EMRFD Message Archive 770

Message Date From Subject
770 2007-05-20 13:43:13 Kevin Purcell NJM2594 Gilbert cell double balanced modulator/demodulator
The NJM2594 from Japan Radio Corporation might be useful to the RF
experimenter

> The NJM2594 is a double balanced modulation/demodulation circuit,
> applied to suppressed carrier modulation, amplitude modulation,
> synchronous detection, FM or PM detection circuit.
>
> Single [-ended] input, [supply] voltage [typ. 4.5 to 9V -- max 14V]
> and simplification of external circuit offers wider applications.

<http://www.ortodoxism.ro/datasheets/newjapanradio/be04016.pdf>

Current consumption is 11mA (14mA max). Rather higher than the 612
and implying a better IIP3.

Conversion gain is low: typically 0dB (plus minus 2dB). Useful for
keeping the gain down before a filter.

An intermodulation figure is given: -60dBm (though the graphs show
high IMD by 5dB) with two -14.4dBm RF signals (1.75MHz and 2.00MHz -
250kHz apart) and a -7dBm LO at 28.25MHz. So the intermodulation
product is -45.5dB wrt a single tone.

A little geometry (if I got it right) gives an IIP3 +8.4dBm (or +6dBm
if you prefer the -55dBm intermodulation figure on the graphs).

This is reasoanble for a moderate current Gilbert cell mixer. For
example, the SA612 IIP3 is around -13dBm with 2.4mA current
comsumption. As IIP3 in a Gilbert cell goes as the cube of the
emitter currents in the mixer and assuming the current is split
amongst biasing and the mixer tree in the same proportion then we'd
expect a difference in IIP3 to be 30 log(10/2.4) or 18.6dB. That's
close to 8dBm-(-13dBm).

The +8dBm intercept is not too far from SBL-1/TUF-1 diode ring mixer
that would have an IIP3 of +11 to +16dBm. Plus the drive power for
the NJM2594 is much lower (saving power consumption in the LO).

Intermodulation and current don't change much with supply voltage so
there is no advantage to running at high supply voltage to increase
the mixer tree current.

Inputs are all unbalanced (different from most other gilbert cell IC
mixers). Impedances are all useful: RF Zin = 600 ohms/3.8pF; LO Zin =
1200 ohms/2.2pF

The balance numbers aren't great (RF leakage typ -35dB; LO leakage
-40dB type but both have -20dB max). You can add external bias tweak
to improve the leakage. That would be useful in a DC receiver (and
with no gain a J310 ground-base amp to set gain and improve isolation
would help too).

The IF output can be taken either from the collector at the top of
the mixer tree Zout = 350 ohms/2.6pF or from a transistor provided on
chip for emitter follower output with a external resistor [510 ohm at
5V though you have some latitude in the value].

Noise figure is not mentioned. Somewhere between "single figures and
the low teens" perhaps? It would be useful to measure this!

Available in surface mount packages: DMP8, SSOP8

Available at DigiKey.

Not used too much in amateur work (yet). The only built reference
I've found is this (odd) single frequency 7MHz SSB/AM transceiver by
JG1EAD where he uses it as a modulator to generate SSB from cheap
7.0445MHz crystals. Pictures but no schematic.

<http://www001.upp.so-net.ne.jp/jg1ead/7044am/7044am.html>

He references a JJ1SLW page but that page seems to be 404

and talking about replacements for SN16913P (the Google Translate
Tool is useful for reading these even if it does make it sound like
haiku and it seems to have a translation limit too).

<http://jg1ead.web.infoseek.co.jp/oldlog/bbslog20021103.html>
<http://jg1ead.web.infoseek.co.jp/oldlog/bbslog20021208.html>

It has been used in a Bat detector
<http://www.mb-labo.com/Batdetector/BD5B_spec.htm>

And mentioned here
<http://qrpforum.de/thread.php?postid=9912>
--
N7WIM / G8UDP
Kevin Purcell
kevinpurcell@pobox.com