MRF286 output power ? UT Source defective parts ??

[KT6UK]

Ok - egg on face

So I built a second version but this time soldering the FET's down to a flat heat spreader and used 22mil shim under the pcb to lift it up to the required Z height.
I also made sure the shim and PCB were right up against the FET's. Exactly as Darrell proscribes

The input match is still about 50MHz low but was easily able to obtain 80 watts out at 5 Amperes in - a great step forward.

Now I'm going to work on the input match and tweak it in.

I still don't understand why it's so important to have a 22 mil spacer rather than a 22 mil deep trench for the FET's - I can only guess the return Z from the edge of the PCB is lower with the spacer rather than the edge of a trench - more later !

73 Tim KT6UK

[VK5PJ]

Hi Tim,

Ok - egg on face
So I built a second version but this time soldering the FET's down to a flat heat spreader and used 22mil shim under the pcb to lift it up to the required Z height. I also made sure the shim and PCB were right up against the FET's. Exactly as Darrell proscribes
The input match is still about 50MHz low but was easily able to obtain 80 watts out at 5 Amperes in - a great step forward.
Now I'm going to work on the input match and tweak it in.
I still don't understand why it's so important to have a 22 mil spacer rather than a 22 mil deep trench for the FET's - I can only guess the return Z from the edge of the PCB is lower with the spacer rather than the edge of a trench - more later !
73 Tim KT6UK

One of the local VK5's built some of the W6QPL(?) boards but used the flange style FET's and has had a similar problem (poor output power) I believe he used a machined heat spreader to lower the FET height as that seemed the natural thing to do.

I have the same boards and FET's as he does but have not assembled mine, hence my interest in following this thread. I was going to have the copper spreader machined but now it looks like I need to find an alternative mechanical layout. or make the trench a few mm shallower than needed and put a thin copper sheet under the board so it can be pushed right up to the FET package like it would be in the original designs?

Regards,
Peter, vk5pj

[VK3QI]

Peter,

You should be able to buy 0.6mm copper sheet from Bunnings that will do the shimming job perfectly.

As I have posted, if the MRF286s are already screwed down, as you push the boards inwards towards the ceramic, they actually "click" into place snug up against the ceramic. I believe that that is the critical issue with the devices - clearly there is some sort of capacitive effect between the edge of the copper on the board and the ceramic body, which is probably well less than 1pF but enough to have some effect on the tuning.

The required dimensions are at the bottom of the page on http://www.nr6ca.org/pdf/xrf286rev3.pdf

Remember Ron VK4DD's warning about pushing down too hard on the FET with spacers in place and popping the ceramic top off!

Certainly when I reworked the boards in this manner, results were dramatically improved.

Cheers

[KT6UK]

Peter, Phil

with regard to your assembly plans.

So I had a quick chat with Darrell about the criticality of return grounding around the FET. He did mention something about him having a video up at one point underlining this matter.

Logically you'd think having a 22 mil slot / trench in the heat spreader and mounting the PCB directly on the spreader AND butted up against the device would be as good as if not better than a 22 mil shim under the PCB.

However while my trench was exactly 22 mil deep it was about 11mm wide thus allowing about 0.5mm float on either side of of the FET tabs.

I was aware of this but figured because the PCB and ground plane was right up against the ceramic it wouldn't matter.

Clearly it does.

It is therefore critical that not only does the PCB butt right up against the ceramic, but the 22 mil shim (or trench) must do likewise.

This leaves two choices 1. use the 22 mil shim and no trench or 2. machine the trench not only to the correct depth, but also to the exact width.

I'm going to re-assemble the PA again using option 1 taking extreme care to ensure shim and PCB are right up against the FET's.

Hope this thread helps others avoid the pit falls I've encountered.

I'll do some more work and see how it turns out.

The only other two factors which concern me slightly is that the PA which works reasonably well is using FET's from a different batch with the mounting slots / flanges removed while the one which is troublesome has full length flanges. But in both cases that are soldered down.

It'll be interesting to see if the revised (correct) assembly method fixes the performance from the PA with these other FET's.

More later.

73 Tim KT6UK

[ZL4PLM]

Tim
heres a pic of my amps

Just so you can see how I built mine

You can see the FET mounting and shim etc ok

rgds

Simon ZL4PLM

[KT6UK]

Hi Simon,
yep that's how it should be done - everything nice and snug. A picture says a thousand words !

That said I notice the load termination resistor connection looks a bit tenuous - I guess the photo was taken before completion ?

One day I may build myself a 'cube amplifier' with four dual PCB's but for now must be content with the plan of record !

73 Tim KT6UK

[VK2JDS]

Hi Simon, i have been working on some of Jim w6pql boards with varying results. getting consistent output from module to module was the problem i have as the combiner was getting hot.
the latest experiment involves soldering the fets into a channel milled into a brass spreader, no shims, and they seem to make reasonable watts considering i am running them off a 24v battery bank.
also i ordered some piston trimmers from DEM to allow the inputs to be peaked a bit more precisely.
what output power were you seeing per module off what supply volts, and how close to being the same were yours? i get about 100 watts per module at the moment, hoping the trimmers will improve this.
73 from Dave vk2jds

[KT6UK]

So I got one of my VE1ALQ amps to produce about 120 watts out with 10 watts of drive, as noted above - not too shabby I guess. It uses the MRF286's supplied by Darrell (VE1ALQ), he had cut the mounting flanges back a little.

So I went ahead and built another one - identical to the first, but maybe even better as I took extreme care to make sure the PCB's and 22 mil shims were right up against the ceramic of the MRF286's. I also cut the flanges back just like Darrell had.

I was expecting great things when I powered it up - instead of 120 watts out I got barely 2 watts out Drain current was just under 7 Amperes.

The MRF286's were bought from UT Source and have a batch number of QQ038006 - anyone else used devices with this batch number and had success ?

The ones from Darrell had a batch number of QQ038806 -not sure if this is relevant.

Maybe we should start a list of batch numbers which work (and those that don't) ??

Has anyone else had this problem ? I'm wondering if the MRF286's are somehow defective.

One clue might be that the input hybrid termination resistor was getting quite warm - didn't happen on the other amplifier, but both MRF's were drawing about the same current of just over 3.5A.

The output hybrid termination resistor was stone cold.

I'll keep plodding on and try and debug this problem - at least I have one amplifier which works reasonably well.

73 Tim KT6UK

[VK4CDI]

Hi Tim
Mine have a batch number 038006......

I will be rebuilding mine as per Darrell's instructions, when time permits......


Cheers

Phil VK4CDI

[ZL4PLM]

Hi Dave

all mine give 200W plus .... 12A at 28V

they are quite voltage sensitive

mine all performed as per Jims design

rgds

Simon

Hi Simon, i have been working on some of Jim w6pql boards with varying results. getting consistent output from module to module was the problem i have as the combiner was getting hot.
the latest experiment involves soldering the fets into a channel milled into a brass spreader, no shims, and they seem to make reasonable watts considering i am running them off a 24v battery bank.
also i ordered some piston trimmers from DEM to allow the inputs to be peaked a bit more precisely.
what output power were you seeing per module off what supply volts, and how close to being the same were yours? i get about 100 watts per module at the moment, hoping the trimmers will improve this.
73 from Dave vk2jds

[KT6UK]

I suspect that I may have damaged one of the FET's at some point - I suspect this for 3 reasons

1. The input hybrid termination resistor was getting excessively warm.
2. I noticed the bias current on one FET had gone to zero and I couldn't tweak it back up.
3. This is the big one, there was a blue flash from inside the ceramic package.

So I figure the FET has gone to heaven.

Now comes the 'fun' of stripping down the amp with a blow torch to get the FET's off the heat sink !

I suspect excessive heat while transplanting the FET's caused this problem.

Anyone have a recommendation for MRF286's besides UT Source ?



Phil - it'll be interesting to see how your amp performs

73 Tim KT6UK

Update,

EDIT - if anyone googles UTsource and MRF286 and stumbles across this thread, I suggest you do not prucahse MRF286 from UT source as they are potentially defective / counterfeit.

So I removed the suspect FET and sure enough gate / source short circuit. Replaced it with another FET from UTsouce of same batch number, was able to set the bias current ok, but still shows the same symptoms - ie 7 A input less than one watt out. Everything looks good around the out matching and combiner.

The FET I installed (carefully with minimum heat) was from a single device amplifier which was producing 40 watts + out. So I figure it was good, unless it got toasted as well during the transfer.

Beginning to think I need to start over with known good FET's !

[KT6UK]

So I've prepared three VE1ALQ amplifers as follows:

1. Using MRF286 FET's from Darren with the tabs cut down soldered onto a heat spreader with a PCB whose final output lines are 0.18" wide by 0.8" long
2. Using MRF286S FET's from Darren, no tabs as it's a S version device soldered onto heat spreader with a PCB whose final output lines are 0.2" wide by 0.8" long
3. Using MRF286 FET's from UT source - these will be bolted down to the heat spreader and will also use a PCB with 0.2" by 0.8" long output line.

In all cases I have the 22 mil shim and PCB right up against the MRF286's and the distance between input and output copper on the PCB's is no greater than 9.9mm.

All three heat spreaders can be fastened down to the same heat sink. PSU is a rigid 28 Volts from a 55 Ampere supply.

Idle current is 400mA per device.

Observations so far

Amplifier #1 with the thinner output lines produces 110W at Id of 10A for 10 watts of drive. The input frequency for best input return loss is about 1280MHz, can't use trimmers to correct
Amplifier #2 uses a later PCB with wider output lines. The input match was slightly high in frequency, tweaking the input trimmers permitted a 30dB input return loss.
Amplifier #3 is awaiting FET's

Notes - I've jigged all these amplifiers up so I can potentially interchange the PCB's.
In all cases so far max power / gain appears to be at about 1250MHz - maybe the output match is off ?
Interesting that Darren widened the output trace width and stated it made no difference. I suspect it might.
Still no where near the 200Watts output - but making progress all be it slowly...

More results later...

[VK2JDS]

have you tried snowflaking the input on one of those amps that produced very little output, i mean radically snowflaking, adding 1cm square pieces and moving them around (carefully not shorting anything out)
i did a testbed knockup using a utsource device on fibreglass pcb and fiddling with the output did very little , the major effect was using large flakes on the input. i cut chunks off the output copper plane using a dremel, and it made stuff-all difference, what really brought up the power was the input copper, and i added lots

Dave vk2jds

[KT6UK]

Hi Dave,

thanks for the notes and the tips on your MRF286 experience.

So far I've only tried fairly small 'snow flaking' around both input and output traces with practically no effect, same goes for the divider / combiner hybrid - I'm pretty sure that the issue of low output is related to input match and phase matching.

I say that because a one point I was measuring 60 watts output, but the output termination resistor (40 watt type) got so hot it de-soldered itself from the PCB. So clearly there's significantly more power to be had !

I also have a single MRF286 device and with 1 watt of drive it produces 35 Watts out. It's assembled and built in the same manner as the dual device PCB, so I figure I can't be doing anything grossly wrong. It'll make a fine driver once I get the dual nailed down.

More experiments this evening....

[VK3QI]

Tim,

Just a thought.

Have you tried using a Spectrum Analyser and probe to see if there maybe some sort of spurious output or oscillation taking place on your dual board?

Under certain circumstances it might be possible for a chip to run large current but appear to give very low output on the required frequency.

Your description of the dump resistor heating so much as to desolder itself is a worrying symptom.

Cheers

Peter Vk3QI

[KT6UK]

HI Peter,

good idea about the spectrum analyzer - but according to my Agilent 8561E, there no signs of instability at any output power level from DC to 6.4GHz. The second harmonic is -32dBc though

I suspect the reason the termination de-soldered was because I was flogging the amp and had no fans blowing air over the heat sink fins

Currently getting 90 watts output with 10 to 20 watts of drive and a drain current of 10 A with 5A going to each FET. Once I get above about 15 watts of drive the output power seems to limit.

So just for kicks I purposely backed the output shim away from the FETs by about 1mm and the power dropped to about 80 Watts.

BTW The driver stage has the FET bolted down with the shims right up against the package whereas the dual board uses MRF286S devices which have to be soldered down meaning in my case there's a very small radius of solder on the heat spreader which stop the shims coming within a few thou (mils) of the package.

Interestingly the driver has lots of gain 1W in gives 35 - 40 watts out - but I'm not using it just yet as I have another external drive source hooked up.

I'm convinced the closeness of the shims to the FET packages is critical - so just for kicks I wrapped a piece of sticky copper tape around the edge of the shim which butts up to the packages on the output side. Locally it took the shim from 22mils to 25 mils thick - but made no difference.

My next experiment is to complete building the MKIII with flanged FET's from UT source and make sure the shims and PCB are right against the FETs - much as they are now.

My MK 1 board with flanged devices soldered down still remains the best board at 120Watts out.

The saga continues....

[KT6UK]

Dave VK2JDS,

when you were tweaking around with an MRF286 on fibre glass board, what sort or power output and gain did you achieve ?

I seem to be running into a wall of about 50 -60 watts output per device regardless of how I assemble /tune.

I'm going to take a close look at the performance of the splitter / combiner, but this of course wouldn't explain the limited 50 -60 watts per device I'm seeing on a single device board.

I'm hoping the next batch of MRF286's from UT source perform a little better.
I'm also going to polish the edge of the 22 mil shim which butts up against the MRF286.

[KT6UK]

Hi Guys,

I built MKIII with tender loving care and installed two brand new devices from UT source, set the bias to 400mA on each device and applied 10 watts of drive.

DC draw was just under 8A and RF output power was a whole 2 watts !

I removed one of the FET's and installed it in place of known good fet on a single device driver PCB I have. The original device was producing 35 watts out for 1 watt in.

With the UT source MRF286 in its place I set it up and measured 1 watt output.

Interestingly I've seem this symptom before with another pair of UT Source MRF286's, but at the time could not be sure it was the MRF's as I accidentally destroyed them before I could confirm the issue.

Now I'm fairly sure the MRF286's from UT source are either counterfeit or defective - they do appear ok DC wise and will drive up with RF applied - but no significant RF output.

I've contacted UT source to see what they'll do about this as I have six of these devices from them, but have only used two.

The batch number is 0318 and they are marked MRF286C, the Motorola M logo also looks a little small.
Does anyone know where to get genuine known good MRF286's ??

Thanks Tim KT6UK

[K5DOG]

Hi Guys,

DC draw was just under 8A and RF output power was a whole 2 watts !

I removed one of the FET's and installed it in place of known good fet on a single device driver PCB I have. The original device was producing 35 watts out for 1 watt in.

With the UT source MRF286 in its place I set it up and measured 1 watt output.

Does anyone know where to get genuine known good MRF286's ??

Thanks Tim KT6UK

Man, wish I had found this thread before I made myself half bald!
I built two of the dual FET boards so far with results of nil...

Similarly to Tim, I can drive the boards to 6-8 amps easily, but output is essentially non existent. I think what I see is just drive power feeding through the FETS. I am using UT Source FETs labeled QQ038006. My first board I soldered down the FETs on a flat heat spreader, but did not use the shim. After emailing with Darrell, I retrofitted with the shims, but no change in results.

After finding/reading this thread, I reworked it again, carefully filing down the excess solder to get the shim precisely against the ceramic (actually, the shim is below the ceramic) . No change in the results.

I don't (yet) think the boards are to blame here. If I were to have some power out, but poor efficiency, then tweaking/snowflaking would probably be in order, but I have nothing to work with. It really feels like the FETs themselves are defective.

Perhaps the older XRF-286's on eBay are less likely to be "counterfeit". Has anyone tried those?

[KT6UK]

Yep - following further inquiries I've made the MRF286's from UT Source and likely most other vendors in Asia are counterfeit. They bias up OK and drive up in current with RF applied but practically no RF output no matter how you install / tune them.

Like you I've wasted significant time and $$ discovering this, having built 3 PCB's The only two which work use FET's from Darrel !

I've sent my remaining 4 un-used devices back to UT source for a refund - be interesting to see if they do the right thing.

I've just e mailed Darrell recommending that he drops UT source as a listed vendor for MRF286's.

I'm sending an order to 'pyrojoe' for some real devices !

End result is that I should have 3 working PCB's - one of which I'll use myself and the other two as spares or for sale !

73 Tim KT6UK