QO-100 / EShail 2 Uplink amplifier development adventures

[Da_Stier]

Two things that are fascinating to me are satellite communications and RF power amplifiers.
So the natural reaction after learing about the QO-100 satellite was obviously excitement and the urge to develop my own gear for it.
For all of you who don't know, what I'm talking about, here you go:

https://en.wikipedia.org/wiki/Es%27hail_2

In short it is a TV satellite from Quatar that HAM operators can use, as long as the spare transponder is not needed.
You can contact it via a 2.4GHz uplink and it will answer in the 10GHz band that can be received by slightly modified Sat LNBs.
For the uplink you need quite a bit of power - in the range of maybe 10W - depending on your antenna and how well you are aiming at the satellite.
Therefore there are quite a few homebrewed or modified amplifiers for the 2.4GHz band around.

I'm not a HAM operator, so I can't ever test the stuff I build, however I enjoy building it more anyway and I have enough test equipment to characterize the gear anyway.
And who knows, maybe I will do all the tests required to be a HAM one day.
I also know enough HAMs that it wouldn't be to hard to do some real world tests anyway.
None the less, I think I can help other people by doing experimentation and development.
After some research I learned that a lot of people modify old cellular equipment to work on the 2.4GHz band, that is used in the QO-100 uplink.
This was a fortunate coincidence, since I had quite a bit of old cellular equipment.
All of this was around 4 years ago.

So I started to do some modifications and building some amplifiers.
I got basically one of each popular amplifier type and modified the hell out of them.
They all work, some better, some not so great, however I learned quite a bit from them.

These are some of the better ones I did:

This is a very popular Nokia Siemens amplifier (Mads tore down a basestation, that used these once)
It is somewhat easy to modifiy and works rather well.
This is the unit as I got it.



This is the unit after my mods.



It has an max. output power of over 90W with a somewhat reasonable efficiency.





However 90W is more than is really needed, even a HAM licence only enables you to use 50W in this band anyway.
So while being a very nice amp, it isn't really what I would call a perfect solution.
Especially since you have to be kind of lucky to find this exact amp ... and at a good price as well...

Still, out of curiosity I wanted to know, how far I can go modding wise, so I did this:





This is a platform for trying different stuff and snowflaking around.
It is definately not made for continous use or any serious use at all.
It is purely for lab work and experimentation.
With this as the final I can almost reach 120W, however the efficiency is pretty bad



Another short but very interesting adventure was the amplifier I took a look at for Andreas Spiess.
He did a video series on his QO-100 setup and fried his amplifier.
I offered to take a look at it for him to analyze what happened and see if I can fix it.
Here is his video for anyone interested:

/>
Short story short, he ran into a thermal runaway situation due to poor biasing of the final transistor in the amp he got.
I tried fixing it, however I couldn't get a replacement for the blown preamp so I could only restore the function of the final.





So at some point, Mads sold me two of these amplifiers.



They came as a rather nice package with a nice heatsink and they modified rather well... atleast so I thought.
I modified it stage by stage and it worked rather nice, except for the finals.
I did a quite substencial rebuild of the shielding enclosure to get a finished unit



I tried something new on this as well, by coating the 3D printed PLA part in a layer of aluminium tape to provide shielding.







The final result was rather disappointing.
The max. output is around 40W ... which is totally fine and a good usable level ... however the efficiency is under 15%, which is horrible.
It would still be perfectly usable, however I dislike the thought, that 85% of the power is just heating the room.
So I thought, since all the driving stages modified rather well, I should take a closer look.
I narrowed down the proplem to lie within the final stage architecture.
It is realizes as a doherty amplifier with very narrow impedance matching and filtering.



A doherty amplifier uses a base amplifier in class A or AB and a peaking amp in class C.
This requires some rather precise phase splitting of the signal, which is done by a microstrip splitter.
This is a very simple and proven way of doing it, however it is also very frequency dependent.

So I had the idea to cut down one of the amps and only use the driving stages as a standalone amp.
In my previous testing, I got around 30W from the driving LDMOS, which would be perfect.

So this is what I did, I cut down the second board and rematched the transistor to be closer to 2.4GHz.



To do so, I used the common "fat line" approach.
If you take a look at the datasheet of a typical LDMOS transistor, you see, that the input / output impedance is rather low.
(in the 1 +- <=1j)
This means that you need to get it up to 50 +- 0j to be a good match at the desired frequency.
Normally you would use a 500k$ simulator to do so, most likely ADS, however without it, the easiest thing is to do it experimentally.
Therefore you start by adding as much of a large stripline, as possible and cut it until you are satisfied.

So I rigged up a pre pre driver and modified one of the power supply board I made for these amps.
This gave me this setup, that was all I needed, to test the amp.



Yes this is kind of a continuity break, since I already started the case at this point, however I don't have any pictures of the previous stage anymore.

It worked pretty well, over 30W of output at over 25% efficiency, which was a good start.
So I went ahead and milled a - in my opinion very nice - case for it.

[Da_Stier]

So I had to split my post into two parts, since it didn't like my extreme amount of attachments.



However, this is were it all got complicated.
While doing some final testing and tuning ... which changes a bit, when mounted in a metal case ... I dropped a SMA adapter into the amp.
This shorted out the preamp and killed the two RF2126 amps used.
This pissed me off quite a bit more than it should have and I paused the project for a week or two.
After this I thought I could just order some new ones and continue.
However the RF2126 is long obsolete and can only be gotten over eBay or similar.
So I thought some from different sellers ... but didn't have any luck with any of them.

So out of despair, I built a little RF2126 amp board to test the parts I got.
This would actually be a rather nice little board on its own, providing around 1W at 2.4GHz.



Unfortunately, the results were pretty bad, they ranged from
- definately not an RF amp at all
- an RF amp but not at 2.4GHz
- an RF amp at 2.4GHz but with only 3dB of gain
- an RF amp at 2.4GHz but it saturated at around 21dBm and died at 21.1dBm

So I was once again pretty pissed, especially since the parts were not that cheap ....
However if anyone knows of a legit source for RF2126 chips, please let me know, it would still be great to find some.

So a few weeks ago I had a conversation with some one saying, that I might be able to modifiy some amp designs, however I couldn't build my own from the ground up.
This got me thinking and I decided to build a new amp all togheter.
This design is supposed to become an open source project in the end with Gerbers, schematics and mechanics for the case being available to everyone.
Befor I do this, I want to get it to work reliably and efficently however.

I searched quite a while for a suitable driver transistor.
You might want to see my RF prototyping thread, since it includes the first test of a possible driver.

https://highvoltageforum.net/index.php?topic=1653.0

After a while I settled on the MHT1008N, which is a 5W LDMOS transistor, that was designed to work on 2.4GHz already.
And the great thing about it is that it "only" costs 15$ in single piece quanteties.
... So I took the bitter pill and bought 10 of them from mouser.

I layed out a board, following but not copying both the datasheet and some rather nice designs by other people for this transistor.
This first revision is only to optimize the RF path so the DC and biasing parts are the bare minimum.
For a final design I wouldn't just slap in a potentiometer for the biasing but do a proper low impedance source with temperature compensation.
(For example an op-amp or a LDO regulator)

MHT1008N.pdf

This is one of the best projects in my opinion.

https://www.dj0abr.de/

These boards have a PHA102+ MMIC from MiniCircuits followed by a MHT1008N LDMOS.
After receiving the boards, I built the first unit.
While doing so I broke my only M2 tap, so the two SMA connectors unfortunately only have one screw each.



After some initial tests, everything turned out as expected and I got around 6W from the amp.



This was kind of expected, since it is a proven design and should work straight out of the box.
So after this I went ahead and build one of the complete boards.
The boards I ordered, include an additional stage, so the first one I built, was actually cut off.
Since I still didn't have a new M2 tap, I did something a little bit different for this unit.







For the final, I used a PTF210301 LDMOS, originally desiged for 2.1GHz cellular.
I did this since I wanted to use parts that are available relatively cheap and these transistors can be had for around 10$ each.

The results were pretty dissapointing once again.
Sure I didn't rematch anything yet, still around 12W output at over 100W DC input was more than bad.
So I did some rematching and got it to around 15W out from 112W DC in so meh, not really.

Since this surprised me, I built a little board only containing the final and doing some testing on this:



The results were kind of as expected, however worse than expeted as well.
The transistor has some very nice 20dB of gain at 2.1GHz, where is was desined for.
However on 2.4GHz the max. gain is around 6 to 8 dB.
I managed to shift the S21 response quite a bit, by adding capacitance to the output.
However this also messes quite a bit with the SWR of the unit.
So the resust were around 12dB of gain at 2.4GHz with no peak at 2.1GHz, however the transistor goes into compression at only 12W of ouput.
So it seams like the PTF210301 is just a very hard transistor to rematch, which is very sad to be honest.

At this point I tried a BLF6G22L40 LDMOS, which is basically identical from most values.
However this is the transistor that is used in the amps from Mads, that retuned rather well.
It is very interesting, how similar tranistors seem to have a completely different internal matching circuit, since some retune rather easy and nice and some are virtually impossible to retune at all.

Right after the change I got around 16W output without any rematching.
After quite some tinkering I got it to 20W output at 112W DC input, which gives me just under 20% of efficiency.
Not great but at least a start.



So the next thing I did were around 3 days of tinkering and trying to improve on this, however I just couldn't get it any better.
So this is basically the point I'm at right now.
So lets recap:
- some celluar amps retune rather nice and achieve respectable results
- some celluar amps are horribly inefficent to retune
- the nice amp I tried to build got fried
- I achieved a rather nice 5W amp from scratch
- the final stage of this is rather bad

So one thing I would love to try is getting some LDMOS finals that are made for 2.4GHz.
There are two suitable ones on digikey, however 150$ a piece is kind of to expensive for my taste.
The devices that would be necessary are the BLF2425M9L30U or the BLF2425M8L140U, both from Ampleon.

https://www.digikey.com/en/products/detail/ampleon-usa-inc/BLF2425M9L30U/5870188
https://www.digikey.com/en/products/detail/ampleon-usa-inc/BLF2425M8L140U/4289056

Unfortunately there are no free samples available for this kind of components.

Another, slightly cheaper alternative would be something like the BLF8G24LS-100VU.
It costs "only" 62$, however it also tops out at "only" 25W.
It would also be a bigger redesign of the PCB, since it comes in a differen package.
Especially since it doesn't have screw mounting plate which means it has to be soldered to a heatspreader.
Which in itself means that it can't be made out of aluminium any more.

https://www.digikey.com/en/products/detail/ampleon-usa-inc/BLF8G24LS-100VU/5035628

So the most promising alternative might be the BLF7G24LS-140,118.
It is still "only" 64$, comes in a screw down package and can provide 30W.

https://www.digikey.com/en/products/detail/nxp-usa-inc/BLF7G24LS-140-118/12617459
And maybe it might be an idea to try our chinese friends for this part, since it is available on Aliexpress for 16$
It is only the solder down version but it would still be a pretty good offer.
https://de.aliexpress.com/item/1945187474.html

So if anyone is interested to participate in this project and or donate a transistor or something like this, drop me a message.
If I happen to get my hands on some nice transistors, I will make a V2 of the amp with a way more polished circuit and a nice milled case for it.
I will also make it open source and maybe sell some kits and or finished units.

Another thing I want to try is to revisit my cut up amp from Mads and integrate a MHT1008N driver stage into the case, replacin the RF2126 drivers.
Mainly since it tuned up so nicely and I invested quite a lot of time into the case.

So this is where I'm at right now.
I hope some of you like my journey through the 2.4GHz RF amplifier world and will continue reading the next part, when it's ready.
If you have any questions or want to discuss anything in more detail, feel free to ask, I'm happy to go into further detail of whatever is of interest.


Greetings,
Michael

[Da_Stier]

Hi all,

today I took a look at the cut up amplifier from Mads.
I really wanted to get this one working, since I somehow like it, propably because I spent so much time on it already.

I started by redoing the electronics.
To do so, I cut up the board some more to make space for a MHT1008 as a driver.
I also redid the whole biasing electronics, since I had to cut off some of it to make space.
This is the unit with the new guts:





So after working on it off and on for a few months and two rage quits, it was finally finished again.
Next I did take a few measurements to see how it performs.
This is the typical setup I use to do so.



The output power is "wasted" in a big dummy load with a -40dB directional coupler.
Since its frequency response isn't all that flat, I calibrate the setup by connecting the signal generator directly to the coupler.
Like this I can program a suitable offset into the spectrum analyzer to give accurate power readings.
Using the spectrum analyzer for these measurements also enables me to see its performance from a harmonic / distortion and stability standpoint.
All of this is rather hard to see on a pure power meter.
I do however like to use my powermeter as a second opinion to validate my readings.

The results were pretty satisfying.
I easily get 20W of output with some headroom.
The maximum power I can get with a reasonable input power is ~29W.
The efficiency is around 30% at 20W output, which is about what I would expect from a somewhat matched linear amplifier.



One interesting thing to note is that the coax I like to use on my workbench - Huber&Suhner Enviroflex 316D - is getting quite hot from continous 20W at 2.4GHz.
According to the datasheet it should handle this amount of power without to many problems.
It did also handle the 90W from the Nokia Siemens amplifier, however it did get hotter than I would like coax to get.



So for some real use, I would definately use some higher power coax.
RG400 or Enviroflex 400 would be my cable of choice for a short run (<= 1m / 3ft)
For these measurements it doesn't really matter, especially since the cable losses are calibrated out anyway.

So after way to much time I finally got a good result from this particular transistor.
It is interesting however, that only the transistor that came with the amp from Mads, performs like this.
The transistors I bought, which have the identical part number - BLF6G22-45 -, don't perform that well.
I don't think that they are fake, since they perform perfectly fine in the band that they are designed for.
I also don't think that there are to many faked RF power transistors out there, since it is pretty expensive to fake them if you want them to look correctly.
(With the white ceramic lid and gold contacts)
Maybe they are a later / earlier revision, that has some slightly different internal matching? 


Greetings,
Michael

[Mads Barnkob]

Wow, I am completely blown away by the project trail of this thread What seemingly was "hey there is this satellite..." and "i got these amplifiers" took us on a extremely long journey of RF hacking!

Regarding the differences in the LDMOS transistors, could it be that its higher quality, special quality, custom matched parts delivered for Nokia Siemens, compared to the "random" public spare parts you can find and buy?

I am looking forward to see you get contact

[Da_Stier]

Thanks Mads, it is a classical case of one thing led to another.
I'm not really disceplined enough to push through a project for many months, especially if something stupid happens.
However I'm stubborn enough to continue after all.

I did just order some of the 2.4GHz LDMOS transistors from Aliexpress and a SMD reflow hotplate from ebay, so there will be a V2.0 at some point 

Regarding the differences in the LDMOS transistors, could it be that its higher quality, special quality, custom matched parts delivered for Nokia Siemens, compared to the "random" public spare parts you can find and buy?

Maybe, who knows, I could imagine, that big manufacturers get these transistors made to their liking.
I compared the recommended layout from the datasheet with the layout on the Nokia Siemens amp.
It is definately different, however it might be electrically identical or atleast similar, it is quite hard to tell from such highly optimized RF circuitry.
Especially since it is also dependent on the exact substrate used, the distance to the lid, etc.



Greetings,
Michael

[Da_Stier]

Hi all,

so I have quite a few updates to share.
I ordered some of the 2.4GHz LDMOS from the aliexpress seller and I am very happy with how that went, since there was still some risk of getting faked or wrong parts.
I also got myself one of these blue SMD reflow hotplates that you can get all over Aliexpress and eBay.
Spoiler it works beautifully and is well worth the 90€.

Unfortunately there are literally no design documents available for the BLF7G24LS-140.
Normally there is an application circuit and some example layout with the matching networks to get you started in the datasheet.
But this transistors datasheet provides literally nothing, I don't really understand why some datasheets provide an example design and some don't.

So to figure out, how the design needs to be, I layed out a small eval board for the BLF7G24LS-140.
For this I only included the minimal amount of parts needed for a basic test.
This is what I came up with.



So next up I machined a few metal parts to assemble the unit.
For mounting the LDMOS, I used a strip of 2mm brass sheet, that I found at the scrapyard.
These are the mechanical bits all machined.



For mounting the LDMOS, I put the whole assembly on the hotplate.
To hold the PCB to the base plate, I used PCB standoffs.
For the standoffs next to the LDMOS, I used aluminium standoffs to prevent them from being soldered to the brass heatspreader.
I taped the assembly to the hotplate with kapton tape.
This was in an attempt to fix it in place, incase I need to move the LDMOS once the solder paste is liquid.
However the tape didn't stick to the machined hotplate surface so it wasn't really necessary / useful.



This is the fully assembled unit



So I was pretty excited and hooked it up to the VNA right away.



But the result wasn't all that great.
The S21 response has its peak from 2.3 to 2.4GHz, however the gain is only slightly over 10dB which is way off.
The avarage gain is given with 18dB so there is some mismatch in the system.



So it looks like I have some experimentation with copper tape to do, in order to find a good match.
This is why I made this board, to be able to experiment with the matching in an isolated system without any other stages.

I also did a quick test to see how the amplifier performs when driven close to saturation.
The results were extremely poor, it delivered 6W output from 9W drive, however all the components are still fine.
I guess it is just severely mismatched and the reflections overlap destructively.


Until next time,
Michael

[Da_Stier]

Hi all,

so I have been working on and off on the amplifier, as my spare time allowed it.
I was working a lot on matching the BLF7G24LS-140 board I did.
However I didn't quite come up with a good design.
In the end I got around 50W output but only at around 30% efficiency which isn't that bad but definately not great either.
(Atleat not for a transistor that should be made to work at this specific frequency.)

So by pure coincidence, I came across the MRF6S23100H transistor.
It is made by NXP / freescale and also operates from 2.3 to 2.5GHz.
It is also obsolete, however the datasheet offers an example layout.

MRF6S23100H.pdf

Since the same Aliexpress seller, I got the last transistors from, had some of the MRF ones for 20€ each, I got some to try my luch on.
I also followed the recommended test layout from the datasheet and got myself some boards.
I wanted to use simple FR4 which is of course not ideal at these frequencies and powers but I didn't feel like spending 200$ for a single Rodgers board.
To do so, I copied all the microstrips precisely as noted in the datasheet.
In the end I calculated the impedance of some of the microstrips for the given parameters of 0.76mm PCB thickness with an Epsilon of 3.5.
Then I chose my FR4 thickness to get the same impedance for the microstrip elements.
This lead to a thickness of 1.2mm.

These are the boards, as I received them:



For connecting the DC lines, I made some 4mm connectors from some 12mm brass round stock.
I think this is a pretty nice formfactor for 4mm connectors.
Let me know, if you know of someone manufacturing something similar. 
They are quite a bit of work to make by hand.





This is the assembly, I used for testing.
I use the 10W max. amplifier I already made as a driver. For the output it is VERY (as I found out later) important to use a good cable.
By far the best ones I found are the Sucoflex 104 Cables by Huber&Suhner but they are pretty expensive.



For easier control, I decided to wire a little bias board to it, that supplies the around 3V of gate bias.
I just cut out the bias section of a BLF eval board I had left.



At the first try it worked pretty ok with around 50W of output at around 35% efficiency.
Unfortunately it only worked a few seconds, before the 5p6 output coupling capacitor blew up and made quite a nice RF plasma arc.
I used a simple 0805 part which is definately not recommended. 

After this I got the capacitors recommended in the BOM from the transistors datasheet.
They work perfectly ...however they are 4.3€ a piece so I made sure I don't lose one in my carpet.

In the end I got around 77W of output with 4W of drive.
It used 5.9A at 28V meaning an efficiency of 46%.
PERFECT! This is what I wanted to see. 





So this is a design I like.
Now I can go ahead and build a single unit consisting of an MMIC preamp, a MHT1008 driver with a MRF6S23100H final.



Greetings,
Michael

[Da_Stier]

Hi again,

so another thing I wanted to see your guys opinion on are the transistors I got.
As mentioned before, I got them on Aliexpress, where you can never be sure what you get.

I suspect they are fakes or atleast not new but I would like to see what you think, maybe someone has an idea?

Here is what I noticed:

1. The bottom side of the Source tab is pretty scratched and dinged up, it just does not quite look new.



2. The lettering on the package looks printed. All other NXP / Freescale transistors I have have laser etched text.
    I do however have some Infinieon transistors that have printed on text, so I'm not sure about that.
    The text also comes off with isopropanol but not completely, it always stays visible.







3. The surface of the ceramic lid is not smooth and kind of like "very very fine sandpaper" as usually but instead has some diagonal lines on it, like it was sanded.
(I'm not sure if you can send ceramic to be honest)



4. The golden parts just don't look quite gold plated but more of a pale yellowish copper color.




Since the devices work fine, I'm really unsure about what to do.
I'm pretty sure they are not new originals but at the same time they work pretty well.
I'm still debating with myself if I complain about them or not, so what do you guys think?



Thanks and greetings,
Michael