Double Balanced Mixer - Very Linear ?

SSB without carrier can be created by an image canceling I/Q mixer pair.

DSB with/without carrier can be generated by a single mixer by adding some DC component on one input.

I do not see how the I/Q could generate some carrier, perhaps by simply bypassing the carrier around the I/Q mixer.

A passive diode ring mixer has a loss of about 6 dB, thus the noise figure goes up. Any decent RF amplifier in front of that can easily take care of that loss.

With square wave LO, you must also consider the spurious mixing products with the 3rd and 5th harmonic of the LO. The offending input frequencies must be filtered out prior to the mixer that might mix with the LO harmonics and produce the IF.

In an up converting receiver, this is easy and is handled in practice with the same LPF that takes care of the image frequency, also removes the offending frequencies.

On a sunny day (Thu, 03 Apr 2014 08:54:46 +0300) it happened snipped-for-privacy@downunder.com wrote in :

I have an Atmel U2790B quadrature modulator chip somewhere, cannot remember the exact specs, anyways pdf says 50 Ohm, 100MHz to 1GHz, 5V 150mW single input, -1dbm output level, Had it in circuit foe a few moments, replaced it with a digital solution, its expensive (17 Euro or there about), but sometimes used by radio hams.

Both, but for different reasons:

- Creating the analytic signal (two copies at 90 degrees difference) makes the requirement of limited baseband. It must be limited at the low end to make the phase shifting possible, 0 Hz does not go. The minimum delay for the filters is a quarter wave at he lowest frequency.

- All non-ideal mixers create undesired products at the output besides the customary sum and difference frequencies. These must be filtered out, which is a challenge at wide frequency ranges. An ideal mixer is a perfect analog multiplier, which remains to be invented to your specifications.

--

-TV

So the receiver would also be tunable in the 0.1 to 100 MHz range ?

This might OK for a wired system, but if you intend to use it on radio waves, i.e. have an antenna connected directly to the 1496, you are going to be disappointed, with overloading and spurious signals all over the place.

You need a frequency selective front end ahead of the 1496.

On a sunny day (Thu, 03 Apr 2014 06:15:21 GMT) it happened Jan Panteltje wrote in :

I forgot to add, I think for frequencies below 100 MHz most goes digital, it is hard to make the RF 90 degrees phase shift on chip over 100 kHz to 100 MHz I'd think, maybe Jimmy knows how to do that. It is much easier to set up some flipflops, and, if you must, use analog switches to generate QAM. I havedunnit for 4.43 MHz PAL modulator that way (UV).

You get that anyway, because you have to turn the diodes on hard to get good behaviour. Mini-Circuits publishes boatloads of data on their mixers, so you can see it easily. If you don't drive the LO hard enough, you get a bunch of intermod between different signals on the RF port, which is much worse than merely finding copies of the IF offset by multiples of the LO.

Making a good frequency plan, one that minimizes those sorts of problems, is a big part of receiver design.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

Yes, although many people don't believe it, mixers work much better when driven hard by square wave LO signals. This is my main objection to the N2PK VNA and the VNWA3 design also. I expect the mixer gain would be much more stable with temperature if they were driven with square wave LO.

If I really did want to fairly thoroughly get rid of the response at the

3rd, 5th etc. harmonic of the LO, (e.g. to build a direct conversion receiver to cover several octaves of RF input frequency, like 100kHz to 1MHz for example), then I would like to try the following:

• Low-pass filter the 100kHz-1MHz RF input with say a 5MHz roofing / anti-aliasing filter

• In a FPGA, generate a fast-ish (>100Mbit/s) digital bitstream from a sigma-delta modulator which has a numerically generated sine wave as its input, so that the frequency-domain representation of the bit stream is a strong tone at the desired LO frequency (in the range 100kHz to 1MHz) and some noise-shaped dithering at >30MHz, then externally re-clock that bitstream with a flip-flop clocked from a low-jitter clock

• Drive the LO port of the mixer hard with the bitstream, and put the low-pass filtered input signal into the linear port of the mixer

Some (very linear) gain ahead of the mixer might be needed to overwhelm any extra thermal noise from the input and low-pass filter that would get mixed into the IF band due to the shaped noise components of the LO signal, however it may also be feasible to reduce that noise by making sure that the mixer input is presented by the low-pass filter with an impedance having a very small real part (for low Johnson noise voltage) or extremely high impedance (for low Johnson noise current) at those (out of band) frequencies where the shaped noise is strong in the LO bitstream.

I wonder if this is an established technique.

Chris

Interesting idea. One of these days I'm going to have to go through the math of noise-shaping, because I don't really know how it works.

(Interesting things that I don't understand bug me like having a pebble in my shoe, so I eventually get round to figuring them out.)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

Fun thread, thanks for starting it. I don't know much about mixers. We do use the SBL-3 from minicircuits which would match your freq. specs

I think they also make some with FET switches rather than diodes.

I wonder if some more knowledgeable RF person would comment on the differences in using a multiplier (like the AD834 that Bill S. mentioned) Versus a switching mixer. The only "problem" I found with "real" multipliers is they like to be driven by higher voltage levels.. (hmm this doesn't seem to be the case with the AD834 +/- 1V FS.)

Well there is the price.

George H.

'search' just does not go well anymore. nothing happens, hangs PC while that background mode is searching, or finds something, and downloads several MB's of pictures/videos [also hanging PC] etc etc. yet only have

26 cookies and a cleared out cache still takes forever, and now that google has gone super commercial and no one bothers to purge dead links...

Thanks for the URLs from the 'index' i downloaded 831 too.

everything looks like 'overkill', set up for communications, and that translates to relatively poor [for my requirements] carrier suppression, linearity, etc.

That's a very interesting idea! To the best of my recollection, I've never seen this before. If you don't mind, I'll mention it to some of the local RF experts, to see what they think of it.

Jeroen Belleman

FWIW, LTspice analyses of the circuit shown in Figure of spec sheet using the MC1496 containing Jim Thompson's NPN 'chip size' CA3046 models show that at CARRIER crossover, zero drive, the spectral noise output is at its maximum of 94nV/rtHz [fairly independent of the modulation level], then at the peak drive of 60mV, this drops to the minimum of around 45nV/rtHz. Simulations confirm your statement that the noise from a mixer will be less if it is abruptly driven ON/OFF.

This is why searches don't work well. Googled Atmel U2790B, got a first page showing two ads, one from Atmel, one from mouser, followed by a list of 'me-toos' trying to provide datasheets, and their own stuff.

Clicking on the ad for Atmel, did nothing for 20 minutes, blank screen! ok paste in Ateml's home address, that worked.

then on the home page for Atmel did a search for the U2790B product line, info, something and got...

Your search - U2790B - did not match any documents. No pages were found containing "U2790B".

Suggestions: Make sure all words are spelled correctly. Try different keywords. Try more general keywords.

I told you my searches just don't seem to work.

LO's should be simple. I envisioned a four quadrant digital waveform and simply pick off the 'cosine' 'sine' carrier. Albeit sq wave. which most of these mixers seem to prefer. 1MHz to 100MHz is not too difficult a range for generating 'clean' LO's

I already have covered the cos/sin for the modulation bands.

But am concerned with that poor carrier suppression spec. Concerned, only as thought experiment. Having 0.2% of unwanted may not be catastrophic, don't know for certain. Just suspect, based upon previous system where

0.01% did cause problems that only suspect something as high as 0.2% will also cause problems.

Do you work in this area?

Have you simulated using LTspice like and/or Matlab like tools?

I've added a feature to our implementation of LTspice, called .tranoise, which combines .tran and .noise analysis into a single operation. I've been using it to determine the best ways to use the MC1496 mixer and really find what happens to all that johnson noise and junction noise and cross products and 1/f from the OpAmps, etc etc

On a sunny day (Thu, 03 Apr 2014 08:20:56 -0700) it happened RobertMacy wrote in :

I have the datasheet, took 10 seconds to find it, dunno from where.

A really long phase sequence filter.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

would you send a copy [assuming pdf] to me?