TheRadioBoard

I'm trying to design a phasing product detector at IF to cancel one sideband. Currently the design is on the drawing board only with no prototyping done yet.

The design uses a 90-degree phase-shifting transformer at the fixed-frequency LO, feeding the upper gates of two cascode mixers. The lower gates of both cascode mixers are tied together and fed from an LC tank tuned to the IF. Rationale for using cascode mixers is their low LO drive, fair dynamic range, and insensitivity to termination impedance.



My questions:

1) What would be reasonable values for C40 and R30? I imagine too-low values would load down the LO and reduce available LO voltage to the mixer. Could I aim for 100k capacitive reactance at the IF, to match the 100k impedance at the LO port? Or should I go even higher?

2) Are there any problems tying the two RF ports together and feeding them from the same RF source (tuned secondary of IF transformer)?

Thanks for any advice.

As as first approach, I'd pick values so that the reactance at your operating frequency are about 1/10 the values of the bias resistors (which are 100k). Hence, you'd be looking at a reactance of about 10k. And so you could make R30=20k so you have some leeway. As for C40, it will depend on the frequency you're running at.

qrp-gaijin,

Designing a single unit Quadrature mixer is a good idea. It's something I've often thought about.

You mention using 90-degree phase-shifting transformer.
By this do you mean a Quadrature Hybrid Coupler?
Or are you hoping to get the 90 deg phase shift with T3?

Some references on Hybrid Couplers:

"Twisted-Wire Quadrature Hybrid Directional Couplers" by Reed Fisher (W2CQH) QST, January 1978

http://www.seboldt.net/k0jd/phase_notes.html

http://www.kitsandparts.com/fluxbb/viewtopic.php?id=130

P.S. After some thought I see you have a Xtal Oscillator with a Phase Shift at a fixed frequency.
That should work OK, although you might need an Amplitude adjustment as well as a Phase trim.


later ........ Zim

I don't know about Quadrature Hybrid Couplers, but you've given me more food for thought.

As you said, I'm assuming (hoping? ) Tr3 will provide the necessary phase shift, based on this previous discussion:
http://theradioboard.com/rb/viewtopic.php?p=30486#30486

Since I'm planning on doing the phase shifting at one frequency only (the IF in a superhet design), my RF phase shift scheme doesn't need to work over a wide range of frequencies. The phase-shifting transformer idea seems nice because it requires only one adjustment (R30) to get the proper amplitude balance between the I/Q signals.

I think R30 will provide amplitude adjustment. How/where would phase trim be adjusted?

I was hoping to get away with just one non-reactive adjustment (R30) for the phase shifting, but I suppose imbalances in the two mixers may require the phase trim you mention.

qrp-gaijin,

Thinking about your circuit, I can see a simpler approach...

Consider firstly the classic divide-by-four quadrature mixer: You have an Osc followed by two flip-flops developing Quadrature square-waves, which in turn drive a pair of switch-mode mixers...

The system works well and has the advantage of no adjustments. Its disadvantage is that it responds strongly to harmonics, but that is no problem in an product detector, as the IF stage should filter out the harmonics.

Now consider an even simpler system: Do all of the above in software!

I assume you are doing the audio processing in a DSP. Assume a sample rate of 50 KHz. You have a analog mixer which converts the IF down to 12.5KHz and a software divide by 4 routine which develops two 25KHz clocks in quadrature. Two software mixers now mix the incoming 50KHz data stream down to two Quadrature streams at 12.5KHz (the mixers need only toggle the sign bit). The above takes only a few lines of code.

An even simpler version of the same thing merely sends each alternate sample to either the I or Q data stream.

I did the above in an earlier version of my present RX. It worked very well and had the advantage that (as the maths is perfect) the side-band suppression is virtually infinite.
It has the disadvantage of an image at twice the IF frequency (eg 25KHz), but that is easily filtered out by the IF transformers.

I think this scheme is used in many of the commercial DSP rigs. e.g. they don't bother with a quadrature mixer before the DSP...

later ...... Zim

> I think R30 will provide amplitude adjustment. How/where would phase trim be adjusted?

To be honest, I'm a little uncertain about this. Most designs use a separate Phase and Amplitude adjustment. Your circuit certainly adjusts both, but whether they both occur at the same point I'm not sure. Perhaps they do.

The important thing is that you are able produce a perfect circle on an X/Y CRO display. I'm not sure if varying the Amplitude can compensate for a Phase error (and vice-versa)...

P.S. After some reading I see that you have already touched on this in another thread.....

......... Zim

Very interesting idea to use a near-zero IF and do quadrature generation and mixing in software with DSP. I'm not sure I completely follow the exact details of the code that would need to be written, but it sounds reasonably simple.

On a similar note, I've successfully used a 10kHz IF combined with a DSP ten-pole brick-wall high-pass filter to completely attenuate everything below 10 kHz, then use software mixing to shift everything down in frequency by 10 kHz to yield baseband audio with the opposite sideband (below 10 kHz) eliminated. The fun part was that to generate the 10 kHz IF, and to eliminate the image 20 kHz away, I used nothing more than an unmodified, below-threshold regenerative receiver and a nearby stray-coupled BFO tuned 10 kHz away from the regen's frequency. (I called this scheme the "audiogenerodyne", and searching for that term on the web will bring up my initial descriptions of this experiment.) This way, the regen becomes a highly-selective RF mixer that can effectively attenuate (by around 30 dB operating at 7 MHz) the 20-kHz-distant RF image and mix the RF down to the 10 kHz IF where the DSP can do its brick-wall filtering and mixing down to baseband. This approach of using a DSP brick-wall filter is probably more computationally intensive than your suggestion of software quadrature generation, but the end result is the same: beautiful, digitally perfect silence as a CW signal is tuned into the suppressed sideband. I had plans to build a more permanent version of this receiver, but they're suspended for now as the raw performance of this scheme is limited by the regenerative front end.

Regarding your comment:

Is it really that easy to build an IF transformer (even at, say, a relatively low 2 MHz IF) that will significantly suppress a 25-kHz-distant image?

My biases are showing, but one reason I am considering a phasing product detector instead of a crystal filter ahead of the product detector is to reduce the need for extreme shielding to prevent signal leakage around the crystal filter; with a phasing approach, there is no ultimate-selectivity-determining IF filter around which signals can leak. On the other hand, if you use another mixer to convert the 2 MHz IF down to a second IF of 12.5 kHz, then that again requires, it seems, sharp filtering at 2 MHz and the attendant need for shielding to prevent signal leakage (i.e. image signal leakage) around the filter.

DSP technology is obviously to the point where it can outperform a discrete-component filter. But crystal filters don't require "extreme" shielding. Putting the IF stage in a shielded box is a good idea, but not entirely necessary. If you lay out the crystals on the bare board, isolate IF input and output, and bypass power-supply leads, you shouldn't see filter blow-by.

73,

_________________

http://kr1s.kearman.com/
http://qrp.kearman.com/

qrp-gaijin,

I found my notes on that "non-quadrature" scheme I was talking about.

Firstly here is my little DSP Receiver I mentioned which used the "div by four" quadrature generator and simple switch-mode mixer.



Here's the block diagram



This is how the software usually works



Next is the "non quadrature" mixing scheme (in software) I tried using the above rig as a test bed (bypassing the Quadrature H/W)...



I used 400 KHz for the "BFO".



Here is how the various images work out. I used a re-tuned 455KHz IF transformer to filter out the unwanted image.



In the end, I abandoned the idea. But it does work, it does avoid the necessity for a Quadrature BFO, and it does give perfect sideband suppression.


later ........ Zim .......... VK3GJZ

Thanks for sharing - very interesting to see. Do you have any idea how much image suppression (of 387.5 kHz) you could achieve with your IF transformer? Also, with a 412.5 kHz first IF, I imagine you needed a pretty tight front-end RF filter to eliminate the RF image 825 kHz away. What did you do for front-end filtering?

The annoying thing about adding conversions in front of a DC receiver is having to worry about these images and spurs.

qrp-gaijin,

Totally agree with your comments.

The "No Quad" software was just a proof of concept experiment, not a practical design.

The image problem was no different to that in a conventional Superhet, with the same constants and solutions. Present day Superhet DSP radios fix the image problem with a roofing filter or by use of a higher frequency A/D converter.

As you say, the Direct Conversion receiver with an audio DSP is a simple and elegant solution.

........ Zim

I started prototyping the crystal-controlled quadrature BFO and have some questions about the appropriate value of the phase-shifting transformer. Some quick calculations revealed that making the 1mH transformer (Tr3 in my original schematic) would require 40-some bifilar-wound turns on an FT50-43 toroid, which would be a pretty tight fit for the small toroid. And at 2MHz, 1mH in the drain might not even be enough to ensure oscillation, requiring more inductance and more turns.

Instead, I reworked the circuit as below, with a 4mH choke in the drain and a separate transformer.



I used a T50-2 toroid in the junk box with about 10 turns primary and 20 turns secondary. C40 was 20pF and R30 was 10k. These were just ballpark figures to see if the circuit would continue to oscillate with the transformer present. It did. (At some point I will also add capacitors between each side of the crystal and ground as required in a proper Pierce oscillator.)

So, now that I've confirmed the circuit will oscillate in this configuration, I need to figure out exactly what the primary/secondary values should be for the phase shift transformer Tr1 (in the above, new schematic diagram). Should it be step-up or step-down, assuming it will be connected to a pair of cascode mixers as shown in my first post at the top of this thread? What would be reasonable primary/secondary inductance values operating at 2MHz? Is the value of C43, the coupling cap between drain and transformer, critical?

Also, what is the name of this transformer configuration? I've tried numerous web searches but I can't seem to find more information about it.

Looking forward to getting my superhet with phasing product detector working. I've downloaded some I/Q-capable SDR software for my PC (quisk) and plan to start by using the PC to do the AF phase shift. Once that's working, I'll investigate building an analog AF phase shift network with op-amps.

The circuit where I saw this idea is this one:

http://www.qrz.lt/ly1gp/SDR/

I initially assumed the transformer RC network was doing all the necessary phase shifting, but instead it seems perhaps C3 and R1 provide RF phase shift while C6 and R2 provide LO amplitude adjustment.

On the other hand I read that quadrature hybrid couplers work over a wider frequency range (up to 1 MHz operating at 28 MHz!) so that might be a better option.