--- In softrock40@yahoogro ups.com
, Art <KY1K@...> wrote:
> >some of these have some good
> >switching speed their on-off time are very different.
> Hi Gian,
> I have thought about switching symmetry a lot. I think symmetry
> improvements COULD make the QSD practical at higher
> the current technology), but probably not at 2 meters.
I have done tests on 145MHz while experimenting with the I7SWX 2
Transformer H-Mode Mixer with IF from 455kHz to 100 MHz, using LVDS
squarer. The FST3125 mixer was working but the conversion losses were
minimum -8dB (LO= RF-IF) for IFs from 4MHz to 70MHz. Transformere
were balun core #43-2403. At 455kHz the Loss was -9dB but probably
due to transformers core. On HF the 2T H-M Mix has a loss of -5dB or
less with downconversion and around -5 with up conversion 70MHz IF.
At 50MHz RF 2T was at -5dBm loss with IF at 28MHz.
The increase of losses is certainly due to internal switch (gating
and sw itself) unbalances, delay, phasing...who knows. When LO
frequency is going up switches gets crazy...hi
RF in = 145 MHz.
IF LO = RF- IF ConvL LO = RF+IF ConvL
MHz] [MHz] [dB] [MHz] [dB]
0.455 144.4 -9 145.5 -9
-8 149 -8
9 135 -8 154 -8.5
10.7 134 -8 156 -9
21 124 -8 165 -10
45 100 -8 165 -10
70 75 -8 215 -12
100 45 -9 244 -15
> Symmetry is perhaps more important than speed, although the current
> generation of analog switches we use are not designed or rated for
> By symmetry, I mean that the t(on) time and the t(off) times are
> identical within each switch (they turn off and turn on at the same
> time relative to the enable line). Also, that each individual
> within a package turns on and turns off at the same speed relative
> the other switches in that package.
You are right. Also components tech figures are given by marketing
and not engineering .... often are what customers would like
to "see".... and simulators are OK for these...hi
> In the case of the current generation of FST switches, we think
> turn on
in ~1.5 nS and take ~4 nS to turn off. If we could delay
> turn on by 2.5 nS in each of the switches in the FST package, we
> might be able to achieve better performance while using the same
Everything is a compromise so we have to take for good the "best
figures" so we can be happy .... I have good friends (well known hams
that can feel and see differences of 0.1dB or less ... for me 1 dB or
more is the same ... I am not in business...hi
> After all, we don't care which 90 degree segment of the input
> waveform we capture, we only care that each individual switch
> captures a different 90 degree slice (without overlap or dead
> of the input wave .
Yes. The problem is that too many "pieces" or "stages" are involved
in those 90 degrees....
> I am not quite sure how to do this easily. But, an rc network on
> enable line
would surely slow down the t(on). But, how do we make
> turn off delay on the enable line respond as though the rc network
> wasn't there?
You need to go inside the FST switches and see what happen ... so you
can put anything you like to balance.
We know that the best digital quadrature generator is the Johnson
counter dividong by 4 the LO ... well, on paper... You maybe be OK
with the 0° and 90° quadrature but if you use the 180° and 270° also
you will see that these complementary waveforms are shifted respect
the others of around 100pS ...
We are all interested oin the Tayloe QSD but maybe we should look at
the limits of use ....
The SoftRock40 V7 was designed because the SDR 1000 and SR V6 were
not performing as expected .... so it is the limit of the QSD
configuration or the not so good quadrature signals driving it or
both or what else? What is generating the high noise of QSD when
going up on frequency?
untill someone will give us a nice set of GAsFet switches with
perfect and fast switching Ton-Toff ... we will have to accept some
limits (HF) ...thenm we can go to SHF. The main problems will
be ...Cases we cannot handle anymore... even the TSSOP are dying
now ...let's collect a lot of those SOIC chips that can last us till
end of century, at least (hope I will resist till the ..mmmm 66