I built a Class D push-pull CW HF TX with 74HC240s providing
oscillation, buffering, an extra inverter on one side, and two stages
of current amplification, eventually driving a pair of IRF530
MOSFETs..

It looks like I struck a good balance even if there was a phase
offset.

This is what happened.

I turned ann inversting 8-section driver 74HC240 from a single-ended
oscillator-buffer-4Xamp (N7KSB style), into this push-pull monster:

| -inverter-2Xamp-16Xamp-...
VXO-buffer-buffer-|
|-------------2Xamp-16Xamp-...

From VXO to the 2+2 sections in push-pull current amplifier, all fits
into one 74HC240, fed at 5Vdc. Then, in a 8X equivalent fanout, I put
2 x 74HC240 on each side, fed at 8Vdc. 16 sections of current
amplification per side provide A LOT of current, already delivered
over 1W RF, and proved more than sufficient for driving a pair of
IRF530's in push pull with individually trimmed positive DC bias,
probably with room for further expansion.

The interesting fact is that there is one extra inverting section on
one side only. On that side the signal should have been delayed by
20-30ns, the propagation time for one section. At 10MHz this should
equal 70-110 degrees. In theory this should mean that both push-pull
sides should spend 110 degrees in high status (MOSFETs conducting),
and 110 degrees not conducting, leaving only 140 degrees for honest-to-
G_d pushpullery. :-(

Woe betide such a push-pull... BUT... with my scope having one
channel KO, I can't juxtapose the events - I can only combine the
results in the output transformer.

The result seems like a pretty symmetric waveform, with steep voltage
peaks on both sides (output balun primary still not tuned). Once
filtered into a sinusoid, I see about 60W DC going in and about 45-50W
RF coming out, with negligible second harmonic.

Is this normal? What did I get wrong? (I.e. got right, but just by
chance). Would I be wasting my time if I sought more push-pull symmety
by means of an unbal transformer? I mean somethign like this:

|-buffer-2Xamp-...
VXO-buffer-unbal
|-buffer-2Xamp-...

Any wisdom welcome!

[OK, I am going to have the scope's 2nd channel fixed anyway... :-) ]

"spamhog" wrote in message
...

I built a Class D push-pull CW HF TX with 74HC240s providing
oscillation, buffering, an extra inverter on one side, and two stages
of current amplification, eventually driving a pair of IRF530
MOSFETs..

It looks like I struck a good balance even if there was a phase
offset.

This is what happened.

I turned ann inversting 8-section driver 74HC240 from a single-ended
oscillator-buffer-4Xamp (N7KSB style), into this push-pull monster:

| -inverter-2Xamp-16Xamp-...
VXO-buffer-buffer-|
|-------------2Xamp-16Xamp-...

From VXO to the 2+2 sections in push-pull current amplifier, all fits
into one 74HC240, fed at 5Vdc. Then, in a 8X equivalent fanout, I put
2 x 74HC240 on each side, fed at 8Vdc. 16 sections of current
amplification per side provide A LOT of current, already delivered
over 1W RF, and proved more than sufficient for driving a pair of
IRF530's in push pull with individually trimmed positive DC bias,
probably with room for further expansion.

The interesting fact is that there is one extra inverting section on
one side only. On that side the signal should have been delayed by
20-30ns, the propagation time for one section. At 10MHz this should
equal 70-110 degrees. In theory this should mean that both push-pull
sides should spend 110 degrees in high status (MOSFETs conducting),
and 110 degrees not conducting, leaving only 140 degrees for honest-to-
G_d pushpullery. :-(

Woe betide such a push-pull... BUT... with my scope having one
channel KO, I can't juxtapose the events - I can only combine the
results in the output transformer.

The result seems like a pretty symmetric waveform, with steep voltage
peaks on both sides (output balun primary still not tuned). Once
filtered into a sinusoid, I see about 60W DC going in and about 45-50W
RF coming out, with negligible second harmonic.

Is this normal? What did I get wrong? (I.e. got right, but just by
chance). Would I be wasting my time if I sought more push-pull symmety
by means of an unbal transformer? I mean somethign like this:

|-buffer-2Xamp-...
VXO-buffer-unbal
|-buffer-2Xamp-...

Any wisdom welcome!

[OK, I am going to have the scope's 2nd channel fixed anyway... :-) ]

Well, if the second harmonic is really low then you must have a good square
wave. On the output side you could determine the phase relationships by
looking at the symmetry of the square wave. Adjust the phase with an rc
network on the input of one of the parallel shortest delay stacks of 74hc's.
I have paralled several 74xx series of chips together before and found that
the propagation delays were within the spec's but varied a lot within those
specs. After you tune your primary you should get a really good sine wave
out of the amp. I would not go through the trouble of unbalancing the
transformer as this could result in some unusual spikes on the output due to
saturation of the core, or cross conduction of the Fets due to di/dvt
issues. One other thing comes to mind is that with a torrid core at high
current and a high frequency, then symmetry of the winding itself is
critical. The spacing of the winding should fill the whole core and be
spaced as uniform as you can make them. Having said that, then, match 2
IRF530's for rds on. (A few tenths of an ohm really makes a big difference
when you are throwing around a bunch of amps.) This will go a long ways in
preventing what looks like an unbalanced output transformer at 10 Mhz.

Sounds like your on the right track. Just a little bit of tweaking to go to
get there.


Rich
In Texas

Thanks Rich!

On Sep 17, 4:55*pm, "Rich Klestinez" wrote:
I would not go through the trouble of unbalancing the transformer

I didn't mean to... The idea was to generate two symmetrical and
opposed signals by putting an un-bal transformer between the VXO and a
first push-pull stage made with just two sections, charged with
rebuilding near-square waveforms.

I matched neither driver sections nor MOSFETs,but I did hear that both
can be a bit variable. Apparently the chip drivers are normally used
without equalization (to the surprise of quite a few authours). As for
MOSFETs, I saw several amp designs with a single separate bias
trimmer, or with a single special driver chip, for several devices. I
have separate trimmers, I didn't bother to check if the 1-mA voltage
was the same for both - but I will. That may have helped equalize the
switching point in terms of RF voltage. As for the different on-state
resistance, I guess it's more critical if one has devices in parallel
- in a push-pull a particuarly low resistance device won't hog the
current.

One other thing comes to mind is that with a torrid core at high current and a high frequency, then symmetry of the winding itself is critical.

Yes, I've been extremely careful. Winding is trifilar, twisted, I
took care to have symmetrical twist pitch and winding spacing.

torrid core

Torrid indeed. Now I have a tiny .82 core and thin wire too. At
normal duty cycle all is fine, after 120s keydown one risks minor
burns - but the MOSFETs barely get warm.

Coincidentally, I just got several 10s of feet of Raychem "high
performance" wire after helping assemble a prototype 4-fuel car (3
liquid fuels + plugin recharger). 1.5sqmm, silver plated stranded
copper alloy, thin 600V radiation-crosslinked high temperature
insulator. I've been verbally told the 1/2sqmm variety can carry 30A
(I asked as I worried it would not bear 3.5). I will try it on a
larger core. I already recovered the ferrite of a 1-kW Hi Gain balun,
but I want to see detailed wire specs first. Once stripped, this wire
should do well in hi-Q coils too.

"spamhog" wrote in message
...
Thanks Rich!

On Sep 17, 4:55 pm, "Rich Klestinez" wrote:
I would not go through the trouble of unbalancing the transformer

I didn't mean to... The idea was to generate two symmetrical and
opposed signals by putting an un-bal transformer between the VXO and a
first push-pull stage made with just two sections, charged with
rebuilding near-square waveforms.

I matched neither driver sections nor MOSFETs,but I did hear that both
can be a bit variable. Apparently the chip drivers are normally used
without equalization (to the surprise of quite a few authours). As for
MOSFETs, I saw several amp designs with a single separate bias
trimmer, or with a single special driver chip, for several devices. I
have separate trimmers, I didn't bother to check if the 1-mA voltage
was the same for both - but I will. That may have helped equalize the
switching point in terms of RF voltage. As for the different on-state
resistance, I guess it's more critical if one has devices in parallel
- in a push-pull a particuarly low resistance device won't hog the
current.

One other thing comes to mind is that with a torrid core at high current
and a high frequency, then symmetry of the winding itself is critical.

Yes, I've been extremely careful. Winding is trifilar, twisted, I
took care to have symmetrical twist pitch and winding spacing.

torrid core

Torrid indeed. Now I have a tiny .82 core and thin wire too. At
normal duty cycle all is fine, after 120s keydown one risks minor
burns - but the MOSFETs barely get warm.

Coincidentally, I just got several 10s of feet of Raychem "high
performance" wire after helping assemble a prototype 4-fuel car (3
liquid fuels + plugin recharger). 1.5sqmm, silver plated stranded
copper alloy, thin 600V radiation-crosslinked high temperature
insulator. I've been verbally told the 1/2sqmm variety can carry 30A
(I asked as I worried it would not bear 3.5). I will try it on a
larger core. I already recovered the ferrite of a 1-kW Hi Gain balun,
but I want to see detailed wire specs first. Once stripped, this wire
should do well in hi-Q coils too.

Awsome but really good wire!

If the Fet's are just barely warm it is my guess that you are driving them
well into saturation and that good. As far as the core getting that hot,
well that's not so good.
I digress. Some years back (70's) I tried to model the BH vs temperature of
some high power cores that I was using at 452 Khz. They were getting very
hot and we couldent figgure out why. Upon observing the BH curve at 452 Khz
at 101% saturation there were bumps and little knots on the curve, they were
really temperature dependant as well. Turned out (remember this was in the
1970's) this was a combination of grain size and a purely mechanical effect
(pizoelectric). We switched to a different material with a very fine grain
domains and the cores ran cool. I nearly wore out my Hemi 256 slide rule on
that one. Still have the darn thing. Nowdays core selection is much
easier for high current high frequency uses. Pick a core empirically do
your design and take advice from the wisdom of the below.

FYI. Adams magnetic products, Tom Corrigan Applications Engineer,
1-800-747-7543 or his direct line 630-359-9445. A really good guy and reps
several lines of cores. Of course there is always the Ferroxcube area
application engineers that are some of the best in the world that I have
worked with in the past year. They will take the time to help you, however,
you should be conversant in the terms of the language of magnetics or you
wont understand what they mean.

Its a challange, but the rewards are great. You will have a transmitter
that will be reliable and reproducible.

rich
in Texas

PS I have retired now (66 Yrs old). Not a ham yet. As soon as I get most of
my honey-doos done I am considering trying for a licence.

On Sep 17, 12:11*pm, spamhog wrote:
I built a Class D push-pull CW HF TX with 74HC240s providing
oscillation, buffering, an extra inverter on one side, and two stages
of current amplification, eventually driving a pair of IRF530
MOSFETs..

It looks like I struck a good balance even if there was a phase
offset.

This is what happened.

I turned ann inversting 8-section driver 74HC240 from a single-ended
oscillator-buffer-4Xamp (N7KSB style), into this push-pull monster:

* * * * * * * * * * * * *| -inverter-2Xamp-16Xamp-...
VXO-buffer-buffer-|
* * * * * * * * * * * * *|-------------2Xamp-16Xamp-...

From VXO to the 2+2 sections in push-pull current amplifier, all fits
into one 74HC240, fed at 5Vdc. Then, in a 8X equivalent fanout, I put
2 x 74HC240 on each side, fed at 8Vdc. 16 sections of current
amplification per side provide A LOT of current, already delivered
over 1W RF, and proved more than sufficient for driving a pair of
IRF530's in push pull with individually trimmed positive DC bias,
probably with room for further expansion.

The interesting fact is that there is one extra inverting section on
one side only. *On that side the signal should have been delayed by
20-30ns, the propagation time for one section. At 10MHz this should
equal 70-110 degrees. In theory this should mean that both push-pull
sides should spend 110 degrees in high status (MOSFETs conducting),
and 110 degrees not conducting, leaving only 140 degrees for honest-to-
G_d pushpullery. :-(

Woe betide such a push-pull... *BUT... with my scope having one
channel KO, I can't juxtapose the events - I can only combine the
results in the output transformer.

The result seems like a pretty symmetric waveform, with steep voltage
peaks on both sides (output balun primary still not tuned). Once
filtered into a sinusoid, I see about 60W DC going in and about 45-50W
RF coming out, with negligible second harmonic.

Is this normal? *What did I get wrong? (I.e. got right, but just by
chance). Would I be wasting my time if I sought more push-pull symmety
by means of an unbal transformer? I mean somethign like this:

* * * * * * * * * * * *|-buffer-2Xamp-...
VXO-buffer-unbal
* * * * * * * * * * * *|-buffer-2Xamp-...

Any wisdom welcome!

[OK, I am going to have the scope's 2nd channel fixed anyway... :-) ]
As I see it you have a condition where both transistors are off at the
same time and both are on at the same time during parts of each cycle.
Adding another gate to equalize the propagation time would be the
first thing I would do. I would want to have my o'scope working
before I started playing with turns on the transformer.

Jimmie

wrote in message
...
On Sep 17, 12:11 pm, spamhog wrote:
I built a Class D push-pull CW HF TX with 74HC240s providing
oscillation, buffering, an extra inverter on one side, and two stages
of current amplification, eventually driving a pair of IRF530
MOSFETs..

It looks like I struck a good balance even if there was a phase
offset.

This is what happened.

I turned ann inversting 8-section driver 74HC240 from a single-ended
oscillator-buffer-4Xamp (N7KSB style), into this push-pull monster:

| -inverter-2Xamp-16Xamp-...
VXO-buffer-buffer-|
|-------------2Xamp-16Xamp-...

From VXO to the 2+2 sections in push-pull current amplifier, all fits
into one 74HC240, fed at 5Vdc. Then, in a 8X equivalent fanout, I put
2 x 74HC240 on each side, fed at 8Vdc. 16 sections of current
amplification per side provide A LOT of current, already delivered
over 1W RF, and proved more than sufficient for driving a pair of
IRF530's in push pull with individually trimmed positive DC bias,
probably with room for further expansion.

The interesting fact is that there is one extra inverting section on
one side only. On that side the signal should have been delayed by
20-30ns, the propagation time for one section. At 10MHz this should
equal 70-110 degrees. In theory this should mean that both push-pull
sides should spend 110 degrees in high status (MOSFETs conducting),
and 110 degrees not conducting, leaving only 140 degrees for honest-to-
G_d pushpullery. :-(

Woe betide such a push-pull... BUT... with my scope having one
channel KO, I can't juxtapose the events - I can only combine the
results in the output transformer.

The result seems like a pretty symmetric waveform, with steep voltage
peaks on both sides (output balun primary still not tuned). Once
filtered into a sinusoid, I see about 60W DC going in and about 45-50W
RF coming out, with negligible second harmonic.

Is this normal? What did I get wrong? (I.e. got right, but just by
chance). Would I be wasting my time if I sought more push-pull symmety
by means of an unbal transformer? I mean somethign like this:

|-buffer-2Xamp-...
VXO-buffer-unbal
|-buffer-2Xamp-...

Any wisdom welcome!

[OK, I am going to have the scope's 2nd channel fixed anyway... :-) ]
As I see it you have a condition where both transistors are off at the
same time and both are on at the same time during parts of each cycle.
Adding another gate to equalize the propagation time would be the
first thing I would do. I would want to have my o'scope working
before I started playing with turns on the transformer.

Jimmie

If I understand you the IRF fets are in push pull. If this is the case then
at no time should both transistors be conductiing at the same time ever.
There should be a dead band between when one transistor conducts and the
other is turning off.