In article , (Tom
Bruhns) writes:

(Avery Fineman) wrote in message
...

Unsymmetric digital signals haven't been found to have much more
even harmonic content than odd harmonit content when viewed on a
spectrum analyzer...until the digital signal pulse width is VERY
short compared to its repetition time.

Huh? Try a 1/3 -- 2/3 ratio. NO third harmonic (or sixth or ninth,
etc); second is the greatest amplitude harmonic at fully 1/2 the
fundamental amplitude. Fourth is larger than the fifth. For a
symmetric square wave, there's no second and the third is only 1/3 the
amplitude of the fundamental. Easy to verify with a spectrum analyzer
or through a Fourier series. As the pulse width goes to zero, the
fundamental and all harmonics go to equal amplitudes. At 5% pulse
width, for example, the harmonic amplitudes decrease monotonically as
harmonic number increases, out to the twentieth, which is a null, but
even the 12th is fully 50% of the fundamental.

Thank you, Tom...but I've done the "SineX over X" thing a few times
already and looked at a lot of spectral displays. Even wrote a
computer program or three to determine the harmonic content of
various arbitrary-shaped defined waveforms (two for corporate use,
one in a shareware/freeware package).

The first null of ALL harmonics occurs at frequencies at the inverse
of the repetition time of the waveform. That turns out to be fairly
true for any waveform, not just a pulse. Trouble is, the harmonic
content NOT at the major null increments varies considerably in
actual practice since the source of RF being multiplied varies in
shape in actual practice.

Discussing harmonic content is so much bafflegab without first
defining the source's waveform shape...even if to coarse levels
of square wave versus sinusoid versus pulse or whatever.

The kickoff message to this thread started with "square wave"
and all I did was further refine that to symmetical square wave.
You aren't going to tell me there's a lot of even harmonic energy
in a symmetric square wave because I can set it up on the bench
and demonstrate it doesn't. Practical test versus theoretical and
results match.

I once managed to phase-lock a pulse generator in order to key
on a signal generator output for exactly ONE RF cycle (35 db on-
off ratio). Interesting broad spectral content, way wide almost
like an SRD output. Impractical, of course, was to satisfy a
couple of others arguing the whichness or the what one day. :-)

This is useful info if you're trying to design a simple amplifier-type
doubler; adjusting the conduction angle to approximately 1/3 will give
you lots of second but little third, making it easier to filter the
output. But if you weren't thinking about it and tried to make a
tripler, and accidentally made your conduction angle 1/3, you might
wonder why you were having so much trouble getting good tripler
efficiency.

Well, I'm not wondering at all since I haven't built any vacuum tube
or FET active multipliers, just the bipolar transistor types that don't
quite fit into the old "conduction angle" simple rules.

Correction: Built a tube crystal oscillator (6AK6) that doubled or
tripled (as desired by tuning) to be the LO for a 6BE6 pentagrid
running at external LO injection. Behaved like a fundamental
crystal oscillator that doubled or trippled in the plate. That was
around 1960 and didn't have the tools to check waveforms on my
lunch hours. :-) Just played with it until it worked...and had
enough LO injection to run the pentagrid at good conversion
transconductance.

Why bother to use frequency multiplier stages NOW when PLLs
and DDSs can have oscillators at their fundamental and those can
be divided down with available parts to a low, stable lock frequency?
It's fine for restoring antiques to use lots of multipliers, but, let's
face it, those can be a pain in the expletive deleted to recreate now.
Others' mileage varies, of course.

I mentioned a simple two-diode "full-wave" doubler because it is an
easy thing to implement and is broadband to start with...no fussing
with tuned-circuit impedances at the start to make it work with a
non-square source waveform. It may not be as efficient as some
would like but it is easy to do compared to direct active doublers.

Len Anderson
retired (from regular hours) electronic engineer person

In article , (Tom
Bruhns) writes:

(Avery Fineman) wrote in message
...

Unsymmetric digital signals haven't been found to have much more
even harmonic content than odd harmonit content when viewed on a
spectrum analyzer...until the digital signal pulse width is VERY
short compared to its repetition time.

Huh? Try a 1/3 -- 2/3 ratio. NO third harmonic (or sixth or ninth,
etc); second is the greatest amplitude harmonic at fully 1/2 the
fundamental amplitude. Fourth is larger than the fifth. For a
symmetric square wave, there's no second and the third is only 1/3 the
amplitude of the fundamental. Easy to verify with a spectrum analyzer
or through a Fourier series. As the pulse width goes to zero, the
fundamental and all harmonics go to equal amplitudes. At 5% pulse
width, for example, the harmonic amplitudes decrease monotonically as
harmonic number increases, out to the twentieth, which is a null, but
even the 12th is fully 50% of the fundamental.

Thank you, Tom...but I've done the "SineX over X" thing a few times
already and looked at a lot of spectral displays. Even wrote a
computer program or three to determine the harmonic content of
various arbitrary-shaped defined waveforms (two for corporate use,
one in a shareware/freeware package).

The first null of ALL harmonics occurs at frequencies at the inverse
of the repetition time of the waveform. That turns out to be fairly
true for any waveform, not just a pulse. Trouble is, the harmonic
content NOT at the major null increments varies considerably in
actual practice since the source of RF being multiplied varies in
shape in actual practice.

Discussing harmonic content is so much bafflegab without first
defining the source's waveform shape...even if to coarse levels
of square wave versus sinusoid versus pulse or whatever.

The kickoff message to this thread started with "square wave"
and all I did was further refine that to symmetical square wave.
You aren't going to tell me there's a lot of even harmonic energy
in a symmetric square wave because I can set it up on the bench
and demonstrate it doesn't. Practical test versus theoretical and
results match.

I once managed to phase-lock a pulse generator in order to key
on a signal generator output for exactly ONE RF cycle (35 db on-
off ratio). Interesting broad spectral content, way wide almost
like an SRD output. Impractical, of course, was to satisfy a
couple of others arguing the whichness or the what one day. :-)

This is useful info if you're trying to design a simple amplifier-type
doubler; adjusting the conduction angle to approximately 1/3 will give
you lots of second but little third, making it easier to filter the
output. But if you weren't thinking about it and tried to make a
tripler, and accidentally made your conduction angle 1/3, you might
wonder why you were having so much trouble getting good tripler
efficiency.

Well, I'm not wondering at all since I haven't built any vacuum tube
or FET active multipliers, just the bipolar transistor types that don't
quite fit into the old "conduction angle" simple rules.

Correction: Built a tube crystal oscillator (6AK6) that doubled or
tripled (as desired by tuning) to be the LO for a 6BE6 pentagrid
running at external LO injection. Behaved like a fundamental
crystal oscillator that doubled or trippled in the plate. That was
around 1960 and didn't have the tools to check waveforms on my
lunch hours. :-) Just played with it until it worked...and had
enough LO injection to run the pentagrid at good conversion
transconductance.

Why bother to use frequency multiplier stages NOW when PLLs
and DDSs can have oscillators at their fundamental and those can
be divided down with available parts to a low, stable lock frequency?
It's fine for restoring antiques to use lots of multipliers, but, let's
face it, those can be a pain in the expletive deleted to recreate now.
Others' mileage varies, of course.

I mentioned a simple two-diode "full-wave" doubler because it is an
easy thing to implement and is broadband to start with...no fussing
with tuned-circuit impedances at the start to make it work with a
non-square source waveform. It may not be as efficient as some
would like but it is easy to do compared to direct active doublers.

Len Anderson
retired (from regular hours) electronic engineer person

Just did an interesting little 'speriment. Standard diode "full wave
rectifier" frequency doubler. Transformer is 16 trifilar turns on an
FT50-43 core (should be a bit more than 50uH per section). One of the
triplet is the primary, and the other two are connected as a
center-tapped secondary. The diodes are 1N4007 -- yep, the 1kV
mains-freq rectifiers. Excitation comes from an HP3326, set to square
wave output, source impedance 50 ohms. 50 ohm load impedance on the
doubler output (input to spectrum analyzer; DC coupled load). HP3326
square wave risetime is about 10 nanoseconds, I believe. Excite at
0.5MHz, +/-2V (4Vp-p) Output waveform observed on a fast scope is
frequency-doubled, close to 50% duty cycle, with fast falling edges
and slow (200nsec) rising edges. Amplitude about 2Vp-p. Strong
spectral output on all even harmonics; all odds suppressed about 20dB
from the low evens, and I'm sure would be much lower with better
matching of the diodes.

Explanation left as an exercise for the reader, but should be obvious
from previous discussion here. I'd guess 1N4148-type diodes would
behave similarly for an input around 100MHz.

Cheers,
Tom

(Tom Bruhns) wrote in message om...
(Avery Fineman) wrote in message ...
... Suffice to say
that a square wave cannot be used with a passive diode doubler; all
the energy is contained in the short transition times and that is rarely
enough to be worth it.

?? Lots of energy in the fundamental; filter to extract the
fundamental and feed it to your full-wave rectifier doubler.
Efficiency can be high if the filter does not cause dissipation in the
source at the harmonics.

Just did an interesting little 'speriment. Standard diode "full wave
rectifier" frequency doubler. Transformer is 16 trifilar turns on an
FT50-43 core (should be a bit more than 50uH per section). One of the
triplet is the primary, and the other two are connected as a
center-tapped secondary. The diodes are 1N4007 -- yep, the 1kV
mains-freq rectifiers. Excitation comes from an HP3326, set to square
wave output, source impedance 50 ohms. 50 ohm load impedance on the
doubler output (input to spectrum analyzer; DC coupled load). HP3326
square wave risetime is about 10 nanoseconds, I believe. Excite at
0.5MHz, +/-2V (4Vp-p) Output waveform observed on a fast scope is
frequency-doubled, close to 50% duty cycle, with fast falling edges
and slow (200nsec) rising edges. Amplitude about 2Vp-p. Strong
spectral output on all even harmonics; all odds suppressed about 20dB
from the low evens, and I'm sure would be much lower with better
matching of the diodes.

Explanation left as an exercise for the reader, but should be obvious
from previous discussion here. I'd guess 1N4148-type diodes would
behave similarly for an input around 100MHz.

Cheers,
Tom

(Tom Bruhns) wrote in message om...
(Avery Fineman) wrote in message ...
... Suffice to say
that a square wave cannot be used with a passive diode doubler; all
the energy is contained in the short transition times and that is rarely
enough to be worth it.

?? Lots of energy in the fundamental; filter to extract the
fundamental and feed it to your full-wave rectifier doubler.
Efficiency can be high if the filter does not cause dissipation in the
source at the harmonics.

On 24 Feb 2004 13:00:57 -0800, (Tom Bruhns) wrote:

Just did an interesting little 'speriment. Standard diode "full wave
rectifier" frequency doubler. Transformer is 16 trifilar turns on an
FT50-43 core (should be a bit more than 50uH per section). One of the
triplet is the primary, and the other two are connected as a
center-tapped secondary. The diodes are 1N4007 -- yep, the 1kV
mains-freq rectifiers. Excitation comes from an HP3326, set to square
wave output, source impedance 50 ohms. 50 ohm load impedance on the
doubler output (input to spectrum analyzer; DC coupled load). HP3326
square wave risetime is about 10 nanoseconds, I believe. Excite at
0.5MHz, +/-2V (4Vp-p) Output waveform observed on a fast scope is
frequency-doubled, close to 50% duty cycle, with fast falling edges
and slow (200nsec) rising edges. Amplitude about 2Vp-p. Strong
spectral output on all even harmonics; all odds suppressed about 20dB
from the low evens, and I'm sure would be much lower with better
matching of the diodes.

Explanation left as an exercise for the reader, but should be obvious
from previous discussion here. I'd guess 1N4148-type diodes would
behave similarly for an input around 100MHz.

Great. So can you guys now agree on the most appropriate waveshape to
generate the maximum amount of even harmonics?
--

The BBC: Licensed at public expense to spread lies.

On 24 Feb 2004 13:00:57 -0800, (Tom Bruhns) wrote:

Just did an interesting little 'speriment. Standard diode "full wave
rectifier" frequency doubler. Transformer is 16 trifilar turns on an
FT50-43 core (should be a bit more than 50uH per section). One of the
triplet is the primary, and the other two are connected as a
center-tapped secondary. The diodes are 1N4007 -- yep, the 1kV
mains-freq rectifiers. Excitation comes from an HP3326, set to square
wave output, source impedance 50 ohms. 50 ohm load impedance on the
doubler output (input to spectrum analyzer; DC coupled load). HP3326
square wave risetime is about 10 nanoseconds, I believe. Excite at
0.5MHz, +/-2V (4Vp-p) Output waveform observed on a fast scope is
frequency-doubled, close to 50% duty cycle, with fast falling edges
and slow (200nsec) rising edges. Amplitude about 2Vp-p. Strong
spectral output on all even harmonics; all odds suppressed about 20dB
from the low evens, and I'm sure would be much lower with better
matching of the diodes.

Explanation left as an exercise for the reader, but should be obvious
from previous discussion here. I'd guess 1N4148-type diodes would
behave similarly for an input around 100MHz.

Great. So can you guys now agree on the most appropriate waveshape to
generate the maximum amount of even harmonics?
--

The BBC: Licensed at public expense to spread lies.