Filter cap options

[KA9EGW]

Moving right along...I have my HV transformer, rectifiers are on order, choke is en rotue...next question is filter caps...curious to know how many folks here use paper-and-oil vs. how man use electrolytics, and if anyone has recommendations for best brands, etc.  Thanks in advance.  73, Brian KA9EGW

[w4bfs]

take a look at the axial cap discussion on this page as a good starting point

a lot of formerly good product is falling by the wayside with its production being moved to a facility with less quality control ....

paper in oil has a finite lifetime as does electrolytic, but it may be very long ...

we need verifiable component data .... good luck in finding it !

[N2DTS]

I use oil filled caps for all high voltage.
They seem to last forever if run within their ratings.
I do NOT like strings of electrolytic caps.

What max voltage are you going to run?
I have a bunch of oil filled caps that need to find homes, most are 1500 volts, but some are 3000.

I could include some with the choke.

Bleeder resistors, step start relay, meters, circuit breaker, HV wire, control relays, chassis, what parts do you have?

For bigger stuff, I often put the transformer on the floor in the rack, or on a rack mount shelf with plastic between the iron and the shelf.

A pair of 813's likes 2000 volts @ 400ma for maximum power output.
That is way over the legal limit at about 600 watts carrier.
What is the modulator going to be?
What is the mod transformer going to be?

[Opcom]

just an opinion: If there is an oil that fits and does the job (enough uF @ the voltage) then it is chosen over electrolytics.

Consider that a stack of equalizing resistors drawing enough current to really help when the caps start drying up and changing value is like a small annoying bleeder. Regardless of if it is a 10mA current through the equalizing resistor, or a 50mA current because it -is- the bleeder too, all its heat is right next to the caps making the aging worse.

The caps all have to be separated because their plastic sleeving won't hold off much voltage if they are all touching.

If a cap opens or just dries up, all the ripple is put across its resistor and the smoke is the warning. If a cap shorts, nothing may be noticed. Maybe a HV fuse ought to be put in series with the hot end of an electrolytic stack.

The only reasons for using a stack of electrolytics for >1KV is for saving cost and weight and for marketing. Like the typical 65 LB 1KW amp that sits on a desk. Give me an oil cap (or parallel bank of them) and a remote bleeder any day. I'll make the space for it. A choke input too if the transformer has enough voltage.

[W7TFO]

+1

73DG

[KA9EGW]

At this point I need to gather up a bunch more funds before I even start looking at plate-modulating this amp...heck I haven't even got all the parts for the RF deck yet and I'm still gathering power supply parts...I do intend to build it able to vary hte grid bias so I can run it in linear OR class C service.

What I have [or have ordered, paid for and in transit] at this point are 2 813's, 4 813 ceramic sockets, the filament transformers, a Johnson 7500-volt 100pF-per-section plate cap, the plate HV transformer, the screen-and-bias transformer, and 4 16kV 1A+ block rectifiers. 

I have a line on a plate tank coil if I decide to go push-pull...I've got the article from July '54 QST wiht a pair of 813's in push-pull, but I question if I can meet current linearity standards that way.

Since SSB is #3 on my 'operating modes' list behind CW and AM [but nonetheless sometimes useful to move traffic] I haven't put a value on linearity yet.  Single-ended might be better?

I freely admit to being a total novice at this; my most QRO [and only] AM rig to date [licensed since 1979] was a Heath Apache...

[N2DTS]

I would go pie net and make coils out of copper tubing, using banana jacks for plugging them in, or do the usual  B+W tank coil assembly.
What about the screens?
Different voltages needed for class C and AB, with the AB wanting regulation.
A pair of 813's as an amplifier is wimpy on ssb and very wimpy on AM.
600 or 700 watts out on CW or AM.

Push pull circuits means you have to make everything balanced and symetrical, and you need neutralizing caps, big ones for 2000 volts. Loading is done with the link, so you need to adjust that, not always easy to have the control in the right place on the panel.
I used chains and sprockets on my push pull 812 rig. Hard to find these days.


Then there are the grids, and the input network.
Plus protective bias and grid leak bias, metering of everything, screen current overload relays and resets, etc.

It gets quite large and complicated if you want an all mode, all band RF deck with all the metering and protection circuits, fixed and grid leak bias, adjustable screen supply with stiff regulation.

My 2x813 rig has everything adjustable, but its big, 6 foot tall...

What is the prime purpose for the amp?

[KA9EGW]

Oh, no, no, no, I won't try to linear-amplify an AM signal.  Hence my desire to switch between Class C [for CW and--eventually--plate-modulated] and AB1 for SSB.  I would guesstimate the amp will spend 85+% of it's life running Class C...and 85% of that below 7.3mc...

[KA9EGW]

I believe I shall go single-ended.  K.I.S.S.

[N2DTS]

For class C, you do not want the screen regulated, and for AM you want it to self modulate (choke or resistor).
For AB1, you need it regulated.
For class C, the voltage is 350 or 400 volts, for AB2 its 750 volts.
Power out for a pair is 450 watts on ssb.

Since the 813 runs AB2, its not going to be very clean I suppose.

A grounded grid amp would be better for cw and ssb, and other tubes would do a lot better.
A single 3-500 is twice the tube of a pair of 813's.
 

[KA9EGW]

I appreciate the input.  Considering that about the only time I get on SSB is to pass traffic I can't unload on CW, and I live less than 35 miles from the Section Traffic Manager and can hit him with 5 watts on 80 cw or 25 watts on 2-meter FM, I think worrying about using these [already paid for] 813's on SSB is a waste of brainpower; what I'm hearing is there's just too many design compromises and too much complexity to get this amp to be "everything to everyone" so to speak.

Taking that into account I don't think I want it to be anything more than my original purpose--plate-modulated real AM...

[N2DTS]

Well, its a LOT easier to build something for one mode.
Its even easier and cheaper if you build it for one or two bands.

The 1967 (and other years) have a one band kilowatt amplifier project using 2x813's.
It does class c and ab1, and you can build RF decks for the bands you want and switch them.
Page 218 in the 67 handbook.

813's and sockets are still cheap enough to do it that way.

Since I only operate 80 and 40 meters, 40 meters 95% of the time, I built some rigs to only do 80 and 40 meters. A switch shorts some of the 80 meter grid coil out for 40 meters, and I use plug in tank coils made out of copper tubing.
Ceramic standoffs with banana jacks mount the coil.
High Q, low loss, dirt cheap.

Here is another design, AM only:
http://www.amwindow.org/tech/htm/813/813.htm

[KA9EGW]

Oh yes--the same one-band kilowatt is in the '63 on page 202...two of them plus a metering panel and a control panel in a tabletop rack; that is the design that inspired me to use 813's to begin with...this has been rattling around the back of my mind since I was gifted a '64 HB in '79 when I was first licensed.  I can't see any use for coverage above 40M, last time I went above 40M was on 10M during the '79-ish sunspot peak...and I see nothing whatsoever wrong with plug-in coils, and I do very much like the idea of avoiding the complexity of bandswitching the plate tank...besides 813's have such a nice glow to them...

[N2DTS]

There are ways to bandswitch easy, the old B+W tank coils, a big coil with a big switch, for the grids you can just wind coils on PVC pipe for each band and switch them in.
With a pie net, you can put parts anywhere within reason, the grid stuff should be seperate, and with many tubes, you do not even have to worry about neutralization.

For a 2 or 3 band rig you can do a shorted grid coil and plug in coils for each band very easy, all it takes is one small switch.

813's do not glow, they light up, but the graphite plate models normaly show no color.

It seems easy, but its a big job to build a rig, lots of chassis, parts, a rack to put it in, front panels, matching meters, knobs, switches, HV wire, hardware, bleeder resistors, relays, sockets, connectors to connect everything, loads of high voltage feed through insulators, variac's, paint, vernier drives, terminal strips,  etc.

Most guys building have collected parts for many years, from back in the day when stuff was cheap and at every fest.

Are you up for all that?

[KB2WIG]

"It seems easy, but its a big job to build a rig, lots of chassis, parts, a rack to put it in, front panels, matching meters, knobs, switches, HV wire, hardware, bleeder resistors, relays, sockets, connectors to connect everything, loads of high voltage feed through insulators, variac's, paint, vernier drives, terminal strips,  etc."

Or not.... ..

[KA9EGW]

My bad--I ought to have said "813's light up very nicely".  Having a complete machine shop, I see no problems with the metalworking aspects.  I don't aim to approach this as a rush job, but you're right, I am one of those guys been collecting parts a long time for this.  I didn't mention several different roller coils one of which would be very useful in a pi-net if I don't go plug-in, but it looks like plug-in will be the way I go...

[KA9EGW]

I HAVE toyed with the idea of gearing that monster roller and the cap together so they both cover their ranges in the same number of turns, a la a Viking II, but for a one- or two- [or at most three] band rig I feel that's adding unnecessary complexity...besides, although I can machine very small-DP [small tooth size] gears from bar stock, I prefer not to have to HI.  

The design is coming together--single ended, parallel toobs, class C...which is what it started out as before all the turns down all the various blind alleys HI.  Too much info out there and too many options--better I stick with what the nice folks on this forum tell me works...

Hope that last didn't sound sarcastic; it was meant literally--the folks on this forum are nothing but nice people in my experience.

It does beg one question though--if Class C plate modulated is supposed to have the screen modulated too, what is gained by having the separate screen supply shown in the Tesla 300 plans?  I've got a pile of big ceramic power resistors in ratings up to a couple hundred watts...

[N2DTS]

Its really a bunch of old grumpy know it all's here, but it does not always come across the internet that way...

Roller inductors are great, and ganged with the tuning works great, that is how my 813 rig is set up, no output band switch.
But you need a nice big roller inductor for AM, and if you run it seperate from the plate tuning, it can be all over the place.
Q is the ratio of inductance to capacitance, high Q is good for harmonics, low Q is good for efficiency, and with a roller, it can be all over the place, a little more C and a little less inductance, or the other way, who knows what the Q is...

Then you need counters, I like having a way to reset or pre set everything.
In the past, I used a bunch of the bigger vernier drives because they had scales on them, but they no longer make the big ones as far as I can tell.

All my building is done with a hand drill, files, and some punches, even a drill press would be nice!
I used to get gears and chains from Herbach? A local place, but they are gone I think, last time I went they were in with a trucking company....

There are places that sell that stuff (for robots and such) but I never researched it.

Screens:
For AM, if you have a stiff screen supply and put the modulation on the plate supply, you do not get good modulation on some/most? tubes as the screen has a lot of control.
Most tubes will self modulate the screens ok with a choke, a resistor, or both.
Plate current goes down, screen current goes up, with a resistor, you get a voltage drop through the resistor which lowers the voltage. Self modulation.

Screens are fragile, so you can never allow too much current to them.
The way I do it is to make a variac supply, choke input, oil filled cap output.
Big resistor in series to the screen (big wire wound pot allows adjustment), then to an overload relay.
I make those, a regular relay with the current flowing through the coil, bypassed with a pot and a small cap.
The pot sets the current trip point, and if the screen current exceeds the set point, it pulls the relay closed, and on closing, connects the screen supply to a resistor to ground to hold it closed.
I put a button on the front paanel that lights up when its tripped, and when pressed, breaks the path to the relay coil for a reset.
I often do bone head things so I need protection.

The series resistor (big pot) in the screen supply allows a wide range of adjustments without exceeding the screen current.
I run about 100 volts extra, then drop 100 volts in the screen resistor, and can tune a lot without tripping the overload.
I can also do tests and adjust things to get the lowest distortion.

The cheap easy way is a dropping resistor off the plate supply.
I do not like that as without current, you get full plate voltage on the screen.
If you go that way, be sure to include an overload relay, as you can toast a screen real fast without one.

A fixed supply also allows you to remove the resistance and set the voltage and run the tube in AB.

Be sure to pick up an older Bill Orr handbook, chock full of great tube stuff.

Besides the choke, I will send you a bunch of stuff I have no use for, use it or move it along to someone else who is building.

[KA9EGW]

If I understand correctly, the pot, the coil and the cap are all in parallel...and the higher the resistance the more current goes through the coil and hence the lower the trip point.  Do I read that right?

[N2DTS]

Yes.

[KA9EGW]

Of course I need a relay that pulls in at a current lower than the rated max screen current 

[N2DTS]

I have an old heathkit power supply that does 0-50 volts at 1.5 amps, with metering that helps build things like this and test them.

[KA9EGW]

I've got a 60V, 30A Motorola variable bench supply for the same reason...the hard part is not specifying components, it was the deciding what form the amp would take.  I enjoy building stuff, no reason I can't build two.

[KA9EGW]

Carrying on with the design process...I downloaded PIEL and it tells me a pair of 813's in parallel, running at 3kV, 333mA [admittedly pushing them very hard, but this is a proof-of-concept exercise] for 1kW in, 700W out, Class C, Q of 12, requires a pi-net of 83pF, 25uH, 252pf.  Sound about right?

[w4bfs]

modulated impedance around 10k Ohm      rf impedance around 5k Ohm

I would try for a lower impedance to make iron choices easier