Date: Mon, 23 Dec 2024 09:39:03 -0500
From: "Bob D." <ajamas.rn@gmail.com>
To: amps@contesting.com
Subject: Re: [Amps] Acceptable ripple on high voltage plate power
supply...
<Check my math....
<For 10% P-P 120 Hz(full wave 60Hz rectified) voltage ripple at max current.
<Ballpark math.
<That's 7% RMS.
<Power is proportional to square.
<Ratio of sideband power will be 0.005
<That's 5W total for two sidebands at a KW.
<2.5W each. -26dB
<MThis is at peak power. S meter reading ratio is voltage so will be -52dB
<on meter.
<If a CWv signal is more than 40dB(on s meter) above noise, an exceptional
<ear, that is looking for them, might hear the sidebands.
<I could be wrong. I do recall reading 10% as the max in an old handbook.
<The above us all at max current draw from the supply. Ripple percent goes
<down with current draw. Power in ripple sidebands goes down with square of
<ripple percent.
<73, -bob ah7i
I take the peak to peak ripple..... then divide it by 2.828 to arrive at
the RMS ripple voltage.
Then take the RMS value of ripple, and divide it by the loaded B+, to
arrive at the % ripple.
IE: Drake L4PS supply...(which uses a FWD, with 8 x 200 uf lytics). 2500
vdc loaded with 800 ma cxr. 207 v P-P ripple. (25 uf total filter C
used).
207 / 2.828 = 73.55 vrms.
73.55 vrms / 2500 = .0294
.0294 x 100 = 2.94% ripple.
To express the ripple in -db, I used as follows.
20 x (LOG .0294) = -30.62 db.
I can relate to ripple when expressed in -XX db, much better than %.
2.94% ripple (-30.62 db) is plane as day on any RF monitor scope. It's
horrid imo. Wavy lines at the top and bottom of the scope...instead of a
straight line solid bar.
I 1st picked up on all this when using my Array solutions power master pep
wattmeter's.
The last segment would...'dither'... but only when > 700 w cxr.... and got
worse as PO increased up to the max of 1290 watts. Normally ( at lower PO
levels), each bar graph segment would light up, and the next segment would
slowly light up, until full brilliance, then the process would just keep
repeating itself. ( It also displays PO numerically).
This...'dithering' of the last segment drove me nuts. Then hooked up
the scope...and sure enough, loads of ripple, plane to see. ( but NO ripple
at all with just the 200w xcvr).
The fix is easy, ripple is just inversely proportional to filter C value.
With 75 uf in there ( 2 x Harbach boards), one at each end, and with 330
uf lytics...... ripple is now 1/3.
It's now just 69 V P-P or 24.4 v rms = .976% or -40.21 db.
Problem solved, and no more dithering segments. IMO, 1% or less ripple is
optimum for ripple on SSB. Dynamic regulation is much improved with more
filter C. When a doubler is used, the caps are only charged up (topped
up) 60 x per second..... vs 120 x per second on a FWB. The theory is,
with caps in the doubler only being charged up 60 x per second, they now
have to hold their charge twice as long.... hence they really need to be
double the uf value to begin with.
On my HB B+ supplies, I'm experimenting with a C-L-C filter setup, with a
spst-NC HD Kilovac ceramic vac relay across the 4H, 127 lb, Dahl
choke....rated for 3 amps CCS @ 15 kv. Also a hb adjustable spark gap
across the choke for safety purposes. Choke is in series with the B+...and
totally electrically isolated. Choke can be shunted...on the fly, to turn
it into one big C filter.
It works but is gross overkill. ( buddy across town had gotten the dahl
choke from the now defunct satellite relay station north of me). With a
2.5 amp CCS load + 6800 vdc, ripple is just a mere 33 millivolts P-P or
11.7 mv rms.
(11.7 mv rms / 6800 vdc = .0000017270) or .0001727% or -115 db
ripple.
The 1st (C1) filter cap knocks the ripple way down.....and the following LC
filter section attenuates at the rate of 12 db per octave... starting with
the initial corner freq being just 1.75 hz.
That -115 db is just for the 120 hz ripple. The harmonics of the 120 hz
ripple, like 240/360/480 hz etc, are attenuated even more.
It was just an experiment, to see if actual measurements jived with the
power supply software....(it does).
Using the software, I found out that the C2 cap has to be at least the same
value as the C1 cap. When C2 is much lower in value vs C1, all hell
breaks loose, with wild up/down oscillations for the 1st 500 msecs.
_______________________________________________
Amps mailing list
Amps@contesting.com
http://lists.contesting.com/mailman/listinfo/amps
|