Single balanced mixer topology without transformer

Hello,
In page 9 of the attached document a very "weird" mixer is presented.
It is a single balanced mixer that has some short of isolation but without using any LC or transformers.
This fact, could possibly lead to a very wideband mixer design, if proper diodes are chosen and if extreme mixer performance is not needed.
I would like your comments on this design, plus any estimation about the bias resistor value and bias voltage.

There are many mixer designs, all of which can work. When Schottky diodes were rare, I used single-ended mixers with good resutls.

Problems are that such mixers need much more LO power and suffer unwanted leakages.Then more filtering and using isolators is necessary.
Good modern mixers are singly up to triple-balanced to improve performance. At low frequency you can use transformers at microwaves there are other options like rat-race or waveguide designs.

jiripolivka said:

There are many mixer designs, all of which can work. When Schottky diodes were rare, I used single-ended mixers with good resutls.

Problems are that such mixers need much more LO power and suffer unwanted leakages.Then more filtering and using isolators is necessary.
Good modern mixers are singly up to triple-balanced to improve performance. At low frequency you can use transformers at microwaves there are other options like rat-race or waveguide designs.

Click to expand...

Yes I have seen some of these, like this one https://www.qsl.net/va3iul/Homebrew_RF_Circuit_Design_Ideas/2304MHz_Mixer_WA2ZZF.gif
The author of the topology I refer, states that the mixer was tested with a LO of +10dbm.

"with 50 ohm RF source impedance at F2, 50 ohm IF load resistor, and
300 ohm LO source impedance at F1 with LO power at +10 dBm, conversion loss
is 13 dB, LO to IF isolation is 17 dB, and RF to IF isolation is 11 dB."

The LO level is not higher than that of an SBL-1. The loss is more than double, but it could be acceptable for such a simple circuit. It is for sure better than a single diode as it provides some isolation, yet the design is really simple.
I have no idea about the load resistor to try and the bias voltage.
What is the effect of the biasing of the diodes in the mixer performance?

Normal diode mixers use unbiased diodes, but in case that small DC bias current is used, the conversion loss can be reduced, if is not available high level signal from oscillator.

Also DC bias (adjustable) can be used in a diode type harmonic mixer, to make an "even only" mixer operate with odd LO harmonics, or an "odd only" mixer operate with even LO harmonics.

vfone said:

Normal diode mixers use unbiased diodes, but in case that small DC bias current is used, the conversion loss can be reduced, if is not available high level signal from oscillator.

Also DC bias (adjustable) can be used in a diode type harmonic mixer, to make an "even only" mixer operate with odd LO harmonics, or an "odd only" mixer operate with even LO harmonics.

Click to expand...

That was very helpfull, thanks.
I have tested it with a 400mhx mxo and a v/u transceiver in the low output power setting. It seems it works ok. Reducing the voltage of the mxo reduces its output power and this greatly reduces IM products, but if reduced too much the wanted mixing products are reduced also. It seems that the LO/RF signals levels must be optimized so that the IM products are kept low, as it should be expected in all mixers I think.
With a 10K bias resistor and 0-12v bias voltage I see no change in the signals. Maybe the current is too low, I think I am going to try a 1K as well.
It would be interesting to see how it behaves on microwave as well...
Any comments or points you like to share about this mixer, or things to try?

If you are dealing with uV level signals, DC bias helps to bias the diodes at the beginning of the knee of the curve with inductive isolation between DC and RF.

Image rejection, f band and width of RF and LO required for IF are all variables as well as THD or IP3 intercept.

Amplified RF signals won't need the DC bias and max input current will be specified.

https://www.minicircuits.com/pdfs/MDB-73H+.pdf is a good example of a simple yet high perfomance DBM with LO,RF from 2.2 to 7 GHz with IF =0 to 1.6GHz low compression, low IP3 and low noise figure.

Modern DBM's may use ceramic substrate and GaAs diodes.

SunnySkyguy said:

If you are dealing with uV level signals, DC bias helps to bias the diodes at the beginning of the knee of the curve with inductive isolation between DC and RF.

Image rejection, f band and width of RF and LO required for IF are all variables as well as THD or IP3 intercept.

Amplified RF signals won't need the DC bias and max input current will be specified.

https://www.minicircuits.com/pdfs/MDB-73H+.pdf is a good example of a simple yet high perfomance DBM with LO,RF from 2.2 to 7 GHz with IF =0 to 1.6GHz low compression, low IP3 and low noise figure.

Modern DBM's may use ceramic substrate and GaAs diodes.

Click to expand...

Thanks for the info, it is really helpfull.
That is what I was thinking about the bias, to help the diodes switch ar low input signal levels.
I have not tested the small signal capability of the mixer yet, I have used a buffered mxo and a v/u transceiver for testing. However I have found that in these larger signals, the bias setting changes the level of the main mixer products by about 5db. Leave it unconnected or ground it and the main mixer products are 5db below. I have used a 1K for the load resistor, but I may change this in the future.

I have not yet measured the isolation between the ports, I intend to do so these days, it would be interesting to see if the author's claimed isolation can be achieved.
IF goes down to 10KHz (measuring capability) and maybe below. The high frequency limit of IF has not been determined yet, but it is above 600MHz for sure as I see products in this frequency.

Like you said, I believe small RFC after this resistor, connected to the diodes bias point, will prevent RF from escaping to the PSU. But this choke will not be so broadband anyway and needs to be changed in different frequencies. Also a shunt capacitor may help (like the biasing of the MMICs).

I was wondering if an IF (load) 50 ohm resistor needs to be permanently connected to the IF port, or the 50 ohm impedance of the next stage of the mixer (eg. measuring equipment or filter) is enough?

It will need impedance matching and 50 Ohm GAT-3+ is suggested on both ports for optimal response.

SunnySkyguy said:

It will need impedance matching and 50 Ohm GAT-3+ is suggested on both ports for optimal response.

Click to expand...

So you suggest to permanently connect a 50R resistor at the load port (IF) of the mixer, regardless of the measuring instrument 50R port impedance?

Wouldn't that parallel the permanent and the instrument resistances to yeld to 25R?

Do I need also to connect the series resistors to ports F1 and F2? Or they are there in the schematic just to describe the impedance of the signals that will feed the circuit?
Just to ensure that the impedances will be 50R for F1 and 300R for F2 (as the paper describes), can I connect two real 50R and 300R resistors in series at these ports?