The AM Forum
March 29, 2024, 12:36:14 AM *
Welcome, Guest. Please login or register.

Login with username, password and session length
 
   Home   Help Calendar Links Staff List Gallery Login Register  
Pages: [1]   Go Down
  Print  
Author Topic: Thoughts about re-purposing the GPT-10K PI network coils and parts  (Read 8525 times)
0 Members and 1 Guest are viewing this topic.
Opcom
Patrick J. / KD5OEI
Contributing
Member
*
Offline Offline

Posts: 8308



WWW
« on: August 04, 2015, 02:26:03 AM »

Some thoughts about using the GPT-10K pi-L components for a smaller amp like a 3CX3000. The parts are on hand. It has taken me a log time to try to figure out what to try. The whole thing is in a PDF in the 3rd post, sorry the diagrams are large, but here is the text and the pictures as best I can attach them.  The purpose is not a huge amp but an indestructible one, as stated.
Logged

Radio Candelstein - Flagship Station of the NRK Radio Network.
Opcom
Patrick J. / KD5OEI
Contributing
Member
*
Offline Offline

Posts: 8308



WWW
« Reply #1 on: August 04, 2015, 02:32:10 AM »

This note and opinions are about looking at the tank circuit for the GPT-10K, with the intention of using its major components in a different and home built amplifier. The GPT-10K can be switched to a 600 ohm balanced or a 50 ohm output. The change-over is done by reconfiguring a dual L section. For 600 Ohms, it is used as a balancing transformer or maybe it could be called a balun.

The GPT-10 uses a 4CX5000A, grounded grid with standard electrode voltages, 7500VDC, about 2A maximum. The goal is to use a 3CX3000, grounded grid, zero bias, 4000-4500V, 2A maximum. One thing stands out, of course the voltage is half but the current is the same. This means the impedance match is different, but the main parts, the huge inductor and band switch, should still be usable.

If nothing else, the 3CX30000 might be run at 4500V/1.2A or 4000V/1.07A. There is no real need for 6KW output except for bragging rights and those are not worth all that much. The goal is an amp that is more or less indestructible and at the same time simple. The 3CX3000 has a 275W grid.

If that is not convincing, other large Triodes are either more costly or are now hard to find. The 3-1000 is a nice glass tube, and a pair would be OK but the cost is comparable to a 3CX3000. The same can be said for other reasonably sized ceramic Triodes used in ham or commercial service.

Figure 1: GPT-10 output stage

The rather low given inductance figures for the GPT-10 pi coil in figure 1 are puzzling. The coil has a total inductance of 8.5uH to cover 2-28MHz, which would call for a lot of C and make the Q very high, too high. The figure is from the manual, with some notes added.

Poring through the manuals a few times uncovered that the amplifier section has not a pi section, but a pi-L output circuit when configured for 50 Ohms output. This would explain the unusually low inductance of the pi coil. The L section, in either configuration, is variable. It is apparently tuned by two long metal cylinders with enclosed blunt ends and driven in and out of the twin 1-8uH L coil lengths together by a geared screw drive. This opinion is from looking at different versions of the manual.

So instead of a simple pi network with only a plate tuning cap and loading cap with a tapped pi coil, it also uses a variable inductor as the -L section, and a capacitor there is used to either balance a 600 Ohm load if the -L coil is reconfigured for balanced output, or to be part of the loading capacitor (sub-loading) if the coils are put in parallel and is used as a 0.5-4uH L section.

Figure 2: simplified circuit for output around 50 Ohms.

A tuning chart also was found giving the turns counter settings and this helped explain much. The chart shows two sets of dial settings for each of 14 frequencies, presumably the high and low dial setting limits inside which a set of adjustments is correct.

Figure 3: Tuning Chart from the manual.

The following seem to be indicated by the chart:
• The OUTPUT BALANCE capacitor C916 (serves as a 'sub-loading' capacitor when the TX is configured for 50 Ohms output) is set to one value or another, depending on whether the amp is below 8 or above 11 MHz. So somewhere in that undocumented range, it is like switching in another capacitor in parallel with the PA LOADING cap C926.
• The combination of the OUTPUT BALANCE cap C916 and the -L coil OUTPUT LOADING setting influences the PA LOADING cap C926 setting. They might as well be combined into one unit for the purpose.
• The pi coil FINAL BAND L902+L903 has the bandswitch and has nine possible inductance values.
• The OUTPUT LOADING 'L coil' L912+L913 is variable but has six chart settings roughly aligned with frequency.
• The OUTPUT BALANCE cap C916 has two settings, more or less what I call 'in' or 'out'. The dial is charted to be at 300 or 003. Might be turns; 3 or 30.
• The PA LOADING cap C926 seems to vary a bit but is kept in a small range by the other variations except for that the 'in' or 'out' setting of the OUTPUT BALANCE cap C916 calls for it to be toward the higher or lower end.
• The PA TUNE cap C927 varies rather linearly with fequency as might be expected, that is, no surprises there.

The other unusual part of this is the plate choke and bypass arrangement. LTspice shows this to be pretty good, at first look.

Figure 4: Plate RF, dB down from “cold” end of pi coil.

Something to consider is the pi coil is at B+ in this circuit. There is nothing wrong with that but worth mentioning. No doubt it un-complicates matters around the plate choke and lowers its voltage burden.

What is actually on hand is a pair of the large pi section coils, each with the integrated band switch. It should be possible to use one as intended for the pi section and reconfigure the taps on the second one to set up the right L inductance for each band switch position. The other alternative would be to use maybe a ¼' copper tubing coil with a switch but no large switch is on hand and the second pi coil is. Having identical shafts they could be mechanically coupled without too much trouble.


* figure 1.png (108.67 KB, 1777x794 - viewed 367 times.)

* figure 2.png (67.58 KB, 1077x789 - viewed 322 times.)

* figure 3.png (96.14 KB, 1079x548 - viewed 306 times.)
Logged

Radio Candelstein - Flagship Station of the NRK Radio Network.
Opcom
Patrick J. / KD5OEI
Contributing
Member
*
Offline Offline

Posts: 8308



WWW
« Reply #2 on: August 04, 2015, 02:33:25 AM »

lastly figure 4 and the pdf file, might be easier to read/make sense of assuming this makes any sense.


* figure 4.png (51.8 KB, 1446x1143 - viewed 332 times.)
* GPT-10 tank for 3x3000 amp.pdf (330.42 KB - downloaded 216 times.)
Logged

Radio Candelstein - Flagship Station of the NRK Radio Network.
N2DTS
Contributing
Member
*
Offline Offline

Posts: 2307


« Reply #3 on: August 04, 2015, 08:15:54 PM »

I see they have the plate voltage on the band switch and tuning cap....
Logged
Opcom
Patrick J. / KD5OEI
Contributing
Member
*
Offline Offline

Posts: 8308



WWW
« Reply #4 on: August 05, 2015, 12:42:23 AM »

There are some advantages to it where the plate choke is concerned such as very low RF voltage across it. The pi coil is very well insulated.  A 35uH choke L907 on the output keeps it at DC. This choke, from a picture in the manual, is a pretty hefty affair.

It's not stated what the impedance is at the junction of the PI and L sections.

I say that because in the 600 Ohm output configuration the two coils L912 and L913 are used as a RF transformer. I'm not really sure how this is working, what is the action. The book explains that the RF voltage from the PA appears at the junction of L912, L913, and C916, and that at the other end of L912 the voltage is 180 degrees out of phase from that of L913. C916 is supposed to adjust the balance in this configuration, rather than being a loading cap. The coils are within several inches of each other but it isn't clear what the coupling is.



* figure 5.png (23.22 KB, 453x496 - viewed 305 times.)

* rfdeckpic1.png (336.26 KB, 1072x907 - viewed 330 times.)

* rfdeckpic2.png (393.14 KB, 946x830 - viewed 300 times.)
Logged

Radio Candelstein - Flagship Station of the NRK Radio Network.
Opcom
Patrick J. / KD5OEI
Contributing
Member
*
Offline Offline

Posts: 8308



WWW
« Reply #5 on: August 05, 2015, 12:43:38 AM »

Can anyone explain how this works and what the impedance transformation would be, and why a similar model, the GPT-10KRL, uses the same values of components in the pi section but does not have the L section at all, yet makes 50 ohms? That would simplify things to not need the L section.


* rfdeck_10KRLmodel.png (53.05 KB, 821x785 - viewed 318 times.)
Logged

Radio Candelstein - Flagship Station of the NRK Radio Network.
w4bfs
W4 Beans For Supper
Contributing
Member
*
Offline Offline

Posts: 1433


more inpoot often yields more outpoot


« Reply #6 on: August 05, 2015, 10:09:44 AM »

look at a 2 inductor balanced coupler circuit ... look closely at the inductors to see if the winding pitch is the same or reversed ...to work as a balanced feeder the line currents must (should) be 180 degrees out of phase with each other and it appears that they are fed in quadrature in this circuit
Logged

Beefus

O would some power the gift give us
to see ourselves as others see us.
It would from many blunders free us.         Robert Burns
Opcom
Patrick J. / KD5OEI
Contributing
Member
*
Offline Offline

Posts: 8308



WWW
« Reply #7 on: August 06, 2015, 03:56:05 PM »

I'm trying too look up "two inductor balanced coupler" and "quadrature feed"  but not getting far.

Those two pictures are the best ones I found in the manuals.  The attachment in this post is of GPT-10 coils offered on eBay, but it is not clear if this is a 'set' from one GPT010 or two coils, matched for winding direction (CCW from the end), from two GPT-10 specimens, because the seller has lots of GPT-10 parts up. I asked what these were and if the 'assembly' was available. There was no reply so far.

From the GPT-10 manual photos, it looks like the coils could be wound in the same direction IF the strap on the ceramic insulator from the near terminal of the right side coil, goes to the rear and to the end of the left-side coil and from there to the balance capacitor. - but it is not certain.

So I am believing the coils are wound in the same direction but could be wrong.




* GPT-10_coil_.JPG (42.55 KB, 640x480 - viewed 288 times.)
Logged

Radio Candelstein - Flagship Station of the NRK Radio Network.
Opcom
Patrick J. / KD5OEI
Contributing
Member
*
Offline Offline

Posts: 8308



WWW
« Reply #8 on: August 06, 2015, 06:02:33 PM »

For the 600 Ohm balanced load connection, if they are wound in the same clock direction, CCW as seems to be shown, but the current passes through them in opposite directions front to back because of the wiring, as also seems to be shown, what would that indicate?

Would this cancel the inductance when they were put in parallel for the 50 Ohm setting?




Logged

Radio Candelstein - Flagship Station of the NRK Radio Network.
Opcom
Patrick J. / KD5OEI
Contributing
Member
*
Offline Offline

Posts: 8308



WWW
« Reply #9 on: August 08, 2015, 12:16:28 AM »

Using any of the pi and pi-L calculators tells me the PI section is not going to do 50 Ohms, not easily and not with a low Q.  How weird the manuals shows otherwise. I could be using wrong plate impedances.

Here's the deal.. such a beautiful and macho pi coil, huge conductors, low loss.. shameful not to use it!


* pi for 3cx3000 gpt-10.png (197.26 KB, 1925x4115 - viewed 315 times.)
Logged

Radio Candelstein - Flagship Station of the NRK Radio Network.
Opcom
Patrick J. / KD5OEI
Contributing
Member
*
Offline Offline

Posts: 8308



WWW
« Reply #10 on: August 08, 2015, 02:20:06 PM »

OK. latest bulletin. I am not so smart but this will work.

I dug out the PI coil and measured it. Why didn't I do that before? Moral: don't believe everything you read.

The full coil is 13.5uH. It's enough for 2Mhz. PI-Calc says that 30Mhz needs 0.9uH. The minimum inductance is 0.255Mhz.

The files tell about this PI coil and the "L" section in the GPT-10. The L coils are the same CCW direction but the current flows in different directions so the L might cancel if put in parallel for a 50 Ohm setup in the transmitter.

The result looks like the PI coil alone will be enough for an amplifier. I still want to see if it can work with a PI-L to keep harmonics down. The point is now the true values of the PI coil are known so the rest of the calculations should make sense too. I hope this whole ordeal wasn't too boring.





* GPT-10 output diagram new.png (185.29 KB, 1391x905 - viewed 295 times.)

* Pi-Calc new.png (45.61 KB, 851x609 - viewed 311 times.)
Logged

Radio Candelstein - Flagship Station of the NRK Radio Network.
w4bfs
W4 Beans For Supper
Contributing
Member
*
Offline Offline

Posts: 1433


more inpoot often yields more outpoot


« Reply #11 on: August 08, 2015, 06:36:39 PM »

when you have components that large, the stray values of inductance and capacitance mount up fast

I was puzzled about the b+ decoupling circuit until I remembered that in the GPT 750 that they had to use bandswitch sections to short out sections of the final inductor for higher freq bands to eliminate parasitic resonances ....bet it the same sort of thing here .... no doubt the results of prototype testing

n.b.  I like the Elmer Fudd U1 cathode signal name .... be vewwy vewwy quiet
Logged

Beefus

O would some power the gift give us
to see ourselves as others see us.
It would from many blunders free us.         Robert Burns
Opcom
Patrick J. / KD5OEI
Contributing
Member
*
Offline Offline

Posts: 8308



WWW
« Reply #12 on: August 08, 2015, 11:09:27 PM »

Yes the large sector shorts three taps at once. I wish I had a GPT-10K power amp & HV section. A thing of beauty if one does not mind lighting the filament..

Wondered if anyone would notice the input node name.
Logged

Radio Candelstein - Flagship Station of the NRK Radio Network.
Pages: [1]   Go Up
  Print  
 
Jump to:  

AMfone - Dedicated to Amplitude Modulation on the Amateur Radio Bands
 AMfone © 2001-2015
Powered by SMF 1.1.21 | SMF © 2015, Simple Machines
Page created in 0.083 seconds with 18 queries.