813 homebrew amp-- go for three or four tubes?

[K1ZJH]

I'm starting a homebrew 813 amp.  It was going to be a two holer, but since the power transformer can handle it, I can expand to three or four tubes.  I'm not sure if going to four tubes will buy a lot.  On the other hand, I have the tubes, sockets and parts...

Has anyone had any luck using four of these tubes on 10 meters, or is that a fool's errand?  I am familiar with tapping the tank coil down for the plate tuning to improve matching to tubes with high plate capacitance.

How important is grounding the metal shells on the base; the sockets will be recessed about 1-1/2 inches.

Pete

[KA2DZT]

I think that metal base is connected to one of the pins which would get grounded.

The more tubes you put in parallel the lower the plate load.  Lower plate loads require more capacitance for both tuning and loading.  The tank coil requires less inductance to maintain the right Q of about 12.

Plate load is the plate voltage divided by twice the plate current.

Fred

[K1ZJH]

Hi Fred

Plate load for two is around 3500 ohms, so I'll be around 1750 or so... which is doable using a Pi L on 80 and 160 to keep the load capacitor size reasonable, and then going over to a Pi Net for the upper bands.  I was more concerned with the combined plate capacitance, which is going to be close to 60 pF    That is value is over the design value for a Pi Net with a reasonable Q at that load impedance.  I am still very curious how others have made out using two or three tubes in parallel on 17 meters and above.

73

Pete

[KD6VXI]

I have seen an abortion of 813s on 'ten' meters.  It used ten tubes!  So yes, 4 will work.

If you want whisper quiet, I'd go for the full four tube amp.  It might not buy you much as far as power out, but it would sure help with plate dissipation and lower the demands for any blower!

Plus, four looks cooler!

--Shane
KD6VXI

[KD6VXI]

One way to get low plates impedance to work, the old L pi L trick.

You can simulate that with the g3sek spreadsheet.

Looks like you'll need about .3 uh of 'step' inductance.

This is a run for 29.3 mhz at 2.5kv and 1.6 kw output.


With a .3 uh step inductor, your C1 is now 32.9 uH.

This is failing to put any parasitic suppressor in the anode leads. I can simulate those as well.

Pictures / screens hots included below.  If your operating parameters are grossly different, let me know and I'll run your numbers instead.  Also, only includes 10 pf stray c.  1=12, at the design frequency.

--Shane
KD6VXI

[K1ZJH]

Hi Shane!

Great info! I was going to use a 0.4 uH inductor in series with the plates to C1, since I happen to be using coils from an Illumitronix tank--and it included a .4 uH coil made from silver plated strap!  I can tap that coil for .3 uH easily.

I'll check your charts for the remaining values. Basically you are using the tube capacity and .3 uH coil for the first leg of the Pi L network?  What is the intermediate impedance (in case it isn't in the charts, which I still have to print out.)

I take it you didn't need parasitic suppressors?  Also, did you bother to neutralize, I was going to copy the W8JI feedback system he used in the AL-572 and other amps, or use a winding on the filament choke for feedback.

Pete

[K1JJ]

Yo Pete -


Fun-Factor:

Quad 813s will make a nice amp and a fun project to talk about on the air. I mean heck, half the fun of home brewing is showing off the project to your buddies, and quad 813s are way-cool! ...    4-1000As followed by 813s are my favorite tubes.


Technical:

I didn't see anyone mention if this amplifier was grounded grid or grid driven, but I assume it is GG since a feedback winding off the fil choke was mentioned.

Yes, the issue on ten meters becomes the 813's plate to cathode capacitance that would probably make a 4-holer 813 linear scream for minimum C1 on 10M.  The L1-C1-L2-C2  L-Pi method is a great solution, especially since you have a nice input silver plated L1 inductor ready for the job..

The neutralization might not be needed, even on 10M, since it is GG, but it will add some additional negative feedback (in addition to the inherent GG configuration FB) which will make the amp cleaner for IMD, so definitely do it if you have the motivation. The NFB will make it slightly harder to drive, but no big issue.

I would use parasitic suppressors - but maybe try it without first to see if they are needed. They may be needed since it is a 4-holer on 10M. All depends on how  short and well-shielded your layout is and the input not seeing the output - as well as "random" parts anomalies that may make VHF feedback paths possible.

Frank/GFZ once gave me an idea to help stabilize the fil/input circuit:  Hang a 2K non-inductive resisitor in series with a .001 cap from each fil lead to ground. (at each fil tube pin) This will help lower the Q for high impedance VHF choke parasitics, if the choke happens to have them. I did it as a precaution on all of my linear amps. FWIW...

The 813 4-holer can use a small breeze with a Variac-controlled muffin fan to be quiet as a mouse.

I say do it!

T

[KD6VXI]

Pete,

The impedance transformed is present, near the bottom of one of the pages.

If you Google G3SEK spreadsheet, you can get yourself a copy.  It's pretty much invaluable.  Let's you design for pi or pi L, let's you enter stray C and the L for the interconnects.

Keep in mind if you want to keep the tuning (Q) as designed.  You need to make that  .3 uH the entire assembly, from anode to Ctune.  Meaning, any strap or wire for the interconnects needs to be measured, with your inductor in circuit, for it to work as designed.  Not saying it won't work if you don't do it this way, but you may end up with a Q of 20 or so, which is what we are trying to get out of anyway.

I //believe// the correct terminology Is an L-Pi.

If you desire an L-Pi-L, I can run those numbers, too.  I believe it wasn't feasible with the image impedance, etc. I was looking at.

Yet another nicety about the L-Pi is that the CTune //should// see about 550 volts under properly loaded conditions.  This is on ten meters, so youll still need a vacuum cap for 80.....  But....  My last 3cx3000 deck worked 10 to 15 meters fine business.  So quad band hf amp isn't out of the question with a decently sized CTune.

--Shane
KD6VXI

[K1ZJH]

Shane

I will be using a Pi-L for 80 and 160 meters, and I assume that L1 will be nearly invisible below 15 meters. Between 30 and 17 meters the design reverts back to a Pi Net.  I found those values yielded the most practical values for the plate and tune caps.  Perhaps that is why Ameritron used a similar tank design in many of their amps?

I was planning on having L1 go directly between the plate RFC and then 68 pF section of a dual plate tuning cap salvaged from a retired Ten Tec amp. I am almost thinking that I will abandon the ceramic plate caps, and use metal heat dissipating ones instead.  That will allow using flat copper strips to connect the short distance between plate caps to the RFC and plate blocking cap. You think that will gain enough to be worth the effort?

Pete

[KD6VXI]

L1 does disappear as you go lower in frequency.  It doesn't really disappear, more absorbs some of C1.  But, it is negligible at 160.

Ive got metal caps on all my tubes except 572Bs.  Those I bought new ceramic ones for, as rf parts didn't have any metal ones.  I prefer metal, I think it adds a bit of heat sink, whereas the ceramic  tends to be an insulator:  keeps the heat in.  I've no proof, just makes sense.

I have also used a normal pi circuit and put a small amount of L between C1 and the pi tank coil, when I just needed a few pf less than the cap was willing to give.  On air variables, bend a plate, not so easy with vac variables.

As to the strap, flat and wide is superior.  And strap, not braid.  You can always bend the strap into a U to attach it to the ceramic caps.....  But again, I like the metal ones, too.

I can't comment on ameritron......  🤔☺️

--Shane
KD6VXI

[DMOD]

See Reply #7 for the Lew McCoy amp.

http://amfone.net/Amforum/index.php?topic=43299.0

One would have to beef up the Power Supply and change the Network tuning circuitry for more than one bottle, but it is a good example of the capabilities of the 813 in GG.


Phil - AC0OB