EMRFD Message Archive 2113

Message Date From Subject
2113 2008-09-23 07:06:08 timshoppa Filtering simple LO's or BFO's?
Several of the examples in EMRFD show low-pass LC filtering
following a simple crystal oscillator or a VFO. These are
straightforward designs, no mixer or multiplier in the LO or BFO
chain.

For example, the VFO in 6.74, the BFO in 6.76, the VFO in 6.77, the
LO in 6.138, etc. At least I think of them as low-pass filters, of
course such a network can be both a filter and a matching network.

Most discussions of balanced mixers make it clear that odd-order
harmonics in the injection are not a problem, and that square wave
drive in fact is the textbook/handbook model. Spurs from square wave
drive are often discussed and usually just accepted as being the
result of using a mixer.

So are the filters in the above examples there to remove even-order
harmonics? When is it important to remove the even-order harmonics?
Is it sometimes important to remove the odd-order harmonics too?

Strangely enough, sometimes I see an oscillator followed by a low-
pass filter, followed by some high-level buffering that I'm sure
will make more even harmonics than the low-pass filter ever removed.

Tim N3QE
2115 2008-09-23 09:51:32 ku4af Re: Filtering simple LO's or BFO's?
2118 2008-09-23 18:40:35 timshoppa Re: Filtering simple LO's or BFO's?
2121 2008-09-23 18:40:36 John Kolb Re: Filtering simple LO's or BFO's?
Note that a square wave has to have 50-50% symmetry to be free
of even hormonics. As the symmetry moves away from this, even
hormonics appear. A very narrow pulse is a good way to generate
lots of hormonics with all components near equal in strength.
The higher the freq goes, the harder it is to keep 50-50% symmetry.

If it is necessary to remove the even order hormonics in the analog
domain, then a low pass filter is the only reasonable way to do it,
suffering removing the odd hormonics at the same time. To remove
only the even's and leave the odd's would require a separate notch
filter tuned to each even you want to remove. Making the notches
track as the osc changed freq would be really horrible.

John

At 06:16 AM 9/23/2008, you wrote:

>Several of the examples in EMRFD show low-pass LC filtering
>following a simple crystal oscillator or a VFO. These are
>straightforward designs, no mixer or multiplier in the LO or BFO
>chain.
>
>For example, the VFO in 6.74, the BFO in 6.76, the VFO in 6.77, the
>LO in 6.138, etc. At least I think of them as low-pass filters, of
>course such a network can be both a filter and a matching network.
>
>Most discussions of balanced mixers make it clear that odd-order
>harmonics in the injection are not a problem, and that square wave
>drive in fact is the textbook/handbook model. Spurs from square wave
>drive are often discussed and usually just accepted as being the
>result of using a mixer.
>
>So are the filters in the above examples there to remove even-order
>harmonics? When is it important to remove the even-order harmonics?
>Is it sometimes important to remove the odd-order harmonics too?
>
>Strangely enough, sometimes I see an oscillator followed by a low-
>pass filter, followed by some high-level buffering that I'm sure
>will make more even harmonics than the low-pass filter ever removed.
>
>Tim N3QE



--
No virus found in this outgoing message.
Checked by AVG.
Version: 7.5.524 / Virus Database: 270.7.0/1684 - Release Date: 9/22/2008 6:39 AM
2122 2008-09-23 19:40:02 Wes Hayward Re: Filtering simple LO's or BFO's?
Hi Tim, et al,

Your analysis is fine. The various amplifiers in that Lichen chain
will, as you suspect, generate harmonics. The amplifiers are driven
fairly hard, at least the last one is, so there will be quite a bit
of distortion. But in this application, it is not a show stopper.
I measured the carrier suppression in the balanced modulator
application and it was adequate. Not spectacular, but adequate.

There was a different reason for the low pass filter in Fig 6.138. I
didn't need a lot of output. A low level of -22 dBm was adequate.
This could be easily achieved by controlling r5. But that
particular crystal oscillator is really lousy for harmonic
distortion. I use the same circuit as a comb generator to calibrate
my spectrum analyzer where the oscillator 2nd harmonic is only about
2 dB below the fundamental. So that circuit is really a bad one
without the low pass filter. The LPF cleans it up fine and it's
simple. Note that I just used molded RF chokes for the inductors.

The oscillator circuit in Fig 6.73 is a nice general purpose block,
but it can oscillate at frequencies other than that related to the
crystal. This was discovered by some builders of the "Mark-II
Universal QRP Transmitter" from QST a while back. The circuit is
not fussy up at higher frequencies, and it's a great one for third
overtone operation. But it was a pain at 7 MHz for a few folks. I
never had a problem, but I was looking at it with instrumentation
from the beginning. The better circuit for the BFO for the EZ90
would be a crystal oscillator into a buffer followed by a LPF. But
the simpler circuit works and should not present a problem. Balance
is not overly critical in this application so long as it is
reasonable.

There are no hard rules here. Ya just do what works, with the
thought that if port to port isolation is a critical mixer parameter,
it may be compromised by distortion in the LO drive.

73, Wes
w7zoi
2123 2008-09-23 21:18:48 Rick Re: Filtering simple LO's or BFO's?
Interesting Discussion. There is another point no one has mentioned. It is well known and
often quoted that a fundamental plus all the odd harmonics is a square wave:

cos(2pi*f*t) + 1/3cos3(2pi*f*t) + 1/5cos5(2pi*f*t) + ...

A square wave is a good way to turn diodes off and on. What is seldom stated is that:

cos(2pi*f*t) + 1/3cos[3(2pi*f*t) + phi] + 1/5cos[5(2pi*f*t) + some other phi] + ...

is NOT a square wave.

In English, what this means is that it's not enough to have the only the fundamental and
odd harmonics. To have a square wave, the phases of the harmonics all have to line up
too. All it takes is a little inductance or capacitance in the path to mess up the relative
phases of the odd harmonics. It will look perfect on a spectrum analyzer, but the
oscilloscope waveform is no longer a square wave, and can create puzzling results when
you apply the waveform to non-ideal diodes.

This is very interesting to explore with MATLAB or some other program that permits
plotting time and frequency domain waveforms made up of sums of harmonic components
with various amplitudes and phases. The waveforms can also easily be sketched by hand
on paper and added up.

When I want predictable results in instrument-grade projects, I use low pass filters and
sine wave drive on all the LO drive lines...but it's fun to solve the puzzles presented by
simple gear too. Some of my most fun projects work better than they should with very
few parts, and defy simple analysis.
2124 2008-09-24 06:35:39 victorkoren Re: Filtering simple LO's or BFO's?
There is another way to "filter" the signal in order to get
fundamental signal and all the odd harmonics in the right phase:
create square wave with 50% duty cycle by limiting the signal in the
time domain.
You can use fast inverters with a DC feedback circuit to adjust the
duty cycle.
I uploaded a file (in a folder with me name) with such a circuit.
You can adjust to 50% duty cycle by changing the value of R2 (replace
it with a trimmer) and tuning it for minimum even harmonics content.
I published this circuit many years ago in EDN in the "design idea"
section.
A similar circuit is sometimes used in some SDR receivers when it is
needed to generate Quadrature LO signals when the LO is at double the
received frequency and using only one stage of /2 divider. In such a
case the Duty Cycle will influence the Quadrature precision.
Victor - 4Z4ME

2156 2008-10-01 05:35:39 Nick Kennedy Re: Filtering simple LO's or BFO's?
I'm doing something similar right now in a BFO for a receiver I'm building.
The BFO uses the basic circuit of emrfd figure 4.31 with a 74HC00 instead of
74HC04 because that's what was in my junk box. I added a blocking capacitor
between the collector of the follower stage and the input gates of the IC.
Then a divider from +5V to ground (4k7, 5k pot, 4k7) and another 4k7 from
the wiper to the gate.

It seems to work pretty well. I just used the scope to match the high and
low times as closely as possible. I doubt that this is all that critical in
a BFO. Output of the circuit is about +10.5 dBm with no filtering. I'll
run it throuh a 3 dB pad and then into the TUF-1 product detector and see
how it does.

73--Nick, WA5BDU






[Non-text portions of this message have been removed]