Double Balanced Mixer - Very Linear ?

The ham radio technology developers have been exploring very linear and wide dynamic range mixers for years, you should start your search there.

Google for 'Tayloe mixer'.

It seems that you need linearity only for the signal path, and not for the carrier path, so that the mixer can be a switching type run at the carrier rate.

Your required baseband width is a serious challenge, if you want SSB. Please remember that a balanced mixer generates double-sideband suppressed-carrier AM, not SSB.

--

Tauno Voipio

This is the KEY, of course. A mixer HAS to be non-linear, at least in one way, to function at all. It can be tricky to achieve that required nonlinearity without introducing other nonlinearities that are non desired.

Jon

If you're willing to bend on your drive and modulation levels (I assume that by 'drive' you mean 'signal' and by 'modulation' you mean local oscillator), then MiniCircuits has some diode ring mixers that might make you very happy.

They're passive devices, so their "operating power" comes from the local oscillator drive.

Figure on over 10dBm of local oscillator drive, and (probably) significantly less than 0dBm of signal to get linearity that this ex-Ham operator/receiver builder considers "very linear".

--
Tim Wescott 
Control system and signal processing consulting 
www.wescottdesign.com

No, it doesn't. An ideal mixer is bilinear, which means that it satisfies linear superposition at both inputs.

V_IF = k V_RF V_LO

If you double either one, the output doubles, and is otherwise unchanged. The amplitudes of mixing products of order +-N go as the Nth power of the input amplitudes. (*)

It can be tricky

That's true, of course, but good switching mixers with square-wave LO drive can do a very good job.

One of these times I'll build a bridge made out of E-pHEMTs such as the ATF58143 and drive it with PECL. Should be pretty good. (Joerg or JL may already have done this.)

I've been using quite a few BFP640FESD SiGe bipolars lately. They're a lot easier to use than you'd think, considering that their fT is like 46 GHz. Their Early voltages are effectively infinite, which is astonishing for such a fast transistor.

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

I used to use the LM1496 chip! :)

Jamie

Any way, getting back on what I said before, a gilbert cell would work nicely.. heres a video..

Jamie

thanks for the suggestion. If someone has a URL, or part number would help. As I said, searches and I don't get along well, got something like

5,000+ useless hits form google and after sitting over 20 minutes at Linear's website gave up.

Yes, I understand communication basics. Besides NEEDING that suppressed carrier aspect, I plan on using two mixers and judiciously combining one mixer's output which comes from cos fc and cos fm [add or subtract depending on whether you want the upper, or lower sideband] with the output from another mixer which comes from sin fc and sin fm voila! SSB, a SINGLE tone, except for the ODD related upper harmonic tones that seem to get in there! I need suppressed carrier because I need those upper harmonics way down. a 5 pole Tchebyshev barely makes it to kill that third harmonic related tone.

Is the bandwidth too much for the baseband, or for the carrier?

Not sure that mixer is specced much above 1MHz, probably will funciton out to 10MHz, hard to believe very good at 100MHz.

Thanks to Jim Thompson who kindly sent me NPN CA3046 'chip sized' models, I've characterized the simplified MC1496 [as it appears in the data sheet/app note]. I now have a complete LTspice model now.

However, the model predicts incredible carrier suppression [well, duh!] but it also probably more accurately predicts the third related harmonic to rear its ugly head at a pretty energetic level. It will be difficult to filter that tone and get it below levels that interfere.

By the way, I've developed what I call .tranoise for use in LTspice [and PSpice] simulations. It enables .tran analysis AND .noise analysis during the same run. Which means you get 'realistic' noise performance through a non-linear circuit, like a mixer. The resulting waveforms are like watching a real scope trace with fuzz AND the spectrum has the realistic noise floor one would expect, inlucding [if you used an OpAmp with pesky

1/f noise] you get appropriate 'bumps' in your resulting spectrum, too. It's like a 'closer to reality' simulation. The zipped model for the MC1496 is over 35MB.

The reason I posted is that the spec for carrier suppression is like 60 dB, not the unrealistic -120dB I now get in simulation, so thought I should look for a better mixer...if there is one.

what's a gilbert cell, anything like the MC1496?

can't view videos at youtube, do you have a part number URL?

Forgot all about minicircuits! they make great stuff. I'll take a look, thanks.

Not sure I followed your definitions regarding signal etc, but I understand the principles behind using diode H bridges as a mixer. Actually designed one from discrete components that went into autopilot electronics modules years back.

The diode mixer might have better specs than the MC1496 topology, I'm not skilled enough in chip design to be able to answer that one. My gut feeling is that it will be hard to beat a well laid out MC1496 structure.

*IF* someone makes one.

Dunno what kind of mixers you have in mind, but square wave switching of diode mixers has always been a big no-no, raises the noise floor by 6dB or something like that.

Square wave switching is usually a big win, actually, because it greatly reduces the intermodulation distortion that occurs when the switches are partially on. This is especially true with diode mixers, but also applies to transistor mixers.

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

Search on Barry Gilbert, who now works for Analog Devices, and has exploited the idea in a number of devices, which you can buy from Analog Devices.

Here's one of them

there are lots of others, mostly more specialised in the direction the OP wants

--
Bill Sloman, Sydney

Really? Then you know something the career mixer designers at the big microwave houses don't.

BTW, IMD results from component mismatch and not "partially on" operation. The double balanced has theoretical infinite suppression of the biggest 3rdIM product.

Nope. If you have a bridge made of perfectly matched nonlinear devices, the even-order products will cancel, but all of the odd-order products will make it to the output.

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

Interesting work. If memory serves, there were various versions of the

1496 that could perform well but the highest I've ever seen it used was up in around 30Mhz where a complete HF QRP transceiver was constructed using that as a mixer.

Anyway, I was at a meeting tonight and one of the guys brought up a subject matter of (Lt)spice having a seminar close to my area in JUNE

13, I may attend that. Should be interesting meet Mike if I can work that in my schedule.

Jamie

yes, that's basically it, but you can construct one using discretes to your requirements.

Also, a ring diode mixer comes to mind.. :)

Jamie

Yep.

Whenever I have a problem like that, I like to look at existing equipment to see what the industry leaders are doing. I suggest you take a close look at the Elecraft K3 radio. Schematic at: (7.6MB) See Page 8 for the mixer. Among the available ham radios, it has one of the better IP3 (3rd order intermod crossover point) of 29dBm @

20Khz and 28dBm @ 2KHz[1], partly because if this mixer. I'll make a guess the what you really want is not linearity, but rather dynamic range, which is the difference between the noise floor and where the gain compresses by 1dB (and goes non-linear). Google for "high performance receiver" for more articles.

You ability to match and balance the FET switches determines your carrier leakage. I suggest you use an integrated package, such as the CBT33257AD You can assume that the devices on a single chip are fairly well matched. If you see any carrier leakage, your problem is the PCB layout.

There are various ways of producing SSB out of DSB. It's usually done with a crystal filter at a specific IF frequency. Trying to do it broadband or at baseband with phase shifters is not going to yield good results. If you must cover a range of frequencies for both the input and the LO, then you'll probably find yourself using two mixers and two LO's, just so that you can use a single frequency crystal filter. You ability to suppress the unwanted carrier will initially be a function of the crystal filter shape factor, but eventually become a PCB layout and shielding problem.

[1]

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
Santa Cruz CA 95060 http://802.11junk.com 
Skype: JeffLiebermann     AE6KS    831-336-2558