The AM Forum

Hi guys. Small time poster but big time lurker here.  ;) After several years of wanting and wishing for more power but not having the means to acquire it I think I am now ready to. I have at my disposal a 3CX3000A7 and basically all the necessary parts to put it too use. My goal is to build a heavy duty 160-10m  amplifier that can be used on AM as well as SSB without worrying about duty cycle and over stressing the tube.It will be grounded grid for simplicity and stability. It will not be run flat out for all I can get out of it. My question is what to use for a filament choke? I have seen the bifilar chokes wound on ferrite rods and torroids however I have no idea the proper mix used. Type 43 perhaps? Fifty-plus amps requires large wire however I have been reading about the idea of floating the filament transformer above ground on an insulator and installing a bifilar choke in the filament transformer primary leads which would require a much smaller choke. The primary and secondary leads would be RF bypassed and the center tap have a choke installed to ground. As I understand it the capacitance of the transformer to ground forms part of the input tuning circuit. Has anybody done this with any success? It would sure make it easier to build the filament choke circuit if it is an acceptable method of creating a wideband HF choke circuit. While I do have some experience with the 3CX3000A7 in commercial FM broadcast service running one on HF is new to me.  ???

Sounds like a nice project.

I have several homebrew linear amps that all use the same type of filament choke. It is simple and least likely to have any resonances....

Run two short insulated wires from the fil transformer to the socket.  Use #4 wire or heavier.  Place as many ferrite sleeve cores over BOTH leads as you can fit.  (Both leads go thru each ferrite core)   6 to 10 will usually do the job, but measure the choke as described below.  There is minimal voltage drop at 50 amps when using these short wires.

I use type 43 cores, about the size of a small egg, but they can be smaller like the size of a thumb.   Depending on wire length and number of cores and size, it is not hard to get 500 ohms of inductive reactance on 160M. You will want about X10 the tube's input impedance at the lowest frequency to be an effective choke.  On 75M and above, the choke will be 1K and higher. Since there are no coil turns on this choke, inter-turn capacitance is not a problem on the higher frequencies above 40M, etc.

To test it before installation, connect a 500 ohm carbon resistor (non-inductive) in series with one of the choke legs and feed a 1.8Mhz signal generator  from one end of the choke to one end of the resistor.  Using a scope, you will see 1/2 of the RF voltage across the choke and 1/2 across the 500 ohm resistor IF the choke truly is 500 ohms.  Calculate whatever split you measure for an exact number.

Mount the choke away from the chassis and you will have minimal capacitance to ground. Use the leads as their own support. #4 wire is pretty stiff for support.


Hope this helps.

Tom, K1JJ

Thanks for the reply Tom. So you are saying to simply run the filament leads straight thru the cores and not wound in a normal torroid fashion.Sounds simple enough. I never thought about that. I suppose it should work as long as it is good down to 160m. A handfull of smaller 43 mix cores is cheap compared to a fist full of rods or a couple doughnut sized torroids and a few feet of #4 is easier to deal with than winding at least #8 as a normal choke. The idea of floating the filament transformer is what I find interesting and would like to see some more info on it but I do  like your idea and may go with it.

Thanks again.

Yes, straight wires run thru the cores - no turns.


Just to be clear, when using this method with the choke in the fil secondary,  floating the fil transformer will not improve anything because there are  .01 caps to ground on that side of the choke anyway. It is really the cold end for RF.  In fact, maybe it would be better to have the fil xfmr mounted on the chassis directly to help bypass any RF that is there.  Correct me if I'm wrong.

BTW, don't make the mistake of running each lead separately thru its own set of cores. This will produce filament hum in a directly heated tube. Axe Frank / GFZ about that error.  The cores always go over BOTH leads.

Also, the cores are not the common donut cores with big holes, but rather the longer sleeve-type cores, about 2" long with a 1/2" hole.   They are not critical as long as we measure the choke before use and make sure it is X10 impedance on 160M as described in the post above.

T

Gottcha 100% Tom. The idea of floating the fil xmfr was something I found on an amps reflector site and was by Rich Measures, AG6K.

Another quick question Tom. Where did you get your cores from? I have lots of cores that are about two inches long or so and large enough but I have no idea what material mix they are. they were simply pulled off old power cords from salvaged equipment and are unmarked. I suppose some experimenting is in order.

Tom

Just a minor point (not to detract in any way from the approach to making a filament choke that you have recommended):

If you perform the test that you described... with a 500 ohm (non-inductive) resistor in series with one leg of the choke... and assuming the inductive reactance of the choke* is 500 ohms at the RF frequency being used in the test:

The amplitude of the voltage measured across the resistor will (as you said) equal the amplitude of the voltage across the choke ... but neither of these will be half of the amplitude of the applied RF voltage source.

They will each be 0.707 of the amplitude of the applied RF voltage source, because the voltage across the choke will be 90 degrees out of phase with the voltage across the resistor.

*I.e. measured from one end of the chosen leg to the other end of the chosen leg... which is the same as the inductive reactance that the choke will present when both legs are used to connect the filament of the tube to the center-tapped filament transformer secondary winding.


Stu

Thanks for the extra info, Stu.   Yes, relative scope readings across the components is a better way of saying it.  

I think old power supply ferrites may have too high of permeability, no?  But you can easily test a sample using clip leads as the wires.  Type 43 are more for RF and can be purchased on the web.  I don't remember the source, but a few places carry the bigger cores that will do the job.  Maybe someone can suggest a place or you can simply do a search.

I suppose using a higher permeability core, higher than type 43, would mean less cores. I don't think there is any problem with saturation since there is equal opposite - canceled current going thru the cores, right Stu?  How would 60 Hz AC affect it?  

Possibly only the unbalanced 1-2A of tube plate current is all the cores need to deal with so they can be just large enough to accept the heavy diameter wire, I dunno.

I never spent any time testing different cores - just used fat type 43's and they worked very FB.


T

This link to the new Palomar Engineers company has on-point information:

http://palomar-engineers.com/ferrite-products/ferrite-beads

Jim
Wd5JKO

Should the pair of filament feed wires be twisted (if possible for that guage size and relatively short length) to help eliminate hum?

I Guess the answer is that if you can't get several full turns it's not worth it.  Maybe for longer runs because of, say socket placement issues it might help.

Type 43 cores are very common and will work well as has been proven. I like to use type 31 below 10MHz. 43's start to roll off somewhere below 20 MHz.

Tom

You are correct regarding the 60Hz filament current. The equal and opposite 60Hz filament currents in the two legs passing through the cores will produce magnetic fields that cancel. Therefore, there will be no saturation issues with respect to the 60Hz filament current.

As you mentioned, the other possible source of saturation will be the DC plate current that flows through the legs (split between the two legs... but the split doesn't have any significant effect).

Large cores (0.5 inch ID or larger) made of type 43 material don't appear to have any saturation problems with DC currents as large as 10A (with only 1 turn... corresponding to the case here), in Steve's (WA1QIX) Class E transmitter designs. So, I don't think that saturation, due to DC current flowing through the legs will be an issue, as long as each core's inside diameter is greater than or equal to 0.5 inches.

[I.e. since there is only 1 turn (wires passing through the core, but not wrapped around for additional passes... the number of ampere-turns is equal to the number of amperes. The DC current-handling capability (to avoid saturation) of each core will be proportional to the inner diameter of the core. A 0.25 ID Type 43 core should be able to easily handle 5A of DC. The inductance of the choke will depend upon the inner diameter of each core and the outer diameter of each core. It will be proportional to the log(OD/ID). The inductance of the choke will also be proportional to the total length of all of the cores. For example, a core with a 1 inch outer diameter and a 0.5 inch inner diameter will have the same inductance per unit of length as a core with a 0.5 inch outer diameter and a 0.25 inch inner diameter. However, the larger core will be able to handle twice as much DC (total) in the two wires passing through it, as the smaller core.]

Stu

Thanks again for the info guys. I like the idea of a single pass thru the cores and will do some experimenting. I work tonight and tomorrow night so it will have to wait a few days but it is certainly something I want to look seriously at.

 As long as we are on this subject, I have a related question.
 I have a Henry 2000d plasma generator amp that I plan to use on 160, 80 & 40.
 In the original configuration, it operated at 27.125 mhz ( I believe that is correct, or whatever the ISM freq is, which is in or near the cb band)----anywho, all has been ripped from the chassis and new caps and tank coils are in the works.
 The filament choke Henry supplied is 12 turns ,1.75 inch diameter, 6 inches in length, fabricated from 1/4 inch copper tubing, with no. 12 copper and a teflon sleeve passing through the center of the tubing, which was then formed into the coil.
 Any thoughts on using this choke ?
 thanks, Ed  AB3HT, ex wb3eii

Similar unit .
 
http://www.mjhaplus.plus.com/g4fph/html/henry.htm

 choke in question can be seen in last two photos.
 73 Ed.

Ed

Using the web-based calculator: http://hamwaves.com/antennas/inductance.html

12 turns

1.75" = 44.45mm average diameter

6" = 152.4mm length

0.25" = 6.35mm wire diameter

1.85MHz frequency

One obtains an inductive reactance of j19.5 ohms

This is probably much too low for your application

At 27MHz, the inductive reactance of this choke is: j279 ohms

Stu

 Thanks, Stu, that's sorta what I figured, good at higher freq, not enough at the other end of the bands.
 I have a rather large reactor from a commercial lamp dimmer, it is a toroid, 4 inch outside, 2 inch inside, 1.5 inch height.
 It is wrapped with 38 bifilar turns, of number 10 copper.
 I have no clue what the core material is.
 Hopefully, this is something I may use here.
 Thoughts?
  Ed.

Ed

Depending upon the core material, this choke may have problems with:

Saturation, when DC cathode current flows through it
Insufficient inductive reactance at RF frequencies
Heating, due to hysteresis, when used at RF frequencies

If the core were type 43 ferrite (not likely for a choke used in a 60Hz lamp dimmer):

It would be good for around 1A of DC without significant saturation (1A x 38 turns with a 2" I.D. core is equivalent to 9.5A x 1 turn with a 0.5" I.D. core... in terms of the peak H-field in the core).

With 38 turns, the inductive reactance would be much more than necessary, at HF frequencies.

With 38 turns, heating due to hysteresis at HF frequencies, in this application, would be negligible... because the RF current flowing in the windings would be very small... and the associated RF H-field in the core would be very low.

However, if the core is made of a material designed for audio frequency use... like silicon steel... then using this choke to block RF would probably be problematic if you need 500 ohms, or more, of inductive reactance at 1.885MHz. However, it might work as a filament choke, if you can "get away" with less than 500 ohms of inductive reactance at 1.885MHz. I.e. the effective relative permeability would be too low at RF frequencies... resulting in a low inductive reactance... and, possibly too much heating due to hysteresis losses. Saturation, due to the DC cathode current, might also be a problem.


Stu

Ed,

    Yer existing choke might work fine on at least 40 with a ferrite core slipped into place.

Dave

stu
 "With 38 turns, the inductive reactance would be much more than necessary, at HF frequencies."
Much more than necessary being just an excess, or perhaps a problem?
   I imagine I could remove the potting compound , and take a great number of the turns off the toroid.
   What is the worst case scenario if choke is used as is?
 I'm trying to get along with what I have on hand, as cash is scarce. (taxes, new roof, kids in school, the whole lot)
  I tried the hamwaves site, and get a 404 message.
dave
 The ferrite core Idea is noted, Along with Tom k1jj's response further above.
 Thanks for the time & trouble.
 Ed

No argument against the suggestions made at all. It is worthwhile to show this coaxial filament choke from the GPT-10K transmitter as used with a 4CX5000A from 2-30Mhz.

Looks like:

4" average diameter (101.6 mm)
12.5 turns
7.14" total length (181.4mm)  [i.e. 7 turns=4"]
0.5" wire diameter (12.7mm)

Inductive reactance at 1.85MHz = 68.5 ohms
Inductive reactance at 3.885MHz = 144 ohms
Inductive reactance at 7.3MHz = 272 ohms
Self resonant frequency: 31.9MHz

http://hamwaves.com/antennas/inductance.html

As an aside:

I wonder how the RF input impedance of a grounded grid, class C, 4CX5000, operating at full specified power output compares to the RF input impedance of a grounded grid, class C, 3CX3000 operating at full specified power output? In class C operation, the differential cathode input impedance (dV/dI) varies throughout the RF cycle. The associated (time-varying) cathode input impedance determines the required inductive reactance of the filament choke.

For a 4CX5000 (Svetlana specification sheet), at a fixed plate voltage of 7000V, it looks like a change in grid-to-cathode voltage from 0V to 100V will result in a change in plate current from 5.25A to 11.5A. Therefore, in that range of grid-to-cathode voltages, the differential cathode input impedance, dV/dI, is roughly 100V/6.25A = 16 ohms. When the grid-to-cathode voltage changes from 0V to -100V, the change in plate current is from 5.25A to 1.25A. Therefore, in that range of grid-to-cathode voltages, the differential cathode input impedance is -100/-4A = 25 ohms. Therefore, it is not surprising (to me) that a filament choke having an inductive reactance of 68.5 ohms at 1.885MHz would be okay.

For a 3CX3000 (Svetlana specification sheet), at a fixed plate voltage of 4000V, it looks like a change in grid-to-cathode voltage from 25V to 125V will result in a change in plate current from 0.75A to 4.75A. Therefore, in that range of grid-to-cathode voltages, the differential cathode input impedance is roughly 100V/4A =25 ohms. This appears to be about the same (but still somewhat higher) as it is in the case of a 4CX5000.



 Stu

With 38 turns, 4" O.D., 2" I.D., and 1.5" length, this toroid would have an impedance of k x 86.4 ohms at 1.885MHz... where k is the effective relative permeability of the core at 1.885MHz.


Reference: http://hyperphysics.phy-astr.gsu.edu/Hbase/magnetic/indtor.html

Even if the effective relative permeability of the core material is very low at HF frequencies, or if the core material saturates as a result of the DC current flowing through the coil, the inductive reactance of this toroid at 1.885MHz will still be at least 86.4 ohms. I.e. the lowest value k can have is 1... which is the relative permeability of air. If k=1, then the core is serving only the purpose of a form on which the coil is wound.

For type 43 material, k is about 600 at 1.885MHz.

Without knowing the core material... and in the event that it is, in fact, suitable for HF operation (which it may not be)... let's assume that k is at least 100.

If so, then the inductive reactance of this choke at 1.885MHz would be at least 8640 ohms.

The problem I have with this is as follows:

If you leave 38 turns on the toroid... then you will probably have enough inductive reactance to use this choke as a filament choke (even if the core material saturates, due to the DC cathode current flowing through the coil)... but I'm not sure whether or not you will see excessive heating, due to hysteresis effects.

A way to partially test this is at the 50 ohm RF output connector of a 100W transmitter (connected to a 50 ohm dummy load). Set the frequency to 1.885MHz. Tune up the transmitter into the dummy load (if tuning up is necessary). Turn off the transmitter. Place this choke in parallel with the dummy load (e.g. using a T connector). Turn the transmitter back on (briefly). Note if there is a significant change in the required tuning and loading (or if the SWR seen by the transmitter changes significantly). There shouldn't be any significant changes. Assuming the required tuning and loading do not change much (or the transmitter's SWR reading doesn't change significantly), then turn on the transmitter for 1 minute. Turn off the transmitter, and (very carefully) check to see if the choke is hot. If the choke is not hot, it is probably okay to use as a filament choke (although you will not have tested for the effect of DC current flowing through the coil).  

Stu

One thing not discussed thus far is the option of resonating the filament choke on the lowest frequency band, and then add shunting inductance to get higher bands.

The attached image is from a W6SAI Radio Handbook, printed in 1951. Here we have a grounded grid 304TL running on 80M, 40M, and 20M using the technique just described. 

The choke that Patrick showed using coax might be ideal to resonate.

As to where to get the ferrite products, see post #8 in this thread.

Jim
Wd5JKO

Interesting thread, as I have been trying to come up w/ a filament choke for my 3CX3000A7 project too!
Wound one on the lathe that had 219uH, but too much voltage drop, and it ran hot.

(http://i180.photobucket.com/albums/x257/fish1_07/am/filament/june_15_2013/final_choke_config_6_15_2013_zps920a08da.jpg)

What I am trying for is a choke that will not saturate or overheat at the legal limit, 160M, AM.

Thanks to Opcom for the photo of the choke from a GPT-10K.
I obtained one to experiment with.

As received, 4.2 uH
(http://i180.photobucket.com/albums/x257/fish1_07/am/filament/oct_1_2014/standard_choke_zps8064ebd2.jpg)

Stuffed full of type 43 rods, 29.1 uH
(http://i180.photobucket.com/albums/x257/fish1_07/am/filament/oct_1_2014/modified_choke_zps79d23ba9.jpg)

73
Frank
KJ4OLL

That GPT choke does not have 1/2 in dia wire.  It's coaxial with a smaller inner conductor of about 1/8 in. Dia and braid of maybe  3/8 in. Dia. But with annular conductivity limited by , say 1/32 in.  Thickness.
RF of course mainly conducts on the outside of a wire but that conductivity decreases expomentially as freq. decreases.
The DC component or very low, say 60 cycle AC component is, of course, dependent on total cross section of a wire.

Hence the reactance is not what it first seems, let alone filament voltage drop.

Yes it's coaxial; I think there's a beauty to that. As far as using it, note the original GPT-10K transmitter circuit applied normal bias and screen voltages to the 4CX5000A as in class AB setup, all grounded for RF. So the workings may be a bit different. I didn't intend to supplant the original suggestions, but would suggest a look at the schematic of that amp (it's online) if its filament choke is to be used for the 3CX3000. I see no issues but there's a lot of things I don't see right off. Thanks for the readings with the ferrite rods added, that is worth a lot!

 Yesterday at the Belton Hamfest I picked up a choke that appears to be an exact match to Patrick's choke pictured in post 20 of this thread. I got it for free.

  If anybody wants that choke, PM me. It is yours for the cost of shipping.

Jim
Wd5JKO

Hello all
I have built some years ago a linear amplifier with a 3cx3000f7 tube, i was using a homemade filament choke with 4 ferrites inside but it was working warm because not enough cooling and not enough coil to work on 80 meters, i plan to use an 52 amp RF Parts filament choke i bought some years ago , but i wonder if this choke should be OK ,52 amps is just the power consuming of the filament; may i have some  advice ?

73's

B&W also made an FC-series choke, a -60 or -90, I don't recall but it is big and is enough for the 3CX3000. The "Jaws" amp has one. I don't have a pic. Probably kind of rare now, as is why I bought a 10K fil. choke. MGS4U.com sometimes has goodies that help with the more fun projects.

Float the filament xformer on glastic.   Put the chokes on the primary.

Now you need a 3 amp choke.   Or 1.5 amp.

--Shane
KD6VXI

Thanks for the reply
@opcom, i have tried the Fc 50 B  W, i have burnt it i guess not enough cooling
@Shane , the problem i have is that the wire between the filament transformer and the choke is quite long ( 80 cm), the power of this transformer is 100 amp and it's very heavy i can not put it close the socket.

It looks as if you already realize you may benefit from more cooling. Getting more cooling air may be easier than getting another choke. A surplus of air is far better than not enough.

Your situation sounds like the perfect opportunity to load these two wires with large ferrite sleeves. Put a string of ferrite as close to the tube socket as possible with both filament leads going thru them.  See my post about this back at the beginning of this thread. (second post)   Put .01 uf capacitors to ground right after the ferrite stack ends (one cap on each wire) - and the long fil lead length remaining back to the transformer will not matter RF-wise.

T

Thanks Tom for the input:
About the .01µf if i understood well, i have to put one condenser between ground and each wire, that mean two condensers?

Yes, that is correct. Also, add one .01 connected between the two wires at the filament socket pins, as usual, to balance the RF flow through the tube filament.

Ok fine!

That is probably the best answer. Cheapest for certain. Wanted to say, be sure it is really a carbon resistor. Some new resistors that look like old carbon ones are really some other kind internally and may be inductive so check the resistor before using it in the test.