HIgh level mixer LO/input isolation in "real world"

[ZL1BDA]

Good Day! I've been trying out a TUF-1H+ mixer in my new home brew rig.
It is driven by a "Norton" rf stage preceded by a tunable BP filter over the HF range.
However I've run into some problems with local oscillator (45-75 MHz) feeding back through the
mixer rf input and the amplifier. The amp is stable and the Lo power is +17 dBm.
I could possibly use more shielding but it is definitely far too high, especially as the specs show
about 55 dB or better rejection.
Of course I am already considering using a 32 Mhz LP filter between the rf and mixer.
However I would really like to know how others have realised similar designs in practice.

FYI I have also determined that the W7ZOI hybrid cascode If amplifier is pretty good once the small mod
to the long tailed pair is used. Also if j309's are used the chain of biasing diodes has to be reduced to
one or two in series. Mine has an extra 9 MHz filter after the If amp and before the product detector.

So anyone with experience of a good layout for a high level mixer???
73's Cliff wright ZL1BDA ex G3NIA

[W1VT]

That is the disadvantage of the "Norton" amplifier--relatively poor reverse isolation.

It can be improved upon by replacing the simple transformer with a directional coupler, but a lot of work needs to be done to get a version that is actually a desirable replacement for the original.

Zack W1VT

[TANAKASAN]

Years ago in Ham Radio magazine there was an article on high performance front ends by Ulrich Rhode. If you can find a copy (I'm still looking for you) then read through that because it gives you all sorts of layout tips. One thing I do remember is the amount of shielding he used including shields between the pins of the mixer.

If I find the article over the weekend I'll tell you which edition it's in.

Tanakasan

[M0HCN]

A diplexer terminating the RF (and IF) ports will help a lot (And will do very good things for your mixer intercept point as well).

Given your high side IF setup, it should be possible to design a simple diplexer to send the IF leakage into a terminating load while passing the HF spectrum, and on the IF side to provide a reasonably broadband mixer termination (Which your 45Mhz first IF filter will not present without help).

Regards, Dan.

[G3RZP]

Tanakasan,

That was back in about 1973 or 4. The Intercept points quoted were very interesting, because the crystal filter specified just could not manage it! I wrote to HR about it and DJ2LR wrote back to say that he was just quoting somebody else's figures - which he didn't say in the article.

The old Plessey Semiconductors SL6440 had about -40dBm or a bit less LO signal at the input port, with a +30dBm input IP3 for a 0dBm input osc level, -1dB conversion gain and an 11dB DSB noise figure. Plus gain if you used a higher load impedance.

Long gone from production, I'm afraid.

73

Peter G3RZP

[TANAKASAN]

OK, found it, Ham Radio Magazine October 1975 pages 26 to 30.

Dr. Rohde uses a grounded gate FET amplifier where as modern techniques favor the use of a diplexer, either way you need to make sure that the mixer sees 50 +/- j0 on all three ports. The screening in the schematic shows that all three ports are isolated plus there's another screen between the source and drain on the FET. There's even a novel screening arrangement between the input and output pins on the first IF filter.

If you post an email address I'll send you a PDF copy of the article.

Tanakasan.

[ZL1BDA]

Thanks Guys!
I have some of Ulrich Rhode's article, but not the actual shielding details. It was also a bit sketchy about the interstage
arrangements I seem to remember. If anyone can point me towards that sort of detail I would be very grateful.
However I can improve the screening across the mixer which should help and yes I am making a proper 50 Ohm termination for the IF
output. At first I was relying on the MIMIC chip I am using as a post mixer amp to match to 50 Ohms. But I'm afraid that I am
going off MIMICS a bit. Despite carefully designing the DC conditions for those I have used I have had about 3 failures in various applications
already.
BTW the front end is built in an old Motorola car phone die cast box on PC stock so overall it is pretty well isolated.

Yes, looking at the Norton circuit, it is pretty obvious that the reverse signal can get right back through the transformer!
Otherwise after more screening I will put in a Pi 32MHZ LPF between the mixer and rf. There is a 7 pole Chebyshev LPF in front
of the rf stage anyway, so I wouldn't expect much signal to radiate. The whole rig is designed as a "partial" transciever.
In that the second 9 MHz filter is diode switched and a separate die cast box contains all modules to generate SSB or CW
on all bands with an output level of about +16dBm to drive an external linear amplifier system.
Just looking at Dan's suggestion again. Very good! I can see that a HPF @ ~ 35MHZ with a 50 Ohm termination on the rf input
might indeed have a good effect, both killing the feed thru and improving the mixer termination at the same time.
I will do some design this afternoon!!!

One other point. Has anyone tried a version of the "all HF band" tunable BP filter as YU1LM and others have published?
Mine works fine in terms of  pass band and tuning range, however I notice that the input impedance goes all over the
place as the filter is tuned. On some frequencies it can be as low as 15 Ohms and as high as 100!
This does make the design of the input attenuator a bit problematical. Has anyone else come across this problem?
Vy 73 Cliff wright ZL1BDA ex G3NIA

[ZL1BDA]

Sorry Tanakasan your post crossed my reply! My email is c.c.wright@clear.net.nz.
Thanks for the offer OM.
Cliff wright

[KB1GMX]

The other problem is that the base Norton amp has not only poor reverse isolation
it also suffers that if either port is not matched the other will not be either.  This
means the TUF1H specs go out the window as those specs are only valid if all
three ports are 50 ohm matched.

Don't have them handy look up the work Chris Trask did with augmented amps
to improve the reverse isolation.

Additional work is in the ARRL book Experimental Methods in RF design, there
is a chapter on high level front ends.  I'd add you maw want to look at the various
switch mode mixers as candidates and the LO needs are lower.

Having designed high IP front ends I can say, if you think your shielding is
excessive then, its just about enough.  Shielding is your friend.

Also at high levels sneak paths like weak shielding on cheap coax,  wires
in the air, or even the DC leads to that circuit will radiate.  Same applies
the the instruments used to measure the results.  A 17 DBM LO is 50mw
and that will radiate/couple to the wrong circuits given a chance!


Allison

[WD4HXG]

As others have pointed out the Mixer Specs depend on
each port being driven/terminated with a 50 ohm RESISTIVE
Source/Load. That is not an easy task even with diplexers
and pads. At best it is typically a tradeoff.

Based on my experience the diplexer will improve matters
greatly if not already in use. Even the noise figure will
normally improve. If using a fixed LO you can use an
impedance matching network to get near enough to
50 Ohms resistive to improve the mixer performance
dramatically. Also make sure the drive power is in
spec at the mixer input. If the LO changes in frequency
then you may be better served with a resistor based
feed to the mixer port from the LO. (Resistors
are much more broadband than LC matching networks)

55 dB LO isolation is a lot. You are talking about over
five orders of magnitude reduction in power. You might
want to look at the mixer's actual graph data if you
have not done so. Marketing in companies highlight
their penultimate performance and often ignore or
neglect to mention that the performance is only valid
at one frequency or a very narrow range of frequencies.
I have used differing mixers over the years and it is not
unusual to see a nice LO isolation spec only to discover
that as frequency increased the isolation spec deteriorated
rapidly with changing frequency.

I usually place a 3 dB pad between the LO and the mixer
as that introduces an additional 6 dB return loss improvement.
The passive mixers I used really loved that pad.
Of course if using a 50 mW LO then you have to increase the
LO output to 100 mW to keep the mixer happy. That carries
an increased power need. In fairness I have run mixers 1 dB
to 2 dB low on driving power and still obtained good performance
over the temperature range found in the shack. 3 dB lower seemed
to impact the conversion loss notably and also the stability
of specs such as conversion loss/noise figure, along with
isolation.

Examine the impedance of the port feeding the RF input
(the output impedance of the RF amplification and filtering
stage). Ideally you want the output to exhibit a flat
return loss. If it isn't than again you can insert a 3 dB
pad to help keep the output impedance at a level which
should keep the mixer happy but pay attention to your
noise figure. If the pad is after the active device in
your rf stage then it should have minimal effect on the
receiver noise figure provided there is sufficient gain in
the rf amp.

Vendors usually spec their products utilizing 50 Ohm
resistive sources and termination. It is up to you to
determine what deviation from 50 ohms resistive
will do to the product performance. This is a caveat
that can really burn a designer when using coax.