Mixers for quadrature demodulation

[hykwei]

Hi guys

I am looking for a high speed mixer to perform the quadrature demodulation (for a 20 MHz signal), I have looked through the web but so far I can only find a AD831 chip that is stated to be suitable for quadrature demodulation in its datasheet.

https://www.analog.com/en/prod/0,2877,AD831,00.html

I just want to know whether there is any other mixer available to perform this kind of function so that I can compare their performance?

Thanks!

 

[FRIED]

NE/SA602
https://www.nxp.com/#/pip/cb=[type=product,path=50812/31466,final=SA602A]|pip=[pip=SA602A][0]

 

[RadioRon]

In theory, any good balanced mixer can do the job. The AD831 is good, but is not specially tailored to use as a quadrature demodulator, so it is no better than, say, a MiniCircuits SBL-1 or some other good (ie. balanced) active or passive mixer.

There are some critical problems in this type of demodulator that call for a better mixer. For example, one key problem is that if you build your own demodulator, it is likely that you will get LO leakage into your mixer RF inputs. When this happens, you get undesired DC output from your mixers and this DC gets in the way when you want to apply baseband gain to the output. Another typical difficulty is in getting an accurate 90 degree phase shift in the LO over a broad range of frequencies, and related to this is the problem of maintaining good amplitude and phase balance between the two channels. To solve these problems, the experts integrate both mixers along with a 90 degree phase shifter onto one IC. By doing this, they insure good matching of the mixers, so that amplitude balance of LO and RF to both mixers and IF output level matching from both mixers is good. They also integrate a clever LO circuit that includes a divide by 2 stage and a digital phase shifter. In this case, the LO injection to the IC is at twice the desired LO frequency. So, when your layout allows signal to leak from the 2xLO input pin to the RF input pin (which is hard to avoid), it has a much smaller affect, which effectively eliminates the DC problem. To show you an example of this kind of IC, take a look at the AD8348 (**broken link removed**.
The block diagram shows how there are two mixers in the one IC, and all inputs and outputs are buffered. Also, the LO circuit includes a divide by 2 and a good phase shifter. This IC, although it does not match your desired frequency of 20 MHz, is an example of what you should be looking for if you want the best performance.

Here is another example of this kind of architecture:
https://www.linear.com/pc/downloadDocument.do?navId=H0,C1,C1011,C1725,P1711,D3850

Having said all this, it is quite possible that your selected frequency is so low that the problems that I mention above are fairly minor. With good pcb layout, you might be able to use a discrete or less integrated design and make it work. I hope you know how to design a pcb for best balance and isolation at 20MHz.

I did a search at Digikey for Mixer IC and then realized that most mixer ICs don't appear to have IF outputs that have bandwidth down to DC, so they wouldn't work well as your demodulator. So, I stand corrected, that we have to be a bit more choosy than I first thought. Your AD831 does have this feature of IF output down to DC.

Others that might work include a range of mixers from Analog Devices including AD8343 and AD8342 although these might be harder to use due to the IF output circuit configuration. I still favor the SBL-1+ from Minicircuits.

 

[RadioRon]

NE/SA602
https://www.nxp.com/#/pip/cb=[type=product,path=50812/31466,final=SA602A]|pip=[pip=SA602A][0]

I don't think the SA602 is a good choice because it doesn't give you direct access to the LO input of the mixer. You would need this to allow injection of a 90 degree phase shifted external LO in one of the two mixers.

 

[hykwei]

Thank you guys for the helps!

Actually, at the beginning, there is a perfect chip from Analog device, AD8333:

**broken link removed**

I think that's the best chip so far that can fit into my application. however, the only problem is that the only package they produce is so tiny for me to solder it onto my PCB. (32 pins, 5mm x 5mm) And I don't know why they don't produce some other packages?

Therefore I started to look some other alternatives...at the moment, an AD607 chip seems to have a similar function, but its highest frequency input is only 12 MHz....

https://www.analog.com/en/prod/0,2877,AD607,00.html

Some times it is really irritating if your design is restricted by the package type.

 

[hykwei]

Originally Posted by RadioRon: I hope you know how to design a pcb for best balance and isolation at 20MHz.

Um...just in case If I make any mistake...can you please also point out some tips on makeing a balance and isolation PCB? Coz I have my PCB layout already and I have only 3 chips, don't want to mess the PCB up and waste any of them?

In my PCB, i just followed the schematic diagram that is shown in its datasheet, and I added some other capacitors at the power supply pin for bypassing...

 

[RadioRon]

Thank you guys for the helps!

Actually, at the beginning, there is a perfect chip from Analog device, AD8333:

**broken link removed**

I think that's the best chip so far that can fit into my application. however, the only problem is that the only package they produce is so tiny for me to solder it onto my PCB. (32 pins, 5mm x 5mm) And I don't know why they don't produce some other packages?

Therefore I started to look some other alternatives...at the moment, an AD607 chip seems to have a similar function, but its highest frequency input is only 12 MHz....

https://www.analog.com/en/prod/0,2877,AD607,00.html

Some times it is really irritating if your design is restricted by the package type.

That chip looks like a very good fit for what you want. I've soldered parts like this before and it wasn't as hard as it seems. Perhaps you should be more daring and give it a try. I would try soldering by first tinning the pads, then putting flux on them, and then finally soldering the IC down using a hot air pencil. I would also buy more than two of these ICs so that if you destroy one you can try again with another.

 

[RadioRon]

Um...just in case If I make any mistake...can you please also point out some tips on makeing a balance and isolation PCB? Coz I have my PCB layout already and I have only 3 chips, don't want to mess the PCB up and waste any of them?

In my PCB, i just followed the schematic diagram that is shown in its datasheet, and I added some other capacitors at the power supply pin for bypassing...

This topic is too large to cover here, but here are some general pointers:

- a full ground plane covering the back side of the board or on an internal layer is mandatory
- route RF and LO traces so that they remain as short as possible,
- route RF and LO traces so that they do not travel near each other side by side (except for balanced pairs),
- route RF and LO traces so that they cross each other as seldom as possible. If they must cross, let them do so at right angles.
- for isolation of two signals, you want distance between them, and you want ground traces or planes between them.
- for balance between any two circuit sections, you should keep the traces and pads the same length, the same shape, and the same impedance. (see note below)
- all circuit ground connections should be passed immediately to the ground plane through a via hole.
- be aware of impedance of RF and LO traces, keep them relatively thin and of consistent width.

In general, the best layout for this type of circuit is where you lay things out as you would draw the ideal block diagram. Each mixer layout should be the mirror image of the other. The in-phase baseband signal layout should be the mirror image of the quadrature baseband signal layout and the I and Q paths should be kept away from each other. Try to imagine the circuit layout as balanced (mirror imaged) around a center line that goes from rf input to baseband output.

If your LO is running at your RF input frequency (ie. if there is no divider on the LO), then keep it away from the RF signal path.

 

[hykwei]

Hi Guys...i have one more question: For some chips like AD8333 and AD8348, there is a 'divide by 2' stage inside the chip so it can automaticaly form a quadrature signal for me, but the quadrature signal that they are generating will have a certain amount of phase errors, for example, for AD8333, the phase error is typically 0.1degree and maximum and minumum error will be -2 degree ~ +2 degree. will that affect my I/ Q reading significantly? If the signal that I am trying to measure is a very tiny phase shift (around 2~3 degree?) so the I and Q changing will be very small!

or shall I implement a proper DDS (direct digital synthesis )chip to give myself a precise sine wave and cos wave? which one will giv me a more stable performance? i.e. the phase difference of my quadrature signal wil be set at 90 degree constantly over a period of time, and will not be drifting...?

Thank you!
hykwei

 

[RadioRon]

I would think that the phase error specified by the IC maker is a steady state or constant phase error (assuming only one carrier frequency). A steady state phase error would generate a small DC shift in the output signal. The signal that you want to demodulate is an AC phase shift, constantly changing, and so it seems to me that the DC error caused by the chip can be compensated for or calibrated out in the baseband gain stage following the mixer if your amplifier has enough dynamic range.

In fact, the larger concern might be that your local oscillator has phase noise exceeding the amount of your desired signal. This LO phase noise might be a much bigger problem to try and work around. My experience with DDS to date, which is very little, makes me wonder if the DDS can deliver better phase noise than an LC oscillator. At 40 MHz you should be able to build a very quiet LO using LC oscillator design. I do not think that the DDS is a better choice simply because you can adjust phase relations between the two outputs. However, this would be a handy way of correcting the mixer phase error.

 

[hykwei]

Sorry RadioRon....one more question!!

cos(wt) ----> DUT -----> Acos(wt + theta)

I am using the quadrature demodulation technique to measure the phase difference between the input and output of my DUT (device under test), so it would be nice if I use the same signal source for both my RF and LO port for some chip like AD8333...so i don't need to create a second signal source..

My question is: is there any chip that can perform the opposite function as "divide by two" circuit in AD8333? like a "zero delayed phase lock loop"? so that i just let my 20MHz signal go through the chip before it enter the 2xLO port of the AD8333....?


Acos(wt+theta) ------> AD8333(mixer) ---------> I and Q
........................................... ^
............................................||
cos(wt) -----> PLL? ---->cos(2*wt)

Will this cause some extra problem, for example, adding more phase shift? Because ideally, one of the quadrature reference signal that is generated by AD8333 (0 degree and 90 degree) need to have zero phase difference between the input signal (cos(wt)) of my DUT....

I mayb havn't explain the question properly...plz reply if you have any question for my description...

 

[mvs sarma]

Have you seen the I_Q Mixers employing 74HC/AC 74 as a divide by 4 with quadarature signal outputs for any clock signal input. So , if it is a fixed clock of say 20MHz you may supply clk*4 at input and you get 0 and 90 degree outputs.

One such application is as designed in DR2D SDR receiver with internal oscillator by YU1LM. ofcourse the clock is conditioned by use of few gates before it is fed to the divider.
PS:
After seeing the level of the design and the complexity of chip for specialized applictions, i feel Hykwei may prefer ICS501 as an osillator to generate 80MHz. this chip appears reasonable cheap and made by IDT

 

[RadioRon]

Sorry RadioRon....one more question!!

cos(wt) ----> DUT -----> Acos(wt + theta)

I am using the quadrature demodulation technique to measure the phase difference between the input and output of my DUT (device under test), so it would be nice if I use the same signal source for both my RF and LO port for some chip like AD8333...so i don't need to create a second signal source..

My question is: is there any chip that can perform the opposite function as "divide by two" circuit in AD8333? like a "zero delayed phase lock loop"? so that i just let my 20MHz signal go through the chip before it enter the 2xLO port of the AD8333....?


Acos(wt+theta) ------> AD8333(mixer) ---------> I and Q
........................................... ^
............................................||
cos(wt) -----> PLL? ---->cos(2*wt)

Will this cause some extra problem, for example, adding more phase shift? Because ideally, one of the quadrature reference signal that is generated by AD8333 (0 degree and 90 degree) need to have zero phase difference between the input signal (cos(wt)) of my DUT....

I mayb havn't explain the question properly...plz reply if you have any question for my description...

In theory a PLL should be able to double the frequency as you suggest, but there will be certain amount of phase error (steady state error) and phase noise (AC error) added to the output relative to the input. I suggest that a simple frequency multiplier (or "doubler") would be a better choice since it will still suffer some steady state phase error but will not contribute very much phase noise, and the circuit is simpler. The frequency multiplier can be made using a common emitter amplifier that is biased into class C operation. A sine wave input to the this amplifier is distorted at the collector so generating many harmonics. The collector circuit is tuned to the second harmonic of the input frequency so that voltage gain is created at Fo=2 x Fin. The output of such a circuit may be "dirty" with harmonics so an additional bandpass filter may be needed on the output to select the x2 signal and exclude others.

There are also frequency doublers that use logic ICs like this one:
**broken link removed**

These circuits will all introduce some phase delay and I don't know of a circuit that is without delay.

 

[hykwei]

Thanks Sarma, RadioRon!

sorry...1 mistake...the AD8333 chip employs a "divide by 4" technique for quadrature signal generation, not "divide by 2"

Does that mean that i just cascade two frequency doubler together to get me a 80 MHz signal? But i suppose that will give me more phase error right?

or...u think create a second signal source wil be an easier choice...

 

[mvs sarma]

Best appears to me , a ready crystal oscillator with differential outputs if possible directly at 80MHz.
youmay perhaps use26.6666MHz crystal with 74HCU04 or an oscillator and generate third harmonic of 80MHz

 

[RadioRon]

Thanks Sarma, RadioRon!

sorry...1 mistake...the AD8333 chip employs a "divide by 4" technique for quadrature signal generation, not "divide by 2"

Does that mean that i just cascade two frequency doubler together to get me a 80 MHz signal? But i suppose that will give me more phase error right?

or...u think create a second signal source wil be an easier choice...

Yes, in theory you could cascade two doublers to get 4xFo however it might be better to do it this way. Generate your 20 MHz signal by using an 80 MHz oscillator which is split into two paths. The first path feeds a divide-by-4 logic prescaler (which outputs 20MHz) and then your DUT. The second path feeds a delay circuit followed by the LO input of the AD8333.
With this arrangement, you can be sure that the two inputs to the quadrature demodulator are phase coherent, and this method would generate the least additional phase noise I think. The divide-by-4 logic will introduce some delay which will be a steady-state delay, and the purpose of the "delay" circuit in the second path is to simply match the delay suffered through the divider so that both paths have equal delay and the steady state phase error is minimized. I think you could create the delay circuit with a sequence of gate logic, where you simply choose the number of gates based on their propagation delay summing up to the propagation delay through the divider.

It will be challenging to find gates and a divider that work at 80 MHz, but I think it can be done. For example, 74HC logic should work, and certainly 74F will work. As an alternative, it is possible to simply leave out the delay circuit and suffer some DC bias on the output of the quadrature demodulator which can be calibrated out.

 

[mvs sarma]

Gates of 74AC86 can operate at above 80MHz

Also 74AC74 can work as divider at that frequency.

Hope the article published by YU1LM (HF SDR S/H sample and hold receiver DR2D) can be studied.