Key difference between class D & E

[steve_qix]

I do not know if that circuit actually worked, at least as advertised, so I cannot give you any advise about it. 

From a voltage standpoint, if you run class E and use an IRFP260, you will be able to run about 10VDC at carrier, assuming 30VDC available for full positive modulation (200%).  That will put the repetitive RF peaks across the MOSFET at around 120V which leaves 80V of headroom before exceeding the device voltage rating.

If you compromise the headroom by running more voltage, the RF amplifier will be unreliable under adverse conditions, which always come about eventually.

With true class D, the RF voltage peaks do not, in theory, rise as high as they do in class E - *HOWEVER* this is only true for a properly terminated RF amplifier.  If the load is bad, the voltage, particularly for a single ended class D RF stage, will rise to the same or higher levels than you will get in Class E.

[W4NEQ]

...
Rules of thumb:  For 4 MOSFETs on 75 meters running 45 volts @ 4A to 5A per module, you need 1000pF per module.  For 6 MOSFETs on 80 meters running 45 volts @ 6.4A to 7.5A per module you need 1500pF per module.  On 160, double the values.  I typically build rigs to cover 2 bands and switch in the additional shunt capacitance when changing to the lower band.  This method works very well....

That is in addition to the fet capacitance and that created by the case via the silpad?  I'm hoping to nail the correct amount for 75 with the spreader / kapton / heatsink, and switch in additional for 160.

How much of the total do you allow for the fet / silpad ?

[steve_qix]

...
Rules of thumb:  For 4 MOSFETs on 75 meters running 45 volts @ 4A to 5A per module, you need 1000pF per module.  For 6 MOSFETs on 80 meters running 45 volts @ 6.4A to 7.5A per module you need 1500pF per module.  On 160, double the values.  I typically build rigs to cover 2 bands and switch in the additional shunt capacitance when changing to the lower band.  This method works very well....

That is in addition to the fet capacitance and that created by the case via the silpad?  I'm hoping to nail the correct amount for 75 with the spreader / kapton / heatsink, and switch in additional for 160.

How much of the total do you allow for the fet / silpad ?

Oh, that's the value of the actual capacitor, not necessarily the total capacitance.  The total capacitance would be the shunt capacitor itself plus the capacitance created by the MOSFETs to ground using the silpad as a dialectric, which I have never measured.

[KF1Z]

FQA11n90 with sil-pad  about 35pf per device case to heatsink capacitance.

For Kapton/copper  heatspreader cap...  your looking at about 4" x 6" of heat-spreader per module...
with 2 mil Kapton...
I did this approach for 75 meters. ( 2" x 6" spreaders)

Unless you just happen to have 2 ,  4" x 6" , 1/2" thick pieces of superbly surfaced copper, it'll be
more expensive than using ATC caps!

The copper HAS to be perfectly flat, and HAS to be thick, or you'll run into problems.
( heatsink of course has to be perfectly flat too)


If I had any extra ATC100C caps left, I'd offer them to you...

I have a few, but may want them for a future project.  :-)
I might be able to scrounge a few used ones...


Of course, if you're planning on making this 75/160m then the Kapton cap must be sized for 75meters, and switch in the extra ShuntC for 160m.

Also keep in mind that the capacitance varies GREATLY with the amount of pressure the copper is pressed to the heatsink with.
( so there's a bit of adjustment to be had)

Bruce
kf1z

[steve_qix]

[snip]

Unless you just happen to have 2 ,  4" x 6" , 1/2" thick pieces of superbly surfaced copper, it'll be
more expensive than using ATC caps!

The copper HAS to be perfectly flat, and HAS to be thick, or you'll run into problems.
( heatsink of course has to be perfectly flat too)

Those are the reasons (particularly the 2nd one) why I've never bothered with anything other than sil-pads under individual MOSFETs, and a discrete component for shunt the shunt capacitor.  It's just plain easier.

[W4NEQ]

I've used 1/4" inch copper before and have been satisfied with the results.  Fortunately, as part of my loop antenna / filter business I have a CNC machine in the shop, so the mechanical details become much easier.  I was planning on milling the surfaces, then polishing them.

Nice photo ... although I haven't figured out yet how you pierced clean holes in the kapton for his tensioning bolts ...

Chris

[KF1Z]

Hole punch

 

[WA1GFZ]

I used a method my Dad showed me on how to make a gasket around an odd shape hole. The end of a screw works well over a threaded hole. A shark XACTO works well also.

[n1ps]

Nice discussion here.  The class D amps do work.  As was pointed out, W1VD has some nice info.

Chris:  Use Steve's designs and methods.  They work, are well thought out and many people use it.  Use his PWM boards.  Your biggest challenge in building these things is the mechanical aspects.  I designed my cabinets in CAD and still inevitably made changes on the fly.  But building these things is half the fun...except when I make dumb mistakes...but that is par for the course for me

Steve:  I'm curious what you (and others like Bruce) have been using for (12V) relays to switch in C for 160.  I wish to upgrade my amp this spring. I'm always looking for "RF relays" at the festers, whether they were designed as such or not.  I always struggle to keep enough RF relays around.  I noted on the photos of your website that you appear to be using some auto relays? 

Also how much extra C should I add onto the drains?  How about Load C?

TNX
Peter

[KF1Z]

Pete,

Steve found some automotive relays at rat-shack that work well. ( I have yet to find any as he describes though)

I use some I found at fairradio.  nice wide contact SPST jobbies.
Doesn't need to be something usually thought of as an RF relay.   

If you have 1000pf per module for  75m, add in another 1000pf for 160

And also need to almost double the loadC

Bruce

( I responded to your PM today wid me new email addy)

[WA1GFZ]

1/8 inch copper is plenty thick enough because the dissipation is so low. 1/2 inch is very over kill. Sil pads, the good ones are nothing more than silver loaded Kapton sheets. I use the spreader for two reasons, heat (minor) and the nice distributed C around each FET with zero lead length. When you look at the waveforms using a 1GHz scope you see all the parasitic inductance effects.

[W4NEQ]

1/8 inch copper is plenty thick enough because the dissipation is so low. 1/2 inch is very over kill. Sil pads, the good ones are nothing more than silver loaded Kapton sheets. I use the spreader for two reasons, heat (minor) and the nice distributed C around each FET with zero lead length. When you look at the waveforms using a 1GHz scope you see all the parasitic inductance effects.

The winged ATC100 caps are certainly easier, but I'm thinking you have to buy 100 of those buggers, and they're pretty pricey for that quantity - a bunch of 470s would be nice  ... a group buy would be required ...  BUT as GFZ commented,

the Kapton sandwich approach seems like it would be the ultimate in low impedance C.  I obtained the high thermal flavor Kapton from McMaster carr.   Perhaps the remaining padder C added for 160 doesn't have to be quite so high Q ?

My original question concerned the essence of the difference between D & E.   I welcome refinements to these rules-of-thumb:
---------------------------------------------------
1. D is class C with square wave drive.

2. E has square wave or saturating drive with a shunt integration capacitor carefully chosen to produce a < 180 degree sine shaped voltage waveform at the drain to allow switching to be done at the low drain voltage part of the cycle.

This capacitor also swamps out higher frequency parasitics AND resonates at the fundamental with the remainder of the output network.

3.  Both classes benefit from an output network which maintains relatively high impedance at harmonic frequencies.

4. E can run single ended, but push pull has significant advantages.
-------------------------------------------------------------------

PS.  Yesterday I visited Don K4KYV in TN.   His antenna farm and shack are nothing less than ultra-cool.  Part homebrew laboratory, part museum. With the possible exception of his Sherwood sync detector, passing the threshold of his converted one-room schoolhouse is like stepping into 1955.

Chris

[KF1Z]

1/8 inch copper is plenty thick enough because the dissipation is so low. 1/2 inch is very over kill. Sil pads, the good ones are nothing more than silver loaded Kapton sheets. I use the spreader for two reasons, heat (minor) and the nice distributed C around each FET with zero lead length. When you look at the waveforms using a 1GHz scope you see all the parasitic inductance effects.

BUT

The thicker the spreader is, the fewer screws needed to hold it down!

I only used 4 for mine..

With 1/8" would require at least 16 if not more... just to keep the pressure even.

And THEN there's the screws for the mosfets themselves.  1/8" isn't enough... as far as I am concerned. 

[WA1GFZ]

my 1995 160 meter final has only .062 copper held in by the 14 FET screws spaced by 3 mil kapton. I monitor the case temp of one FET and may hit 110 degrees F after an old buzzard on a hot day. A FET dissipating 50 watts is not a problem.
Unlike the QRO SS linear I'm working on that will have 3/8 copper under the FETs.
In thermal design terms heat will travel about 10 times the thickness with most concentrated about 3 times the thickness. A linear is only about 50% efficient so a much bigger deal. That is the price of an all mode machine that covers 160 through 6m. The FETs dissipate 250 watts PEP.

[WU2D]

Frank,

Did you get to attend the class given by Dr. Sokal in 1998 at Raytheon? I was at ADI at the time and the IEEE sponsored two great classes in Lexington, one on the history of spread spectrum which was awesome and Dr. Sokal's class on higher classes of amplifiers.

MIke WU2D

[WA1GFZ]

No I did not know about his class but I had many emails with him where he shared lots of notes.
I only had one eyeball with him at Boxboro.
I just bought 5 - 2400 watt PEP MRI modules cheap on ebay that I plan to rebuild to move them from 80MHz down to 160m. I've done this at 300 watts so now will have a number of 300 watt stages going through combiners. I'm working on a deal for some high performance heat sinks to keep things cool. Presently planning to use two modules at reduced voltage to build a bullet proof 160 through 6m linear.

[W1VD]

Back after an extended hiatus ... been busy ...

The current mode Class D information on the W1VD website is a bit out of date ... I'll update with the newer design in the next couple weeks. The most significant difference between the 'old' and 'new' is the output transformer ... which was changed from a 'conventional' primary/secondary transformer to a transmission line transformer and balun. While the original design worked well on 160 and 80 meters, high leakage reactance in the conventional transformer seriously affected performance on 40 meters. Changing to the transmission line transformer / balun cured that problem and excellent efficiency (92+%) on 40 meters was attained. The 160 and 80 meter decks were subsequently retrofitted with transmission line transformers / baluns with a slight improvement in efficiency (95%).

In terms of differences between Class D and E ... the main operational difference is the ability to operate into a range of load impedances.

Class E typically includes as part of the topology an adjustable tank circuit which (if the tuning components have enough range) allows for operation into loads with a VSWR of 2, 3, 4:1 or maybe even more.

Class D typically doesn't have an adjustable tank circuit - just a low pass filter to remove harmonics. This means that the transmitter requires a load fairly close to 50 ohms so that the FETs see the proper drain impedance via the output transformer. Some VSWR (less than ~ 1.5:1) is tolerable but it will be seen as a shift in the drain voltage/current ratio. If one normally uses an antenna tuner as I do, this isn't an inconvenience since the load will then be close to 50 ohms.

Basically the 'tuning duties' are carried out in the antenna tuner with Class D and in tank circuit with Class E. Folks that run Class E and an antenna tuner have even more adjustments to make!



              

[n1ps]

Basically the 'tuning duties' are carried out in the antenna tuner with Class D and in tank circuit with Class E. Folks that run Class E and an antenna tuner have even more adjustments to make!             

That kind of sums it up does it not?  TNX Jay!

Bruce....yes I thought those relays looked like auto types.  I'll look around...maybe Steve got a bunch on a deal.  I try not to buy much from RatShack...they tend to be expensive and not too good.  I'll look at Fairradio....I also rarely use true RF relays....but relays that tend to have flat strips and not wires tend to make good RF types. Also TNX for the advice on adding C for 160.  Those values were what I had considered using.  Yes got UR PM reply...TNX.

On the spreaders...I think I am missing the point on WHY the use of such.  I did use copper spreaders in a linear project for mobile use to dissipate heat. 

TNX

Peter

[KF1Z]

Peter...

Fairradio  relays I use:

https://www.fairradio.com/catalog.php?mode=viewitem&item=6350

Bruce

[n1ps]

TNX Bruce!  I'll order a couple.   Also need DPDT...will look around the FR site.

p

[WA1GFZ]

I bought a bag full of nice relays from Dave Schneider a year or so ago. The price was right. He had a boat load of them.
Transmission line transformers rock. Philips has the all time best application notes on transmission line transformers and power combiners. Single shielded teflon wire of different sizes is a lot cheaper than low Z coax. #24 is very close to 25 ohms. 25 ohms is required for a 1:2 step up to 50 ohms. 2X 50 ohm cables in parallel is you want to run higher power.

[KD3CN]

Steve found some automotive relays at rat-shack that work well. ( I have yet to find any as he describes though)

I believe this is the relay Steve used:

http://www.radioshack.com/product/index.jsp?productId=2062477

73, Karl

[W4NEQ]

...
On the spreaders...I think I am missing the point on WHY the use of such.  I did use copper spreaders in a linear project for mobile use to dissipate heat.  ...

TNX

Peter

My reasons for this approach are that, in order to suppress the likelihood of parasitics, the shunt capacitor must be very high quality, providing a low impedance to ground for higher frequencies and so far I am unable to obtain the ATC100C style magic caps in small quantity (although they are supposed to be getting back to me.)

And the use of a copper heat spreader sandwiching Kapton against the grounded aluminum heat sink makes an exceptionally hi Q capacitor.

Not that thermal issues are expected to be an issue with this design, but my experience has been that bolting the transistor case directly to copper thermal mass without any insulator significantly improves heat transfer - to that piece at least.  The copper then has more surface area to then transfer heat through the "improved thermal" type of .002 Kapton on to the heatsink.  The copper is expensive, I think the Kapton was $19.
It's more complex mechanically, but that's not an issue for me.   I'll see how it works.  Others have had success with it. YMMV.

Chris
 

[ND9B]

Basically the 'tuning duties' are carried out in the antenna tuner with Class D and in tank circuit with Class E. Folks that run Class E and an antenna tuner have even more adjustments to make!
              

Is the adjustable tank circuit an option that just happens to be in the popular E design? Or, is there a technical reason that class E transmitters have to have an adjustable output tank?

I see the K7DYY class D design has a fixed PI filter for each band that doesn't require tuning.

For that matter, I never understood why linear tube outputs always require tuning, and solid state outputs don't. Can anyone explain that?

ND9B

[WA1GFZ]

Tube linear needs the tank to flywheel the single ended drive to a 360 degree sine wave output. SS is push pull analog so there is drive to the output over 360 degrees. I bet SS IMD would improve with a tuned circuit in the output.
Tuning on the output of a class e rig allows user to dial in the waveform over a wide frequency / load range. Single frequency fixed load and output power would not require variable components.