Hi group,
An associate is proposing a mixer design that I'm having a hard time
understanding. I did some Google searching and haven't found an answer, so
perhaps some kind sould can help.

The basic porposal is to generate a clock at twice the LO frequency, divide
it by two and use some sort of differential mixer circuit. This is for
wideband (broadcast) FM reception. He says that the key is the use of
differential circuits within the mixer. Of course, this is intended to be
used with quadrature demodulation.

Sitting down and thinking about it, I could see how it would work if one
were to take the rising edges of the 2X LO clock for one phase and the
falling edges for the 90d phase, but it seems like doing it that would way
would likely result in the I+Q not being perfectly orthagonal due to things
like different delays in rising vs falling edge logic, imperfect duty
cycle, etc. And even then it wouldn't need to be differential.

Well, if anyone has a suggestion for a mixer architecture that does what he
says (ie uses 2x LO clock and requries differential circuitry) then I'd
appreciate the help.
Regards,
Matt

I seem to recall a design using EXOR gates to divide the signal by two -
I cant remember the details. I recall however that the duty cycle of the
input clock was critical I have successfully used a Johnson Ring divide
by 4 Counter a few times to generate quadrature outputs from a 4X clock
signal. The duty cycle is not critical in this case as the clock counts
on a rising (or falling) transition of the input.
I am not sure what you mean by "differential mixer" Does this mean two
mixers with the inputs fed 180 deg out of phase? Of course at 100 MHz
you would have to use ECL logic.

Richard


MattH wrote:
Hi group,
An associate is proposing a mixer design that I'm having a hard time
understanding. I did some Google searching and haven't found an answer, so
perhaps some kind sould can help.

The basic porposal is to generate a clock at twice the LO frequency, divide
it by two and use some sort of differential mixer circuit. This is for
wideband (broadcast) FM reception. He says that the key is the use of
differential circuits within the mixer. Of course, this is intended to be
used with quadrature demodulation.

Sitting down and thinking about it, I could see how it would work if one
were to take the rising edges of the 2X LO clock for one phase and the
falling edges for the 90d phase, but it seems like doing it that would way
would likely result in the I+Q not being perfectly orthagonal due to things
like different delays in rising vs falling edge logic, imperfect duty
cycle, etc. And even then it wouldn't need to be differential.

Well, if anyone has a suggestion for a mixer architecture that does what he
says (ie uses 2x LO clock and requries differential circuitry) then I'd
appreciate the help.
Regards,
Matt

I seem to recall a design using EXOR gates to divide the signal by two -
I cant remember the details. I recall however that the duty cycle of the
input clock was critical I have successfully used a Johnson Ring divide
by 4 Counter a few times to generate quadrature outputs from a 4X clock
signal. The duty cycle is not critical in this case as the clock counts
on a rising (or falling) transition of the input.
I am not sure what you mean by "differential mixer" Does this mean two
mixers with the inputs fed 180 deg out of phase? Of course at 100 MHz
you would have to use ECL logic.

Richard


MattH wrote:
Hi group,
An associate is proposing a mixer design that I'm having a hard time
understanding. I did some Google searching and haven't found an answer, so
perhaps some kind sould can help.

The basic porposal is to generate a clock at twice the LO frequency, divide
it by two and use some sort of differential mixer circuit. This is for
wideband (broadcast) FM reception. He says that the key is the use of
differential circuits within the mixer. Of course, this is intended to be
used with quadrature demodulation.

Sitting down and thinking about it, I could see how it would work if one
were to take the rising edges of the 2X LO clock for one phase and the
falling edges for the 90d phase, but it seems like doing it that would way
would likely result in the I+Q not being perfectly orthagonal due to things
like different delays in rising vs falling edge logic, imperfect duty
cycle, etc. And even then it wouldn't need to be differential.

Well, if anyone has a suggestion for a mixer architecture that does what he
says (ie uses 2x LO clock and requries differential circuitry) then I'd
appreciate the help.
Regards,
Matt