Simplest Tx

[Stretchyman]

OK, seeing as my 'Simple SiC TX' post has been somewhat led astray, I thought I'd start a new post..

Having fiddled and tweaked most designs over the years I've found nothing is quite like the LuLu.

I still see posts showing other designs (bless you DM!) but there's nothing to compare to the LuLu.

If you want simplicity and efficiency there's nothing to touch it.

I've done little more than change some of the components to modern variants, the design is essentially the same.

I must apologise if I dont have the same ethics as some determined to use 'Junk box' parts that no-one other than the author posseses!

I will use the latest and best components out there, they're readily available from ALL the major component suppliers, cheap too!

As with anything RF, you cant just put it together and expect it to work, no matter how simple it is, the main 'problem' being one of the 'layout' of the components.

ANY RF design will NOT work on 'Veroboard' and the simplest method is to build above a copper groundplane, 'Ugly style'.

However the best way is to produce a PCB which simplifies the whole process and makes it repeatable.

OSHPARK is highly recommended for this as their quality and turnaround is good.

Learning to use a layout program will take some time, I use EasyPC and have for years as they're a local company.



PCB Layout




Schematic.



I'm no longer using IRF devices as they're too many clones around (of the IRF510) which either dont work and with a Vds of 100V you can only use 14V max and therefore limited to 10W carrier power.

With the SiC and Gan devices you can use 200+V so there's literally no limit to the output power of a single device.
With 14V you'll be able to acheive a 20W carrier and everytime you double the voltage you quadruple the power.
However this is when the Class E is tuned for maximum power and efficiency is fairly low (around 80%).
For maximum efficiency a 5W carrier (@14V) is about right. So at 28V you will acheive a 20W carrier and >95% efficiency.
So with 60V or so you will get 100W with the same high efficiency, clearly this is the way to go if you have lots of volts and want high power.

Whilst the SiC devices are decent the GaN devices (if a little pricey) are amazing and can be driven with <8V on the gates and the same design will go to above 14MHz.




To finish off we need to modulate the RF generator.



There seems to be some 'Mystique' around how this is done and quite simply there are 2 methods.

1, The 'Modulation Transformer' method.

The humblest of modulation methods and fine as long as powers are kept low (<50W carrier) otherwise the 'Mod Tranny' gets a bit silly, size wise.

These so much misinformation about modulation transformers and which one to use etc and the simpest answer is, build your own.

There's 1000's of transformer manufacturers worldwide who stock standard bobbins and laminations for making transformers.

Find one who stocks 'E160' bobbins and lams and use those.
The Mod tran needs to be wound with (approx) a 1:2 ratio, I wont go into the reason for that here but trust me it's right as I've wound dozens and experimented to find the best one too.

I now get them manufactured by the same company I purchased the parts from, a lot less hassle and a decent price too.

So using an audio amplifier, I'm not going to recommend any other than to say for $10 the TPA3116D2 designs all over the web from CH are fine and will push 100W in PBTL mode are class D and very efficient.

Drive the primary of the transformer with the amp (in PBTL aka mono mode), just like it would drive a speaker, the secondary of the mod tran goes between the modulating voltage and the RF generator, this enables the voltage to get waggled up and down with the audio.

Decent BASS response is an issue here as unless the modulation transformer is absolutely massive you will have to limit the low frequencies to 300Hz otherwise the MT will saturate, and genearate lots of heat and distortion.
Limiting yourself to a 20W carrier with the E160 size recommended you'll be fine to 100Hz however.

2, The 'PWM' method.

A somewhat more efficient method that has one 'minor' drawback.

However even the efficiency of this method is a bit of a mute point if youre using a class D amp as the modulator in method '1'!!

Apart from the 'Problem' of having to design/build your own PWM modulator, beause the PWM is in series with the supply and will ONLY allow the voltage to shift between zero and Vcc, you will have to run the carrier at half the supply, you don't get a voltage 'Swing' as you do with a Mod tranny. So if youre limited to 14V then the carrier wil run at half of this 7V and the power ONLY a quarter, so 5W in this design.

Clearly for low voltage systems this is NOT the way to go, but for high power systems with lots of volts it's fine and has no limits regarding BASS response etc and is far more linear & highly efficent, being class D.

There's a little a little trick you can use here and as long as you're happy with a 5W carrier (only having a 14V supply) or have a 24-28V supply, you can directly PWM Modulate the RF by using one of those TPA3116D2 amps run in PBTL mode. Simply connect any of the speaker O/P's directly to the RF, it will have half the supply present and waggle up and down with the audio. Simple eh!

See; https://www.youtube.com/watch?v=lMMSw0kFHvw

For DEMO.

Please if I may ask this thread is kept with simplicity in mind.

Regards

Stretchy.

[Echoqueen]

Hi!

That is just awesome. Excellent post and video!

Is Output Z=50R?

Thanks
EQ

[Stretchyman]

Yes!

All RF things are 50R, they don't have to be, just convention as most cable, connectors etc are 50R too!

Please let me know if you have anymore questions.

I'll post up some 'scope images of likely good (and bad!) waveforms as you optimise the tuning with a scope initially.

Str.

[Charlie_Dont_Surf]

I realize that the title is "Simplest TX", however, this is a bit too simple for me. Here are a few things that concern me upon initial inspection.

No gate resistors in the circuit

Widely agreed upon switch-mode principles would say that small-value resistors in series with the gate are recommended. This is for two main reasons:

   a) To de-Q (reduce the Q) of the gate circuit and reduce the severity of any ringing.
   b) Protect the transistor gates and the gate driver IC in the event of a fault.

Transform, the maker of the GaN device in the schematic, strongly recommend gate resistors in their datasheet and application notes. According to the datasheet of the FET driver in this schematic, the On Semi NCP81074A, the series resistance of its outputs is ~0.5 Ohms. This is insufficient. Consider adding 10 to 30 Ohms in series.

You may consider adding ferrite beads in series with the gates as well to reduce transients and keep RF out of the FET driver. Transform recommends these.

While I am mentioning protection, this has nothing to do with the gates but it's also worth saying that fuses and MOVs are not shown in this schematic and are a very good idea too.

https://www.transphormusa.com/en/document/datasheet-tp90h180ps-900v-gan-fet/
https://www.transphormusa.com/en/document/recommended-external-circuitry-transphorm-gan-fets/


Power supply sequencing is important for GaN

I don't know about the SiC device shown in the schematic but if you are using the GaN device, you must pay attention to power supply sequencing at turn on and turn off. If you do not follow the correct steps, the transistor may work initially but there is a very good chance that it will fail (destruct) soon enough. That time period between first use and destruction will vary with the drain to source voltage, the gate to source voltage, the temperature, etc. It could be microseconds later or it could be hours later, based upon my experience.

The basic turn-on sequence for GaN is as follows:
   - Apply a gate voltage well below threshold (even a negative voltage is a good idea) to bias the GaN device.
   - Apply the desired drain-to-source voltage.
   - Increase the gate voltage to bring to the desired operating point.
   - Turn on the RF.
   
The basic turn-off sequence is exactly the reverse order. The sequencing can be implemented manually or with one of the many bias sequencing semiconductors on the market. In case you think I am pulling your leg and think that this procedure is unnecessary, here's the same information from Qorvo:

https://www.qorvo.com/design-hub/blog/how-to-bias-gan-transistors-without-damaging-the-device-video-tutorial


EDIT 4/6 - I looked at this a bit more. This is not a typical GaN device, rather, a hybrid of a typical enhancement-mode GaN and a traditional enhancement-mode MOSFET. This means that turn-on and turn-off sequencing are not like regular GaN. I communicated with Transform and they recommend first applying gate voltage then drain voltage, so not completely unlike what I wrote above. Thus, regardless of composition, the concern over sequencing remains.


The circuit in the schematic does (says) nothing about proper sequencing. It is either left to chance or to the user to implement. 

Things to consider:
   - How fast does the +5 V regulator and the Cardinal CPP oscillator come up to full voltage? It might be faster than the modulated drain voltage. If so, that is undesirable.
   - Of note here, the NCP81074A datasheet lists a UVLO period of 10 microseconds. Then what happens to its output if the Cardinal isn't ready with its output when the UVLO period has expired? It depends upon what the Cardinal does.
   - If the Cardinal and the NCP81074A are delivering a voltage well above the FET threshold voltage while Vds is still coming up to final value, that is also undesirable.
   - How fast do all the supply voltages come to final value and do they do so at their same individual rates under all conditions?

If you plan on making one of these, I recommend checking things out with a storage oscilloscope so that you can capture the turn-on and turn off effects, make sure that the sequencing is correct and check for ringing without having to do a repetitive waveform. Turning things on from "cold" is not always the same as from "hot".

Schematic.

[Stretchyman]

I've made hundreds of these and they work fine. Voltages and drive come up in sequence and there's no need for a gate Res as it just slows down the drive waveform. I use SiC in this design.

Regards.

Stretchy.

[Charlie_Dont_Surf]

I've made hundreds of these and they work fine. Voltages and drive come up in sequence and there's no need for a gate Res as it just slows down the drive waveform. I use SiC in this design.

Regards.

Stretchy.

OK.

Your write up calls this a Class-E design but your schematic indicates to "tune the LPF to the second harmonic". I have not simulated your LPF to check the impedances so forgive the question:

Do you mean to tune the LPF to optimize rejection of 2fo and pass 2fo or do you mean tune the filter to optimize passing 2fo and rejection of fo (which is not really a textbook Class-E amplifier)?

[Stretchyman]

You can't pass and reject at the same freq!

The LPF had nothing to do with class E but as with any single ended design the 2nd harmonic is a little high so adding the cap and adjusting the inductance it can be notched out.

I don't bother with 'Simulation' as I use a network analyser to make sure it Really works!

Regards.

Stretchy.

[Charlie_Dont_Surf]

You can't pass and reject at the same freq!

That should have been "Do you mean to tune the LPF to optimize rejection of 2fo and pass fo or do you mean tune the filter to optimize passing 2fo and rejection of fo (which is not really a textbook Class-E amplifier)?"

Never mind. I figured it out myself.

The LPF had nothing to do with class E but ...

Uhhh, no. Class E has to have a filter to reject harmonics, by definition.

I don't bother with 'Simulation' as I use a network analyser to make sure it Really works!

OK, dude. So you start every project with no prior knowledge to work from by walking into the lab and randomly grabbing any old parts off the shelf, put them together and arrive at a solution by complete trial and error. For everything. And you stumble through life that way. So happy for you!

Obviously everything gets checked out and tuned on a VNA, spec an, etc. in the end, regardless of whether you simulate it or not.

[Charlie_Dont_Surf]

...there's no need for a gate Res as it just slows down the drive waveform. I use SiC in this design.

That's akin to saying, "I removed the brakes from my automobile. They only served to slow me down." or, more particular to the subject at hand, "I bypassed the fuse. It was limiting my transmit power."

There is a time, place and purpose for everything.

Gate resistors regulate the rate at which the gate and output circuitry can add or drain charge from the parasitic capacitance of Cgs, Cds, etc. This is why power MOSFET datasheets include specs like Qgs, Qgd and Qg. Because Q=CV and i= dQ/dt.

The rate of discharge of those parasitic capacitors is, ultimately, the slew rate - the rise and fall time. And we know that faster rise and fall times often go along with more overshoot and ringing.

So somewhere there is probably a happy sweet spot of gate resistance that doesn't make the edge rates so slow to be unusable at the desired frequency, while also limiting overshoot and ringing and offering some protection to the gates and the gate driver.

And oh look: I just noticed that the Cree SiC device has an internal 26 Ohm gate resistance already. Gate resistance that high is not inherent in SiC technology; they added some resistance to the device inside the package. I'll give you one guess as to why they did that. This reinforces my point. Adding more externally would help protect (given the caveats above) but obviously the need is reduced in this case. I still don't like the idea of blowing up the internal gate resistor and wrecking the whole device when I could have done something about it.

You can leave out the external gate resistors. That's your choice, not mine.

[JimIO]

"I won't query your remarks any further as you have nothing useful to add."

I tried to say that in fewer words but my post got moved.

https://www.hfunderground.com/board/index.php/topic,65499.0.html

[redhat]

And oh look: I just noticed that the Cree SiC device has an internal 26 Ohm gate resistance already. Gate resistance that high is not inherent in SiC technology; they added some resistance to the device inside the package. I'll give you one guess as to why they did that. This reinforces my point. Adding more externally would help protect (given the caveats above) but obviously the need is reduced in this case. I still don't like the idea of blowing up the internal gate resistor and wrecking the whole device when I could have done something about it.

I don't think this is quite true.  The gate resistance is usually caused by the lossy-ness of the semiconductor material and the very fine binding wires used to connect the die to the outside world.  You'll notice on most datasheets this is derived by measuring the AC resistance of the junction at a known frequency and amplitude.  The larger the die and its gate structure, the larger the parasitic gate capacity, and the higher the gate AC resistance.  This is dependent on which silicon technology and topology is being used, obviously.

I can't speak with any authority on other technologies, but the metalization of the gate structure in SiC is one of the reasons for the higher gate resistance than comparable Si devices.

+-RH

[Stretchyman]

Hopefully that's the end of my thread getting hijacked as did the other.

If you have a simpler design or one as good and not too tricky to build then post up otherwise, please, don't bother.

Also please don't post up any useless theoretical simulations, Yes we've all got a cracked copy of ADS or whatever but barely worth the time and effort for such simplistic design/build, may as well go straight to PCB.

If you have a better design, built it, market it and tell us about it otherwise don't bother.

Ta

Stretchy.

[Charlie_Dont_Surf]

Hopefully that's the end of my thread getting hijacked as did the other.

If you have a simpler design or one as good and not too tricky to build then post up otherwise, please, don't bother.

To the moderator: does anyone, other than a moderator, have "ownership" over threads?

Also please don't post up any useless theoretical simulations, Yes we've all got a cracked copy of ADS or whatever but barely worth the time and effort for such simplistic design/build, may as well go straight to PCB.

You don't get it, do you? I wanted to see what the response at the harmonics were. I wanted to see what you were doing, whether you were seeking to minimize 2fo or something else because it wasn't clear to me. What better way to get the answer since you couldn't be bothered than to simulate it? And the result validated exactly what your schematic indicated, that the response peak is at around 45 meters and also that 2fo and 3fo were attenuated  So, what's the problem again?

(Shaking my head.)

If ADS is so "useless" then why do you have a copy of it?

Since you are so sure that sims are "useless", you had better tell the Qualcomms, the TIs, Huaweis, the Ericssons (and for that matter the General Motors, the Toyotas and the Whirlpools) of the  of the world. They'd be very surprised to know that all that money they have invested in design simulation is a waste.

Keep in mind that every time you use Ohm's Law, every time you calculate your expected power output, every time you determine the expected inductance on an Amidon core, you are using a mathematical formula that describes a physical phenomena. That's all simulations are, math models. So Ohm's Law is useless too, huh?

If you have a better design, built it, market it and tell us about it otherwise don't bother.

I've got something better. I've got one of your 40 W TXs that I bought secondhand that blew up on me about 20 seconds after I turned it on. I could literally see flames through the air vents on the enclosure. You really know how to leave a great first impression with your "superior" hardware.

I'll start a whole thread on my investigation into that explosion, so that you get to come in and "hijack" it.

[Charlie_Dont_Surf]

And oh look: I just noticed that the Cree SiC device has an internal 26 Ohm gate resistance already. Gate resistance that high is not inherent in SiC technology; they added some resistance to the device inside the package. I'll give you one guess as to why they did that. This reinforces my point. Adding more externally would help protect (given the caveats above) but obviously the need is reduced in this case. I still don't like the idea of blowing up the internal gate resistor and wrecking the whole device when I could have done something about it.

I don't think this is quite true.  The gate resistance is usually caused by the lossy-ness of the semiconductor material and the very fine binding wires used to connect the die to the outside world.  You'll notice on most datasheets this is derived by measuring the AC resistance of the junction at a known frequency and amplitude.  The larger the die and its gate structure, the larger the parasitic gate capacity, and the higher the gate AC resistance.  This is dependent on which silicon technology and topology is being used, obviously.

I can't speak with any authority on other technologies, but the metalization of the gate structure in SiC is one of the reasons for the higher gate resistance than comparable Si devices.

+-RH

Hmmm. Yes, it is measured at 1 MHz but they indicate "resistance", not "reactance" or "impedance". Though 1 MHz is probably more application-specific for most users than just a DC resistance.

26 Ohms at 1 MHz works out to 4.14 uH (assuming it's purely reactive). A product like this can get away with that much gate inductance. If this were a microwave transistor, the thing wouldn't work at all. 

[ChrisSmolinski]

To the moderator: does anyone, other than a moderator, have "ownership" over threads?

Thread drift can make it difficult for others to later find topics of interest based on the subject, so try to create a new thread/message when needed. But correct, there's no "ownership" per se by the originator. A moderator might split a thread if needed and/or move messages to another board, if appropriate.