[Amps] L-PI amp network

[jim.thom jim.thom@telus.net]

Date: Sat, 17 Dec 2022 17:15:35 -0500
From: Steve Bookout <steve at nr4m.com>
To: amps at contesting.com
Subject: Re: [Amps] L-PI amp network

<Jim,

<Thanks a bunch for your input.

<I was close.

<I had used .65 for the current and a "K" factor of 2.0 as I'm 99% CW.?
<It's been DECADES since I was above 28.1 MHz.? I could have easily
<chosen 1.8.?? I also had '9999' for the choke and .1 for suppressors.? I
<also had '5pf' for strays, where you used '3pf'. At least, I wasn't
<totally lost.? I was counting on being able to use GM3SEK's calculator.

<Having said that, I followed along with your numbers, and I can ALMOST
<get what you got.? I'm looking for where my entered data is different
<from yours.? At one point, I hit the comma instead of a decimal period.?
<Boy, did that mess things up.? Had a heck of a time finding the problem.

<Not sure how you got a Q of 19.? I had entered a Q of 12 and that was
<reflected in the C1/C2 info from the get-go.

<How much of a big deal is it going to be if I use 1/4 copper tubing??
<It's what I've got.?? I've got many feet of .015 x 2.00 inch copper, but
<even if I split it, it's too thin to self support.? Do you have a good
<source for a suitable copper strap?

<I'm lucky as I have a 14 inch metal lathe I use to wind these coils.? In
<the past, I've used #4 solid copper (3/16 inch) as I had a bunch of it.?
<Makes good solid coils, but it work hardens, making changes a PITA.? I
<DO put the lathe in neutral and turn it by hand...

<Steve, NR4M

##  You can't use a K factor of 2.0  since that's for class C.  You can't
run any high MU GG triode in class C on CW or you will get severe key clix
on CW.  It has to draw a small amount of idle current for CW use, like
10-25 ma.  On my hb amps, and also my 4 x L4B's, I can adjust the bias V /
idle current ..on the fly.  No point in sucking 180 ma of idle current on
an 8877, when using CW mode. This is why I don't use zener's for bias.  A
string of 6A10's or 10A10's is dead simple, and they have a 400 / 600 amp
surge rating, you won't blow em up, nor overheat them.  A 1N5408 runs
fairly warm with 1 amp CCS flowing through the string.  A spst toggle can
tap the string for CW use.  Then tap the other end for required bias for
SSB.   I'm not a data expert, but I believe u can reduce the idle current
way down for data / rtty / FT-8 modes.

With increased bias V.... you then require a tiny bit more drive..to
overcome the extra bias V.

Enter ZERO for all the suppressor parameters.
The xtra stray C is added on both line 43..and also line 53.  I
guestimated 5 pf for line 43 and 3 pf for line 53.   I don't have a 8877
tube + socket to test, and measure.

Without a digital LCR meter, you are dead in the water. It's ur number one
tool when building amplifiers, or anything else.  Then you can map out both
the tune and load caps in fine increments and make up a cheat sheet.
 Ditto with measuring coil values.   I also use it to measure ALL the stray
C.  The 8877 is 10 pf from anode to grid.  Once in the socket the anode to
grid C will rise a bunch...due to the proximity of the lower anode fins to
the chassis below em.   You will also get more stray C from rear and side
cabinet walls..and also the top lid.  This is why you have to be careful
when mounting the tube, when amp is used on upper HF bands, like
17-15-12-10-6m.  Using the lcr meter, between anode and chassis, then you
know exactly how much stray C  you have..and can enter into the spreadsheet
on line 43.   If tube is mounted in the back corner, u will get stray C
from both the sidewall and also rear wall. In that case, the tube's anode
has to be further away from the side / rear walls.  C between anode and
side / rear wall is inversely proportional to distance.  IE:  1" spacing is
X pf.   1/2" spacing will be 4X pf.  Same concept as any air variable cap.
On that 3CX-6000A7 6M amp, I had scott keep the anode a full 2.5" away from
the left sidewall, and also 2.5" away from the rear wall.  Any amount of
stray C will screw up everything. Good news is, stray C is easily
measured.  Once known, plug it into the software.

Copper strap was obtained locally at my metal supermarket.   They have 3' x
7' sheets of copper, in thickness from .015"   all the way up to .125".
.032" is ideal, and easily holds it's shape.  I had em shear in 3' lengths,
in 3/4" widths..and also 1"..and also 1.5" widths.   You can buy copper
strap from Georgia copper in .032" thickness..and in several different
widths.  https://www.gacopper.com/032-strap.html

They also have it in 12" lengths in 1" widths.

Copper flashing works good too...and is cheap.

The huge advantage of strap coils is...they are easy to snip off, say 1/4"
at a time, and repunch the hole with a roper whitney or eq punch tool.
They come with dies from 3/32" up to 9/32"...in 1/32" increments. 7 x dies
included.  You can't drill  copper strap, punch it.  I use the same punch
to punch 3/16" (6/32")  in .125"  thick 6061-T6  plates all the time ( for
10-32 screws)...and also  5/32" holes for 8-32 screws.  Dirt cheap on
amazon, HF, etc.

When u snip off the end of the strap coil, u simply twirl the entire coil
a few degrees, so the newly punched hole fits it's termination on the cap.
What u are doing when twirling the entire coil is.... you are reducing it's
diameter a tiny bit.   That's a bitch to do when using tubing.

If you are going to use tubing colis, download K6STI's 'coil' program.
It's dead on vs measured coils. It will only work down to a min of 2
turns.  It will do fractional turns, like say 2.75" turns etc.  BUT you
have to know where each end is gonna terminate.  On his coil software, you
can change the diam of the coil to anything you want..and also it's total
length. and also factor in the leads on each end of the coil  (more tubing,
and usually flattened on the extreme ends).  It only works for tubing or
wire coils, NOT strap coils.  It's very close though, when using strap
coils. The strap coils use wider material, but have minimal spacing between
turns, so it comes out a wash.

When I mentioned a Q of 19..that was withOUT the 1st coil !!  That was just
a normal PI network.  Which is why I don't use a normal PI network on
15-6m.

1/4" tubing runs damned hot on 15M..if you use a dead cxr for say 10-12
seconds, let off the footswitch and touch the coil, it will take ur finger
prints off.  Ok for cw / ssb..and that's it.

Tubing coils only conduct current on the outside of the tubing, never the
inside (like water inside a pipe).   Tubing coils also don't conduct on the
inside of the completed coil. The current bunches up on the outer
circumference.   IE:  1/4" OD tubing has a  .785" circumference..and only
1/2 of that handles the actual current.... = .3925"..which is miniscule.
 IF strap is used, it will have to be .3925" wide to handle the same
current. 3/4" wide strap will handle 90% more current  vs  1/4" tubing.
1" wide strap will handle  2.55 X more current vs 1/4"  tubing.

Current handling capability decreases as to the square root of the ratio of
the 2 x freqs.
IE:  A coil on 28 mhz will only handle 1/2 the current as the same coil
material on 7 mhz.(28 / 7 =4.   sq rt of 4 = 2) And circumference is
directly proportional to diameter.  3/8" tubing will handle 50% more
current on any freq..vs 1/4" tubing, etc.

For 10m, both coils are very small  (.8 uh and .91uh)  When coils are that
small, that's when strap coils make sense.  You can use almost zero spacing
between turns on a strap coil. A tubing coil will require at least it's
tubing OD for the spacing between turns. (You can see the effect on both
UH....and also the coil Q asap, on the coil software. UNloaded  Coil Q is
just the XL of the coil  /  ESR of the coil.  The coil software will spit
out Q, XL, and ESR.

OK, here is the interesting part that folks always forget.   XL is directly
proportional to freq...since XL = 2 x PI x F x L

ESR, (effective series resistance) or RF resistance  is NOT proportional to
freq.  ESR on 28m is only double  vs 7 mhz.  Meanwhile the XL on 28 mhz is
quadruple vs 7 mhz.

What this means is the coil on 7 mhz might have an UNloaded Q of say 500 on
7 mhz.  Same coil now has an UNloaded Q of 1000  on 28 mhz.  Joe ham thinks
his 28 mhz coil is superb, since it has an UNloaded Q of 1000.  In
actuality, the 28 mhz coil has DOUBLE the ESR  vs the 7 mhz coil.  With the
same RF current, squared X double the ESR, the 28 mhz coil now has TWICE
the watts dumped into the coil..... vs 7 mhz.

When we talk about loaded Q (total network Q) of a PI net, higher loaded
Q's result in directly proportional higher tank circulating currents.
Current in the main coil with a Q of 19 is gonna be a lot more  vs a loaded
Q of just 11. 19/11 = 73% MORE current with a Q of 19.

When I switch from SSB to say a dead cxr, the average plate current on SSB
( with a lot of both inboard + outboard processing)  is exactly 1/2 of the
key down plate current. With a dead cxr, average plate current has now
DOUBLED..and the amount of watts dumped into the tank coil has now
QUADRUPLED.  It's just I squared x R.  In this case it's RF current
squared x  ESR of the coil.  Now you can see why so many amplifiers burn up
bandswitches, crack the ceramic, melt the polystyrene on airdux, etc, when
used on FT-8..which appears to be all the rage these days.

The way I deal with heat is..... don't generate as much of it in the 1st
place..which implies running the lowest loaded Q as practical.  On the
lower bands, a loaded Q of 7-10 is ample....resulting in more than
sufficient harmonic suppression.  The tube in GG already has 6-7 db of 2nd
harmonic suppression, before the PI / L-PI  / PI-L / L-PI-L  network is
installed.  My ants are not resonant on their even harmonics, and my loaded
40m yagi, due to it's loading scheme, resonates on 24 mhz..and not 21 mhz.

Ok, after getting the loaded Q as low as practical, next up is the coil
loss..and the only way to minimize that is to use bigger OD tubing, or
wider strap.  Current handling capability is directly proportional to
circumference of the coil..which is directly proportional to the tubing
diameter used.  With strap coils, it's proportional to strap width.

The 11m ops have not figured out the L-PI  dual coil setup..yet.  They
instead use too high a loaded Q, then try to reduce the loaded Q  by
reducing the plate load Z by overdriving the tube. The problem is, with
their typ 5-10 amps of plate current, their 3 x turn tubing tank coils run
so hot, that in some cases they will crack the ceramic on both the ceramic
vac tune + load caps.

BTW, you can always take your .015" x 2" wide strap, cut it lengthwise,
then double up, to double the thickness to .030".   That's been done
several times before, and works good. You can solder the mating edges in a
few places. It's the outer coil  layer, that's handling all the current.

Jim  VE7RF