EMRFD Message Archive 13347

Message Date From Subject
13347 2016-12-07 23:27:40 itsbeen29years LO signal quality for Active JFET Mixers

I have been experimenting with the S7C Simple Superhet from EMRFD, chapter 12.  I have tried using different VXO designs to try to broaden the tuning of the receiver, but keep use of a VXO design ( option 1;  George Hobbs, Universal VXO using a BJT with an output buffer; Option 2:  or a basic JFET VXO, for example KD1JV "72 Part 30 meter Receiver"..  For the BJT VXO, I have used a tuned output buffer, and  can get a reasonably stable LO signal, that is almost sinusoidal.   When I use a JFET VXO design, I get a nasty looking periodic signal.   My question:  does the LO signal have to be very clean and sinusoidal, feeding an active JFET mixer?   I haven"t been able to detect much audible difference between the two FETs in a simple listening test.    Its been an interesting voyage, experimenting with different VXO designs, I can get about a 20 kHz tuning range, using 3 parallel 11.059 MHz crystals, and 2 series inductors in a Super VXO configuration.

Rgds,  Nigel


13348 2016-12-09 08:06:39 winston376 Re: LO signal quality for Active JFET Mixers
Nigel.   Depends on what your performance spec is.   If you're looking for the best possible MDS of a receiver, balanced with a high IP3  etc.  I would be using a DBM and an  Si570 based VXO......Done that already and it works very well.   If you're only experimenting, then that's fine but a fet mixer is going to go into saturation very quickly and you will be producing alot of IMD and spurs that you will need to contend with irregardless of whether you have a sinusoidal or square wave VXO fed into a fet mixer . Done this too: multiple xtal (sequentially, as many as 10 (ten)  at 24 MHz ) in a VXO config---results were a substantial stability degeneration beyond a relatively limited frequency range. There is no free lunch.  Engineering trade-offs are what you will need to balance. What are you trying to accomplish??
13349 2016-12-09 09:56:26 Nigel Maund Re: LO signal quality for Active JFET Mixers
Thanks so much Winston,

Well, in answer to your question, I am really experimenting with the basic superhet and exploring how well I can get it receiving so I can compare its performance and see how it works.

I was intrigued by the JFet active mixer.  I see that this little circuit has formed the basis of a few journeys (Ashar Farhan building a bidirectional SSB transceiver with a JFet mixer; the "Clackamas" Transceiver, and further evolving to the "CRX1", etc.

I'm going to clean up the vxo signal with a filter and see if I hear the difference.  Then it may be interesting to try a double balance diode mixer.

I have an old DDS which I may also test ...

Eventually I'd like to test the noise figure and MDS of the system.


Sent from my iPhone

13350 2016-12-09 14:20:00 kb1gmx Re: LO signal quality for Active JFET Mixers
The S7C has a caveat.... you must supply enough LO drive or the fet conversion gain will be very poor.
How much is enough more than 4-5Vpp seems to work for me.

Waveform quality is really asking how much noise, jitter and harmonic content is important?  The 
answer is its all important if you going to get to the best possible MDS.

FYI the MDS is basically the point where NOISE limits what you can hear.  A RX without RFamp
MUST have a low noise mixer and LO to get to the best numbers.  The S7C is a interesting radio 
but like many minimum art(fancy way of saying simple) radios its not likely to beat or match an 
effort that though enough parts and circuit design to get the ultimate result (or best compromise).
As design and presented its a pretty good radio using easy parts and simple circuits and uses 
low operating power.


13351 2016-12-09 17:30:10 Bill Carver Re: LO signal quality for Active JFET Mixers
Nigel, an analog multiplier is the "perfect" mixer. Circuits using mixers like the MC1496, if not overdriven by too big an LO, will have very clean outputs. They might work better in situations where the input is unfiltered. However, in an RX where we're only "interested" in one of the products, and the antenna signal is heavily filtered, that wideband cleanliness may not make much difference....if the LO is low noise. Because in a "perfect" mixer (multiplier) amplitude noise and phase noise on the LO will be transferred onto the IF output signal.

Most mixers are driven with more LO. First because it reduces the conversion loss, and it may also reduce the intermodulation distortion, which we can hear on the one product we're interested in. When overdriven the mixer isn't as sensitive to amplitude noise on the LO: it's "scrubbed off" by the saturated mixer. This is carried to an extreme in switching mixers like the FSA3157 that PA3AKE used: the LO is a square wave: it is totally insensitive to amplitude noise on the square wave.

Diode balanced mixers, when driven at their data sheet LO level (+7, +10, +13 dBm, etc) have the diodes driven into saturation. They are somewhere between a perfect multiplier like the MC1496, and the FSA3157 switching mixer. They are less sensitive to amplitude noise on the LO.

But the is a big difference with diode mixers: the LO should not have any even harmonic content. The second harmonic is potentially the largest, and the most difficult to filter out. But if you don't the operating point of the diodes is shifted....because even harmonics shift the average of the LO away from the zero-crossing (zero voltage) points at the fundamental frequency. IE, it's a mathematical reality.  Some "hardcore" designers, like Ulrich Rohde, back in the 1970s, always put a lowpass filter on the LO to his diode mixers, and they were "elliptic" filters which had an attenuation notch that he placed at the second harmonic.

And if you say "OK, I'll square up the LO signal", you must filter out all even harmonics BEFORE you run it through a squaring circuit. If you do not, you will not get a perfect square wave, 50%/50% duty cycle. That square wave will have even harmonic content, undoing most of the benefit of the squaring process. And, if the LO has amplitude noise on it, a squaring process will turn amplitude noise into phase noise. And even a saturated mixer operated with that square wave will transfer that phase noise to the i.f. signal.

There is no free lunch. There are tradeoffs in conversion gain, third-order intercept, etc. Different applications might work best with an MC1496, receivers needing high intercept probably do better with a switching mixer. Like juggling balls, as you work toward "perfection" in the mixer, you also have to make the LO "perfect".

Bill W7AAZ

13352 2016-12-10 00:27:02 Ashhar Farhan Re: LO signal quality for Active JFET Mixers
The dual jfet mixer is a pretty good performer. It has +5dbm IIP3, and a bit of gain too. It has no need for broad band termination unlike a diode mixer. The only trouble is that the output is not as broadband as a diode mixer though you could make it so. for even more interesting jfet front-ends, head over the KK7B's page on his university and look at the single jfet switching mixer front-end, now that is something...

- f