This was a most unusual early radio design

With IQ demod, near-zero IF is almost as good as zero. The low-pass before the A/Ds implements the passband. But you know that.

Since you have I and Q you try for zero (beat, but not zero BW)... then a modified-Costas loop nicely controls the PLL/VCO.

I'm not sure an SDR beats dual-conversion for image rejection.

I recently worked a system with 1st IF at 300MHz, 2nd IF baseband. ...Jim Thompson

Yes, because that avoids a notch in the audio crossing the zero point?

A big problem with most of these receivers is lack of good front-end filters - wide open in fact. All the out of band signals reduce your A/D ENOB and consequently SNR.

Dual conversion with a really high first IF works fine if the mixers are strong enough. A weak mixer or a low first IF will give you a forest of spurs.

Cheers

Phil Hobbs

There isn't really an image - there's some carrier bleed through and folding around DC if things aren't exactly right. Image rejection in that sense is as good as the phase and amplitude matching of the mixers and the VCO divider. If all that's done on a single chip like the AD9364 the numbers are really good. It's also possible to do some phase and amplitude tweaking in software if you're not in a rush.

AD834 and I got it backward... 300MHz input, 17MHz 1st IF, then baseband... I should look at project schematics before I leap ;-) ...Jim Thompson

Watchmakers did the same with "jewels".

Now the audiophools do the same with tubes. ;-)

Which always overshot the proper tuning, since it looked for the peak level before removing power from the drive motor. Then it would coast to a stop. The 'wonder' was why Delco ever built them.

Yup. And, it's not just a problem for SDR receivers. A lot of today's multi-band ham transceivers have "broad as a barn" front ends with little filtering. They tend to degrade very badly in the face of strong not-in-the-desired-band signals, suffering from severe intermod and image problems. Pager, public-safety, etc. signals will often "blow open" the squelch.

An older-generation single-band HT will often remain quiet (when there's no real signal) and sensitive (when there is), in "RF hostile" environments, when the newer generation of HTs (SDR and otherwise) are unusable.

Hmm, the younger generation won't remain quiet...where have I heard that before?

Cheers

Phil Hobbs

Here's some of the story - the origins of the Japanese semiconductor industry:

Yes. But with enough attention to IP3, you can stay linear over a very wide dynamic range.

The VK3YNG sniffer (for DF in the 2m band) uses a now unobtanium dual-gate GaAsFET in the front end that allowed a 70dB front-to-back ratio at high linearity - so his sniffer was the first we had that was mostly immune to the adjacent pager band - and it handles almost 140dB dynamic range in the first RF stage (under a microvolt to several volts!)

That's why I'm interested in Phil's PHEMT/SiGe cascode - it potentially offers incredible IP3 figures at VHF because of the huge Early voltage of the SiGe - perhaps replacing the unobtanium GaAsFET?

(Hat tip to Phil to alerting me to the existence of such devices!)

The problem with high first IF is that you need an even higher LO with hard phase noise requirements.

With an IF of 300 MHz and intended reception range 0.1-200 MHz, the LO must tune 300.1 - 500 MHz. A lower frequency oscillator would typically have a better spectral purity than a 300+ MHz oscillator.

Old style equipment was amateur-band-only and even a HT would contain a filter only as wide as the amateur band. More expensive equipment even had narrower filters that were tuned in gang with the receiver frequency.

Today's equipment has "general coverage" of a very large range around the amateur band (e.g. a 2m HT receiver receives 108-174 or even wider, a 70cm HT does 400-520 or more), and there is no place for input filters anymore.

This is normally not true. The low-pass before the A/D sets an initial bandwidth of the baseband IF, but the actual receiver bandwidth is reduced from that using a software-defined filter. That may not even be centered around zero, to avoid the problems around zero Hz.

I've often derived the LO by mixing a lower frequency VFO or synthesizer wi th the (fixed) IF_1 - IF_2 oscillator. (I'm usually doing phase-coherent me asurement gizmos in the lab rather than communications receivers.)

That way the low frequency phase noise cancels out. It's much easier to get rid of the spurs on the resulting LO signal than in the IF. The LO is big, it's narrow, and you know exactly where it is. Offset PLLs or a bunch of M ini Circuits modules and cables with a 1:1 PLL at the end will clean it all up pretty well.

I have a more complicated version of this spread out all over my optical ta ble right now. It takes two 40-90 MHz VCOs (X and Y scan) and generates a L O at 2f_x+2f_y+f+f_IF, all phase coherent so I can measure the optical phas e on the sample. Following an octave tunable VCO with an offset loop, frequ ency doubling, and then mixing with another doubled VCO produces the desire d output, all right--along with a very impressive collection of spurs that all cross the desired LO. Fortunately, they're narrow enough that a 1:1 PLL cleaned it all up amazingly. (Tuning stepwise makes them all miss the LO.)

Yes, the real instrument will use DDSes.

Cheers

Phil Hobbs

One reason for using a wider than necessary pre-ADC filter is to let some wideband band/thermal noise into the ADC to function as dithering. In this way, the ADC doesn't need to have a huge number of bits. The wanted signal+dither noise is then low/bandpass filtered in software to the wanted final bandwidth in which each sample has more bits than the physical ADC.

The cleanup offset with a low phase noise VCO (locked to a reference comb) with a narrow band filter at the offset frequency is a nice trick. I think George Lohrer (PTS) was the first to introduce the idea in commercial synthesizers (at General Radio?). I know he licensed his implementation to several other companies (eg Rockland).

A wider filter also keeps the interesting signals away from the group delay hump near cutoff.