use of the existing FST switches.

Actually, no. Same switches, same cmos technology. The problem is power and monetary issues. Faster switches are made, but they need active biasing to get the speed we need, so they draw 10's of ma supply current (or more) for each switch. The phemt switches that are available are very expensive and require a negative going control signal due to the nature of the beasts. So, interfacing is easier said than done. The current cmos switches we use are very cheap and switch fast and use little power-the next step up is hardly attractive in terms of cost and difficulty to interface.

I suspect if the Tayloe detector was available commercially, there might be some slightly better switches. But, probably won't happen soon.

I have 2 reasons for suspecting the need for an rf amp stage is minimal for 20 through 10 meters...

1) The existing switches don't add noise and already pass sub microvolt levels. With a threshold that low, the need for an additional rf amp is pretty minimal...even on 10 meters.

2) The switches turn on 1 nS after the control line goes high, but remain on for about 4 nS after the control signal goes low...which means they are on for 3 nS longer than they should be. Having them on when they should be off is just as bad as having them be off when they should be on. At 30 Mhz, each switches on time should be 37.5/4 = 9.4 ns. Instead it is on for ~ 9.4 + 3 nS, which is 12.4 nS. The 3 nS difference is is a 33 percent error!!! It's minor at 10 Mhz, but becomes quite significant at 30 Mhz.

I think a very modest gain (no more than 6-10 db), but spectrally clean high level rf stage and correcting the timing error induced by the switches would bring the softrock to life on 10 meters.

Regards,

Art

Ford Peterson

Art wrote:

2) The switches turn on 1 nS after the control line goes high, but
remain on for about 4 nS after the control signal goes low...which
means they are on for 3 nS longer than they should be. Having them on
when they should be off is just as bad as having them be off when
they should be on. At 30 Mhz, each switches on time should be 37.5/4
= 9.4 ns. Instead it is on for ~ 9.4 + 3 nS, which is 12.4 nS. The 3
nS difference is is a 33 percent error!!! It's minor at 10 Mhz, but
becomes quite significant at 30 Mhz.

So what you are saying is that the control signal needs some shaping too. The period needs to remain 30MHz in your example but the pulse width needs to be 3nS less. Not easy to do, which is always the case when it comes to perfection. And of course, you would then no longer be using a nice 50% duty cycle waveform any more, which by definition will be FULL of spurs.

Thanks for the comment & 73

Ford-N0FP
ford@...