Thanks guys, I appreciate all the help.

The filter is a bandpass filter (the IF is 45MHz), though, so don't I really
need a triplexor if I really want to wideband terminate the mixer "nicely?"
(Low-, band-, and high-pass outputs.) Although my understanding is that
"triplexor" in this context can consist of a single inductor for the low-pass
port and a single capacitor for the high-pass port, which certainly is quite
doable.

---Joel

On Jan 3, 7:09*pm, "Joel Koltner" wrote:
Thanks guys, I appreciate all the help.

The filter is a bandpass filter (the IF is 45MHz), though, so don't I really
need a triplexor if I really want to wideband terminate the mixer "nicely?"
(Low-, band-, and high-pass outputs.) *Although my understanding is that
"triplexor" in this context can consist of a single inductor for the low-pass
port and a single capacitor for the high-pass port, which certainly is quite
doable.

---Joel

?? So, "triplexor" to me means something with three bands, which you
obviously won't do with just a high pass and a low pass. In any
event, the goal is to keep a reasonably constant load on the mixer at
all frequencies where it may have significant output. The distortion
generated by a mixer depends on the load it sees.

If you go straight into a crystal filter from the mixer, it can be
problematic to get a constant load impedance, since the impedance
looking into the filter changes so rapidly in the sharp transition
between passband and stopband. For example, if your passband is 44.99
to 45.01 and the mixer output has strong signals at 44.96 and 44.98,
how can you, with a practical LC circuit, make sure those signals are
terminated in an impedance which maintains low distortion, while still
passing the signals in that range to the sharp filter? Note that 3rd
order intermod between those two out-of-band signals I suggested lands
right in the middle of the crystal filter bandpass. That's why you
want to be sure to terminate the mixer in an impedance that doesn't
lead to excessive distortion (in particular, distortion products that
can land in the passband of the crystal filter). I think a reasonable
answer is to learn what range of termination impedance causes
problems, and what range is OK. For example, it may be bad if the
mixer is terminated in an "open", but of relatively little consequence
if it's terminated in a "short." That gives you a handle on how to
achieve the lowest practical distortion--which is generally the whole
point of controlling the load impedance seen by the mixer over a wide
frequency range.

Cheers,
Tom

Hi Tom,

"K7ITM" wrote in message
...
?? So, "triplexor" to me means something with three bands, which you
obviously won't do with just a high pass and a low pass.

What I meant was... output of mixer to parallel connectors of (1) the crystal
filter (bandpass) , (2) series inductor/resistor (low-pass), and (3) series
capacitor/resistor (high-pass), hence, a triplexor.

But I realize now that a duplexor is just as viable, consisting of a parallel
connection of (1) the crystal filter (bandpass) and (2) a parallel LC
resonsator in series with a resistor (bandstop).

In any
event, the goal is to keep a reasonably constant load on the mixer at
all frequencies where it may have significant output. The distortion
generated by a mixer depends on the load it sees.

Yeah, I see what you mean. I should do some actual measurements to see just
how much improvement is possible with a proper wideband termination...

Thanks for the help!

---Joel

On Jan 4, 6:54*pm, "Joel Koltner" wrote:
Hi Tom,

"K7ITM" wrote in message

...

?? *So, "triplexor" to me means something with three bands, which you
obviously won't do with just a high pass and a low pass.

What I meant was... output of mixer to parallel connectors of (1) the crystal
filter (bandpass) , (2) series inductor/resistor (low-pass), and (3) series
capacitor/resistor (high-pass), hence, a triplexor.

But I realize now that a duplexor is just as viable, consisting of a parallel
connection of (1) the crystal filter (bandpass) and (2) a parallel LC
resonsator in series with a resistor (bandstop).

In any
event, the goal is to keep a reasonably constant load on the mixer at
all frequencies where it may have significant output. *The distortion
generated by a mixer depends on the load it sees.

Yeah, I see what you mean. *I should do some actual measurements to see just
how much improvement is possible with a proper wideband termination...

Well, as the second paragraph of my previous posting hints, a 50 ohm
termination at all frequencies may actually not be optimal. There's
no law of physics, as far as I know, that says the mixer's lowest
distortion (especially distortion products that fall in the following
filter's passband) happens when it's loaded by 50 ohms. If you set up
to make some measurements, you may find that you actually get lower
distortion at some other load impedance. Perhaps optimal for the
mixer's performance would be a 50 ohm load in the filter passband, to
maximize signal output, and, say, a short at all other frequencies.
I'm not saying it IS that way for any given mixer, and perhaps not for
ANY mixer, but it's a question worth pondering if you're looking for
the best possible distortion performance.

But then you'll find the next problem: the input impedance of the
crystal filter will change dramatically, very quickly, in the region
of its passband -- and likely will be capacitive on one side of the
passband and inductive on the other side. Assuming it's a reflective
design with low internal dissipation, by definition it must reflect
out-of-band energy (and pass in-band energy to the load at the other
end of the filter). Given that situation, how do you design a circuit
that will maintain the load impedance you want for your mixer, for
frequencies a little below the filter passband, in the filter
passband, and a little above the filter passband? If you try to do it
with inductors and capacitors, where do you get parts with high enough
Q to allow the required extremely rapid change of impedance near the
filter's center frequency? Or do you just accept that the distortion
won't be optimized? Or do you throw away some of the signal and put
in a bit of attenuation (a 50 ohm pad) between mixer and filter? Or
do you go looking for the "Holy Grail": a buffer amplifier that runs
on low power and offers good input and output return loss, a third
order intercept that doesn't degrade further what the mixer has
already done, and a good enough noise figure? The buffer amplifier
can solve the matching problems (at least if you're happy with a
constant load, e.g. 50 ohms, on the mixer), but it's not trivial to
find an amplifier that will do what you need without introducing
problems worse than the cure.

As you think about all this stuff, it becomes easy to see why nobody
has yet built the perfect receiver. ;-)

Cheers,
Tom

Thanks for the help!

---Joel

"K7ITM" wrote in message
...
As you think about all this stuff, it becomes easy to see why nobody
has yet built the perfect receiver. ;-)

Hey, I was at a conference some five years or so ago now where some high-level
muckety-muck from Intel got up there and claimed that within a few years we're
be connecting antennas straight to ADCs and radios would henceforth be 100%
digital... :-)

Of course, it is a bit easier to build a good radio when you're operating in,
e.g., the cell phone bands and by law you control the spectrum (no big
intereferes), you manage the power of all the transmitters dynamically
(limited self-interference), etc.!

On Wed, 5 Jan 2011, Joel Koltner wrote:

"K7ITM" wrote in message
...
As you think about all this stuff, it becomes easy to see why nobody
has yet built the perfect receiver. ;-)

Hey, I was at a conference some five years or so ago now where some
high-level muckety-muck from Intel got up there and claimed that within a few
years we're be connecting antennas straight to ADCs and radios would
henceforth be 100% digital... :-)

No, that just means the potential is there for "the perfect receiver".
Even if the hardware is "perfect", the software still has to be written.

Of course, it is a bit easier to build a good radio when you're operating in,
e.g., the cell phone bands and by law you control the spectrum (no big
intereferes), you manage the power of all the transmitters dynamically
(limited self-interference), etc.!


Like that classic FM broadcast receiver in the old GE Transistor Manual.
A tunnel diode acting as an oscillator and mixer, it drops to an IF about
200KHz, where there is a pulse counting detector. It works because one
can live with the image frequencies, and one knows where all the wanted
signals will be.

Michael VE2BVW

On Jan 5, 3:10*pm, "Joel Koltner" wrote:
"K7ITM" wrote in message

...

As you think about all this stuff, it becomes easy to see why nobody
has yet built the perfect receiver. *;-)

Hey, I was at a conference some five years or so ago now where some high-level
muckety-muck from Intel got up there and claimed that within a few years we're
be connecting antennas straight to ADCs and radios would henceforth be 100%
digital... :-)

Of course, it is a bit easier to build a good radio when you're operating in,
e.g., the cell phone bands and by law you control the spectrum (no big
intereferes), you manage the power of all the transmitters dynamically
(limited self-interference), etc.!

Well, I guess it was about four years ago now we introduced a 100kHz
to ~35MHz receiver that practically does that. There are gain stages,
attenuators and some selectable filtering in front of the ADC, but our
customers want to be able to "listen" to the whole band at once, so
you can switch the filtering all out if you want. I did most of the
hardware, up to the signal processing. We've gotten feedback from
some customers that it's the best receiver (for that purpose) that
they can buy. It's particularly spur/residual-free: I added some
copper tape to one to see just how good I could make it, and the worst
residual is about -144dBm at the switching frequency of one of the
switching POL regulators (around 600kHz). Worst residual above 1MHz
is about -154dBm. I think it's fair to say that's pretty hard to do
with a general-coverage superhet design. But neither the hardware nor
the software that goes with it are inexpensive enough to worry that
it's going to replace other ways to do it any time soon--and the
performance doesn't quite equal what you can do with a really good
single-signal superhet design. Check in again in a few years,
assuming that there's enough economic incentive for ADC designers to
give us a little bit better parts. There are some claims out there
for really stellar ADC performance with Josephson junctions, but they
require cryogenics, and from what I've heard not all the claims are
substantiated...

By the way, the gain that's available in front of the ADC in this
design is mostly there because customers expect to need the gain,
based on previous experience. In a perfect world where they really
understood what's needed, I could have gotten by with maybe 10dB
maximum available voltage gain between the antenna and the ADC. I'm
more worried about how to gracefully handle big signals--much more
worried. What do you do about the plethora of short wave broadcast
signals, several of which can each be up to perhaps 0dBm out of your
antenna, or the fellow just down the street (or on the same ship,
etc.) who keys up a transmitter and feeds +20dBm to your receiver --
WHILE you want to keep listening to the signal that's only -110dBm at
your receiver?

Cheers,
Tom

"K7ITM" wrote in message
...
On Jan 5, 3:10 pm, "Joel Koltner" wrote:
What do you do about the plethora of short wave broadcast
signals, several of which can each be up to perhaps 0dBm out of your
antenna, or the fellow just down the street (or on the same ship,
etc.) who keys up a transmitter and feeds +20dBm to your receiver --
WHILE you want to keep listening to the signal that's only -110dBm at
your receiver?

At work one of the products we sell to the military consists of a handful of
electronically adjustable notch filters for precisely this purpose -- the
output is fed to (someone else's) SDR.

....although our dynamic range isn't the 130dB+ that you'd need for your later
example... (at least in some reasonable bandwidth...)

Your almost-all-digital HF receiver there sounds quite impressive. Do you
give the user the option to adjust the switcher's frequency away from 600kHz
if they happen to really want the best sensitivity right there? (...this
seems to be the common approach with many a ham HF rig...)

---Joel