Topics

QCX(-like) for 2m #alignment #vfo

Hans Stoll
 

Currently there seems to be no inexpensive 2m (VHF) CW QRP transceiver (kit) on the market. Is there any chance to have a QCX for 2m band?

73 de DO9HC

James Zelazny jr
 

Maybe this circuit can be modified for what your looking for ?
https://hamprojects.wordpress.com/2015/06/29/vhf-beacon-and-aprs-tracker-english-version/

Hans Stoll
 

I think this is only for FM.
There are a lot of cheap FM transceivers for 2m but no for CW / SSB.
Could the QCX not work on 2m?

@CurtisM
 

Interesting idea. We would need a 28 mhz qcx, that we don't have yet, to have a reasonable scheme to mix to 144 MHz. Then a low cost PA to get at least a couple watt, plus a reproducible 144 MHz filter on transmit. A better idea still is a pll that can make 144 Mhz since Rx would be easier to filter. Direct mix to audio? Wonder if a transformer like qcx is possible at 144 mhz, we might be happy with less image rejection. 

Curt

Shirley Dulcey KE1L
 

There are two problems with trying to directly translate the QCX receiver design to 2 meters. One is that you can't directly generate quadrature clocks at that frequency with an Si5351A local oscillator by either of the popular methods: using two outputs of the Si5351A (as done in the QCX) or using an x4 divider (as done in the Softrock and other designs). The other is that the analog switch ICs used in the QSD don't work well at that frequency.

A single conversion design where the 144 MHz band is downconverted to some lower frequency in the receiver is probably a better bet. Since it's a CW-only design we don't need coverage of a large segment of the 2 meter band, so downconversion to 17m should be fine. (That would be a better choice than 20m because there are fewer strong signals at the IF frequency to contend with.) The main advantage of a 10m IF is that it's easier to get adequate image rejection, but 17m should be good enough and could use a lot of the existing QCX design.

The transmitter doesn't need quadrature LOs so it can use a 144 MHz LO. The Si5351A can handle that. Getting a class E final to be repeatably buildable by a home builder without fancy instruments might be tricky, as would meeting the more stringent spectral purity requirements on 2m. A simpler but less efficient method would be to use a linear amplifier module for 2m.

I would love to see somebody demonstrate that it can be done!


On Thu, Sep 5, 2019 at 10:30 PM wb8yyy via Groups.Io <wb8yyy=yahoo.com@groups.io> wrote:
Interesting idea. We would need a 28 mhz qcx, that we don't have yet, to have a reasonable scheme to mix to 144 MHz. Then a low cost PA to get at least a couple watt, plus a reproducible 144 MHz filter on transmit. A better idea still is a pll that can make 144 Mhz since Rx would be easier to filter. Direct mix to audio? Wonder if a transformer like qcx is possible at 144 mhz, we might be happy with less image rejection. 

Curt

zl1ujg
 

I suspect going too high in frequency will degrade phase noise significantly. This would noise-up the audio band in RX, degrading receive sensitivity, thereby requiring extra front-end receive gain, to over-ride  the degraded noisefloor, and potentially noising up the 2M band for other users, while transmitting... 

Hans Summers
 

Hi Shirley

Agreed the FST3253 certainly won't work at 2m frequencies. 

Leaving aside the issue of spectral purity - what is the upper limit of the Si5351A in quadrature mode? Why do you think that it would not do quadrature at 2m? Maybe you are right - do we know for sure? 

73 Hans G0UPL

On Fri, Sep 6, 2019 at 6:04 AM Shirley Dulcey KE1L <mark@...> wrote:
There are two problems with trying to directly translate the QCX receiver design to 2 meters. One is that you can't directly generate quadrature clocks at that frequency with an Si5351A local oscillator by either of the popular methods: using two outputs of the Si5351A (as done in the QCX) or using an x4 divider (as done in the Softrock and other designs). The other is that the analog switch ICs used in the QSD don't work well at that frequency.

A single conversion design where the 144 MHz band is downconverted to some lower frequency in the receiver is probably a better bet. Since it's a CW-only design we don't need coverage of a large segment of the 2 meter band, so downconversion to 17m should be fine. (That would be a better choice than 20m because there are fewer strong signals at the IF frequency to contend with.) The main advantage of a 10m IF is that it's easier to get adequate image rejection, but 17m should be good enough and could use a lot of the existing QCX design.

The transmitter doesn't need quadrature LOs so it can use a 144 MHz LO. The Si5351A can handle that. Getting a class E final to be repeatably buildable by a home builder without fancy instruments might be tricky, as would meeting the more stringent spectral purity requirements on 2m. A simpler but less efficient method would be to use a linear amplifier module for 2m.

I would love to see somebody demonstrate that it can be done!


On Thu, Sep 5, 2019 at 10:30 PM wb8yyy via Groups.Io <wb8yyy=yahoo.com@groups.io> wrote:
Interesting idea. We would need a 28 mhz qcx, that we don't have yet, to have a reasonable scheme to mix to 144 MHz. Then a low cost PA to get at least a couple watt, plus a reproducible 144 MHz filter on transmit. A better idea still is a pll that can make 144 Mhz since Rx would be easier to filter. Direct mix to audio? Wonder if a transformer like qcx is possible at 144 mhz, we might be happy with less image rejection. 

Curt

Hans Stoll
 

I though the Ultimate3S has similar design and this can operate on 2m.
I'm not an expert ...

Shirley Dulcey KE1L
 

I remember reading somewhere that the upper limit for quadrature was somewhere around 110 MHz. But I haven't tried it myself, and I don't have test gear with enough bandwidth to verify whether it works or not. My best oscilloscope is only rated to 100 MHz.


On Fri, Sep 6, 2019 at 2:37 AM Hans Summers <hans.summers@...> wrote:
Hi Shirley

Agreed the FST3253 certainly won't work at 2m frequencies. 

Leaving aside the issue of spectral purity - what is the upper limit of the Si5351A in quadrature mode? Why do you think that it would not do quadrature at 2m? Maybe you are right - do we know for sure? 

73 Hans G0UPL

On Fri, Sep 6, 2019 at 6:04 AM Shirley Dulcey KE1L <mark@...> wrote:
There are two problems with trying to directly translate the QCX receiver design to 2 meters. One is that you can't directly generate quadrature clocks at that frequency with an Si5351A local oscillator by either of the popular methods: using two outputs of the Si5351A (as done in the QCX) or using an x4 divider (as done in the Softrock and other designs). The other is that the analog switch ICs used in the QSD don't work well at that frequency.

A single conversion design where the 144 MHz band is downconverted to some lower frequency in the receiver is probably a better bet. Since it's a CW-only design we don't need coverage of a large segment of the 2 meter band, so downconversion to 17m should be fine. (That would be a better choice than 20m because there are fewer strong signals at the IF frequency to contend with.) The main advantage of a 10m IF is that it's easier to get adequate image rejection, but 17m should be good enough and could use a lot of the existing QCX design.

The transmitter doesn't need quadrature LOs so it can use a 144 MHz LO. The Si5351A can handle that. Getting a class E final to be repeatably buildable by a home builder without fancy instruments might be tricky, as would meeting the more stringent spectral purity requirements on 2m. A simpler but less efficient method would be to use a linear amplifier module for 2m.

I would love to see somebody demonstrate that it can be done!


On Thu, Sep 5, 2019 at 10:30 PM wb8yyy via Groups.Io <wb8yyy=yahoo.com@groups.io> wrote:
Interesting idea. We would need a 28 mhz qcx, that we don't have yet, to have a reasonable scheme to mix to 144 MHz. Then a low cost PA to get at least a couple watt, plus a reproducible 144 MHz filter on transmit. A better idea still is a pll that can make 144 Mhz since Rx would be easier to filter. Direct mix to audio? Wonder if a transformer like qcx is possible at 144 mhz, we might be happy with less image rejection. 

Curt

Hans Summers
 

Hi Shirley
 
I remember reading somewhere that the upper limit for quadrature was somewhere around 110 MHz.

That's interesting. I'm not sure it is correct... there is a lot of misinformation written about the Si5351A. Not with bad intentions. Just because it is a complex chip and the documentation is poor so there are a lot of opportunities to misinterpret the datasheet. There are also a lot of inferior libraries floating around which exacerbates the problems. 

As far as I know, nobody else has done quadrature on the Si5351A until the QCX... so I don't know who would write about 110MHz anyway. 

The way the quadrature works using the internal even integer mode MultiSynth dividers with configurable phase offset in number of VCO quarter-cycles, it logically should work all the way to the top end of the Si5351A's frequency range, and not be limited to 110MHz. 

Though I have spent an enormous amount of time with the Si5351A chip I still feel it has further secrets waiting to be discovered. 

73 Hans G0UPL

Hans Summers
 

Hi Hans
 
I though the Ultimate3S has similar design and this can operate on 2m.
I'm not an expert ...

The Ultimate3S http://qrp-labs.com/ultimate3/u3s can produce output on 2m and even on 222MHz - the Si5351A can manage this easily (it is rated to 200MHz but can typically be pushed further). 

But the QCX http://qrp-labs.com/qcx question is a slightly different one. The Si5351A in the QCX generates two outputs in quadrature (90-degree phase offset). The question is about whether the Si5351A can continue to generate quadrature outputs up to 2m. 

73 Hans G0UPL

Kārlis Goba
 

The problem can be split in two parts. My two cents:

TX: the existing QCX architecture uses Si5351A to generate square wave LO and directly pass it to a class-E amplifier. As mentioned, class E operation can be tricky to tune without good instruments at such high frequencies. Alternative is to use class C final stage as done in U3s. Some efficiency will be compromised at the benefit of much greater ease of construction/alignment. This can be accommodated more or less by modding an existing QCX PCB (and requires modified firmware).

RX: this is harder. To keep the simplicity of the existing QCX design the RX should rely on a QSD detector. As mentioned, the analog switch used in QCX will not work directly at 144 MHz even if the Si5351A provides clean quadrature output. There are no easy and cheap alternatives to that. One way is to use 2 identical 'standard' (e.g. diode ring) mixers instead of the QCD, but that adds to the cost and requires serious redesign of the QCX board. Another, perhaps better alternative, would be to add a downmix stage in the RX chain and keep the existing QSD. No need to focus on an IF that lies exactly in an amateur band - anything can be used, and better yet with a monolithic crystal filter which can be readily sourced for 10,7 MHz, 21 MHz or 45 MHz. That still requires modifications, but could be accommodated as a small separate PCB in addition to the base QCX.


--
Karlis YL3JG

Guido PE1NNZ
 

There are (more or less) pin compatible alternatives for FST3253 that are faster like the PO3B14A.

This Potato chip :) has a total switching timing of 5ns, which in theory should cover 200 MHz.
With the regular FST3253 in my QCX, I have 50Mhz working and hear in the past weeks quite a bit of FT8 activity on 6m; for this I have modified the firmware and injected antenna signal directly into the QSD (without using T1 and inserted a 1k resistor in the bias). My impression is that the QSD is working up to about 80 MHz, then somewhere when tuning further up, I hear a change sharp click, still hear noise but it no longer detects any sensible signal. To receive 2m with FST3253, undersampling at 144/3 Mhz could be an idea.

Just some thoughts and experiences.
Guido

ajparent1/KB1GMX
 

None of the Bus switches used are even close to fast enough for 2M nor quiet enough.

I tried fast switches at 6M and compared to a DBM they were terrible
and the system as a whole still too noisy for effective 6M weak signal work.
My HB radio has a NF under 4.5dB and is optimized for weak signal work.
I could not get that low with bus switches and had to resort to balanced fet
switching mixer.

To be effective on 2M you need low noise mixer and a RF amp (LNA).
A J310 fet and a pair of DBM can do that. 

Then the tx side needs work as the 74HC00 BS170 will not do much
power at 2M.

Basically stuff that works at 1-30mhz does not work above that or does
so poorly.

The basic idea would work but then anyone thats on 2M knows there
is not a lot of CW anymore though during contests there is.

Allison

Kārlis Goba
 

The PO3B14A is indeed a nice find but I'm afraid it won't really make much of a difference. I'm basing my view on what Chris Trask reported in his Mixer musings (https://www.mikrocontroller.net/attachment/146369/Mixer_Musings.pdf). He measured the switching performance (albeit in a different configuration than a QSD) for FSA3157 (see Table 1), which theoretically has t_on=3.4ns and t_off=2.1ns, so kind of close to 5ns of the 'Potato' chip. However, the mixing/switching performance starts to drop once you reach about 100 MHz. He lists several other alternative switches (SPDT though), but they might all have very similar performance (hasn't been exhaustively tested though). PO3B14A also stands out with quite high 'on' resistance unfortunately. 

So, experiments might be possible, but the prospect is not that bright I think.

Regarding your click at 80 MHz, it seems that you might hit the t_on+t_off speed limit of FST3253, which could be around 12ns.

Now, mixing on the 3rd harmonic is a workable idea, though it would require a RX preamp to cover the additional losses.

--
Karlis YL3JG

Guido PE1NNZ
 



On Friday, September 6, 2019, Kārlis Goba <karlis.goba@...> wrote: 
So, experiments might be possible, but the prospect is not that bright I think.
 
Probably true, but worth a try. The  Reduced noise figure performance probably could be compensated with a tiny VHF pre-amp.

Regarding your click at 80 MHz, it seems that you might hit the t_on+t_off speed limit of FST3253, which could be around 12ns.

Yes, that is probably what it is, I have no idea how it performs at 70 Mhz but at 50 Mhz it is seems sensitive enough for actual use.
Actually, I forgot about it, but at higher frequencies (167 Mhz) this FST3253 QSD is just working again.. Absolutely no clue why this is possible (maybe t_on/off times starts overlapping partially).

ajparent1/KB1GMX
 

OK, lets address a few things..

For any kind of IQ format radio the very nasty part of switching times is when
they become near and approach a significant portion of the sampled frequency
the level of phase distortion is greater.  Reminder an IQ (phasing analog or DSP)
requires two mixers to base band.  They must behave well and be identical.
That  is complicated by a 2 phase clock usually from a source divided by 4 and
the trick QCX uses will likely not work at 144mhz and parts that run at 300mhz
are not cheap and tend to be small format SMT.

The Tayloe or quadrature sampling mixer was to take advantage of inexpensive parts
at HF or lower frequencies where it shines.  At VHF and up high dynamic range 
and many other factors are taken in a different context and often not needed.

For all the effort a pair of minicircuits ADE-1+ DBMs will be cheaper and easier to use
and cover .5mhz to 500mhz at a list price is 3.25$ each!  Also the LO then is easier
to meet.  Its still SMT.

FYI any mixer will require a decent preamp/LNA before it for decent 2M weak
signal performance.  That is a given based on a lot of design history in the market.
You need it for NF performance, MDS and to keep the signals from the mixer 
from getting back to the antenna.

How do I know?  Besides being an engineer I also build radios for myself
and have been on 6/2/70-cm as a ham and up to 6ghz in the commercial
space.  I have radios built for 6 (more than 5),  2 (2 SSB radios), and 432mhz
(both SSB) not including many transverters. A well equipped bench is a 
real aid to troubleshooting and tuning.

Add to that all of the technical requirement for purity and radiated emissions 
are more stringent at frequencies above 30mhz.

FCC97.307:
(e) The mean power of any spurious emission from a station transmitter or external
RF power amplifier transmitting on a frequency between 30-225 MHz must be at
least 60 dB below the mean power of the fundamental. For a transmitter having a
mean power of 25 W or less, the mean power of any spurious emission supplied
to the antenna transmission line must not exceed 25 µW and must be at least
40 dB below the mean power of the fundamental emission, but need not be
reduced below the power of 10 µW. 


In English that meas a performance level 17DB better (50X) than whats acceptable
below 30mhz!   Simple kits would not achieve that and the cost to do that is higher.
For the but I'm not in USA crowd, most are under similar requirements or more
stringent.

Further parts that perform well at below 30Mhz just don't at 2M.  An example is
IRF510 is maybe a 1$US and works well to 6M, at 2M your forced to use  
RD15VHF at 5$US each.  Low power devices as well go from pennies each
to 30-60 cents each for reasonable performance.  Coils same issue and
cheap caps are totally out.  The board design has to be far more tightly
controlled and some sections will have to have a shield assembly over them.
At 2M SMT is the best way to go and is the accepted best practice for
high performance.   In the end for a market that is typically 1/100th that
of HF sales it is  a lot of testing needed to design and market an
acceptable radio.  For kit makers the first useful price break for the
expensive parts is typically over 100pcs and for some part more like
500.  Single source parts are rarely cheap.

So a 2M QCX would easily be a 199$US radio for CW only at maybe 3W.
And being CW only its not going to be all that popular as those people
on 2M and needing a 2M radio for transverter to 432 and higher wants both
SSB and CW but the power can be very low (under 10 milliwatts).

Easy way to get on 2M is a transverter from 20M or 10M, base QCX with out BS170s
will do that well.  The transverter is still more effort and cost than the QCX alone!

This is why many kit makers look at 6M and higher (2, 222, 432, 900, 1296...)
and offer nothing as the market is small and design and manufacture costs
are high.

This is why every time some one asks will there be a 2M version...
I always ask what are you willing to pay?  
    (keep in mind a new Ft818 is 800$ and a used FT817 400$ and both do HF to 432mhz!)
How many will you buy?
Would you build it if its SMT and parts count is in the 300plus peices plus hand wound coils (not all toroids)?
If it is a kit can you tune and test it with your gear?
  *subtext how many can measure power at 2M, SWR, RX sensitivity???

FYI: the only design I've seen that can work as well at 2m as it does at 7mhz is the KK7B
MicroR2 and the matching MicroT2.


Allison

Bob Macklin <macklinbob@...>
 

I think that running the QSD around 10MHz and using a converter in front of it would be a better solution. Maybe even running the QSD around 5MHz or 2.5MHz would be better.
 
I think some people try to do too much with too little.
 
The Elecraft units use a scheme that converts the incoming signal to about 5MHz. You could run the QSD as a 5MHz IF and do the tuning the was the Elecraft rigs do. Crystals are available for all bands below 10M. They are stocked by Mouser and DigiKey.
 
Bob Macklin
K5MYJ
Seattle, Wa.
"Real Radios Glow In The Dark" 

----- Original Message -----
Sent: Friday, September 06, 2019 2:40 PM
Subject: Re: [QRPLabs] QCX(-like) for 2m #alignment #vfo #vfo

OK, lets address a few things..

For any kind of IQ format radio the very nasty part of switching times is when
they become near and approach a significant portion of the sampled frequency
the level of phase distortion is greater.  Reminder an IQ (phasing analog or DSP)
requires two mixers to base band.  They must behave well and be identical.
That  is complicated by a 2 phase clock usually from a source divided by 4 and
the trick QCX uses will likely not work at 144mhz and parts that run at 300mhz
are not cheap and tend to be small format SMT.

The Tayloe or quadrature sampling mixer was to take advantage of inexpensive parts
at HF or lower frequencies where it shines.  At VHF and up high dynamic range 
and many other factors are taken in a different context and often not needed.

For all the effort a pair of minicircuits ADE-1+ DBMs will be cheaper and easier to use
and cover .5mhz to 500mhz at a list price is 3.25$ each!  Also the LO then is easier
to meet.  Its still SMT.

FYI any mixer will require a decent preamp/LNA before it for decent 2M weak
signal performance.  That is a given based on a lot of design history in the market.
You need it for NF performance, MDS and to keep the signals from the mixer 
from getting back to the antenna.

How do I know?  Besides being an engineer I also build radios for myself
and have been on 6/2/70-cm as a ham and up to 6ghz in the commercial
space.  I have radios built for 6 (more than 5),  2 (2 SSB radios), and 432mhz
(both SSB) not including many transverters. A well equipped bench is a 
real aid to troubleshooting and tuning.

Add to that all of the technical requirement for purity and radiated emissions 
are more stringent at frequencies above 30mhz.

FCC97.307:
(e) The mean power of any spurious emission from a station transmitter or external
RF power amplifier transmitting on a frequency between 30-225 MHz must be at
least 60 dB below the mean power of the fundamental. For a transmitter having a
mean power of 25 W or less, the mean power of any spurious emission supplied
to the antenna transmission line must not exceed 25 µW and must be at least
40 dB below the mean power of the fundamental emission, but need not be
reduced below the power of 10 µW. 


In English that meas a performance level 17DB better (50X) than whats acceptable
below 30mhz!   Simple kits would not achieve that and the cost to do that is higher.
For the but I'm not in USA crowd, most are under similar requirements or more
stringent.

Further parts that perform well at below 30Mhz just don't at 2M.  An example is
IRF510 is maybe a 1$US and works well to 6M, at 2M your forced to use  
RD15VHF at 5$US each.  Low power devices as well go from pennies each
to 30-60 cents each for reasonable performance.  Coils same issue and
cheap caps are totally out.  The board design has to be far more tightly
controlled and some sections will have to have a shield assembly over them.
At 2M SMT is the best way to go and is the accepted best practice for
high performance.   In the end for a market that is typically 1/100th that
of HF sales it is  a lot of testing needed to design and market an
acceptable radio.  For kit makers the first useful price break for the
expensive parts is typically over 100pcs and for some part more like
500.  Single source parts are rarely cheap.

So a 2M QCX would easily be a 199$US radio for CW only at maybe 3W.
And being CW only its not going to be all that popular as those people
on 2M and needing a 2M radio for transverter to 432 and higher wants both
SSB and CW but the power can be very low (under 10 milliwatts).

Easy way to get on 2M is a transverter from 20M or 10M, base QCX with out BS170s
will do that well.  The transverter is still more effort and cost than the QCX alone!

This is why many kit makers look at 6M and higher (2, 222, 432, 900, 1296...)
and offer nothing as the market is small and design and manufacture costs
are high.

This is why every time some one asks will there be a 2M version...
I always ask what are you willing to pay?  
    (keep in mind a new Ft818 is 800$ and a used FT817 400$ and both do HF to 432mhz!)
How many will you buy?
Would you build it if its SMT and parts count is in the 300plus peices plus hand wound coils (not all toroids)?
If it is a kit can you tune and test it with your gear?
  *subtext how many can measure power at 2M, SWR, RX sensitivity???

FYI: the only design I've seen that can work as well at 2m as it does at 7mhz is the KK7B
MicroR2 and the matching MicroT2.


Allison

ajparent1/KB1GMX
 

Running QSD at lower frequencies is not advantageous for VHF
and up.  For image rejection its best to keep the IF at 10% or less 
for 144 that's 20M and for 432, 28mhz works well enough.  
For HF anything goes but even at 28mhz an if at 2.5mhz is
the minimum or then the filters to keep the spurs down get
to be a pain.

Crystals are not needed with fixed IFs as modern synths can do that easily.
Though its not a requirement to do one over the other.  The concern at
VHF and up is frequency accuracy and stability.  The higher you go
the more difficult it is.

QSD works as well at 5, 10 or even 25mhz as it does at 1mhz so the choice
is then image responses and spurious product responses (birdies) at the
system level.

With all that I've been bench testing a up converting design with IQ running
at 36mhz using DBMs (see KK7B designs for R2 and friends).  When designed
for a single frequency these image rejecting (phasing) systems work as well
at 1GHZ as they would at 1mhz,  so 36mhz was chosen as a convenient
(common crystal) frequency.  The RX would then convert the input from
160 through 10M to 36mhz where its detected and a single VFO/Synth is
used for the up conversion.  The scheme allows the IQ to be optimized
for  a single frequency (IF) and requires no tuning once set and no
compensation for phase/amplitude that can very with tuning directly. 
As I'm doing it all analog it saves power as very low power and fast
DSP are not easy, though we are getting there.

Testing says its clean and with 1/2 octave filters it doesn't take many to cover 
the range.  TX would require the same base circuits running is reverse.

Elecraft does a lot of the IQ work in DSP and that can be does as well as
once you convert to base band or some very low IF (12-50khz)  its easier.

Allison

Bob Macklin <macklinbob@...>
 

My plan is to use the QRP Labs receiver module to replace the 4.915 crystal filter IF in the Elecraft design. Running the QSD on a 19.66MHz crystal.
 
And using the Elecraft HF front end ahead of it.
 
Then for VHF I will use conventional transverters with a 28MHz IF for the VHF frontend.
 
Bob Macklin
K5MYJ
Seattle, Wa.
"Real Radios Glow In The Dark"

----- Original Message -----
Sent: Saturday, September 07, 2019 8:04 AM
Subject: Re: [QRPLabs] QCX(-like) for 2m #alignment #vfo #vfo

Running QSD at lower frequencies is not advantageous for VHF
and up.  For image rejection its best to keep the IF at 10% or less 
for 144 that's 20M and for 432, 28mhz works well enough.  
For HF anything goes but even at 28mhz an if at 2.5mhz is
the minimum or then the filters to keep the spurs down get
to be a pain.

Crystals are not needed with fixed IFs as modern synths can do that easily.
Though its not a requirement to do one over the other.  The concern at
VHF and up is frequency accuracy and stability.  The higher you go
the more difficult it is.

QSD works as well at 5, 10 or even 25mhz as it does at 1mhz so the choice
is then image responses and spurious product responses (birdies) at the
system level.

With all that I've been bench testing a up converting design with IQ running
at 36mhz using DBMs (see KK7B designs for R2 and friends).  When designed
for a single frequency these image rejecting (phasing) systems work as well
at 1GHZ as they would at 1mhz,  so 36mhz was chosen as a convenient
(common crystal) frequency.  The RX would then convert the input from
160 through 10M to 36mhz where its detected and a single VFO/Synth is
used for the up conversion.  The scheme allows the IQ to be optimized
for  a single frequency (IF) and requires no tuning once set and no
compensation for phase/amplitude that can very with tuning directly. 
As I'm doing it all analog it saves power as very low power and fast
DSP are not easy, though we are getting there.

Testing says its clean and with 1/2 octave filters it doesn't take many to cover 
the range.  TX would require the same base circuits running is reverse.

Elecraft does a lot of the IQ work in DSP and that can be does as well as
once you convert to base band or some very low IF (12-50khz)  its easier.

Allison