Class E/D of 40 Meters

[K9ACT]

I changed the filter to agree with the Butterworth calculator with a 10 mHz cutoff and it looks much better but I am still looking for advice on what the cutoff freq should be.

The good news is that changing nothing else, it actually works and I made a few contacts with it yesterday and will run the net with it today. People were seeing 10db increase on mod peaks. I guess this is because the modulation is almost 200% according to my monitor.

I am running at 20V and 4A carrier and getting about 40W out so the efficiency is only 50% at this level but nothing is getting hot. On 80 meters it was putting out 200W with 30V @ 7A but I will run it this way for a while before pushing any harder.

Bruce might be able to hear the net. We do get to the East Coast. 7.295 1700Z.

Thanks for all the help.

js

[kf1z]

Wayne,

I think you are right...

as a matter of fact, I thought you ( someone?) asked Jay about it, and that was the response.
I saw this either here, on amfone, or via email;... I don't remember which.
OR for that matter what exactly he said about it! ( some help huh?)


Using a butterwrth calc, The values for the inductors come out the same as Jay's figures, but the capacitor always comes out at exactly 2x Jay's value.

That works the same with Butterworth and Chebyshev filter.

So, I've taken to using the butterworth T calculator, dividing the cap value by 2 and adjusting it later if need be.

Really doesn't matter much WHICH design is used, and could go with more elements too if one wanted to.
It IS after all just a low-pass filter!

50 ohms impedance, cutoff frequency JUST a little higher than the band edge.

I think Jay used 2mhz and 4mhz as the cutoff frequency for the 160 and 75m designs .

[K9ACT]

Wayne,


So, I've taken to using the butterworth T calculator, dividing the cap value by 2 and adjusting it later if need be.

Turns out that's sort of what I did but I forgot to quantify it. I put variable across the fixed cap and tuned for cleanest sine wave. As the cap was barely meshed, it didn't seem to be worth the trouble but I might have been surprised if I actually measured it.

>50 ohms impedance, cutoff frequency JUST a little higher than the band edge.

This bothers me a bit as I assume the cut off freq is the 3 db point so there would be a 3 db loss at the band edge which does not seem practical. That would be the point to begin the roll off not to be well into it.

Am I missing something?

js

[Gilly]

Found some of Jays old emails and he stated that the filter was a "half wave " type .....
Have never heard of this type of filter before but I'm sure he is right though.....
Probably would not matter where cutoff is really .... 10Mhz would be fine as the 2nd harmonic is greatly reduced because the output stage is push pull.....etc:

OK on your efficiency ...its poor at only 50%....something must be getting HOT
Have you checked the duty cycle......
The output balun ... is that warm ?
If there is nothing obvious try increasing the voltage a bit ... I'm sure something will be hot....

Wayne

[kf1z]

Wayne,


So, I've taken to using the butterworth T calculator, dividing the cap value by 2 and adjusting it later if need be.

Turns out that's sort of what I did but I forgot to quantify it. I put variable across the fixed cap and tuned for cleanest sine wave. As the cap was barely meshed, it didn't seem to be worth the trouble but I might have been surprised if I actually measured it.

>50 ohms impedance, cutoff frequency JUST a little higher than the band edge.

This bothers me a bit as I assume the cut off freq is the 3 db point so there would be a 3 db loss at the band edge which does not seem practical. That would be the point to begin the roll off not to be well into it.

Am I missing something?

js

Some info I found yesterday says to compensate for the 3db point, base your calculations on a frequency that is 1.2x higher than your operating frequency.

so ....

7.3mhz x 1.2 = 8.76mhz

[kf1z]

ok, this may explain the odd capacitor value....


The formulas used are:

L= Xl / ( 2pi x Fc)

So, (for Jay's 75 meter design)

L= 50 / ( 6.28 x 4)

L= 1.99 uH


AND

C = 1 / ( 2pi x Fc x Xc )

So,

C = 1/ (6.28 x 4 x 50 )

C= 1/ 1256
C= 796pf

[kf1z]

So Jack,

Based on the "new" information....

I would suggest...

use 8.75 to 9 mhz to base a butterworth filter ...

Or use the formulas I just gave ( and still use 8.75 to 9mh as the corner freq)

Which would come out to

L1. L2 = 0.91 uH
C= 364 pF

[kf1z]

The butterworth T design assumes the "C" is 'shared' between L1 and L2
so you use 25 ohms as Xc instead of 50 ohms .
So, the cap value WOULD come out twice as high in value.

[K9ACT]

I think for now I am a happy camper.

What I ended up with is a 50W carrier and over 150% and a good enough signal to run the net on 40. Most people don't even notice the difference other than missing the kerchunk when keying.

The efficiency is only about 50% but nothing gets hot and it seems to be reliable enough for a backup to my tube rig. This is all I really wanted so,"mission accomplished".

For the record, all I did to get here was to make a new LPF and remove a few turns from the output transformer secondary.
The latter may not have been necessary but I had to ad them to make it work on 80.

I would love to hear from Jay but it appears that just changing the LPF on his design will get it on 40.

[K9ACT]

Here is an update on my 40 meter project.

I am currently running the Class D rig with about 125W input and 65W out. That's only about 50% efficiency compared to about 95% on 80 meters but it sounds great on the air and is a good back up for the tube rig.

I have not tried it yet at the 200W level yet because of the efficiency problem.

The only change I made for 40 meters was the output filter.

One area I am not clear on is that people say you can't drive 2 fets with one DD414 but that is exactly what I am doing. What am I to look for as reasons why it can't be done? Efficiency? Ignorance is bliss?

Jack

[kf1z]

""
One area I am not clear on is that people say you can't drive 2 fets with one DD414 but that is exactly what I am doing
""


Take a look at your fet gate waveform while transmitting...

This MAY be why your efficiency is so low.

Bruce

[K9ACT]

Someone pointed out today that Jay had some comments about 40 on his build page.

Specifically that type 61 material works better than 45.

I can not find a source for FB 1020 -61.

Can anyone help out here?

jss

[kf1z]

I never found any

The closest thing I can think of is use more, shorter cores.
Like FT-114A-61 those are 1.14 inches long 0.74 ID and 0.545 long

[K9ACT]

I never found any

The closest thing I can think of is use more, shorter cores.
Like FT-114A-61 those are 1.14 inches long 0.74 ID and 0.545 long

Good idea but it turns out that Amidon has the 1020-61 but they are not on their web site. I ordered enough for the 200W deck and my driver for the 8000 transmitter.

I sort of forgot about this driver I made last year and thought I would see how it works on 40. It is basically the 100W VD design with single fets and drivers. With an 18V supply I get about the same output as I did on 80 changing nothing but the VFO freq and output filter and still get over 90% efficiency.

So I guess this proves that the reason for the poor efficiency on the 200W deck really is trying to drive two fets with one driver. It will be interesting to see what difference the core material makes on either or both of these decks.

It is philosophically interesting to think about total efficiency here because the drivers consume more power than the final in absolute terms. The drivers draw 4A at 12V for 48W and the finals draw 1.8A at 18V for 32W in and 30W out.

jack

[K9ACT]

Getting frustrated trying to make my Class D work on 40 as well as 80, I did some cut and try engineering and I thought I would share what I learned.

Where I ended up was with about 50% efficiency, about 70W out max and lots of distortion as a function of carrier level and very frustrating to operate.

For the record, I did change to 61 material for the transformer but it made no difference that I could tell.

WA9UJR pointed out that things might be a whole lot easier with a Class E configuration and got me to ponder Steve's schematic again. I didn't particularly want to start a whole new project so I did as I usually do and took a stab at some creative putzing with what I had.

Leaving everything as it was, I removed the LPF and built an output network as in QIX except for 3Uh instead of 5uh. I put this between the output transformer and the dummy or antenna and got satisfaction immediately. The loading does not work quite like a Pi network but it does seem to provide the same function and I get a nice clean sine wave no matter what I do to it. It loads the antenna just fine and the efficiency is about 80% to 95% depending on stuff I do not understand yet. A ran the net at 100W two days and everybody loves it.

I am now experimenting with the transformer turns and ratio but have come to no conclusion. Steve calls for 1:1 and single turns and I started at 4:5 turns and removing some for testing purposes.

One question is why does it work so well with such a variance from what is called for?

Another is, does just changing the output circuit make this a Class E?

And for the record, I am still driving two fets with with single drivers.

js