[hpsdr] Alex

[Graham Haddock]

Hi Frank:
Reply comments below.
--- Graham


FRANCIS CARCIA wrote:
> I would avoid using T50 cores at 100 watts. Might get away with 
> stacking 2 or 3 cores.
> Red for low end and yellow 40 meters and above.
> T68 work well up to around 200W. Also allows larger wire to keep the 
> resistance low.
The Elecraft K100, 100 watt PA successfully uses single T-50 cores at 
the 100 watt
level.  It does appear that they are using lower permeability cores than 
most designs
to lower the losses and heating.  I will try that approach in my 
transmit LPA filters for the the
first prototype, and let people know what happens.

> Check RF current rating of caps might need to parallel multiple units 
> to keep them cool.
I will be using industry chip caps specifically rated for RF power.  
AVX-SQ or AVX-U Series

> Stay away from XR7 dielectric caps.
Amen, brother.

> Relays use the ones about the size as a SRA1 mixer can. Then can take 
> the voltage.
The relays will be Omrons spec'd for RF service.

> Add at least a gas tube for static protection. I would think some sort 
> of a clipper needs to be ahead of an expensive A/D. Biased Transorbs 
> isolated by low C fast diodes at least.
I am still thinking my way through this issue.  Between any antenna 
input and the input to the
MERCURY, there are at least 3 Hittite switches, that will probably fail 
with a 30 volt
pulse.  I am not sure a gas tube is going to protect them.  I will put 
one in anyway to to
limit total energy.  After that I still need something that will clamp 
at less than 10 volts
to protect the solid state switches.  And it should not clamp at less 
than about 2 volts.

 From the input of the MERCURY to the A-to-D converter there is always a 
30 dB
attenuator chip and a 20 dB gain block in the path.  I suspect Phil will 
need some
kind of 10 volt or less voltage clamp to protect the attenuator chip.


> I'm a bit worried about a D/A making a clean signal. We know what a 
> D/A in a DDS does, how is this different?
See Phil H's response on this topic.


--- Graham / KE9H

==



 1185551135.0