Class D

I am back after a little hiatus to give a quick update on the transmitter. The new V2 transmitter board has proven to be flawed after some testing. It has been tested and shown that the long clock traces are rounding the waveforms a bit too much. I am going to update this post with the new V2.1 which will correct the issue.

In addition to the clock trace problem, the new NCP 8-SOIC surface mount drivers are a bit more demanding in their heat dissipation requirements than I thought. The new V2.1 board will require some small heastinks on the drivers to prevent overheating. The drive voltage has also been decreased to 15V instead of 18V because it has been proven to be just as efficient operating at 15V and the drivers get a little less hot.

So the changes in V2.1 will be:

  • Shorter clock traces
  • Heatsinks required on drivers
  • Voltage changed to 13.7V or 15V on the drivers
  • Inverter chip removed (integrated inverted inputs on drivers used instead)

Stay tuned for an update. And please post a comment if you have anything to offer, I’d like to get an idea if anyone is still reading this blog. I will continue to update it regardless, however.


I am sending the V2.1 to fabrication and they should be done in a few weeks; I will then begin the testing. I believe they will work much better. I still think there may be an efficiency problem with the output match. I might need to change to using coax in the balun instead of wire. I am not sure how this is done though, so I will do some research.

Here is a photo of the V2.1 board:

The latest version of the transmitter board has just finished production and will be arriving soon.

The new version has several improvements over the last. It is now using the NCP drivers and using one of those per MOSFET. The layout is a bit different as well. In order to reduce trace length as much as possible, the MOSFETs are now mounted below the board at a right angle. The new heatsink will be a square tube AKA cooling aggregate. I have not completely decided on this part yet, but fischerelektronik

. . .has some very nice heatsinks for this purpose. This will go right under the center of the board and the MOSFETs will mount upside down onto it.

I will also be using a PWM to modulate from now on. RF2017 (eBay) from Greece has now supplied me with an updated PWM module that is capable to modulate 500W carrier.

I have collected some more datasheets and created an excel spreadsheet for the parts list (which used to be in a text file and was limited). Sorry, the new PCB file has been removed from the archive. If you would like the PCB, please contact me or Tech Ingredients to order. You can still find the old PCB file and Gerber files in the archive if you want to send to fab yourself. You may modify the file using Sprint Layout.

You can download the updated archive here:

You may notice that there are some datasheets in the archive containing shortwave receiver ICs. I am also building a receiver around the KT0915, but that project is on the back burner. The chip is interesting though because it can receive all the way from 150KHz to 110MHz with FM or AM demodulation and no gaps. Making this project into a transceiver may be possible if there is some type of RF relay, but I know very little on the subject and will have to do more research. Either way, the chip has the best sensitivity out of all the SW receiver all-in-one chips I could find. I want to keep this simple, so I do not plan on constructing a receiver from scratch. The chip is very easy to use. You just need a microcontroller (like an Arduino or PIC) and the chip itself. The chip will output a line level signal with no external components. A PCB is in order and will be experimented with when my main project is complete.

OK, so the project is finally completed and I’ve done some tests. Of course, this project will never be 100% complete. The transmitter does work and at a high efficiency of 95%, but only below 5MHz. It should work great for the 80 meter and 160 meter ham bands, but my goal of 40 meters (7MHz) has not yet been met.

I need to do more work to determine the weakest link. It is likely that it has to do with the drivers. I need to take some measurements with a scope to determine the problem. I will keep this post updated with the changes I make to get it working on 40m.

So far, there is one notable change I made that differs from the schematic I posted in my earlier post. I ended up needing 18V into the drivers rather than 8V. So I upgraded the power supply to 24V and used a 13.7V reg for the control board and then the two 18V regs for the drivers.

My efficiency is now as follows:

FrequencyVoltageAmpsRF WattsHeat WattsEfficiency

Update 02/27/19

I’ve done more testing and have concluded that it is not possible to do more than 5 MHz (5.5 MHz tops) with this PCB. The TC4452 drivers are just not going to do it and I can’t change them out for something better as the ‘better’ ones use a different pinout. Looking on a scope I can see the rise time on the DDS module to be about 7ns regardless of frequency (3 – 8 MHz). However, looking at the output of the TC4452(s), it is about 20ns rise and 25ns fall time. I think this is too large for it to be efficient at 6 to 8 MHz. Rise time is critical for these types of amplifiers because the MOSFETs operate at the highest efficiency when they are fully on or off. If they are part way on (during the rise time), they generate a lot more heat.

Another thing I tested to try to improve the efficiency was changing the duty cycle to 40%/60% rather than the normal 50%/50%. This apparently reduces the chances of cross conduction (the time when both sets of MOSFETs are on at the same time). We don’t want both sides on at the same time because this is a push pull amplifier. It will result in poor efficiency and possibly popping a FET if there is too much cross conduction. After changing the duty cycle to 40/60 (ie 40% on time and 60% off time for each side respectively), the efficiency of the amp went up to 80% from the original 74% shown above. That is not nearly enough of a change to call it ‘working’ at 7 MHz though.

In order to get good efficiency at 7 – 8 MHz, I am going to have to redesign the board. Two alternative drivers that should work better at these frequencies are the IXDD614 and the NCP81074A. In addition to changing the drivers, I will use one driver per FET. The NCP only comes in a surface mount package, so I am a bit concerned with heatsinking, but I have seen this one used in a picture. I think it will work well.

So stay tuned. It will probably be a while until the next update here because I have to design and order the new PCB, but I’ll be sure to update on how that one is working at the higher frequencies.

OK, here is part 2 – or part of it anyway. I am still currently constructing the amplifier, but I will upload the video on Tech Ingredients very soon. I will continue to edit this post with more details as the project progresses.

So let me start out by showing you the schematic (right click and view to see the full size image):

You can see that I am using four MOSFETs and four drivers here. In my actual build, I decided to use four drivers and eight MOSFETs. It doesn’t matter however because the fudnamentals are just cloned to the extra devices. Another discrepancy is that I am using two turns on the primary of the balun and one turn on the secondary. The balun core consists of four FB-61-1020 glued together with silicone. I also purchased four FB-43-1020. I’ve gotten mixed answers in my research regarding which one will work better so I got both to try. In addition, I also got some c2m0280120d MOSFETs in addition to the ones shown in the schematic because they were recommended, but I think the ones I chose in the schematic should be better (they have much higher current and lower RDS on).

Here is the PCB:

Just to note, all of this and more can be found in the archive download I listed in the part 1 post.


So what we need to make this working is:


1. Audio amplifier ––320-3346

2. Heatsink –

3. A fan – I am using a 12V crossflow fan that can be found on Amazon or eBay.

4. 1:1 toroidal transformer (1000VA is what I am using, but it’s probably overkill) –

5. PCB (I have a few in stock, but have gerber files too)

6. Electronic components – there is a list in the archive described in part 1.

7. The ferrite cores (described above) –

8. MOSFETs (included in parts file)

9. Low pass filter – got it from rf2017 on eBay.

10. RF DDS generator + LCD control board – from rf2017 on eBay.

Send rf2017 a message. He was the one who helped me out with this design. He has his own designs also. His are all pre-built, so if you are looking for something similar to this already made, then check him out. None of his designs are as high power as this (as of now). He also has a Youtube channel you should check out here. I based my design off his amps, so he deserves some credit.

Another site that was helpful during this build was:  It is a class E transmitter instead, but there is a lot of info there that is helpful anyway.

Alright, I’ll continue to update this as my build progresses. . .

In my first radio project here we are going to be building a class D AM transmitter. Ever since I learned that an individual was capable of building a radio station and that all they needed was a transmitter, I got addicted. I was about 13 years old when I figured out that all you need to make an AM transmitter was a crystal oscillator and an audio transformer. After having success with that first project, I was left excited, but wanting more. The range of the thing was only a couple feet. I wanted to build an amplifier, but I didn’t know how. At this point I didn’t even know there was a difference between an audio amplifier and an RF amplifier, and of course, I also didn’t know the classes of amplifiers. I tried a couple things, but ultimately gave up due to lack of knowledge.

So here I am today. I am writing this article for someone like my former self. You want to build an AM radio transmitter, so you look online and find a bunch of 0.1 watt part 15 devices. AM radio is quickly becoming an antique and letting this weird realm of waves fade away as the HF ham operators die off is not a good thing.

I am bringing it back here. In this article we will build a 1000 watts AM class D transmitter. I have chosen the class D design as opposed to the more efficient class E because it has a very wide bandwidth. The transmitter we will build will have the capability of covering 3.5 – 8 MHz and 0.5 – 1.7 MHz (AM broadcast band) with a few modifications. I would really recommend the HF version over the broadcast band one just because you will have a very hard time building an efficient antenna for the broadcast band. An AM broadcast dipole antenna has to be 290 ft at a minimum in length and 145 ft off the ground. A vertical ground plane antenna has to be 145 ft tall, but a loading coil can be used at the base to increase the “length”. You can see the size problem here. The horizontal axis is wavelength above ground and the vertical is signal strength.

(BTW: This image comes from VOACAP. It is a very nice site for calculating propagation. See here.)

An antenna tuner can be used to tweak a smaller antenna, but you really can’t tune something less than half the size without a considerable drop in efficiency. This is especially important at these power levels where we will be producing a lot of heat if the antenna is not tuned properly. In addition to the size limitation, the range of an AM broadcast transmission will be much less than an HF transmission at equal power levels. This is because AM radio waves bounce off a different layer of the ionosphere. The lower D layer will not bounce signals as far. In the diagram below, you can see the day and night propagation for AM. For HF it’s as if it was always at night.

Awwwl thingggs considuuurd, you will have a lot more fun on the HF bands than on the medium wave (AM broadcast) band.

So here in the first post (part 1), this is what I will give you. Here is an archive containing some PCB files and everything else you’ll need to build this transmitter. Gerber format is included (this is what can be sent to a PCB fab). I have also included the original .lay file which is editable in Sprint-Layout. If you do not have sprint layout, I highly recommend. It is a very easy way to produce PCBs without having to learn CAD. And learning CAD is not something you want to do unless you are doing it professionally.

The archive also contains a parts list (8_FET_Parts.txt) with links to purchase. If anyone wants a jump start with this project, I can send you a PCB. All you’ll need then are the components, a heatsink, an audio amplifier, a toroidal power transformer and a signal source.

The transmitter we are building is a high level modulated solid state class D transmitter with SiC FETs. So what does this mean exactly?

This means we are taking a normal audio amplifier (like your car stereo) and hooking it up to one winding of a toroidal power transformer (like what is used in a linear power supply). The other winding then has the DC voltage fed through it and into the MOSFETs. This is how we accomplish the modulation. To generate the actual RF signal, we feed two square waves that are 100% out of phase into the input of the FET drivers (TC4452). This creates a push pull effect. Half of the MOSFETs are on while half are off and vice versa. The FETs we are using are C2M0160120D. These are the best SiC FETs we could find. These are the type of transistor that is normally used in switching supplies or audio amplifiers. These are not RF devices. This is another advantage of the HF band. You do not need to use costly RF transistors at these frequencies. 8MHz is really pushing the frequency high for these types of transistors and that’s why we chose to use SiC FETs here rather than the standard FETs that you see many hams using (like the FQA11N90). GaN FETs are the fastest switching non radio frequency transistors, but they are pretty expensive. The Cree SiC FETs are the best bang for buck and fastest non-RF devices.

Here is a basic diagram of what we are trying to accomplish:

OK, that’s enough for now. I’ll give a detailed procedure for the construction of the device in part 2.