[AMPS] big amps/low parasites
Mon, 25 May 1998 13:17:41 -0700
I'm home today, the first time in a few weekends (it's also Memorial Day, a
gov't holiday). Thought I'd share a few experiences about the 200 KW
push-pull amp that is about to be turned on.
We cannot run the HV yet, as the cable contractors pulled my HV cables (100
feet long each, pre-made by Caton) in 105 foot conduit. Duh. Also, moving
the amplifier to the area where 480 VAC (for filament power supplies) and
water are available is scheduled for the coming week, using a forklift.
Meanwhile, we got all the finger stock on the doors, the casters on the
bottom (amazing that we spent almost 4 days and in the machine shop three
times just to get the casters - wheels - on this 1/2 ton amplifier to work
properly. Never buy your wheels from a distributor who doesn't give you a
drawing of their exact dimensions. These had a sticker on them - made in
My filament ramp up power supply was heat tested this weekend, with
success. Using a saturable reactor to control a 480 Volt single phase line
to the filament transformer, i have a controlled ramp generator for the DC
bias to the reactor. It needs about 1 Amp DC at 22 volts on the control
coil for full power on the filaments. The ramp takes 9 minutes, to meet the
tube manufacturers requirements. Any quicker runs the risk of shorting the
filament to grid, as things have long thermal equilirbium times to expand
and stretch inside the big bottles. Ditto for ramp down, plus a 15 minute
blower on the bottom of the tubes to cool the 5 KW filment stems.
Grid bias pulser will be finished in a day or two, I hope. It goes from
-900 cutoff to about -100 to -200, to establish class A. Duty factor of the
amp is 5%, so dissipation is low. The grid bias establishes the amplifier
pulsing and DF. Plate voltage and screen voltage stay on all the time.
Now for the interesting part, to this group anyhow:
Built a test load for the beast, as it drives a 500 ohm, not 50 Ohm load.
Actually each tube of the P-P amp drives a 500 Ohm. It is really a pair of
single ended tubes, driven from a 180 deg grid circuit, in opposition. The
output pi's drive short lengths of line to the next stage, which consists
of a pair of EEV or Philips/Amperex triodes, operated as cathode followers.
These drive the load, which is a gap in a beam pipe in a proton storage
ring. They have very low output Z, the reason for using grounded plate
Back to the new amplifier, it is the intermediate stage. The final amp is
working already, having been there for 15 years. I am replacing six
4CW25,000A's with the new IPA stage.
Feeding my test load from a network analyzer (8753C hewlett packard) out
the back end (it has N connector at the far end of the globar load
resisitor in a tapered coaxial line), I inject sweep to the PA network.
Using the HP high Z probe (active, 1 Meg Ohm input, 0.7 pf loading) I
measured the response of the PA output from 1 to 500 Mhz, then concentrated
on 1 - 60 Mhz, as the dominant parasitic resonance was found. Similar to
the ones that hams have seen at 100-150 Mhz, I have mine at 45 Mhz. Bigger
tubes, higher C, all contribute. A GDO confirms the big one, sucks out the
GDO effectively. On the sweep, it appears to be Q=120, but the 1 Meg probe
is loading it only slightly. The main tank at 2800 KHz is Q<10. The
parasitic is formed by the tube plate capacitance, the strap over to the
blocker, the main tank tuning capacitor (a large vacuum variable). The
blocker is not a factor, as shorting it out with copper makes no change in
the mode. It is a special 0.02 uF 20 KV ceramic disk made for me by High
Energy. The pi network has little effect on the parasitic, as it can be
touched, varied, even unhooked, and the GDO still shows the 45 MHz thing.
There is a 60 Mhz related to the pi, but this is far out from the plate,
not a concern as it is lower Q and definitely not of the magnitude of the
45 Mhz mode.
Tested a number of techniques to suppress 45 Mhz in the plate output
connection. Since we have not fired up the HV, we are not installing the
ciruits until we see how stable it is. For instance, the neutralization of
the amp could be adequate to high HF. Haven't even checked the screen
bypass inductance, for it can be used to effectively self-neutralize the
tube at some point in the low VHF range. Also, we have a 100 Ohm swaping R
on the grid. It actually terminates the input circuit; the amp has little
grid current (none - class A remember). It helps to load the input to
prevent VHF instability too.
After building big ones and small ones, and even trying mu metal (sorry
Rich, couldn't find any stainless shim stock), I have narrowed it down to a
couple of tricks for an adequate parasitic suppressor. Remember, these are
contingency circuits, to be installed if it decides to really want to run
One is a 1/2 inch wide ribbon, which is copper, wound around a 10 Ohm
globar, one of the shorter ones. This gives about 20 dB of suppression to
the 45 Mhz in the circuit. It went in place of the wide, low inductance
copper strap from the plate to the blocker, that my co worked did such a
beautiful job of making. He shuddered at the contraptions that I was
forcing into his beautiful layout!
We tried two of the above things, but the suppression got worse by 5 dB.
Tried a pair of Ferroxcube ferrite cores next to the strap, which really
were quite effective. These are high mu (>2000), so they saturate easily. I
went to the trouble of trying #43 Fair Rite EMI cores, and several other
lowerer mu materials, but none had the effect like the original cores.
The best so far, was to make a haripin loop of copper, to present higher
reactance to 45 Mhz, but pass 2800 KHz readily, and shunt the loop with a
pair of 10 Ohm globar (Cesiwid 667AS series) resistors at the center. This
knocked the 45 Mhz response down 25 dB, and by adding a pair of 3C8
ferrites inproximity to the copper loop, (not through the cores, just
nearby) I gained 5 dB. All this had NO effect on the 2800 KHz resonse, it
did not drop even 0.01 dB on the sweep. Now, we have to be aware the
harmonic current will be heating things. But 5% DF gives us some relief
Well, gotta go now, will be back with more progress in a few days.
FAQ on WWW: http://www.contesting.com/ampfaq.html
Administrative requests: amps-REQUEST@contesting.com