Double Balanced Mixer - Very Linear ?

good

operate

unwanted

I would suggest an ADL5391 of an AD835.

?-)

The effect of carrier leakage isunknown at this time. I will simulate shortly, may have much less effect than I thought. But it seemed that *if* the LO leaks, it will be difficult to configure a couple as a fairly good LSB or USB mixer too.

My experience with minicircuits has been good. I keep forgetting about them. Will check.

not in this system

will get the data sheet, some how. just went to intersil.com and sat there with some type of vague background picture for 20 minutes, had to bail out. maybe one of the sources will supply a data sheet more quickly.

Thanks for those. And, the data sheets appeared fast to get!

Trying to get the spice models, well that never happened, finally gave up.

I didn't have any trouble getting the AD Spice model for the (much slower) AD734.

Of course, when I got it, it didn't work very well and my expert friend in London ended up editing out a questionable line, after which the model worked fine.

Sounds typical of Analog when it comes to their models. Years ago almost ALL of their complex OpAmp models [looked like brute force representations] were too large to run on 'student PSpice'.

Assuming the questionable line is left intact and commented out, would you share those results?

No. My expert friend knew exactly what I was doing, and his editorial work presumably reflected that knowledge. What worked for me may well not work for other people.

We thihnk it's the same die, just bonded out differently.

Shouldn't be hard to use in linear mode, though perhaps not with Phil's linearising trick.

Clifford Heath.

'Tweren't mine. I just read about it in the LM13600 datasheet long ago, and liked it.

You could make a real diode-linearized Gilbert cell using two HFA3046es--one for the lower diff pair and diodes, and the other for the cross coupled upper stage. You don't care if the upper and lower stages track each other, just themselves.

If you want, you can use the spare transistor on each die as a temperature sensor/heater, to keep the gain constant.

Cheers

Phil Hobbs

Oh, and be sure to use the tiny package--the SOP ones oscillate when you use them as diodes at any significant fraction of ICmax.

Cheers

Phil Hobbs

...snip...

Clifford Heath.

THANKS! You just provided for me to get that data sheet in less than 30 seconds, what would have taken over 45 minutes, plus tons of unwanted cookies and images stored in memory ALL which MUST be purged.

It was fortuitous that Aioe locked me out, gave me time to drop the spice model included in that spec into a gilbert cell modulator. Wow! gangbusterrs, Clifford! Barely rolls down at 100MHz, and don't understand this, but had more gain in the same implementation as the 2N2369's

The pinout on the gilbert cell is incredibly perfect for the application too. If this thing does work up to 2.5GHz at least, have an application up there to try, too.

Looks like you found the winner! ...now, uhm, cost, availability? ;)

Thanks! I usually just pull the contacts out for potential future use, never considered using one as an 'oven controller'! genius! Just have to be careful about gradients acoss the chip, but still, extreme temp range operation is sometimes useful.

Thanks for the 'heads up' on the oscillation. I would have gone with larger package just because my fingers are larger than the point of a safety pin. Also, probably would have found more of the SOP packages available.

I always use

Even the layout on the silicon was designed as a Gilbert cell, so it's not surprising they paid attention to the pinout too.

Thanks for those, too. I'll bookmark for future.

I now have an LTspice model for the gilbert cell, HFA3101, an 8 pin SOIC package, where the 8 NPN transistors are used [according to a note in the spec] with two each in parallel at the bottom legs. Next, I'll make an LTspice model for the 'packaged' version per Harris/Intersil's AppNote describing all their package parasitics. [I hope whoever posted that question about intra-pin capacitance can get a copy. Very good ] Confirmed Jim Thompson's answers to their posting and gave details to help a person expand on their models a bit.

I don't understand this. I can only see HFA3101 in SOIC-8, and HFA3046 in SOIC-14. Only the HFA3127/3128 are in QFN. What chip and package were you referring to?

Right, it was the 3127 I was thinking of. You can put a 10-ohm bead in series with the bases, but it's a sad waste of bandwidth.

Cheers

Phil Hobbs

I googled "harris HFA3101 gilbert cell" and it is the first thing...

thanks for this one, too.

I got it earlier, along with the chip transistor models and the package parasitics. So now I built an LTspice model for the HFA3101 SOIC packaged gilbert cell.

Do you know if the 'nuhfarry' NPN models in the data sheet accurately predict noise? Are transistor models inherently accurate as noise models? Comparing noise from the HFA3101 to the MC1496 built with CA3046 chip transistors; the HFA3101 has around 50nV/rtHz for the same circuit as the modulator application in the MC1496 data sheet, which had around

Sorry, I don't know. Like I said, I was just passing it on. I originally looked at this Harris IC back in Metricom's end days, as Agilent/HP had eol 'd the IQ modulator we were using. But a total train wreck of a chap7 put an end to my investigations. There weren't many good choices back then.

The HP IQ mod is almost in the center of the board, to the lower left of th e 12 MHz TCXO:

Metricom, Los Gatos?! Did you know Bob Friday? I worked on shielding the Ricochet modem over the 'forbidden bands' AND did a survey of the whole Bay Area for ambient energies in the 2.5GHz band. THAT was very educational. Have spectrum analyzer copies somewhere.