Topics

Do I need any Heat Sink compound on these mating surfaces.


Nick Gregory G0HIK
 

Sorry for the Newbie question :)

I'm a bit paranoid as I've let the "Magic Smoke" out of microwave projects in the past.

I've deburred everything and polished with fine wire wool and Brasso to get the best matting surfaces. I've not used a heat spreader before. I know that I dont use Heatsink compound on the copper to aluminium faces. But what about the transistor ? They seemed to have the remnants of compound on them when they arrived.
I guess if I do it's not the usual stuff,  is there something called Artic Ice available from the local PC store for Overclockers ?

Thanks

Nick G0HIK


Robin Szemeti - G1YFG
 

A little heatsink compound would be good, the machined surface is not bad, but will benefit from some gap filling, at the microscopic level.  If you have some CPU style compound, I'd use that under the transistors, and I'd probably just use the plain old white stuff under the copper/aluminium as the cpu past can be a bit "stiff" and you may have trouble compressing it down over a larger area. The white stuff is also cheaper.  You've done a good job with the cleanup that's for sure!

To the others waiting for a spreader: apologies, as I am still waiting for raw material, my guy went on holiday but is rumoured to be back today, the joys of living in the sticks!  3 have gone out, more will follow as soon as I can pick up a slab of copper, apologies to those still waiting.



On Tue, 20 Oct 2020 at 10:41, Nick Gregory G0HIK via groups.io <nickg0hik=googlemail.com@groups.io> wrote:
Sorry for the Newbie question :)

I'm a bit paranoid as I've let the "Magic Smoke" out of microwave projects in the past.

I've deburred everything and polished with fine wire wool and Brasso to get the best matting surfaces. I've not used a heat spreader before. I know that I dont use Heatsink compound on the copper to aluminium faces. But what about the transistor ? They seemed to have the remnants of compound on them when they arrived.
I guess if I do it's not the usual stuff,  is there something called Artic Ice available from the local PC store for Overclockers ?

Thanks

Nick G0HIK


--
Robin Szemeti - G1YFG


militaryoperator
 

To the others waiting for a spreader: apologies, as I am still waiting for raw material, my guy went on holiday but is rumoured to be back today, the joys of living in the sticks!  3 have gone out, more will follow as soon as I can pick up a slab of copper, apologies to those still waiting.
Robin Szemeti - G1YFG




On a related topic, did those noise boards ever materialize Robin? 

Ben


Neil Smith G4DBN
 

The Arctic Silver stuff needs a bit of pre-planning as it settles over time, so really you need to assemble the device to the spreader and  put it through a few heat cycles, then re-torque it.  It is very rare that I don't solder the device to the spreader using low-temp Bismuth/Tin/Silver paste.  Bolting them down needs great care so the bottom remains flat. I usually flatten my devices off using 2000-grit wet/dry glued to a ground surface plate and lubricated with IPA, using a figure-8 pattern and rotating the device regularly.  I then lap the bottom of the groove with some wet or dry glued to the edge of a machining parallell to get the bottom of the device groove reasonably flat. Not aiming for a polished finish, just a good wettable surface that will take solder well. I usually remove the plated layer on the device underside right back to copper when soldering.  When I use Arctic Silver, I use 600 grit and leave the surfaces quite matt, and take care not to over-torque the clamping screws.

I am always nervous about not using solder to fix devices to spreaders because of the additional inductance in the source lead and the uncertainty in the length of the RF path.

For large spreaders, I machine the faces, lap the surface and heatsink with 1200-grit paper and WD40, then lap them together with a bit of 5 micron diamond or carborundum paste, degrease, then use ordinary white silicone grease (or Arctic Silver if I'm feeling extravagant), torque it down, run a few heat cycles and torque it down again, then run full power and check the temperature of the device, spreader and heatsink.

I am a bit over-picky though. 90% of that is probably unnecessary, but the GaN device I'm doing right now cost over £400 and I REALLY want it to stay cooooool.

Neil G4DBN

On 20/10/2020 10:41, Nick Gregory G0HIK via groups.io wrote:
Sorry for the Newbie question :)

I'm a bit paranoid as I've let the "Magic Smoke" out of microwave projects in the past.

I've deburred everything and polished with fine wire wool and Brasso to get the best matting surfaces. I've not used a heat spreader before. I know that I dont use Heatsink compound on the copper to aluminium faces. But what about the transistor ? They seemed to have the remnants of compound on them when they arrived.
I guess if I do it's not the usual stuff,  is there something called Artic Ice available from the local PC store for Overclockers ?


Nick Gregory G0HIK
 



 Thanks for all the information Neil, that is going to take some absorbing for a newbie.
I started this afternoon with some CPU paste under the devices. But on switch on, it was pulling a lot of current and the small load was getting warm, so it was taking off.
When I stripped it down, I noticed the compound was only touching in a very small strip of the transistor mount.
I cleaned it up and reassembled, screwing down and resweating the transistor tabs several times. This stopped it from oscillating, but I know the mating surfaces are not great as the left transistor is getting quite warm even on standby. 
I'm also not sure about the output power, I'm getting 60 watts out for 1 watt in, which makes me think it may still be taking off as that is a lot more than the modification notes suggest.
Nick G0HIK


Neil Smith G4DBN
 

You need to make certain that you have a low-inductance path from the source to the ground plane under the gate and drain connections.  That's why I cheat and solder the PCB underside and the device base(s) to the heatspreader.

Neil

On 20/10/2020 21:28, Nick Gregory G0HIK via groups.io wrote:


 Thanks for all the information Neil, that is going to take some absorbing for a newbie.
I started this afternoon with some CPU paste under the devices. But on switch on, it was pulling a lot of current and the small load was getting warm, so it was taking off.
When I stripped it down, I noticed the compound was only touching in a very small strip of the transistor mount.
I cleaned it up and reassembled, screwing down and resweating the transistor tabs several times. This stopped it from oscillating, but I know the mating surfaces are not great as the left transistor is getting quite warm even on standby. 
I'm also not sure about the output power, I'm getting 60 watts out for 1 watt in, which makes me think it may still be taking off as that is a lot more than the modification notes suggest.
Nick G0HIK
_._,_._,_



Nick Gregory G0HIK
 

Yes Neil,

I've reread your message several times and have a bit more of a grip of the information now.

As I progress, I'll post how I get on.

Nick G0HIK


Nick Gregory G0HIK
 

Thanks for the info Robin and the heat spreader of course.
I'm learning that there is a lot to learn :)
But it good fun and satisfying.

Nick G0HIK


Robin Szemeti - G1YFG
 

Don't panic if one is hotter than the other, unless you have replaced the bias circuits, they are supposed to run like that.   This is a "Doherty" amplifier, one device is carrier and biased Class A/B, the other device is peaking, and biased well into Class C.  The Class C device often barely carries much current ...



On Tue, 20 Oct 2020 at 22:03, Nick Gregory G0HIK via groups.io <nickg0hik=googlemail.com@groups.io> wrote:
Thanks for the info Robin and the heat spreader of course.
I'm learning that there is a lot to learn :)
But it good fun and satisfying.

Nick G0HIK


--
Robin Szemeti - G1YFG


Nick Gregory G0HIK
 

If I modify the bias so they are both running in A/B are they a parallel pair then?
This would be an interesting  demonstration to myself.
I see the figures I'm getting (1w in 60w out) ties up with the datasheet for gain. The modify docs don't suggest this amount of gain, hence me wondering if it is hooting .

Nick G0HIK


Chris Bartram G4DGU
 

Nick,

Running a Doherty amplifier with both devices biased evenly won't put the devices in parallel. What it will do is to reduce the amplifier's efficiency, and thus increase heating. That will result in increased channel temperatures in both FETs, as they are thermally coupled, and will reduce the lifetime of the amplifier and it will also screw-up the linearity. There's a lot of literature available on-line regarding the Doherty amplifier configuration - it's by no means new, the principle dating back to the 1930s, and it isn't a complicated concept. Much better to do that than carrying-on blindly ...

Moving the frequency of operation of an essentially fairly broadband amplifier by a few percent is pretty unlikely to cause any gross instability, but I understand your concerns. I approach this from a different direction, having a fairly well equipped lab., but it's possible to get a pretty good idea of whether there is any problem by getting local reports, looking at the way in which the output varies, or by spending £100 on a Pluto, loading SATSAGEN on a computer, and using the spectrum analyser, plus some input attenuators, and probably a directional coupler to look properly at the output spectrum.

The people who design power amplifiers for cellular and mobile service know exactly what they are doing and have excellent resources, but their design targets will be necessarily different to ours. We can often modify their work, but it requires the background knowledge to do so effectively.

73

Chris G4DGU

On 21/10/2020 08:46, Nick Gregory G0HIK via groups.io wrote:
If I modify the bias so they are both running in A/B are they a parallel pair then?
This would be an interesting  demonstration to myself.
I see the figures I'm getting (1w in 60w out) ties up with the datasheet for gain. The modify docs don't suggest this amount of gain, hence me wondering if it is hooting .


Nick Gregory G0HIK
 

Hi Chris,

Thanks for the information.

I'm gradually building up test kit suitable for this type of work. I have thought about a Pluto, I'll put it on my Shopping / Christmas list 

I have varied the input to the amp and it windups smoothly, so there does not appear to be instability. I was not suspecting the design, more my construction. I had noted instability with transistors mounted with compound, I do not notice any now with it bolted direct to the heat spreader. But I still want to improve the contact as per Neils information.

Today my directional coupler arrived and I can measure the power more accurately.

1.25 watts in gives 50 watts out @ 4amps
2.5 watts in gives 70 watts out @ 6 amps

Nick G0HIK


Robin Szemeti - G1YFG
 

Those drive figures and output levels sound "about right" for 28V LDMOS at 2.3GHz ... LDMOS is very gainy compared to bipolar, My 23cm amp will produce 250W out for 2W of drive ... from a single 48V LDMOS device, so 70 out for 2.5 watts of drive at 2320/28V does not sound at all unreasonable.


On Wed, 21 Oct 2020 at 20:52, Nick Gregory G0HIK via groups.io <nickg0hik=googlemail.com@groups.io> wrote:
Hi Chris,

Thanks for the information.

I'm gradually building up test kit suitable for this type of work. I have thought about a Pluto, I'll put it on my Shopping / Christmas list 

I have varied the input to the amp and it windups smoothly, so there does not appear to be instability. I was not suspecting the design, more my construction. I had noted instability with transistors mounted with compound, I do not notice any now with it bolted direct to the heat spreader. But I still want to improve the contact as per Neils information.

Today my directional coupler arrived and I can measure the power more accurately.

1.25 watts in gives 50 watts out @ 4amps
2.5 watts in gives 70 watts out @ 6 amps

Nick G0HIK


--
Robin Szemeti - G1YFG


Nick Gregory G0HIK
 

Thanks for the confidence boost Robin

Nick G0HIK