The AM Forum
March 28, 2024, 10:37:24 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: Optimizing pi output networks —what’s the target??  (Read 4429 times)
0 Members and 1 Guest are viewing this topic.
K8DI
Contributing
Member
*
Offline Offline

Posts: 423


« on: September 05, 2021, 01:08:12 PM »

So, I’m gathering some parts and plans to do this broadcast to ham conversion of an old RCA transmitter. Stock, for BC use, it has a pi, a 2nd harmonic filter and an L setup, with fixed capacitors and tapped inductors. Most ham rigs will use variable capacitors and a band switched inductor; some use a roller. 

I can play around with the PIEL calculator, or others, and see I can move around the calculated values by changing the Q. What Q is a target, and why?  If I build an output network with a roller, how does one optimize it/tune it when you can just add a turn or two and then tweak the caps and it’s back in resonance?

The real question is in the title…what’s the target one aims for? Why?

Thanks for your help, as always, in increased by my knowledge…

Ed
Logged

Ed, K8DI, warming the air with RF, and working on lighting the shack with thoriated tungsten and mercury vapor...
W1ITT
Member

Online Online

Posts: 573


« Reply #1 on: September 05, 2021, 01:46:16 PM »


the Q of a network is a seesaw that plays losses versus frequency rejection over a given range.  I seem to recall that I aimed for a Q of about 15 in the 4-1000s and GS35s that I've built.
My favorite way of setting up an output network to get into the ballpark is to tune it backwards, and cold, with a network analyzer.  For instance, if your calculated plate load impedance is 2500 ohms, the job of the pi-net is to transform that to 50 ohms.  Being a passive network, it doesn't know its front from its back. (I had a boss like that once.)  The method is to place a 2500 ohm non-inductive resistor from the plate of the tube to ground.  There's little power involved so a half watt resistor is fine.  Leave the tube right in the socket.  Yes there will be series inductance in the resistor leads but this is HF so we'll happily neglect it.   Set the inductor for the target inductance that you calculated during design.  Then, with the network analyzer looking back into the output  (you may have to jumper across any internal T-R switching)  tune for the prime center of the Smith Chart at your favorite frequency with the marker.  Record your setting and move on to any other bands.   When you get finished, don't forget to remove the load resistor, undo any jumpering around the T-R switch and disconnect the network analyzer.
I originally started doing this decades ago with Hewlett-Packard network analyzers that cost a lot of money.  With the NanoVNA available for under $50 for the 2.8" screen and under $110 for the 4" screen, most anyone can afford one.
I have used this technique on my own legal limit amps as well as some 300 KW short wave broadcast rigs way far overseas.  It gets you very close to final parameters and it saves a lot of spitzensparken and foul odors in the hamshack.
73 de Norm W1ITT
Logged
KD6VXI
Contributing
Member
*
Offline Offline

Posts: 2648


Making AM GREAT Again!


« Reply #2 on: September 05, 2021, 02:36:10 PM »

Pi-L program or (better) G3SEK spreadsheet is the hot ticket.

I shoot for a Q of 8 or so, although at 10 meters with some tubes (high Court) this becomes problematic without throwing a little L between C lock and the tube.

With a low Q you end up with lower 2nd harmonic suppression, but you retune less.  So if you like to MO e around the 75 meter ghetto without having to retune, a low Q is preferable.

If you need high 2nd harmonic suppression then a higher Q is better to shoot for, although you may have to retune when moving from the low end O the ghetto to the high end.

I shoot for low Q.  I like not having to row.

--Shane
KD6VXI
Logged
Chuck...K1KW
Contributing
Member
*
Offline Offline

Posts: 167


« Reply #3 on: September 05, 2021, 06:32:44 PM »

I shoot for a higher Q of at least 15 to 20 for the following reasons:

1) Better harmonic suppression which very important when driving multiband antennas like log periodics and tribanders.  I have heard many 2nd harmonics from stations on 20M on 10M!  I also have heard many 160M and 75M AM stations on their 2nd harmonics!  

2) When running a linear amplifier in Class B I run a Q of 20 to 25 and with the resulting higher stored energy in the pi net tank, in band distortion products are lower by 4 to 8 dB.

3) The theoretical added loss is very low especially with a high Q inductor in the pi net.  With a Q of 20 it is under 0.06 dB with an inductor having a Q of 200 or more.  You will never notice this!  If you have a roller inductor you can experiment by observing heating with different Q values.  Unfortunately roller inductors are not high Q inductors.  Most of them I have measured are in the 90 to 150 range.


Chuck K1KW

Logged

73, Chuck...K1KW
W4AMV
Member

Offline Offline

Posts: 672


« Reply #4 on: September 05, 2021, 08:30:47 PM »

It is useful to note that whatever loaded Q you decide to choose, it is set by that first capacitor, C_tune,  closest to the tube plate and given by XC_tune=Rplate/QL.. So for each band of interest, determine that C_tune reactance and capacitance first. Then the remainder of the elements of the pi net, L tank and C_load are trivial to find but more important, they are unique. The pi-net is now no more complicated then a simple L network. You can use a Smith Chart to make the calculations or use the back of a napkin. Please do avoid spinning the C_tune and C_load as well L tank all around the chart to obtain whatever Z transform you desire. You will be continually altering the circuit loaded Q, bandwidth and ultimately the network efficiency.

Alan
Logged
VE7RF
Member

Offline Offline

Posts: 56


« Reply #5 on: September 08, 2021, 11:51:46 AM »

Use the  GM3SEK  PI  /  PI-L  spreadsheet.    It's the gold standard..and factor's in  EVERYTHING.... like  stray tube C, and any stray C from anode to chassis. On my  3CX-3000A7 gg amp, the tube is  24 pf  from  anode to grid.... which increases to 33 pf, when plugged into the grid ring / socket.  The extra 9 PF is from the proximity of the lower anode fins..to the chassis below it.  It also factors in any stray L.... between the anode ..and the  C1 tune cap.  Any uh in there makes a HUGE  difference.   The 33 pf tube C  + this  stray L  form a step down  L network, transforming the plate load Z  to a much lower value.

We use that effect to make a PI network with a lower Q.   IE: transform the plate load Z  down to a lower value... low enough, that a practical PI net can be built  with a lower Q.   Typ this trick is used on the upper HF bands, and also 6M.   On my  3x3 amp, I use a RI-50  edge wound roller coil  (  3/8" ribbon, with 2 x huge contacts pinch each side of the ribbon. For  20+17+15m,   I used a hb  3/8"  silver plated  tubing coil, ( aprx  2 uh) on the end of the roller.   A modified  HV contactor makes the  17/15m  tap on the tubing coil.    IE: roller used on 16-30m bands.  When on  20-17-15m  ( no  10 + 12m on this amp), the roller is cranked down to  0  uh, leaving just the tubing coil in the circuit.

I also installed a  .66 uh  coil (  3/8"  tubing,  7 turns on a 1.5"  ID) )   between  plate block caps..and  C1 vac tune cap.   Plate load Z drops way down on 15M.   I did this since the 15M tap is  re-used on 17m.   My  Q on  15m is just 8.    Which rises to  a Q of 13.... when on 17m.

Even on band switched amps, the extra coil is inserted between plate block caps and the C1 vac tune cap.



When u enter the Q  on the spreadsheet, it's for  OVERALL network  Q....  which is the  sum of the input  AND  output Q.

On older  ARRL books, etc, they used only the  INPUT  Q  ( based on the  C1 cap).   Newer  arrl books use overall  network  Q.

IE:  old books  will use a Q of  10.     New books will  use a Q of  12.    With an overall network  Q of 12,  it equates to an input Q of  10...and an output Q of just  2.   ( 10+2 = 12). 

Depending on material used for the coils..and  plate current used..and  Q  used,  I avoid high Q networks.  Even tubing coils will run way too hot for my liking.   With a  Q of  20+,  u may as well toss in lighter fluid and a match.

On my  drake  L4B amps ( I have 4 of em).... when on  15m band..into a DL, and a 1200 w  cxr, the coil runs  HOT.  Funny thing is..the un-used portion of the 1/4"  tubing coil runs at room  temp.  The  1/4" tubing coil is used on  20-15-10m on the L4B.    I would have thought, when on 15m, that the remaining un-used portion of the tubing coil would  heat sink the  used portion..it doesn't.   It's localized heat.  15m portion runs hot...right up to the tap point.  1/2"  beyond the tap point ( 1/2" into the un used portion)..and it's room temps...go figure.

Owen duffy claims with a 2k ohm plate load Z..and  a Q of   6-8,   2nd harmonic suppression is 34 db..which rises to 36-38 db, if a  Q of  12 is used....which doesn't increase  much more, if a Q of  15+ is used.   He also claims the tube ( in Class A)  already has  7.5 db of  2nd harmonic suppression...and close to that figure, if Class  AB is used.   Orr's last books are correct..with the chart depicting  harmonic suppression  vs plate load Z 
 (Input Q of  10...overall Q of  12)    2nd harmonic suppression on a PI   will vary from  28db...up to 38 db.... when the Z varies from 5 k ohms down to 1 k ohm.   And that doesn't include the 7 db from the tube itself.

W8JI  claims an easy  -50 db on the  Ameritron  AL-1200..when tested on 40M into a DL.  The AL-1200 is a simple  PI on  40-10m.  Go figure.

I won't use a PI-L..major pita on a multiband amp, with no benefit..and a lot of drawbacks.   The touted increase in harmonic suppression is only there, if a DL is used..so the harmonics see  50 ohms.    In my case, my 80m loaded, rotary dipole doesn't resonate on it's  2nd nor 3rd harmonic...nor  4th or 5th.   My  loaded  40m yagi   doesn't resonate on  20m.... nor on  15m. ( it resonates on 12M).

I operate  mainly  5.7 khz wide  ESSB..and dabble with  AM.  On 40m LSB, the 5.7 khz wide signal is now  double that, or  11.4 khz wide...on 20m.   SWR on 20m is beyond sky high.   20m  USB  ops are not about to even decipher a  LSB signal, that has been stretched to 11.4 khz wide.   And it's gonna be  triple  5.7 khz..on 15m...and quadruple that on 10m.   AM  +  CW, that's another matter, esp  CW.

Try listening to urself on a 2nd  RX...on your various harmonics    using both  LSB..and also USB. 

Lower loaded  Q means less tank heat..and  less  re-tuning, when  QSY'ing...on any band.

end of  rant.

Jim   VE7RF
Logged
VE7RF
Member

Offline Offline

Posts: 56


« Reply #6 on: September 08, 2021, 12:06:14 PM »

One important note when using the  GM3SEK  PI /  Pi-L   spreadsheet,  if using the sheet for stuff like tuned inputs, it works superb..but with some caveats.
Only overall network Q can be entered.  And on tuned inputs for 50 ohm input Z tubes, the sheet will not handle 2 x equal Z's..it will crash.  The simple work around is to enter  50 ohms on the input side (enter where u would normally enter the plate load Z....then  enter  50.001 ohms for the cathode ( enter on line for the ant  Z).   All other parameters are set to  ZERO...except for the plate choke....set it to '99999' so the plate choke has  zero effect.   The  PI / PI-L spreadsheet  was designed for PI  output networks, hence the various parameters, stray tube C, plate choke value  etc.   I just fool the sheet, when using it for tuned inputs. 

With an overall network Q of  4 used,  the input  Q will be  2...and the output  Q  will also be  2  ( 2+2 =4).   This explains some of the irregularities over the years with tune input Q values in some construction articles.

Be careful  with any PI  tuned input on a GG amp.  The bifilar fil choke is effectively in parallel  with the  C2 on any tuned input.   C2 will have to be increased ( esp on lower bands)  to compensate for the XL of the fil choke.    IE: the  fil choke's  XL  will cancel out an equal amount of  XC   on the C2 cap. (C2 cap faces cathode,  C1 cap faces the xcvr).   Rich measures used the oem 9 uh fil choke on his  160m conversion on a SB-220.  The C2 cap has to be increased by aprx  900 pf...when on  160m.

Jim   VE7RF
Logged
K8DI
Contributing
Member
*
Offline Offline

Posts: 423


« Reply #7 on: September 21, 2021, 11:12:22 AM »

After reading the different takes on Q and playing with both the PIEL program and the G3SEK spreadsheet,  and more importantly actually having the BTA1R1 on a cart in my garage...

I've decided to aim at a Q of about 10 to 12.  The two calculators give different values, as the G3SEK one accounts for more variables (other inductances and capacitances in the amp).  I found on 1885, that G3SEK says I need an input capacitance of 320pf, and output of about 1500, with the PIEL being 200 and 1015pf. I did not run other frequencies on the G3SEK one yet, I did the PIEL ones first, and that 200 input falls to 26pf for 20 meters.  So that is a working range for the tune capacitor, 25 to 350 or so.  Output is going to range 135 to 1500 ish.  So....

I have in my junkbox a 7.5kV rated bread slicer at 125pf.  I could use it with a padder for 160 (and 80 if needed). But the 7500 volts leaves me slightly nervous, 3100v plate plus 100% modulation is 6200v.  I do not have an appropriate output/load capacitor in hand.

I acquired with the transmitter an oddity, though.   Some background:  these were the main and backup for an AM station until about ten years ago. The guy I got them from said they were on the air until shortly before he got them. then.  They were configured so that one RF output ran into the other transmitter, which has a Large contactor inside it to switch the antenna between the two.  Also installed only in that one was a vacuum variable. It has an RCA part number (as well as Jennings) but there's no mention of a variable in the manual. It had been removed (along with all the cabling between the two transmitters) ten years ago when he transported them, so its original connection is uncertain. I can guess but...  This capacitor is a Jennings UCSX 700-15, 700pf, 15kV. Voltage wise it would be great for the tune capacitor, but its value is twice as big as I need.  Voltage wise, it is complete overkill for the load capacitor, and its value would require a padder to be added.

First, any idea why this RCA labeled part was being used? Anyone aware of any documentation?  Second, should I use it? or try to trade/sell it and get something that lines up better?  Third, I have heard tales that vacuum variables will be wrecked if you turn them too far. How far is that, how many turns to get to the end, etc?  I've no hands-on time with vacuum variables, and I don't want to break it.

Ed



* IMG_5704.JPG (566.21 KB, 2016x1512 - viewed 200 times.)
Logged

Ed, K8DI, warming the air with RF, and working on lighting the shack with thoriated tungsten and mercury vapor...
w9jsw
Two shots of Whisky
Contributing
Member
*
Offline Offline

Posts: 592



« Reply #8 on: September 21, 2021, 03:20:25 PM »

I am building a 4-400 RF deck using a pair of 25-500pf vacuum variables I bought for a very good price. My band switch has been setup to include the appropriate padding for 160/80 for antenna load. I am using those larger 16kv Russian doorknobs in 500pf increments. Will use 2 for 80 and 4 more for 160. That gets me to 1500-2000pf for 160 and 500-1000pf for 80. The problem with padding is you then need a pretty capable multi-deck band switch. No free lunch, it seems. Large vacuum variables with both big C and big V are expensive and bordering on unobtanium.

John
Logged
K9MB
Member

Offline Offline

Posts: 359


« Reply #9 on: September 21, 2021, 09:18:13 PM »

I am building a 4-400 RF deck using a pair of 25-500pf vacuum variables I bought for a very good price. My band switch has been setup to include the appropriate padding for 160/80 for antenna load. I am using those larger 16kv Russian doorknobs in 500pf increments. Will use 2 for 80 and 4 more for 160. That gets me to 1500-2000pf for 160 and 500-1000pf for 80. The problem with padding is you then need a pretty capable multi-deck band switch. No free lunch, it seems. Large vacuum variables with both big C and big V are expensive and bordering on unobtanium.

John

Hi John,
I am curious why you are concerned about getting high voltage breakdown caps for loading?
Generally, if you are loading a nominally low impedance antenna, a 500 volt cap should suffice.
I think that I have a spare 3x680pF cap from an HP generator that will give you more than 2000pF at more than 500 volts, which should suffice.
Also, any good HVAC relay rated for 370vac will work if you want to switch in more capacitance.
If you want higher voltage, I have some Russia. Vacuum relays rated more than 15 RF amps at 3kv continuous.
You can pad with receiver variables also, by the way. Maybe not as sexy as a vacuum variable though. 😉
Let me know if you want any if this and email me via QRZ email and I will give you a deal.
Also have a bunch of 4-400A tubes and some 4-250s I likely will not use. Bought too many so have spares.
4-250As are NOS.
73, Mike
Logged
w9jsw
Two shots of Whisky
Contributing
Member
*
Offline Offline

Posts: 592



« Reply #10 on: September 22, 2021, 05:48:02 AM »

I need the caps to handle the current, not the voltage.

I will send you an email on the tubes.

John
Logged
VE7RF
Member

Offline Offline

Posts: 56


« Reply #11 on: September 25, 2021, 01:06:22 PM »

I am building a 4-400 RF deck using a pair of 25-500pf vacuum variables I bought for a very good price. My band switch has been setup to include the appropriate padding for 160/80 for antenna load. I am using those larger 16kv Russian doorknobs in 500pf increments. Will use 2 for 80 and 4 more for 160. That gets me to 1500-2000pf for 160 and 500-1000pf for 80. The problem with padding is you then need a pretty capable multi-deck band switch. No free lunch, it seems. Large vacuum variables with both big C and big V are expensive and bordering on unobtanium.

John

Typ one will use the biggest load cap  you can get ur hands on..then use the least amount of padding.
Since you know the power output..and the load ( 50 ohms), it's then easy to calculate the current through the load cap.  Once the load cap value is determined in software,  we then calculate the  XC of the load cap ( I use an online calculator for this function, like those on pasternacks site).  The load cap is in parallel with the 50 ohm output.  RMS V  / XC of the load cap =   RMS current through the load cap.  IF padders are used, a similar calc is done, except that the XC of each padder  is used for the formulae.   IE: air variable, or in this case, the 500 pf vac cap  handles some of the total current, with the padder(s)  handling the rest.

This is a non issue, even for  air variable  broadcast style caps.   The problems start up when  different types of padders are used.   The HH-58   500 pf caps have lousy  temp coefficients, and will drift.  The  HH-58  also have lousy current handling capabilities, esp on the lower bands.   A much better cap is the  HT-50  series.  HH-58 caps are only  .6 kva rated.    HT-50 caps  are  10 kva rated.   HT-57 caps are 35 kva rated. 

For coupling or  bypass, if caps drift, no big deal.   But padders are part of a tuned input..and any drift is a disaster.

IMO,  I would save the  500 pf  vac cap for another project.  A  4 x section broadcast variable  like the  17-540 pf per section  variety  would be ample, and require minimal if any padders for 80m. ( .030" spacing is rated for  1100 v).   I would not recommend using 500 V rated air variables.  The  V rating is  marginal with any 1.5 kw  amp...and the spacing between plates is miniscule.

Before using any vac cap for any application, it would be highly advised to hi-pot test it  1st.     Buddy has several  50-4000 pf  @  5 kv jennings ceramic vac load caps, and 9 of em fail the hi pot test miserably.  Some vac caps will hi pot test a lot higher than their stamped  ratings.

Years ago,  Fair radio in Lima  Ohio was selling 2000 pf @ 10 kv glass jennings vac caps.   There were nib.  I use one of em on my hb  3x3 amp..for loading.  It has to be padded for 160m use.   I use a modified  HV solenoid contactor for that job, but just about anything will work.  That amp uses RI-40  roller coil, so no bandswitch used.

IF a bandswitch is used, the 160m position is never used ( entire coil used on 160m)...and the   160m bandswitch position is instead used  to switch in the padding for 160m load cap.  Now that's dead simple and works..... provided the load cap has enough C  for 80m  to begin with.

I'm assuming your  proposed  4-400 amp is a single tube ?   If so, the plate load Z  will be double that of a 2 x 4-400  amp.   And with double the plate load Z, the main coil will be double the value of a 2 holer.....and both the  tune and load caps  will be  1/2 the calculated values for a 2 holer. 

On a side note, I use HB adjustable spark gaps  across both the tune.....and also the load caps.   This is to protect the expensive vac caps I used for tune and load.  Easy to make, using solid brass machine screws + brass lock nuts.   Where the 2 x screws face each other,  solid brass acorn nuts  ( home depot)  are used.  They are domed..and perfect for this application.   One machine screw  goes onto the chassis.   The other machine screw is bonded to the  hot side of the cap.   A feeler gauge is used to set the gap.... which is well below the vac caps breakdown V..... but just above the normal  peak V encountered.   IF the acorn nuts are damaged, they are easily replaced.

Now if something stupid happens, like wrong ant, no ant,  wide open / dead short, caps are protected, and also the expensive bandswitch..if a bandswitch is used.   It will also protect TR relays, both mech and also vac types.
Logged
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 0.079 seconds with 19 queries.