Re: FW: LF: FETs getting sick - why not use tubes?

[M0FMT <peteg8fmt@gmail.com>]

Hi all

The Decca PA seems to be scattered with spike limiting diode from memory Transils / Transorb TVS diodes. (don't know the values) I usually put one in the drain set above the peak drain wave voltage seen on the scope and below the peak operating value of the FET  So far so good in single ended amps.

Jim M0MBU did a short treatise see http://www.wireless.org.uk/guard.htm.

The Decca TX "guard circuit" by Jim Moritz, M0BMU.

73 petefmt

On Mon, Apr 3, 2017 at 5:39 PM, <g3kevmal@talktalk.net> wrote:

 

 

Sent from Mail for Windows 10

 

From: g3kevmal@talktalk.net
Sent: 02 April 2017 15:28
To: marcocadeddu@tin.it
Subject: RE: LF: FETs getting sick - why not use tubes?

 

HI PA AMP folks

 

Surprised more LF/MF ops do not use the DECCCA 5501 design, class D mosfets in bridge configuration with series resonant output then matching ferrite transformer to match either 50 or 75 ohms.

I have been using this method now for several years and todate not had a failure.

This design is narrow band but only needs the series resonant cct switched for band change

Ie SMOKE FREE DESIGN

 

Good luck all with your PA projects and also agree Tube PA,S are hard to beat. I use Dentron PA amps on the HF bands, These amps have a pair of 8875 tubes and live a long time.

73 gl de MAL/G3KEV

 

 

Sent from Mail for Windows 10

 

From: marcocadeddu@tin.it
Sent: 02 April 2017 14:34
To: rsgb_lf_group@blacksheep.org
Subject: R: LF: FETs getting sick - why not use tubes?

 

Hi Tom, Wolf and the group,

 

thanks for feedback, suggestions and nice pictures of the tube TX! :-)

very nice to see!

If I had not room restrictions probably

will ... [I should have hidden in some place some 6146B and PL519s as

spare parts of older rigs ;-) ].

 

The pushpull FET PA served without problems for at least 5-6 years..

withstanding several kind of insults (mismatchings.. missing antenna..

wspr15 etc..) this time the first FET shorted out after a ceramic

trimmer in the directional coupler arched and it was the first of a

long list: replace/short/replace/..

There is no bias (simply a couple resistances in serie to gate and from

gate to ground) driving the pushpull with a IXD609 via a transformer.

The PA is broadband and is followed from an external LPF for LF or MF.

 

Hope next FET will survive enough time to check what is going wrong...

 

73 Marco IK1HSS

 

----Messaggio originale----

Da: DK1IS@kabelmail.de

Data: 2-apr-2017 14.29

A: <rsgb_lf_group@blacksheep.org>

Ogg: LF: FETs getting sick - why not use tubes?

 

Am 02.04.2017 um 13:09 schrieb Dr. Wolf Ostwald:

> hi Marco !

> FETs die in many peculiar ways. Often they just become half

conducting

> on the Source-Gate path. If they are on a common bias source, the

> Source voltage will open up all FETs connected. Leads to immediate

> death of all of them. Use individual bias on each of them. I had

that

> happen a few times b4 i learned my lesson. Of course capacitive

> blocking of DC from the input transformer is essential too. But i

> think these thoughts have been published here b4.

> I am in the process of putting a single tube to use on 472. But its

a

> time consuming enterprise.

> 

> 73 de wolf df2py

Hi Wolf and group,

 

nice to hear that someone else is thinking about this approach! I´m

content with my homemade tube PA for LF and MF which has provided

reliable service since nearly 4 years now. Only some thoughts about

this

concept - I hope not to bore all those hams who are happy with their

semiconductor PAs:

 

Years ago I had a MOSFET PA for LF, Class B push-pull with 250 W RF.

It

worked well at constant conditions, but when I had to retune the

antenna

due to larger QSY or made antenna experiments there always was the

danger of blowing up these nervous semiconductors. After 4 or 5 times

changing the MOSFETS I decided to build a new PA - with tubes! Looking

a

little bit anachronistic this PA is absolutey good-natured. Designed

for

broadband service on LF and MF it makes no problems when changing the

antenna coars tuning  from one band to the other even when the fine

tuning is´nt done yet. With my former MOSFET-PA this would have been

impossible.

 

I wanted to have a linear PA - this usually means class B. You have to

decide between narrow band and broad band (like an audio-amp) design.

For narrow band you can use a single-ended PA but you have to add a

resonance circuit. For broad band you should use a push-pull PA and

have

to build a suitable output transformer. I opted for broad band design

because it is usable for LF and MF without changes at the PA. With

this

design and sin-driving I reach a total harmonic distortion of about 5

%

at 700 W RF on a pure resistive dummy load. With the usual narrow,

narrow band antennas on LF and MF you don´t need additional filters!

 

Concerning the tubes: If you take the common TX tubes with plate

voltages of several kV all output circuits have rather high

impedances,

that means large coils for the resonance circuits resp. large

transformer windings and very high voltages - potentially a

construction

problem. This led me to the choice of 2x 4x PL519 in push-pull, a

rugged

colour TV line output tube with low plate voltage and high plate

current. In this way I came down to a plate-to-plate resistance of

about

1 kOhm at 600 V DC plate voltage, where you easily can build a ferrite

broad band output transformer down to 50 Ohms. A disadvantage of this

concept is that you have to give individual bias to each tube, that

means for the first start-up you have to align 8 potentiometers

carefully to nearly equal cathode currents for all the tubes. But

according to my experience this alignment remains stable over a long

time. I have inserted 1-Ohm-resistors in each cathode line and have

brought the voltage drops to 8 cinch connectors, where I can monitor

the

DC component (with external filtering) as well as the real time

current.

With 4 tubes in parallel per branch of course you have to take care

for

self oscillations. The extensive use of bypass capacitors, ferrite

beads

and parasitic chokes in the plate lines is mandatory as well as good

grounding concepts are. The tubes don´t pull control grid current

(this

would even be true in class C!) but you need 3 or 4 W RF input power

due

to all the ohmic loads at the tube´s control grids caused by the

individual bias paths. On the other hand this certainly helps to avoid

oscillations. You can see some pictures of this PA at www.QRZ.com/db/dk1is.

 

By the way: why not to try these tubes at class D? With DC plate

voltages of perhaps 1200 V you should get a nice QRO-PA ...

 

Wolf, you are right: building such a PA from scratch is a time

consuming

enterprise. I didn´t count the working hours but according to my lab

log

the whole project took about 9 months - an adequate time for a new

baby!

It was a great experience anyway.

 

Good luck and 73,

Tom, DK1IS