Have done many experiments with Mini Circuits ringmixer SRA-1H. It is a highlevel mixer with 50mW LO power.

Here is the hole circuit. RF port and IF port is terminated in 50 ohm broadband termination. 3rd order IP on mixer input was ca +27.5dBm

Here the mixer is terminated in a series resonnance circuit on the input and a narrowband (50ohm) termination with xtal filter on the output. The IP3 has dropped to 12.5dBm, mainly because of the bad IF port termination, RF port is not so critical although it is wrongly terminated

Here the input is couplet via an input dividing filter, and the output is correct terminated via a 50 ohm amplifier to the the spectrum analyser and the IP3 is still +27.5sBm. On the mixer output it is important to use correct termination. Tried the same method as for CDG2000 (Radcom summer 2002), diplexer and 90 hybrid. It looks promissing, but after the hybrid two equal xtal filters are required. Between the diplexer and 90 I have tested the amplifier with two transformers. It has good dynamic and can handle the output signal from the mixer without reducing the IP3. (LA7MI)

Input filter curve, -3dB @ 18MHz for each leg

To make these experiments you need a two-tone generator and a spectrum analyser, more constructional details are shown below

Highlevel IF buffer for several purposes (LA7MI)

In the book written by William Sabin og Edgar O. Schoenike "Single.sideband systems & Circuits" from 1987 is shown a circuit with good dynamic and low noise. The book was publisched in 1987 .

The most suitable device is 2N5109 with 50mA collector current- Have suggested component values for 1-30MHz. The transformer winding ratios decides amplification and input impedance. It was tested with 1:4 for both transformers, it was found suitable for 50 ohm impedance and 12dB gain. LA7MI used a Mullard two-hole core - bought as surplus from Birkett in England, permittivity was measured to =180. The polarity is very important, otherwise you'll make an oscillator. It was quite tricky to do it right! He believes it is better to use a two-hole core than standard toroid core for good coupling between the windings. (LA7MI)

Wanted to test the amplifier with a different toroid than what was already described. Couldn't find any 2N5109, and the 2N3866's were surplus and regarded as unreliable, so I used 2N4427. Also used some low-performance capacitors. Gain is 12dB, it drops 1dB @55MHz. The current gain for 2N4427 seem somewhat low for 50mA current. The toroids are Philips 3122 134 90783 (ferrite grade 4A4 permeability 500) with RG58/U braid used for the single winding to achieve sufficient coupling. See page n12. It was somewhat difficult to solder the transformers the right way around, and smaller cores or beads would have been an advantage. The problem with this circuit is somewhat high return gain, suppose it is better for W4ZCB's Zwischenbasis amplifier described on page c19, but should be of lesser importance according to LA7MI. ((LA8AK))

My comments about RF transformers where it is one turn consisting of a brass tube, is this really neccessary, could a coax braid be used just as well?
The ends could be soldered together, and you need no end plates (LA8AK)

Redifon high-level mixer

LA7MI Stein Torp writes that he has tested a mixer which was used in a Redifon (GR-345) mobile transceiver - advertized in Wireless World in 1965, but used sort of modern components. The diodes in the construction shown are smd version matched twin 2800. The input transformer is a pentofilar wound type with 5x 3turns (0.25mm enam cu wire), see drawing for the type of Mullard two hole core used in the transformers (=200, sold by Birkett).T2 and T3 are wound with 2x 3turns 0.3mm wire. With input on 80m and 10.7MHz IF (LO = 14.5MHz) 3rd order IP on the input was measured to +20dB, and the strange is that output termination is not critical at all. It is possible to terminate it with an xtal filter and still obtain +20dBm, which is not at all possible with a ringmodulator. Measured the 1dB compression point to +10dBm. With some more LO excitation it is possible to increase the IP3 above +35dBm, see the notes below.
Diode-pairs (Ferranti ZC5800) and transformers in the original version seems to have been building blocks manufactured by Hugh Aircraft Corp., but the component designations does not make sense.

This mixer is also mentioned by G3RZP Peter Chadwick in ART6 (written by G3VA) pp131-132 and ART7 pp140, but two diodes are incorrectly drawn, and he doesn't mention that the mixer is not sensitive for termination impedance.

Changing the oscillator drive circuit to 74HCU04 the IP3 increases to +25dBm. But the diodes may be driven even harder, so R1 and R2 are reduced to 120 ohm and IP3 increases to +30dBm.
At last he thought it was worth it to try to increase the voltage for 74HCU04 to 7V, hoping that the manufacturer's data could be forced up a few volts, and the IP3 increased to +35dBm. It is no problem using an xtal filter at the output (see the note for 18.558MHz xtal filter below), but it seems to be an advantage with a lowpassfilter on the input, with lower impedance on the higher frequencies, all this would totally spoil the performance of an ordinary ringmixer.

An improved buffer which also can adjust symmetri of oscillator signal to minimize LO leakage

The 18.556MHz crystal ladder filter is shown in Radcom, Technical topics Dec. 95 pg 71.
See constructional details and more xtal filters on page n23

Reference: RA3AAE anti-paralell diode mixer (harmonic mixer/product detector with diode pair) on page c21

Most topics are received in letters from LA7MI.
He has a soldering iron, but no email, and is reached by telephone nr
He will be very pleased to receive comments, the postal address is: LA7MI Stein Torp, Tollbodalmenning 34, N-5005 Bergen, Norway


Last update: 2004.10.26