I've been playing around with some homebrew superhet radios, and I'm finding
that a significant amount of energy ends up at my 1st IF frequency that seems
to be coming from the LO. While I expect to see energy from LO+/-IF end up at
IF, of course, I've checked the LO+/-IF spurs (the LO is coming from a
PLL-based synthesizer), and in general it seems that a lot more energy ends up
at the IF than what the spurs alone would suggest. I remember a talk I
attended where the presenter mention that one of the biggest problems with
building receivers was "the LO getting into the IF," so I'm thinking this is
what he meant? Are there other less obvious paths for the LO getting into the
IF than just the LO+/-IF spurs?

The signal right at the IF eventually gets turned into DC and hence filtered
out, so in theory it doesn't really matter that much, but in practice with
very weak signals eventually the IF feedthru is stronger than the weak
signals, so it limits how much amplification I can provide and hence limits
the ultimate sensitivity of the receiver.

Thanks,
---Joel Koltner

In article ,
Joel Koltner wrote:

that a significant amount of energy ends up at my 1st IF frequency that seems
to be coming from the LO. While I expect to see energy from LO+/-IF end up at
IF, of course, I've checked the LO+/-IF spurs (the LO is coming from a
PLL-based synthesizer), and in general it seems that a lot more energy ends up
at the IF than what the spurs alone would suggest. I remember a talk I
attended where the presenter mention that one of the biggest problems with
building receivers was "the LO getting into the IF," so I'm thinking this is
what he meant? Are there other less obvious paths for the LO getting into the
IF than just the LO+/-IF spurs?

The signal right at the IF eventually gets turned into DC and hence filtered
out, so in theory it doesn't really matter that much, but in practice with
very weak signals eventually the IF feedthru is stronger than the weak
signals, so it limits how much amplification I can provide and hence limits
the ultimate sensitivity of the receiver.

Local-oscillator "blowthrough" is almost inevitable in non-balanced
mixers. Even in single- and double-balanced mixers it's fairly
common - it's an indication of less-than-perfect balance. Commercial
DBMs usually quote an IF-isolation spec. For example, the popular
Mini-Circuits SBL-1 mixer seems to have LO-to-IF isolations of
anywhere from better than 65 dB (at HF) down to around 30 dB (at UHF).
If you're trying to tune a weak signal (say, 80 or 90 dB weaker than
the LO signal) then the residual LO feedthrough can cause the sort of
swamping you're seeing.

Homebrew DBMs with hand-wound toroids, and less-than-well-matched
diode sets may have significantly worse LO-to-IF isolation.

With diode-and-transformer DBMs, performance can degrade significantly
if the mixer ports don't "look into" a nice broadband 50-ohm
impedance. If, for example, you connect the IF port directly to a
bandpass filter, the filter's impedance at the LO frequency is likely
to be very much different thatn 50 ohms, and this may affect LO
feedthrough.

The higher-performance superhet designs I've seen, generally take care
to ensure that each mixer port sees the proper termination impedance.
In the case of the LO port, the easiest way is to generate a
stronger-than-necessary LO signal (maybe 13 dBm) and then feed it to
the mixer through a 50-ohm pad (6 dB in this case).

In the case of the RF and IF ports, you can either use a resistive pad
with modest attenuation (perhaps 3 dB), or use a diplexer. For the IF
port you might want to feed the mixer into a common-gate or
common-base broadband amplifier stage, run at a current level which
gives it a 50-ohm input impedance... and then feed the output of this
amp to your IF bandpass filter.

I've seen one fairly expensive device (a Tek spectrum analyzer) which
had horrendous RF-spur problems. Turns out that its front end uses a
simple non-balanced diode mixer with essentially no RF-to-IF or
LO-to-IF isolation. The LO-to-IF isolation isn't a problem due to the
nature of the analyzer design... but the analyzer uses a relatively
low swept-IF frequency range, and the incoming RF signals can blow
right through and be detected as if they were IF mixing products.
It's an OK analyzer for bench testing of many radios, but is pretty
much useless for trying to look at VHF over-the-air signals.

--
Dave Platt AE6EO
Friends of Jade Warrior home page: http://www.radagast.org/jade-warrior
I do _not_ wish to receive unsolicited commercial email, and I will
boycott any company which has the gall to send me such ads!

On 11 Jun, 18:54, "Joel Koltner" wrote:
I've been playing around with some homebrew superhet radios, and I'm finding
that a significant amount of energy ends up at my 1st IF frequency that seems
to be coming from the LO. *While I expect to see energy from LO+/-IF end up at
IF, of course, I've checked the LO+/-IF spurs (the LO is coming from a
PLL-based synthesizer), and in general it seems that a lot more energy ends up
at the IF than what the spurs alone would suggest. *I remember a talk I
attended where the presenter mention that one of the biggest problems with
building receivers was "the LO getting into the IF," so I'm thinking this is
what he meant? *Are there other less obvious paths for the LO getting into the
IF than just the LO+/-IF spurs?

The signal right at the IF eventually gets turned into DC and hence filtered
out, so in theory it doesn't really matter that much, but in practice with
very weak signals eventually the IF feedthru is stronger than the weak
signals, so it limits how much amplification I can provide and hence limits
the ultimate sensitivity of the receiver.

Thanks,
---Joel Koltner

Screening is the best solution. Put the various stages in separate
screened enclosures.

Leon

Hi Dave,

Thanks for the information; it's quite helpful!

I was (am) committing the faux pas of feeding the mixer output directly into a
narrow bandpass filter...

For example, the popular
Mini-Circuits SBL-1 mixer seems to have LO-to-IF isolations of
anywhere from better than 65 dB (at HF) down to around 30 dB (at UHF).
If you're trying to tune a weak signal (say, 80 or 90 dB weaker than
the LO signal) then the residual LO feedthrough can cause the sort of
swamping you're seeing.

So... say I'm using a 45MHz IF, trying to tune 414MHz using the SBL-1 and an
LO of 414-45=369MHz (low-side injection) at the SBL-1's recommended +7dBm.
Presumably I'll see a 7dBm - 30dB (LO-IF isolation) = -23dBm signal at the IF,
but it should still be at 369MHz, right? How does the LO "bleed through" to
the 45MHz IF... and what power level should I expect to see there?

---Joel

Before writing my comments, I read the comments of two others who
responded to your query.

On Wed, 11 Jun 2008, Joel Koltner wrote:

I've been playing around with some homebrew superhet radios, and I'm finding
that a significant amount of energy ends up at my 1st IF frequency that seems
to be coming from the LO.

At this point I'd like to know what your problem is with "significant
amount of energy..." compared to whatever your ultimate goal is.

While I expect to see energy from LO+/-IF end up at
IF, of course, I've checked the LO+/-IF spurs (the LO is coming from a
PLL-based synthesizer), and in general it seems that a lot more energy ends up
at the IF than what the spurs alone would suggest.

Again, what specifically are you "seeing" (measuring?) compared to what
you think you should be getting? Could your PLL synthesizer be dirtier
than you think?

I remember a talk I
attended where the presenter mention that one of the biggest problems with
building receivers was "the LO getting into the IF," so I'm thinking this is
what he meant? Are there other less obvious paths for the LO getting into the
IF than just the LO+/-IF spurs?

One way I would think about this is to ask if you looked at known circuits
that work and ask yourself what are you doing that is different from known
circuits that work. We also had some posts maybe 1-2 years ago where a guy
was working with chips and circuits and computer modelling (IIRC) and he
was unhappy that he was not getting (with real circuits) what his computer
modeling program told him he was supposed to get.

The signal right at the IF eventually gets turned into DC and hence filtered
out, so in theory it doesn't really matter that much, but in practice with
very weak signals eventually the IF feedthru is stronger than the weak
signals, so it limits how much amplification I can provide and hence limits
the ultimate sensitivity of the receiver.

I do homebrew but with tubes and can tell many stories about what should
have been a straightforward project but electrical performance was
unacceptable. QST has had, in the past, articles on why ham-built copies
of ARRL circuits don't work and of course all the blame goes on the ham
and not ARRL but there is a lot of missing information in the handbooks,
too. I've learned a few tricks by the crash-and-burn, smoke tests with
smoke and no function, the "guess and pray" techniques, and the "dumb
looks"-after-the-smoke response. Very roughly, I'd say 50% of my projects
work the way I hoped they would work, the rest go to the glue factory.



Thanks,
---Joel Koltner

"A" wrote in message
.com...
At this point I'd like to know what your problem is with "significant amount
of energy..." compared to whatever your ultimate goal is.

I'll get something in the ballpark of an -90dBm spur at the IF when the noise
floor is down around, say, -120dBm. Hence I have a harder time recovering
signals at, say, -100dBm even though they still have a decent SNR (and my
commercial receives have no difficulty at all hearing them). This is measured
on an Agilent 8563 spectrum analyzer.

Could your PLL synthesizer be dirtier than you think?

I've done some wideband sweeps of it, and there are some spurs that are only
~ -70dBc. It's obvious when you choose a channel that suffers from these
higher-level spurs, though... it'll add 20dB or more to the IF spur. I
ditched my homebrew PLL-based synthesizer for a good HP box borrowed for
testing, though (its spurs are more like -90dBc worst case), and the problem
is still there.

One way I would think about this is to ask if you looked at known circuits
that work and ask yourself what are you doing that is different from known
circuits that work.

The usual problem is that very few circuits found on the Internet actually
come with performance data -- unless you build them yourself, you really have
no way to know if they're just as bad or worse than your own efforts!

We also had some posts maybe 1-2 years ago where a guy was working with
chips and circuits and computer modelling (IIRC) and he was unhappy that he
was not getting (with real circuits) what his computer modeling program told
him he was supposed to get.

It would almost be alarming if those circuits did work, in real life, within,
say, 0.1dB of their simulated results rather than the more typical 1-5dB
that's often still considered "good agreement!"

---Joel

On 11 Jun, 22:52, "Joel Koltner" wrote:
Hi Dave,

Thanks for the information; it's quite helpful!

I was (am) committing the faux pas of feeding the mixer output directly into a
narrow bandpass filter...

For example, the popular
Mini-Circuits SBL-1 mixer seems to have LO-to-IF isolations of
anywhere from better than 65 dB (at HF) down to around 30 dB (at UHF).
If you're trying to tune a weak signal (say, 80 or 90 dB weaker than
the LO signal) then the residual LO feedthrough can cause the sort of
swamping you're seeing.

So... say I'm using a 45MHz IF, trying to tune 414MHz using the SBL-1 and an
LO of 414-45=369MHz (low-side injection) at the SBL-1's recommended +7dBm.
Presumably I'll see a 7dBm - 30dB (LO-IF isolation) = -23dBm signal at the IF,
but it should still be at 369MHz, right? *How does the LO "bleed through" to
the 45MHz IF... and what power level should I expect to see there?

---Joel

DBMs should be terminated in 50 ohms over a wide bandwidth. MCL has
recommendations concerning this.

Leon

On Jun 11, 5:52*pm, "Joel Koltner"
wrote:
Hi Dave,

Thanks for the information; it's quite helpful!

I was (am) committing the faux pas of feeding the mixer output directly into a
narrow bandpass filter...

For example, the popular
Mini-Circuits SBL-1 mixer seems to have LO-to-IF isolations of
anywhere from better than 65 dB (at HF) down to around 30 dB (at UHF).
If you're trying to tune a weak signal (say, 80 or 90 dB weaker than
the LO signal) then the residual LO feedthrough can cause the sort of
swamping you're seeing.

So... say I'm using a 45MHz IF, trying to tune 414MHz using the SBL-1 and an
LO of 414-45=369MHz (low-side injection) at the SBL-1's recommended +7dBm.
Presumably I'll see a 7dBm - 30dB (LO-IF isolation) = -23dBm signal at the IF,
but it should still be at 369MHz, right? *How does the LO "bleed through" to
the 45MHz IF... and what power level should I expect to see there?

Ah - here you hit something on the head. I didn't know you were asking
about UHF, but now I can contribute!

SBL-1 balance isolation gets substantially worse as you move from HF
to VHF to UHF. Looking at the spec sheet, as you move from 3MHz (where
I use the SBL-1) to 369 MHz (where you use the SBL-1) the isolation
gets 34 dB worse.

*AND* the typical and guaranteed numbers are only if all the ports are
terminated nicely.

So while my gut feeling was originally "LO-IF isolation, that's never
been a problem for me at 3.5 or 7 MHz with the SBL-1" now I look at
the spec sheet and think "wow, isolation is a much bigger problem at
UHF than anything I've ever dealt with in the SBL-1 at HF"!

If you can arrange the IF stuff at the output of the mixer such that
you have a good flat termination match and such that you hit an IF
filter that will get rid of the LO before any substantial gain, I
think you'll come out way ahead of just switching to a different
variety of mixer.

Tim N3QE

In article ,
Joel Koltner wrote:

Hi Dave,

Thanks for the information; it's quite helpful!

I was (am) committing the faux pas of feeding the mixer output directly into a
narrow bandpass filter...

For example, the popular
Mini-Circuits SBL-1 mixer seems to have LO-to-IF isolations of
anywhere from better than 65 dB (at HF) down to around 30 dB (at UHF).
If you're trying to tune a weak signal (say, 80 or 90 dB weaker than
the LO signal) then the residual LO feedthrough can cause the sort of
swamping you're seeing.

So... say I'm using a 45MHz IF, trying to tune 414MHz using the SBL-1 and an
LO of 414-45=369MHz (low-side injection) at the SBL-1's recommended +7dBm.
Presumably I'll see a 7dBm - 30dB (LO-IF isolation) = -23dBm signal at the IF,
but it should still be at 369MHz, right? How does the LO "bleed through" to
the 45MHz IF... and what power level should I expect to see there?

I'd say that you could be seeing the 369 MHz LO signal in your IF
pathway through a number of mechanisms.

First: some of it may be getting past your bandpass filter. If
you're using a crystal filter, you may find (if you look) that the
filter's stopband attenuation is neither wonderful nor flat at some
frequencies. Crystals often have secondary resonances at (or near)
the harmonics of their fundamental frequency - that's how overtone
filters work - and your LO of 369 MHz isn't all that far from the 8th
harmonic of your IF frequency.

You'd probably need to sweep your filter's response (wideband
oscillator plus an RF detector or a spectrum analyzer) to determine if
this is occurring.

If this is the problem, you might remedy it by adding a one- or
two-stage passive lowpass filter (maybe LC, maybe just RC) somewhere
after the mixer and before the final detector. This would attenuate
out the residual LO signal that the bandpass filter doesn't. Maybe
run the mixer output into that grounded-base/source 50-ohm buffer amp
I suggested, gain a few dB, and then run the signal into a slightly
lossy low-pass filter whose output presents your bandpass filter with the
termination impedance it expects?

Second: you may be getting parasitic transmission of the LO signal
*around* the bandpass filter - either capacitive or inductive. If
your bandpass filter is a tuned LC type rather than a crystal filter,
there might be coupling between the inductors.

Third: you might be getting some LO pickup by the gain stages in the
IF path, direct from the local oscillator - wire-to-wire coupling or
something like that.

Fourth: you might be getting LO wandering around on the power supply
lines or in the ground paths/planes - insufficient power-supply
decoupling, shared power-supply wires, shared ground paths, and so
forth.

I get the impression that careful attention to layout, shielding, and
decoupling are very important for higher-performance receivers. Some
builders seem to prefer to place the LO, mixer, and filter in separate
shielded sub-compartments (soldered together from double-sided PC
board material if nothing else), to carry the signals between these
compartments on shielded coax, and to use feedthrough caps and ferrite
beads on the power wiring. This helps minimize the various forms of
parasitic coupling between stages.

--
Dave Platt AE6EO
Friends of Jade Warrior home page: http://www.radagast.org/jade-warrior
I do _not_ wish to receive unsolicited commercial email, and I will
boycott any company which has the gall to send me such ads!

See at end...

On Wed, 11 Jun 2008, Joel Koltner wrote:

"A" wrote in message
.com...
At this point I'd like to know what your problem is with "significant amount
of energy..." compared to whatever your ultimate goal is.

I'll get something in the ballpark of an -90dBm spur at the IF when the noise
floor is down around, say, -120dBm. Hence I have a harder time recovering
signals at, say, -100dBm even though they still have a decent SNR (and my
commercial receives have no difficulty at all hearing them). This is measured
on an Agilent 8563 spectrum analyzer.

Could your PLL synthesizer be dirtier than you think?

I've done some wideband sweeps of it, and there are some spurs that are only
~ -70dBc. It's obvious when you choose a channel that suffers from these
higher-level spurs, though... it'll add 20dB or more to the IF spur. I
ditched my homebrew PLL-based synthesizer for a good HP box borrowed for
testing, though (its spurs are more like -90dBc worst case), and the problem
is still there.

One way I would think about this is to ask if you looked at known circuits
that work and ask yourself what are you doing that is different from known
circuits that work.

The usual problem is that very few circuits found on the Internet actually
come with performance data -- unless you build them yourself, you really have
no way to know if they're just as bad or worse than your own efforts!

We also had some posts maybe 1-2 years ago where a guy was working with
chips and circuits and computer modelling (IIRC) and he was unhappy that he
was not getting (with real circuits) what his computer modeling program told
him he was supposed to get.

It would almost be alarming if those circuits did work, in real life, within,
say, 0.1dB of their simulated results rather than the more typical 1-5dB
that's often still considered "good agreement!"

I don't recall what the guy was upset about. I don't think (IIRC) that we
had any comments regarding what, quantitatively, he was looking for vs.
what he was getting, quantitatively, but he was definitely upset.

I don't have anything like the fancy gear you have and my criteria are
more or less "if you can hear it and its OK with you, then its fine" (for
receivers anyway). On transmit, I have enough tuned circuits in various
places that my spurious signals should be good enough for the FCC. And,
I'm using quite clean free running oscillators at VFO, LO points. Its
really wonderful to be able to see sine waves on any one of my three
Tektronixs scopes, thought. I'll add that if I don't have at least some
minimum number of tuned circuits, that "sine wave" develops very visible
"distortions." Maybe someday I'll be interested in quantitating them.

In the meantime, good luck on your endeavors. And, thanks for the fills on
the technical stuff. Can't help you any more than that.

---Joel