The AM Forum
July 14, 2024, 12:25:12 PM *
Welcome, Guest. Please login or register.

Login with username, password and session length
 
   Home   Help Calendar Links Staff List Gallery Login Register  
Pages: [1] 2   Go Down
  Print  
Author Topic: GOODBYE CLASS "E": WELCOME CLASS D REVERSE "1/D"  (Read 18837 times)
0 Members and 1 Guest are viewing this topic.
WD5JKO
Member

Offline Offline

Posts: 1997


WD5JKO


« on: April 12, 2022, 10:28:20 PM »

Lots of South Americans on 7160 AM...

Looks like some really good circuit development, both tube type, and solid state.

Quite a bit of stuff on Class D Reverse, getting 400Watts out on 40m at somewhere around 95% efficiency, and using RF drive instead of digital drive.

https://lu5hah.blogspot.com/

http://lu6dcs.blogspot.com/

http://lu1agp.blogspot.com/

Amplificador RF Clase D Inversa 40m 450W (CMCD - Current Mode Class D Amplifier 40m):
https://youtu.be/Qs-iScjrtaA

Looks like 40M AM in Argentina is alive and prospering.

Added a spreadsheet, rename with .xlxs at the end instead of pdf.

Use Google Translate to switch to your preferred language..

Jim
Wd5JKO

*  (40.42 KB - downloaded 286 times.)
Logged
K9MB
Member

Offline Offline

Posts: 359


« Reply #1 on: April 23, 2022, 11:04:50 AM »

Hi Jim,
Thanks for sharing the information on South American Class D operation, etc.

I have a couple of questions.

1. This all looks like the transmitters that W1VD developed several years ago called CMCD (Current Mode Class D) as opposed to VMCD (Voltage mode Class D?
It looks like the same design.

CMCD is fascinating in a lot of ways and designs have also managed to cover two bands by switching inductors and capacitors.

The only thing that troubles me about Class D is the necessity for the load to be close to 50 ohms or the design output impedance of choice.
Most just feed the transmitters into a tuner to effect a match to their real world antennas.

The 92+% efficiency of the amplifier is impressive, but the effects of the tuner are not stated.
Cheap tuners may lose between 10 and 20%  and very good ones seldom are better than 90% efficient.
If a 93% efficient transmitter with 400 watts input has an output of 372 watts into a 50 ohm dummy load and is then put through a tuner to match a 100ohm antenna at  a great 90% efficiency, the power into the antenna is 372 x .9 =335watts
335/400 = 84% efficiency for the current mode Class D amplifier.

Are my calculations valid?

If so, then a Class E transmitter which inherently is 90% efficient (360 watts output) would be considerably more efficient than a system using Current Mode Class D.
I now that Class E is more difficult to adjust and requires better FETs with higher voltage ratings, but the idea that CMCD is clearly a much better system is not as convincing for me in a real world application after looking at the full implementation of the design.

Please discuss. I do not know enough to do more than ask questions here.
73, Mike K9MB
Logged
M0VRF
Member

Offline Offline

Posts: 175


« Reply #2 on: April 23, 2022, 03:59:00 PM »

Your right, it's nowt new.

I'm sticking with class E, works for me and covers a good portion of the band without tuning.

Plenty of high voltage FETs around both GaN and SiC.

However good info!

JB.
Logged
K9MB
Member

Offline Offline

Posts: 359


« Reply #3 on: April 24, 2022, 10:43:44 AM »

Your right, it's nowt new.

I'm sticking with class E, works for me and covers a good portion of the band without tuning.

Plenty of high voltage FETs around both GaN and SiC.

However good info!

JB.



Thanks for the comments.
I have a question on the SIC and GaN devices.

The SIC devices seem to have very low input and output capacitance and low reverse transfer capacitance, so they look impressive there, but the heat transfer from junction to case is several times higher on the cheaper versions I have looked at.
The C3M0280090D is a commonly used SIC device by WolfSpeed that is a good case in point.
The gate resistance (Rgs) on this device is 22 ohms, while older silicon devices like the FQA11C90 have Rgs values less than 4 ohms.
The input capacitance is only 220pF, which is 9 times smaller than the 11c90, but the increased Rgs cancels a lot of that advantage-right? It might still be faster to load and unload the gate, however in spite of the much higher series resistance of the gate.

The transfer rate of heat from junction to case is 22 degrees C/watt compared to 0.47 degrees/watt on the old 11c90 FET.
It appears that any heat generated inside the device faces a much higher transfer gradient in this C3M0280090D SIC device compared to the venerable FQA11C90 device.

I found a very nice SIC at Mouser made by LittleFuse that has very low Rgs and good heat transfer while also having low capacitance and high rise times, etc, the
LSIC1MO120E0160

The only downside to this device is the cost is more than $12 each. 11c90 can still be purchased for less than $4 each.

The GaN devices are very impressive, though voltages seem low and also the cost is pretty high on them.

Choosing the best devices for a particular design seem less cut and dried for me, therefore.
I would appreciate your ideas on these tradeoffs and why you advocate the SIC Nd GaN devices, given their own inherent limitations and higher cost?
73, Mike



* 9F9331D6-1413-4BC2-B03B-7CF4762D38A4.jpeg (365.26 KB, 2365x1286 - viewed 351 times.)

* A7BDEB8C-5AFF-43EB-9975-10C55195F7E6.jpeg (553.13 KB, 2397x1696 - viewed 344 times.)

* 31DB847C-53D1-4853-941C-8AFF5D9B0B3A.jpeg (670.04 KB, 2490x1721 - viewed 346 times.)
Logged
M0VRF
Member

Offline Offline

Posts: 175


« Reply #4 on: April 24, 2022, 11:13:30 AM »

I've never bothered to use the Fairchild FETs that most folk seem hung up on. Way too hard to drive on 40M as I use SOIC drivers and need a low Qg. The GaNs from Transphorm are even better and have a pair running on 20M using the same drivers.

I've published my designs on here for quite sometime with next to zero interest as I'm only interested in 100W carrier designs at 48V and this lot want multi KW.

The Cree C3Ms are fine, use the lowest Qg (280090) ones for 40m and 80m + 65090s for 160m.

Stick to the ClassD design that's floating around with tuned O/P on the drivers, I think it's 2 half bridges in series? Or the W1VD design as it's fine.

You won't go far wrong using either.

JB



Logged
K9MB
Member

Offline Offline

Posts: 359


« Reply #5 on: April 24, 2022, 12:48:29 PM »

Thanks for the information.
I will search your posts to save you the trouble of repeating yourself.
No doubt- the 100 watt world is far different than the 1kw world.
Scaling up a fly will not produce a flying titan, snd explains the rather thick legs on elephants.

The Class D you recommend,
I have previously posted on has the 50 ohm limitation mentioned.

You have not addressed my questions regarding the high Rgs and low heat transfer rates in the cheaper SIC devices, either.
Perhaps you discussed it elsewhere?
I will research that.
Thanks again.
73,
MB
Logged
M0VRF
Member

Offline Offline

Posts: 175


« Reply #6 on: April 24, 2022, 05:18:55 PM »

Re the SiC devices. I have nothing to compare them against and the class E/D designs certainly at my power level generate so little heat anyway can be passively cooled.

Everyone in the real/commercial world have been using SiC devices for years, there's no room for obsoleted nostalgic devices!

All my designs are technically class E, i.e. the O/P has the L/C bandpass config. However the B/W is around 400KHz in total so covers the band and the antenna has a less than 2:1 vswr over the same range and have never bothered to use a tuner.

There's a really no need for wide B/W system anyway as most sit on a single frequency or a narrow portion of an A.M. band segment.

JB
Logged
K9MB
Member

Offline Offline

Posts: 359


« Reply #7 on: April 24, 2022, 08:37:11 PM »

Re the SiC devices. I have nothing to compare them against and the class E/D designs certainly at my power level generate so little heat anyway can be passively cooled.

Everyone in the real/commercial world have been using SiC devices for years, there's no room for obsoleted nostalgic devices!

All my designs are technically class E, i.e. the O/P has the L/C bandpass config. However the B/W is around 400KHz in total so covers the band and the antenna has a less than 2:1 vswr over the same range and have never bothered to use a tuner.

There's a really no need for wide B/W system anyway as most sit on a single frequency or a narrow portion of an A.M. band segment.

JB

Ok, I understand that at 100 watts, heat transfer gradient is a minor thing.
I am thinking it is apples and oranges when looking at a rig like I want to build capable of 700 watts carrier output.

When you mention the “real/commercial world, are you talking about RF generation or for inverters and other pulse power applications?
Also, I have never followed “everyone” automatically, so I am immune to the shame I might rightly feel for using “obsoleted nostalgic devices”.
I began my designing career in 1978 and it continues today, if in different fields, but my approach has always been to use older or newer technology, depending on which met the criteria I had set for the project. New is not always better in the key ways that define the best overall solution for a particular design.

I believe that the real reason that FQA11c90 FETs are stiil used is not nostalgia, but practical factors like stability and ruggedness in building large Class E transmitters.
I do not believe that guys like Steve are Luddites that fear change, but are onstead practical engineers that weigh the advantages of different approaches and if no clear path exists, they choose the reliable older solutions until evidence can justify dropping them for a new less tested solution.
Right now, in our country, this kind of thinking would have us all dump our gasoline fueled cars for electric ones and generate all that electricity with wind and solar power. Good luck when the sun goes down and the wind calms…
No engineer worth his salt would ever consider going out on such a thin limb when the consequences of failure loom so dangerously below such a fall. In Radio experimentation, only an inconvenience looms for us after making chuckle headed design choices so we can safely be much more reckless… 😉

I can easily see why GaN and SIC are vastly superior up to at least 200 watts, however, specially when less drive is important, and for higher frequencies like 40 meters or even 20 meters where GaNs become very attractive solutions even if voltages need to be reduced.

Thanks again for your counsel. You obviously have a lot of experience building these FET rigs and experience is the very best kind of knowledge.
73, MB

So are you operating in VMCD or E, then?

I am learning, so not everything is clear.
Logged
M0VRF
Member

Offline Offline

Posts: 175


« Reply #8 on: April 25, 2022, 02:36:29 AM »

Yes, the 'Real' world being voltage conversion generally.

My designs are technically class E but show some D type characteristics.

A dual FET push pull circuit I'm about to publish is able to cover 5-7MHz with realtively flat power, obviously class D and then changing a single 'tuning' inductor changes the B/W to 200KHz, obviously more class E like, most strange.

Design wise, I'd just start experimenting and let us know what you find.

Have a look at the designs from KQ6F..

JB.
Logged
K9MB
Member

Offline Offline

Posts: 359


« Reply #9 on: April 25, 2022, 10:03:16 AM »

Thank you for the run down on your new design. Sounds like you are on to something there. Some call things like that luck, but they probably said that about Edison when his lab tried more than 1000 things before discovering that a carburized bamboo  fiber in a vacuum was by far the best filament material in his vacuum pumped bulbs.
No Monday Morning quarterback even dared to suggest that he knew all along that this setup would be the beginning not only of electric lighting, but later thermionic vacuum tubes.
Activity and intuition often trump logic based on what we “know”. 😉

I have studied KQ6F and W1VD designs.
I find the arguments for current mode Class D to be compelling, and I am thinking to build a 40 meter transmitter using that design mode.
As you say- the 11C90s begin to lose their edge above 75 meters and are very marginal at 40 meters.
Keeping carrier voltage down to 35-40 volts using some of the Transphorm GaN devices in CMCD looks attractive for a 350-400 watt 40 meter transmitter. That seems like a lot of power at 7mHz.

Definitely going to stay with Steve’s 11C90 design for the 160-80 meter transmitter that will run 50 volts at 15 amps for the carrier and should provide almost 700watts carrier output.
Logged
M0VRF
Member

Offline Offline

Posts: 175


« Reply #10 on: April 25, 2022, 02:01:07 PM »

Sounds a good plan. Any old thing will work below 4MHz, even IRF640s!

Yes I'd stick to GaN for 40m, can be driven with 6V and using the 'tuned' I/P design prob fine wi 5V.

Always best to use the highest voltage practicable at the drain as the losses are less obviously. I'm limited to a 24-30V carrier so even generating 100W needs 4 devices.

Good luck on your quest.

JB.
Logged
steve_qix
Contributing
Member
*
Offline Offline

Posts: 2593


Bap!


WWW
« Reply #11 on: May 02, 2022, 08:58:54 AM »

Your right, it's nowt new.

I'm sticking with class E, works for me and covers a good portion of the band without tuning.

Plenty of high voltage FETs around both GaN and SiC.

However good info!

JB.



Thanks for the comments.
I have a question on the SIC and GaN devices.

The SIC devices seem to have very low input and output capacitance and low reverse transfer capacitance, so they look impressive there, but the heat transfer from junction to case is several times higher on the cheaper versions I have looked at.
The C3M0280090D is a commonly used SIC device by WolfSpeed that is a good case in point.
The gate resistance (Rgs) on this device is 22 ohms, while older silicon devices like the FQA11C90 have Rgs values less than 4 ohms.
The input capacitance is only 220pF, which is 9 times smaller than the 11c90, but the increased Rgs cancels a lot of that advantage-right? It might still be faster to load and unload the gate, however in spite of the much higher series resistance of the gate.

The transfer rate of heat from junction to case is 22 degrees C/watt compared to 0.47 degrees/watt on the old 11c90 FET.
It appears that any heat generated inside the device faces a much higher transfer gradient in this C3M0280090D SIC device compared to the venerable FQA11C90 device.

I found a very nice SIC at Mouser made by LittleFuse that has very low Rgs and good heat transfer while also having low capacitance and high rise times, etc, the
LSIC1MO120E0160

The only downside to this device is the cost is more than $12 each. 11c90 can still be purchased for less than $4 each.

The GaN devices are very impressive, though voltages seem low and also the cost is pretty high on them.

Choosing the best devices for a particular design seem less cut and dried for me, therefore.
I would appreciate your ideas on these tradeoffs and why you advocate the SIC Nd GaN devices, given their own inherent limitations and higher cost?
73, Mike



Hi Mike, very good insights on SiC devices !!!  The die heating is why I have not gone further with these devices.  Looking at the SOA curves for any SiC devices I've seen, the standard MOSFET SOA curve (such as the FQA11N90) shows the device can handle nearly 10x the current in an overload situation at high voltage than a similarly spec'ed SiC device before the die melts down.

I have built a couple of 1kW class E RF amplifiers with SiC devices, and they work fine, but the reliability is not there.

The reality is that overloads happen - antennas fall down, feedlines arc over, human error (keying up with no antenna, a shorted antenna, etc.), components fail (or arc over), etc. etc. etc.

I do think that if one de-rates the SiC devices significantly (maybe 1/8 of their published ratings) to leave PLENTY of headroom for overloads and other electrical hazards, that a reliable SiC based transmitter could be had.  However, the cost would be fairly high as compared to using other devices.

I'll admit that my expectations relative to reliability are very high.  I expect anything I build to be able to tolerate normal hazards that can and will occur over the life of the transmitter (which should be many decades), and not ever, ever have to replace a MOSFET, driver or any other such component.  Rare TransZorb failures due to unpredictable external factors are acceptable (such as antenna arcing while transmitting), and I have had several of these over the past 20 years, which is about when I started using class E transmitters on a daily basis.
Logged

High Power, Broadcast Audio and Low Cost?  Check out the class E web site at: http://www.classeradio.org
K9MB
Member

Offline Offline

Posts: 359


« Reply #12 on: May 03, 2022, 11:16:02 AM »

Thanks, Steve. I have always been skeptical of new devices that are untested by time and experience in specific applications.
I respect your own approach to over spec everything after carefully looking at “what could go wrong? - and the consequences expected from likely faults that can occur.

I have for years poured over data sheets looking for devices that had all the benefits of the ruggedness and heat transfer capability of the FQA11N90 and also low input capacitance, low reverse transfer capacitance low gate resistance and fast rise and fall times and have found many that have some things that are better, but none seem to have that obtuse L SOA curve that defines the the available current stability under wide ranges of Drain voltage and temperature.



Almost all devices are SOA plotted only for 25 degrees C and that only happens when the devices are still in tubes on the shelf. I found one company that plotted at 25C and 85 C and the plots were vastly different so that peak voltages and currents per device were sharply reduced at 85degrees C.
This magnifies the heat transfer spec that defines degree rise per watt tramsferred from tje chip to the case. The 11N90 has a 0.45degree/watt and most of the populat SICs are more like 2+ degrees/watt.

It is a complicated thing, however, because the input capacitance and gate resistance define the amount of energy needed to drive the device and can anybody guess what becomes of that drive energy?
The higher the frequency, the more times per second the gate capacitor must be charged and discharged through that gate resistor, so chip heating will go up rapidly with clocking speed.
Also- rise time defines the amount of time when the device is linear and acting as a resistor- not as a pure switch.

The 11N90 is slow by modern standards and the 2500pF input C make a lot of coulombs required for servicing those gates. Fairchild did not spec the gate resistance, though I expect it to be in the 3-4 ohm range, so there is some heat produced. What it dies well, as you say is dissipate that heat effectively.

I found one device by LittleFuse that has a lot of good characteristics and a comparatively good SOA curve compared to otjer devices I have seen. They cost $11 each in 10s, which is not bad. They may be worth building a 40 meter current mode class D amplifier with. Maybe two blocks of 4 each with each device having a driver might work. The gate resistance is less than one ohm and input C is 800pF, so less heat in driving it at 7mhz. Might be worth trying it out….No killer devices there, but this SIC might be a good choice at 40 meters- at 160-80 meters, tue Fqa11c90 is still better and cheaper…

LittleFuse  LSIC1MO120E0160

https://www.mouser.com/datasheet/2/240/Littelfuse_Power_Semiconductor_Silicon_Carbide_LSI-1308072.pdf


Logged
N9NEO
Member

Offline Offline

Posts: 282


« Reply #13 on: June 01, 2022, 07:57:05 PM »

Class D reverse looks an awful lot like class F inverse.
Logged

I'd rather be anatomically correct than politically correct!
WD5JKO
Member

Offline Offline

Posts: 1997


WD5JKO


« Reply #14 on: June 02, 2022, 11:14:25 AM »


LU5HAH has updated his website with some new stuff..

https://lu5hah.blogspot.com/

Google translate can be used to switch the language to your native language.

Jim
Wd5JKO
Logged
KJ7AGO
Member

Offline Offline

Posts: 7


« Reply #15 on: July 18, 2022, 01:37:38 AM »

Thanks for those links.
I've been experimenting with CMCD amps on 20m for a while now.  
Basically a copy of W1VD's boards but with GanFETs.

Latest build was with six Chinese Inn650D02's.  Very similar to GS065011 device but a lot cheaper.

14.1 Mhz, 285W RF 90%
14.1 Mhz, 71V, 350w RF, 76% eff , 90C fet temp
13Mhz, 450W RF, 87%

Efficiency was derived from multi-run calorimetry method.
The NCP81074A drivers seem to top out at a little over 15Mhz.
The GanFET devices are tiny and the board is small (70mmx100mm).
The heat sink is a 1/8" sheet of copper that protrudes through the PCB to the bottom of each GanFET where it is soldered.  A copper heat spreader covers the bottom of the PCB.  


Sorry for the bad pics.
KJ7AGO

Some links:
https://www.ee.columbia.edu/~harish/uploads/2/6/9/2/26925901/c15.pdf
http://educypedia.karadimov.info/library/along_2003.pdf


* sixgan.jpg (175.25 KB, 1600x1200 - viewed 379 times.)

* six gan heat sink.jpg (154.27 KB, 1600x1200 - viewed 352 times.)

* Modular GaN INN650D02.jpg (95.9 KB, 1358x625 - viewed 375 times.)
Logged
M0VRF
Member

Offline Offline

Posts: 175


« Reply #16 on: July 18, 2022, 05:05:47 AM »

Great design and good to see a PCB!

Like me you don't leave enough room for the chunky stuff!

Those flatpack GaNs help with the neat layout, I've used TPHs from Transphorm and being TO220 have to be bolted thru a cutout.

I thought of mounting the flatpacks under the PCB and mounting them sat in a milled section in the heatsink.

I'm inspired!

Good work!

Best Regards

JohnB.
Logged
K9MB
Member

Offline Offline

Posts: 359


« Reply #17 on: July 18, 2022, 11:01:08 AM »

That is a beautiful circuit board layout and the output is impressive at 20 meters.
Great demonstration of where GaN really shines and SMT shrinks.
Now if someone could formulate a ferrite with hyper flux density capability and temperature stability and low Rs in an FT50 size, we can really miniaturize these high power reverse Class D transmitters.
While we are at it, why not a fusion reactor power supply the size of a brick to round it out?😉
  Ok, some things must wait, but clearly- you are demonstrating that we have come a long way from big hollow state systems of the past, for RF generation and it is indeed very exciting
Very interesting work that I hope you will share more details on here.
73, Mike K9MB
Logged
KJ7AGO
Member

Offline Offline

Posts: 7


« Reply #18 on: July 18, 2022, 11:19:44 AM »

Thanks, yeah, I chronically make the board too small:)

The DFN package devices are cheap, have very low gate capacitance numbers, and can be hard to work with (but aren't really).
They make switching at 15Mhz pretty easy though.  
It ends up being more about the drivers.
And your clock symmetry and duty cycle become way more important than lower frequencies.
Gotta watch the junction temp, as Rds(on) goes up rapidly with temp.
Using the medium voltage GanFETs gives you a nice margin for error but they don't have the super low Rds(on) that the lower voltage devices have...but you can parallel them easily.
They have been remarkable rugged in development.

Chuck KJ7AGO  


* boardbottom.jpg (39.72 KB, 1358x625 - viewed 293 times.)
Logged
KJ7AGO
Member

Offline Offline

Posts: 7


« Reply #19 on: July 18, 2022, 04:06:07 PM »

Brain dump on 20m GanFET amp.

I'm an RF noob, so nothing that follows should be taken as gospel.  Correct me if I'm wrong please.  Danger; amp builds can be addicting (photo). The  CMCD was easier to build and tune (for me) than the class E's I built.  Something about the intentional miss-tune of resonator and clock in class-E messed with my head.  The device utilization factor benefits of CMCD are real which translates to power density.  ZVS is very easy to conceive and achieve with CMCD (for me).  It doesn't run away, it tolerates open circuit and high Z OK, runs at lower peak voltage.  Pretty friendly circuit for beginners like me.

SMT/QFN packages:
Working with SMT FETS with hot air is easy once you trust that they will float into position and the solder really is under there somewhere:)  I use a low temp solder to attach the heat sink and it has been rugged and relatively easy to take on and off.  I'll try to find a picture of the heat sink.

Heat Sink:
The bottom cooled QFN devices are really meant for a stitched PCB and sink on the opposite side of the board, probably in a throw away context.  I wanted real copper heat sinking so I've made holes in the PCB and machined the copper to fit up through the board flush with the top side.  A low temp solder lets you attach and detach the copper without disturbing the GanFETs.  The heat sink is common  source which is convenient.  See attached photo (older 4 fet design but same concept);

Tank:
Make sure you are solid on your tank design concepts or you will have hot caps or warm inductor, either way your efficiency will suffer and you won't know why:)  Ask me how I know.
Good primer: https://www.electronics-tutorials.ws/accircuits/parallel-resonance.html
I think I have a loaded Q of 14 on the last board.  You have to control the circulating current. Low ESR components are critical as frequency goes up.
My son made a graphing calculator with all the formulas and variables so you can watch the anticipated losses of the tank as you design.  Calculator output maps to real life fairly closely. Take a look (I'll rope him in if you have any questions about it):
https://www.desmos.com/calculator/gonkn8ylun



Tuning:
Get your components close but shoot for slightly low in frequency.  VNA tests with the parts installed is ideal. Spread the inductor to fine tune.  There is no way to avoid some trial and error.  Tune for best wave forms (not minimum current).   An efficiency meter like WA1QIX had is really useful for tuning.  

Balun:
The t-line balun has been trouble free and easy to design and make.  Make sure you have enough choking inductance and your mix is appropriate for your band.  I'm using 61.  If you were shooting for ultimate efficiency I think you would probably use powdered iron (as JohnB , Nigel, and others have demonstrated) .  The t-line characteristic impedance is chosen to be the median of  load and amp... so 25 Ohm coax for my board.  Magnetics has been a huge learning experience for me.  RF magnetics over simplified; the more deeply you activate magnetic domains the more heat you will generate in the core... read the cross section flux density charts.
If you have a bad load match the heat will appear in the balun as well as the GanFETs/heat sink.

Clock circuits:
The Johnson/ring counter front end circuit has been reliable.  My duty cycle circuit is simple and works by feeding an RC slope to the gate drivers, adjustable with a pot.  Problem is, the Schmitt triggers on the gate IC's are not identical so they end up switching at slightly different times, which changes with both frequency and temperature. I'm going to modify this part of the circuit so the RC ramp is fed to an inverter so the gate drivers all see the same square wave.

Gate drive:
The NCP81074A drivers are a great little IC.  They work  up to 15Mhz where they then fade away and get really hot.  The ability to use different pull up and pull down resistors is a useful feature at this frequency.  I use 4R up, 2R down.  Gate ringing is real and  illustrates the inherent speed of the GanFET as it triple switches on a ringing gate.  I tried to use cheap Chinese resistors here and it was a big mistake.  They literally burned up, but only after causing hard to find anomalies. Good quality, low ESR all the way...
Side note:  anyone try to source the  NCP81074A the last two years?  Ouch.

Chuck KJ7AGO


* heatsinkfingers.jpg (3114.18 KB, 4032x3024 - viewed 320 times.)

* heatsink1.jpg (3969.15 KB, 4032x3024 - viewed 355 times.)

* danger.jpg (3795.25 KB, 4032x3024 - viewed 347 times.)
Logged
K9MB
Member

Offline Offline

Posts: 359


« Reply #20 on: July 18, 2022, 08:54:01 PM »

Steve used a 74AC04 IC for his one shot pulse width generator to get to tue 40% he wanted.
Is this what you meant by using “inverters” instead if trying to use the schmitt trigger in the driver?
Seems like the 74AC04 or equivalent would be better matched, plus the lower heat levels would not be a factor like in the driver chips…
I guess porcelain NP0 caps and high stability resistors are needed too?
At 14Mhz, not a lot of time to not be sure…😉

https://www.mouser.com/datasheet/2/308/74ACT04_D-2309712.pdf

https://www.onsemi.com/pdf/datasheet/nc7nz14-d.pdf

Logged
KJ7AGO
Member

Offline Offline

Posts: 7


« Reply #21 on: July 19, 2022, 02:10:01 AM »

Duty cycle adjust circuit:
See pic..
The pot is used as a variable resistor, together with the cap, to change the slope of the rising clock signal.  Somewhere on the ramp up the Schmitt trigger fires.  The diode makes the falling edge square.  Notice in the pic there is an inverter after the RC circuit.  I've been feeding the rounded clock signal to the driver bank.  Slight variations in manufacturing, temperature, distance, parasitics, etc, means each driver fires on a slightly different part of the slope... much better to square it up before it hits the drivers.

For tank caps, ATC 100b have been the best.  I've also used silver mica but their measured ESR has been worse than the NPO ceramics.
Low ESR Rf rated NPO ceramics for coupling caps and filters.  Parallel as much as you dare.
Every time I try to economize or cut corners on the caps I end up regretting it.

Chuck KJ7AGO

Resistors have been much more tolerant, but I learned to use name brand parts.


* duty cycle.png (11.42 KB, 680x481 - viewed 373 times.)
Logged
KJ7AGO
Member

Offline Offline

Posts: 7


« Reply #22 on: July 19, 2022, 02:14:30 AM »

@ JB  MOVRF,

"I've published my designs on here for quite sometime with next to zero interest as I'm only interested in 100W carrier designs at 48V and this lot want multi KW."

Not true my friend, I've followed your designs with great interest:)

Chuck KJ7AGO
Logged
M0VRF
Member

Offline Offline

Posts: 175


« Reply #23 on: July 19, 2022, 11:25:02 AM »

Oh, that's nice!

Will probably do the same with 4 flatpacked GaNs for low voltage (48V) 100W design.

Re duty cycle adjustment if your drive your FET driver with a sine wave, you can go from 50% down when driven with less volts. So wi NCP81074A's from 3V to 5V p-p will give 40-50%.

I've seen some incredible complex circuits in here to achieve the same but you don't have to use any components at all!
Logged
KJ7AGO
Member

Offline Offline

Posts: 7


« Reply #24 on: July 19, 2022, 03:00:01 PM »

@MOVRF
Sine wave duty cycle adjust:
Clever!  Slope is built in. (stops to imagine the front end circuit..)
Do you use the sine wave for anti-phase too?  Positive, negative?

48V amp with QFN GanFETs:
I started by trying to build 12V boards for convenience.  It was much harder to design for any kind of power.
Then I went to 48V since it's considered the threshold for 'low voltage' around here.  Power supplies are easy to source.  It is more practical than 60V.
But efficiency (and power) goes up with voltage so it crept up.  Plus the GanFETs available push you that way.
The really low Rds(on) devices are all sub 100V which doesn't give you enough headroom.
There are some 200V die package only, but their gate capacitance is comparatively huge.
If you focus on the C(iss) for higher frequency, you end up back at higher voltage devices with higher Rds(on).
Running the same 300W six Gan board at 160V give truly insane power density (haven't tried it yet, have to change all the caps)

The Chinese INN650D02 are $1.50 right now (but somewhat obscure)
The GS065011 are less than $5
New players almost every six months it seems.

I've got old four GS065011  5x6 package boards with NCP81074A drivers, I'd love to send if you want to give these tiny things a go (see pics).  The more testing and experience the better..

Chuck Kj7AGO


* 4gan2.png (4624.81 KB, 2024x1187 - viewed 373 times.)

* 4gan3d.jpg (77.12 KB, 1358x625 - viewed 313 times.)

* 4gan1.png (3113.07 KB, 1547x1076 - viewed 336 times.)
Logged
Pages: [1] 2   Go Up
  Print  
 
Jump to:  

AMfone - Dedicated to Amplitude Modulation on the Amateur Radio Bands
 AMfone © 2001-2015
Powered by SMF 1.1.21 | SMF © 2015, Simple Machines
Page created in 0.071 seconds with 19 queries.