[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Fwd: H-Mode Mixer emails



Hello, All!

I have sent to the list a very interesting set of emails
from GianCarlo Moda, I7SWX regarding H-MOde mixers.  I have
done this because I think if they can be made to work at 144
MHz, they would prove to be outstanding as a front end for
Linrad for 144 MHz EME.

There are several messages, I have sent them in
chronological order.

Many thanks to Giancarlo, who found a post of mine with
questions on the H-Mode mixers and composed these very
informative messages!

http://www.qsl.net/w3sz/Fig_2.pdf
http://www.qsl.net/w3sz/Fig_3.pdf

Are Giancarlo's figures referred to in his emai.

73,

ROger W3SZ

Giancarlo's messages follow: 
 
----- Message Forwarded on 05/29/03 -----
From: Giancarlo Moda 
To: w3sz
Subject: H-Mode Mixer
Date: Mon, 26 May 2003 09:55:14 -0700 (PDT)
Dear Roger,

My name is Gian and I am I7SWX.

I read on a web your message informing about the
publication in RadCom April 2003 of my version of the
H-Mode Mixer and about your question: "What the
differences are between the H-mode mixers and Leif's
design (if any) I am not smart enough to see, except
that the transformer configuration appears different
with the H-mixers."

The two mixers are different. If I do recall properly,
the  Leif is a single balanced mixer and it was using
a 74HC4066 IC.

The H-Mode Mixer was developed by Colin Orrabin,
G3SBI, around 1993.

Let me tell you a short story on the H-Mode Mixer. In
the 80s, Ed Oxner, at that time working for
Siliconix,developed a the doubly balanced FET ring
mixer, using an integrated array of MOSFET, the
Siliconix SD8901. This circuit was an upgrade of the
well known DBM with 4 schottky diodes. I am sure you
can find the diagram searching on the web or looking
in some handbooks. Although Ed’s mixer offered an IP3
in excess of +30 dBm, it had a loss of 8 to 9 dB,
still certainly a superior performance versus the
classic Double Balance ring Mixer with diodes. The
main deficiency of this circuit is the fact that, as
the RF input signal level increases it has a
significant effect on true gate-to-source voltage
needed to switch the FET “on” or keep it switched
“off”, increasing the IMD. This aspect requires larger
local oscillator amplitude and we may still have
linearity problems due to the difference, in the FET
Ron resistance, between negative and positive RF
signal states. All these matters were at the origin of
G3SBI H-Mode Mixer development, where the source of
each FET is now grounded, so that the RF signal
switched by the FET cannot modulate the gate-source
voltage. The first H-Mode Mixer used the SD5000 FET
array, it had avery high IP3, >+50dBm and an
attenuation similar to Oxner mixer. Also it was
electrically quite complex. In 1998, I presented the
idea of using the newly announce Fast Bus Switches
FST3125. This idea was tested first by Bill Carver,
W7AAZ (ex K6OLG) and proved the mixer was simplified,
reproduceable and gave good IP3 performances, around
+45dBm, and a low attenuation, -4.5 to -5 dB. This
solution permitted the development of the CDG2000 with
a simple and reproduceable front-end. The IP3
differences, with the H-Mode Mixer are certainly
associated with the type of transformers used,
commercial or home made. 
Now we have a very HOT Mixer and all the limitations
are showing up on the amplifiers, filters and passive
components like transformers and coils.
Today a new receiver project should have the H-Mode
Mixer as "THE MIXER" in the G3SBI or I7SWX version. 

If you will look at the Oxner mixer and at the Yaesu
FT1000 and FT920 first conversion mixer, you will
"recognize" the similarity between the two.

It would be interesting to see the differences
replacing the original Yaesu mixer with the H-Mode
one. I am sure you will discover IMD/IP3 limitation in
ather parts of the receiver, amplifiers, second and
third mixers.

The performances of the H-Mode Mixer are not easy to
measure, unless you have proper instrumentation.
I suggest to "risk" and try....you will appreciate the
results. 
I have applied a mod to my Icom IC751. I have replaced
the third mixer, a dual gate mosfet, with an I7SWX two
transformer H-Mode Mixer and some additional pieces to
adapt input/output impedances. The total attenuation
of this mixer is around -8/9dB. The Receiver's IP3 is,
measured at 10MHz input with a very clean two tones
generator, with -10dBm signal is now:
a) tones  spaced 20kHz = +15dBm and at 10kHz =
+1/0dBm.  Previously was around -10dBm at 20kHz
spacing. As you can guess, the IP3 limitation is now
clearly on the second mixer, then on all the other
items I previously mentioned.

When one does a receiver design has to use the
technology/technique he likes.... he can still carry
on with the ring diode mixers but should look at the
H-Mode Mixer first. Here we are not talking about
NE602 or NE612...

I hope I clarify your question(s)....

73

Gian
I7SWX
F5VGU

Please let me update you with some info on my IC751. 
At 20kHz tone spacing the IP3 is: with wide IF xtal 
filter (3kHz)at 455kHz = +15dBm, with narrow (1.9kHz) 
= +20dBm. 

A friend, Iz7ANL, has replaced the first mixer of his 
IC781 with the G3SBI H-Mode (home made transformers 
and I7SWX 74AC85 squarer as in the I7SWX H-Mode 
Mixer)Tone spacing IP3 dBm Original 
25 kHz +25 +15 
20 +20 +10 
10 +10 +0.5 
5 -5 ? 

He then discovered self-oscillations on the post mixer 
amplifier between the two roofing filters. He will 
replace the second mixer (ring DBM) with the I7SWX 
mixer, as this one has the IMD problems. 

The H-Mode Mixer is for HF. It can cover the 50 MHz 
band. I did a "quick" test on 144MHz has too high 
attenuation, -20dB, but I made a mistake as I tested 
it only at the IF output of 10.7MHz; I was stupid not 
to test it with IF of 70 MHz. Probably, it could be 
used at 144MHz if the IF conversion is between 70 and 
45 MHz, with down conversion, LO = (144MHz - 70MHz). 
This will give an LO band from 74 to 76 MHz (144-146) 
or 74-78 MHz (144-148). The problem of high LO is that 
the limit is due to the Ton and Toff timing of the 
3125. To cover the 144 MHz you could try the above LO 
frequencies and for the transformers use a balun #61 
(125mu) and maybe 4 to 2 turns for each winding. It 
would be a very interesting test as the input to 
output bandwidth of the FST3125 is over 300MHz. Maybe 
you could find some List members available to do the 
tests. I have not much time as I my time is quite 
limited and I can "fiddle" only at weekends, when xyl 
"permits it"...hi. 
Also, at the time I am working with the 3V version of 
the I7SWX H-Mode Mixer and a new squarer (when time 
permitting), so I cannot try the mixer on 
144MHz....but ...you never know...hi! 

If you do something I would appreciate feedbacks. 

BTW, IZ7ANL, Nicola, did previously replace several 
front end switching diodes with PINs (expensive HP); 
he had and improvement on IMD of +3dBm. Now he had a 
failure of one of this diodes and he discovered (I 
told him at the time of the mixer tests that first you 
have to improve the mixer then go after the other 
parts but I do not believe on the diode replacement) 
the improvement was only due to the attenuation of 
-3dB the PINs have inserted into the circuit. I am 
sure if you “pump” 10 mA or more of current in the 
switching diodes the RF circulating in the diodes will 
not do much modulation. 

You may improve IP3 for a few dBm with the balance 
adjustment circuit on the squarer. 
I am not in the condition of measuring the IP3 of the 
H-Mode Mixer as I have only an HP 141T Spectrum 
Analyzer (IP3 = +15dBm). I have sent the original 
mixer to a friend to do the testing but he still quite 
busy....maybe in a month or two he should be able to 
perform the measures and we should know how good it is 
versus the G3SBI one. From my point of view an H-Mode 
mixer with +45 or +41dBm does not make much of a 
difference as all the IMD/IP3 limitations are now in 
the amplifiers, filters and passive components, as 
confirmed with the CDG2000 project. 

I am adding, herewith, the two circuit diagrams of the 
two xfrmrs version, with and w/o balance adjustment. 
You can distribute them. 

73 

Gian 
I7SWX 
F5VGU 
W1-I7SWX 

Herwith is the test Bill Carver, W7AAZ, commented to 
the CDG2000 Group regarding the H-Mode Mixer (G3SBI), 
following a question by one of the group members. 

73 

Gian 
I7SWX 
F5VGU 
W1-I7SWX 

To: CDG2000@
From: "William Carver" 
Date: Sat, 31 Aug 2002 10:34:17 -0600 
Subject: Re: [CDG2000] Mixer 




The FST3125 mixer is an elegant part of the CDG2000 
design and one 
cannot help but consider its applicability to other 
projects. I've not 
measured FST3125 performance above 14 MHz, nor in 
upconversion, but 
there is no question that it will mix: almost ANYTHING 
will "mix". The 
real issues are (1) what is the intercept and (2) what 
is the conversion 
loss? 

0.5nS rise/fall time at the AT THE FET SWITCH GATES 
(not the external 
terminal) would make switching time 10% of the LO 
period and increase 
FST3125 conversion loss. Ed Oxner, of Siliconix, 
worked out the 
mathematics predicting the conversion loss in an 
application note in the 
1980s, but I cannot find my copy of it. 

I measured conversion loss of FST3125 at 4.8 dB with a 
roofing filter on 
it. A diode mixer has about 1 dB higher conversion 
loss, so it would not 
take much degredation of the FST3125 to make that 
factor "a wash" 
between them. 

LO asymmetry in any mixer produces a first-order mixer 
product 
indistinguishable from a third order IMD product. I've 
tried 
compensation for internal gate-drive delays without 
effect and such 
compensation is unlikely to apply over a broad band of 
frequencies. 
Although just speculation, it is not inconceivable 
that a 100 MHz LO 
would degrade the FST3125 intercept by 10 dB, at which 
point it matches 
the intercept of +30 diode mixer. 

So although the FST3125 H-mode mixer is a beatiful fit 
in the CDG2000, 
it would not take much degredation of the FST3125 in 
up-conversion to 
make it less beautiful. A packaged diode mixer is 
expensive and requires 
a lot of LO drive, but "it's only money", and square 
wave LO drive power 
CAN be generated by six 74AC04 sections in parallel, 
and its symmetry 
adjustable to precisely 50-50 to optimize intercept. 
See "Sources of 
Intermodulation in diode-ring Mixers", H.P Walker, 
B.A., M.Sc (of 
Hewlett Packard Ltd, Scotland) in Radio and Electronic 
Engineer vol 46, 
No 8, May 1976. 

Without measurements one can only speculate: only 
construction of a 
breadboard and measurement can resolve differences 
this small. I7SWX has 
some HIGH SPEED 3.3V FST312x switches and wants to 
apply it to 
modification of a Japanese (upconversion) rig. I 
expect to breadboard 
and test an upconversion FST mixer in the next few 
months and can post a 
report if there is general curiosity. 

Make no mistake: I have +30 diode mixers in my junkbox 
and have built 
square wave LO, but I have an FST3125 H-mode mixer in 
my downconversion 
receiver! 

Regards - Bill W7AAZ 




73,
Roger Rehr
W3SZ
FN20ah
http://www.qsl.net/w3sz


LINRADDARNIL