Good site to calculate resonating caps for fixed chokes

[VE7RF]

Check this site out.  Handy  calculator for resonating chokes.   Enter only 2 of the 3 items.
IE:  enter desired freq (hz) and choke value (H).... it spits out the required resonating cap. (uf)
Enter choke value and resonating cap value..and it  spits out the resonant freq.
Enter resonating cap value and freq....and it spits out required choke value.

Microwave oven oil caps are typ .85uf  to  .9 uf..and  rated at 5 kv.   They are easily wired in series for higher  B+ values....and lower uf values.   

https://goodcalculators.com/resonant-frequency-calculator/

Later.... Jim   VE7RF

[Opcom]

Very useful. I use a spreadsheet with the formula in it for a simple reason that chokes change inductance with frequency.
Some people will want to tune for idling current in a linear amp, others tune for bleeder current if the bleeder's small.
As the current increases, the ckoke value decreases, raising the resonant frequency.
All of this allows for quickly calculating a range of current (inductance values) and capacitances in which the ripple is best rejected at the chosen current level.

example:
Assume an 8H choke is rated 1A. It may be 10H at a bleeder load of 100mA.
If resonated to 120Hz (100Hz) at the lower curent, the full load resonance frequency will be a bit higher.

Some have postulated the rule of thumb of adding 5% to the ripple frequency when resonating, since low current is where the regulation needs the most help. It's obviously not always true, depending on the quality of the choke.

It's all splitting hairs I guess, but why not strive?

full load: 8H on 120Hz resonates with 0.219881041 µF

bleeder load: 10H on 120Hz resonates with 0.175904833 µF

At full load of 1A, the choke will be at 8H and with 0.175904833 µF across it, the resonant frequency will be 134Hz. It will still help regulation, but the best effect will be at the 100mA bleeder load. The reasonable value of 0.18uF could be chosen making 118Hz. at the 100mA level.

[VE7RF]

Very useful. I use a spreadsheet with the formula in it for a simple reason that chokes change inductance with frequency.
Some people will want to tune for idling current in a linear amp, others tune for bleeder current if the bleeder's small.
As the current increases, the ckoke value decreases, raising the resonant frequency.
All of this allows for quickly calculating a range of current (inductance values) and capacitances in which the ripple is best rejected at the chosen current level.

example:
Assume an 8H choke is rated 1A. It may be 10H at a bleeder load of 100mA.
If resonated to 120Hz (100Hz) at the lower curent, the full load resonance frequency will be a bit higher.

Some have postulated the rule of thumb of adding 5% to the ripple frequency when resonating, since low current is where the regulation needs the most help. It's obviously not always true, depending on the quality of the choke.

It's all splitting hairs I guess, but why not strive?

full load: 8H on 120Hz resonates with 0.219881041 µF

bleeder load: 10H on 120Hz resonates with 0.175904833 µF

At full load of 1A, the choke will be at 8H and with 0.175904833 µF across it, the resonant frequency will be 134Hz. It will still help regulation, but the best effect will be at the 100mA bleeder load. The reasonable value of 0.18uF could be chosen making 118Hz. at the 100mA level.

Be careful  with chokes.   Went through this  with  hammond/  dahl division just a while ago.   They  run rated current through the choke when measuring the inductance.  they showed me the megabuck  test setup they used.    A plane jane  digital  LCR meter ( like my  B+K  875B)  will  read a bit on the low side. 

SOME chokes  will decrease in value when a lot of current is drawn.... some more than others.... depends who made the  choke.

IF the  choke is to be resonated..esp at  120 hz,  it's  gotta be specially made.   When  you resonate the choke at  120 hz ( or say resonate at  118 hz with just bleeder...and 120 hz  with rated normal  current)   the  circulating loop current in the  choke is sky high..and is greater than the  DC  current  flowing through the  choke !   

My  dahl  4H  choke is rated for  3 amps  CCS..and the  damned thing weighs a whopping  127 lbs  ( it uses the same hypersil  C core as the dahl 10 KVA
 CCS  plate xfmr).    DC resistance is just  9.56  ohms.   I  use the choke  in a  C-L-C  config and choke is  NOT resonated. 

I tested the  choke with my small 0-60 vdc 5A  lab supply...(in current limit mode)..and ran  3A  CCS  through it for  2.5 hrs.    After  1.5 hrs, it was stone cold....  so  checked  again an hr later.   Ok, after  2.5 hrs, i was  2  steps  BELOW barely luke warm.   I got in there with my fingers..and  could barely feel a temp rise on the outer  winding.   And I think a 2-3F   deg  temp rise  over  ambient..using my fluke  62  IR.  Shut it down..and measured  dc resistance..which had risen from  9.56 ohms.... now increased to 10.01 ohms.....after  2.5 hrs.

Then I find out abt the  120 hz  circulating current..IF choke resonated at  120 hz.   It's higher than the dc  current.  Typ it would be  4-5  amps  of  circulating  current flowing through the  choke..then  through the mating resonating  cap..back to other choke terminal.... +  the  3 amps of dc  coming out of the choke.   IE:  choke has to really be rated for  7 amps  CCS, when resonated... to  rate it at  3A  dc.   Also the  hi  pot test on resonated chokes is  typ through the  roof.... vs  using the  same  core  as a plate xfmr.   IE:  resonant choke  is  severely  de-rated  for  B+.   

Ok,that explains why a  4H,  3 amp  rated choke  can vary from 127 lbs ( for  120 hz  resonant use).... vs  not resonated.... or resonated at > 120 hz, like
 145-180 hz...which weighs a helluva lot less.

W8JI worked out that even if the  choke is resonated at  175 hz,  it will  still have  a huge impact on  120 hz  ripple redux.

I also noticed,  (using that online calculator I posted).... that the  collins  screen  supply resonated  3.5 H choke  on the collins 30S1 amp,  is  resonated dead on at  120 hz.    But the anode  supply (8H) choke is  resonated at  145 hz. 

Can you simulate the  circulating current through the  resonant  choke..at both  120 hz...and also resonated at higher freqs....using   Spice..or  LT  spice ??   
The glass audio  folks were well aware of the higher circulating current  vs  dc current effect on resonant chokes.

On my latest   C-L-C   experiments,   I installed the   127 lb dahl at bottom of rack... but floated  it.    Choke sits on a pair of  upside down  4"  wide  aluminum  channels..drilled and tapped  with  1/4-20  threads.  ( choke will only take a max of 1/4"  bolts).   Extreme ends of channels have  2"  diam   x 2.5"  tall red glastics   on the undersides of the channels.   ( glastics  have embedded  3/8"  threads on each end of em).   Bottom of the 4 x glastics  ists on  2 laminated sheets of  new  3/4"  plywood...  painted gloss black.   Plywood  sits on  2x2" angle alum.... one angle  running  front to  back on each side of bottom of rack.     Used the  cum along to get the  choke into bottom of rack.

I calculated the  total V stand off at 60++  kv.   I also installed a  pair of  3/4 x 1/16 thick  angle alum on the choke terminals.... perhaps  5"  long each..and parallel...with an aprx  3-4" parallel gap  (whatever the choke terminal spacing is).   Installed a hb spark gap between the angle alums..made of  1/4-20  brass bolts...and brass lock nuts to set the gap.   Where the brass bolts  face each other, I used  solid brass acorn  nuts.  Then set the gap to  aprx 200-300  volts.   IF the acorm nuts get damaged..they  are easily replaced.   Also installed a kilovac  SPST-NC  hd  ceramic vac relay...  tungsten contacts, designed for  DC  switching applications, not RF.   50A  rated  @ 25 kv.  12 vdc  coil.   Relay is so In can shunt the  choke..and just have the paralleled caps on either side of the  choke..and turn it into a big C filter.   

Choke  installed in the  B+...between the  filter cap on either side of it.    I'll post a pix  if I can figure out  how to post a pix on this site.

Jim    VE7RF

[WBear2GCR]

<snip>

W8JI worked out that even if the  choke is resonated at  175 hz,  it will  still have  a huge impact on  120 hz  ripple redux.

Jim    VE7RF

Tom Rauch, W8JI, is one heck of a lot more knowledgeable than I will ever be.

But "resonating" a choke is really obscuring what is happening electronically. What
we get is an Elliptical filter.

So, typically there is a monster (super high Q) dip at the "resonating frequency", followed by
a rise up, where it meets the slope, going higher in frequency, that would have been
present without the "resonating capacitor".

So, I fail to understand how setting the frequency higher has much benefit. The reason is that
below the "resonating dip" the filter has an essentially flat bandpass, a sharp-ish 3dB rolloff point,
then a cliff drop (whose depth depends entirely upon the values chosen, trimmed and any variation
in the choke's value). So, if you design the filter to be higher in resonated frequency, the desired
(120Hz.) target ends up in the flat passband section. No benefit.

Some benefit if one aims low, below the target by maybe 10% or so, then one is still maybe
in the dip, and worst case, back on the "stock" roll-off curve of the filter. C-L-C being 18dB/octave.

W8JI, probably was showing something specific to a specific circumstance?

Btw, random values of "resonating capacitor", they do not work. The filter really needs to
be calculated, then trimmed into performance in the actual working circuit, imho.

           _-_-bear

PS. the site indicated above is ok for figuring out the basic resonance for and L & C in
parallel, but that will only get you somewhere in a ballpark. One needs to either or do both
a SPICE model, with impedances present, since the impedances and other capacitances will
effect both the Q of the filter and the actual frequency it works at - that dip is very very
sharp and sensitive to variations, and R values - and to actually measure the result in the
actual circuit to see that indeed one gets the resulting curve, Q, and depth of notch (that's
what it is, a high Q notch filter) that is desired!

PPS. I spent a lot of hours on a similar elliptical filter problem as an audio low pass on LF woofer.
The goal was to create a rolloff of a woofer that was naturally rising due to both the inherent
response of the woofer and the partial horn loading of that driver. I was able to set the notch
resonance ok, but had to carefully select the other C values and watch the effects of the series
R of the choke(s) as well in order to get the results that LTSPICE showed possible. I'd have had
zero chance without the simulations...

[VE7RF]

<snip>

W8JI worked out that even if the  choke is resonated at  175 hz,  it will  still have  a huge impact on  120 hz  ripple redux.

Jim    VE7RF

Tom Rauch, W8JI, is one heck of a lot more knowledgeable than I will ever be.

But "resonating" a choke is really obscuring what is happening electronically. What
we get is an Elliptical filter.

So, typically there is a monster (super high Q) dip at the "resonating frequency", followed by
a rise up, where it meets the slope, going higher in frequency, that would have been
present without the "resonating capacitor".

So, I fail to understand how setting the frequency higher has much benefit. The reason is that
below the "resonating dip" the filter has an essentially flat bandpass, a sharp-ish 3dB rolloff point,
then a cliff drop (whose depth depends entirely upon the values chosen, trimmed and any variation
in the choke's value). So, if you design the filter to be higher in resonated frequency, the desired
(120Hz.) target ends up in the flat passband section. No benefit.

Some benefit if one aims low, below the target by maybe 10% or so, then one is still maybe
in the dip, and worst case, back on the "stock" roll-off curve of the filter. C-L-C being 18dB/octave.

W8JI, probably was showing something specific to a specific circumstance?

Btw, random values of "resonating capacitor", they do not work. The filter really needs to
be calculated, then trimmed into performance in the actual working circuit, imho.

           _-_-bear

PS. the site indicated above is ok for figuring out the basic resonance for and L & C in
parallel, but that will only get you somewhere in a ballpark. One needs to either or do both
a SPICE model, with impedances present, since the impedances and other capacitances will
effect both the Q of the filter and the actual frequency it works at - that dip is very very
sharp and sensitive to variations, and R values - and to actually measure the result in the
actual circuit to see that indeed one gets the resulting curve, Q, and depth of notch (that's
what it is, a high Q notch filter) that is desired!

PPS. I spent a lot of hours on a similar elliptical filter problem as an audio low pass on LF woofer.
The goal was to create a rolloff of a woofer that was naturally rising due to both the inherent
response of the woofer and the partial horn loading of that driver. I was able to set the notch
resonance ok, but had to carefully select the other C values and watch the effects of the series
R of the choke(s) as well in order to get the results that LTSPICE showed possible. I'd have had
zero chance without the simulations...

Dunno where Tom  got his humongous ( -54db) ripple redux with  resonating the choke at  175 hz....and a small oil cap  after the resonant choke.  The fellow who everyone was trying to help had a henry amp, like a 2k4... which had a replacement   8 H  choke  he installed.   The fellow was gonna use a .1uf  resonating cap... ( resonates  at  aprx 178 hz.)   I said he has the resonant choke freq  way too high..and to stick in a 2nd  .1 uf cap  (resonates at 125.82 hz)   Then tom reads  everybody the riot act about NOT resonating the choke at  120 hz.   Tom tried resonating a choke at exactly 120 hz  years and years ago..... and the supply went nuts, blew the HV leads  clean off, and he just about got killed.   He never resonated a choke at 120 hz again..ever.  If he did use a resoant choke setup, it was always resonant on the high side of 120hz....  like  145-180 hz.

John Lyles who designed the  BE  line of  single phase  5 kw  FM broadcast TX  PA's....  used a resonant choke setup..... and used a pair of 5 kv microwave oven .9 uf  oil caps in series....=   .45 uf.   Caps were  sent to the  choke manufacturer who  built the  choke around the series pair of oil caps.   John said it resonated, under normal full bore load  exactly at  120 hz.    The choke manufacturer also installed a slightly bigger tap at  aprx  5.62 H...so the same oil caps could be  re-used.... with a 50hz  setup... ( and choke resonates at  50 hz). 

I can understand Tom's  concerns,  but I sure as heck can not fathom  how he came up with  -54 db ripple redux, by resonating a 8H  choke at 178 hz.... followed by a single  16 uf  oil  cap.   I believe the  no load B+  on the  fellow's  henry amp was aprx 3900vdc....  with just a big bleeder load.

What I do these days is figure out on software ( usually  PSUD-2)   what the  P-P  ripple is..under load.  Then  divide by  2.828  to get  RMS  ripple.  Then take  RMS value..and  divide by loaded  B+  value.   Then use LOG of that result..then multiply by 20.      For some perspective, my oem drake  L4PS  outboard supply is just a FWD...and  with a 800 ma load,  B+ is down to  2.5 kv.  8 x 200 uf caps  in series  =  25 uf  C  filter.   207 volts  peak to peak ripple... = 73.2 vrms.   (73.2 / 2500 = .029)    (20 Log .029) =   -30.67 db 

Story I got from hammond /dahl a  while ago was..... if the choke is gonna be resonant at   120 hz  ( or  100hz).... it's gotta be  built like a tank...to handle the circulating current on top of dc current..and also the V handling of the choke... IE  has to hi pot test really high... like  4-5  X   B+ voltage.

I suspect collins  resonated at  145 hz...  to relieve stress on the choke..and  still sorta do the job.... and not be  heavy and bulky.

Interesting to note on John Lyles  resonant choke setup.... that it  did not meet spec  for ripple redux on the harmonic freqs...like  240-360-480-600 hz  etc.   He had to install a 2nd choke  + another oil cap  to meet  spec.  (2nd choke NOT resonated). 

I like my  C-L-C  setup  better.  Kills  everything  in  one shot...without having to resonate the choke.

Jim    VE7RF

   

[K1JJ]

About 2 months ago I was trying to reduce the ripple in a cap input 2000 volt power supply. I just happened to have the right values to resonate near 120 Hz and installed it.  I found a small reduction in ripple, but not much. Not worth the effort.  I was disappointed after a lot of tries with different component values.   I then switched over to a standard  C-L-C  configuration and the ripple was virtually gone. Regulation was good - depending on cap values, of course.  

I have since stayed with C-L-C when ripple is an issue.   An L-C  is OK too, though a step up in ripple. In any case, regulation is usually good with strapping chokes and big caps.  

The common brute force BIG input single C supply works, but will always have some ripple unless we use very big cap values.   I have a single C input 4KV supply with a 140 uF  10KV cap. Very nice all around. Has a 30H BC choke that can be switched in and out, but I keep it as a single C input most of the time. There's nothing like horsepower.  No free lunches.

T

[VE7RF]

About 2 months ago I was trying to reduce the ripple in a cap input 2000 volt power supply. I just happened to have the right values to resonate near 120 Hz and installed it.  I found a small reduction in ripple, but not much. Not worth the effort.  I was disappointed after a lot of tries with different component values.   I then switched over to a standard  C-L-C  configuration and the ripple was virtually gone. Regulation was good - depending on cap values, of course.  

I have since stayed with C-L-C when ripple is an issue.   An L-C  is OK too, though a step up in ripple. In any case, regulation is usually good with strapping chokes and big caps.  

The common brute force BIG input single C supply works, but will always have some ripple unless we use very big cap values.   I have a single C input 4KV supply with a 140 uF  10KV cap. Very nice all around. Has a 30H BC choke that can be switched in and out, but I keep it as a single C input most of the time. There's nothing like horsepower.  No free lunches.

T

I agree.  BUT what I discovered so far is the  C-L-C works best  if the  C2 cap is  double the value of the  C1 cap.   Bare minimum is both caps are the same value.    With  ssb / cw loading, testing every config I could implement  using  PSUD-2,  it's a total disaster if the C2 is a lower value than the  C1  cap. You get this wild V oscillation's in the output  B+, when the load is instantaneously slammed on...OR  off.   ( But fine for  FM broadcast or any static load).

With both caps the same value,  it's pretty good, but still has a bit of  oscillation.  With  C2  cap a bit higher value than  C1 cap... the  oscillation  is  greatly diminished.   With  C2  value  double the  C1  value, there is  no oscillation,  just a gradual roll off in  B+.   And ripple is non existent. ..poof, gone...and ditto with all the harmonics of  120 hz too. 

My problem with resonating the choke, like in a  henry amp... is.... the  harmonics  will sail right on through the  resonating cap.    The smaller the choke value, the bigger the resonating cap has to be..and the harmonics really pass through.   At the telco I worked at for 34 years, we used a 2 uf cap to  block the -52 vdc.... and pass the 90 vac  @ 20 hz  ringing generator (rings the phone out in the field).   You only need (2 uf  / 12 ) = .167 uf  to  pass  240 hz  2nd harmonic....and all higher harmonics.     That just might be why the fellow who tom was trying to help out...was advised to stick with a .1uf resonating
cap... ( with his  8 H choke)...so the 240 hz harmonic was at least partially  attenuated  due to the slightly higher  XC of the .1 uf cap.

IF he had used a .2 uf (125 hz)..or  a .22uf cap (resonates at 120 hz), the 240 hz and up harmonics would sail right through.... and be attenuated only by the  16 uf oil cap. 

My setup is  beyond gross overkill.   But it's a good experiment, since I had the parts on hand.   253 lb dahl choke, 2 x FWB's in parallel,   625 uf -  4H  dahl hypersil 127 lb choke- 1250uf.   Dynamic and  static  regulation is superb, zero ripple.   That's on a 5700 vdc  supply.

In the  7700 vdc config, its just  400uf-4H-400uf.   Again static and dynamic regulation is superb...  no ripple.

I installed a spst-NC  ceramic vac relay across the choke terminals...and also a hb  adjustable spark gap  for safety.  Entire choke is floated from bottom of rack.   ( choke installed in the B+).   With choke shunted, it's one big C filter.   I had caseloads of  10,000uf @  450 vdc lytics..so used em.

Interlocks  everywhere..and also  HV meters in  HV supply cab #1 ( plate xfmr and 2 x FWB's  +  HV fuse #1)  and also cab #2 (  contains lytics, and choke..and also the  fan cooled 50 ohm glitch assy..and also HV fuse #2.).   RF deck in cab  #3 (with another  HV meter).

50 ohm glitch assy consist of  4 x 200 ohm, 225 watt  type AS  globars in parallel. (18" long  x 1.5"  diam).   These are the energy absorbing type, each resistor rated at 120 kv and  119K  joules.    Buss  HVU-3  fuses used just before the glitch assy.....and also between plate  xfmr sec..and input to  both paralleled  FWB's   ( in one leg of xfmr sec).   And yes,  you can cro bar the output, fuse opens off in < 2 msecs, event  over.      Also uses a magnetic hydraulic breaker on the 240  input to the HV supply, 'instant trip type'...with the oil removed from the internal reservoir.   

Jim   VE7RF

[WBear2GCR]

Interesting aspects here!

Seems to me that IF the problem is harmonics of 120Hz, then the best route would
be to use a second(ary) filter designed to attenuate above 240Hz. Aka, adding
poles to the existing filter to increase the ultimate roll-off.

Also, I wish I had the LTSPICE freq response graphs to post up here so that it would
be easy to see what I am talking about. But... consider a standard C-L-C being an 18dB/oct.
rolloff.  So, assuming the -3dB point @ 100Hz (for example) then an octave higher we're
down 18dB. Another octave up and 36dB. The elliptical filter will only make a high
slope rolloff (say centered at 120Hz) which then rises back up to meet the existing
3rd order slope! So, anything much wider than the octave (or less) around the dip/notch
ends up being the original filter slope. This is why it doesn't make much sense to put the
"resonated" choke at something like 145Hz! The benefit at say 240Hz. will be only a few
dB over the stock 3rd order filter.

So, adding another L & C would give a 5th order ultimate rolloff. Aka 36dB/octave.
The curve might look like a brick wall/cliff after say 105Hz, dropping to between -50 to
-60db @ 120Hz, then rising to meet the 5th order curve an octave higher at 240Hz which
is now -36dB down. Not bad. (one can add another notch at 240Hz. but that may or may
not be desirable...)


Tom, the reason that the ratio of 1:2 that you found for the C values works is that it
forms a more ideal filter. If you run the C values in a graphic simulator for 3rd order filters,
you'll see that as you change the ratios, it runs though a variety of desirable and undesirable
filters. Get the wrong values and the stop band slope starts to look less like a 3rd order and
more like a second or 1st order...

                 _-_-bear

[Opcom]

Can you simulate the  circulating current through the  resonant  choke..at both  120 hz...and also resonated at higher freqs....using   Spice..or  LT  spice ??   
The glass audio  folks were well aware of the higher circulating current  vs  dc current effect on resonant chokes.

Jim    VE7RF

The other points in your post are very well taken. In my previous post it was to demonstrate that inductance changes with load, but it was not a good example, so i'm glad you posted all of that information.

I have done what you asked about in LTspice for 120Hz and 125Hz some time ago when trying to figure out some transient stuff. Tried 180Hz and didnt like it but forgot why. Worse ripple IIRC.

The items measured are called out on the waveform images. The 4700V 2A power supply schematic is also shown.
It includes a pulser to study what happens transient-wise going from 300mA to 2A then back to 300mA, so that segment is shown so disturbances are included.

Using an 8H choke with 0.2uFcapacitor to resonate. circulating current 1.1A with 2.05A DC output.
I chose 8H because the henry generator used 8H in a resonant supply, that's all.

I did a similar one with a 2.5H choke and 0.65uF cap. circulating current 3.3A with 2.03A DC output. Images not uploaded for that at this time.

Also 30H and 0.0586uF, not uploaded at this time. Less remarkable, but large output voltage transients with pulsed load with or without resonating, choke is so big. 30H looks better for AM linear, 8H or maybe less looks better for SSB or TTY. Or QSK but that's crazy trouble and off topic.

[WBear2GCR]

Well, ok!
Good!

Whose SPICE model is that?

Since I happen to have a KWS-1 and I am told that apparently they are "known"
to blow up chokes, this model is of some interest!

Wondering what happens if one moves the choke to the ground leg of the PS?
The current will remain the same, but the voltage across the choke?

It would be good to use the Spice model to see the freq response of the filter section,
with load and without.  the signal source would need to be a swept sine)

Love the "dit and dah" generators!!
That and the load "switch"

Cool.

                               _-_-bear

[VE7RF]

Well, ok!
Good!

Whose SPICE model is that?

Since I happen to have a KWS-1 and I am told that apparently they are "known"
to blow up chokes, this model is of some interest!

Wondering what happens if one moves the choke to the ground leg of the PS?
The current will remain the same, but the voltage across the choke?

It would be good to use the Spice model to see the freq response of the filter section,
with load and without.  the signal source would need to be a swept sine)

Love the "dit and dah" generators!!
That and the load "switch"

Cool.

                               _-_-bear

I drew this out on paper....and installing a choke in series with the B-   does not look too good.   Even if it's a L-C ......   or a resonant L-then C.... or a C-L-C...or a C-resonant L- C  .    Any V drop across the  choke, with choke in series with the B-   will result in un wanted bias  yo-yoing on a  GG  Triode tube.    And if  safety diodes used in  the RF deck...and also a separate   HV supply, the  safety diodes  between B- and chassis  in each box.... will have effectively shunted the  choke!   

Installing the  choke in series with the B+... but isolated from chassis, will  relieve all the  HV stress  on the choke, no different than  a series string of 450 vdc electrolytics  in a SB-220.

An adjustable spark gap across the choke terminals will protect the choke from any mishaps.   Isolating the choke from the chassis  will eliminate any V breakdown.   I tried an experiment an inserted a  new hammond 1H  @  1 kv rated choke, in series with a 2650 vdc supply, zero issues...as long as the choke is floated from chassis on glastics, or some other type of insulator.   Again, an adjustable  spark gap was used.

Jim    VE7RF

[WBear2GCR]

My only concern with floating the choke is that it could become "hot" with B+ voltage
making it a deadly place to touch.

As far as I know, turning the whole filter section "upside down" and putting it in the
"return" ought to perform identically to the situation where it is on the hot side.

Why it would not do that? That needs some explanation and thought.

                _-_-bear

PS, the idea that the "higher harmonics" would sail through? That is false, IF the
filter performs as or nearly as the simulation shows. The attenuation after the notch
is no less than it would be without the "resonating" cap. The only explanation
would be that the Q of the "resonating" components is essentially poor and the depth
of the notch is less than desired, so the filter does not follow the 3rd order rolloff
curve subsequently... (which is what a C-L-C is...)

Ok see below:

Generic Elliptical (Cauer) filter, not the notch and rebound, but how many dB down
it meets the "curve"?

And the comparison to a 4th order, C-L-C-L, the green trace.

[Opcom]

Well, ok!
Good!

Whose SPICE model is that?

Since I happen to have a KWS-1 and I am told that apparently they are "known"
to blow up chokes, this model is of some interest!

Wondering what happens if one moves the choke to the ground leg of the PS?
The current will remain the same, but the voltage across the choke?

It would be good to use the Spice model to see the freq response of the filter section,
with load and without.  the signal source would need to be a swept sine)

Love the "dit and dah" generators!!
That and the load "switch"

Cool.

                               _-_-bear

Sorry for the late reply. That pulsating power supply for dit/dah is my doing. Was looking at what happens in power supply topologies when the load is abruptly changed. One with dits is attached as a text file, you can rename to .asc and it should run.
You can add a dah by adding another dit gen with a single longer pulse. rise and fall times of the keying are also setable in those voltage sources. Another one with a soft start is attached which is a little different. These might provide some ideas to modify for your own simulations. The beauty of ltspice is that you can not only use a transistor to switch thousands of volts and amps, but also drive the base with same.. LOL Just use the generic model and don't pick a part. It doesn't matter because the only goal of these is to cause behavior in, or to the power supply.

[VE7RF]

My only concern with floating the choke is that it could become "hot" with B+ voltage
making it a deadly place to touch.

As far as I know, turning the whole filter section "upside down" and putting it in the
"return" ought to perform identically to the situation where it is on the hot side.

Why it would not do that? That needs some explanation and thought.

                _-_-bear

PS, the idea that the "higher harmonics" would sail through? That is false, IF the
filter performs as or nearly as the simulation shows. The attenuation after the notch
is no less than it would be without the "resonating" cap. The only explanation
would be that the Q of the "resonating" components is essentially poor and the depth
of the notch is less than desired, so the filter does not follow the 3rd order rolloff
curve subsequently... (which is what a C-L-C is...)

Ok see below:

Generic Elliptical (Cauer) filter, not the notch and rebound, but how many dB down
it meets the "curve"?

And the comparison to a 4th order, C-L-C-L, the green trace.

My  floating choke is at the bottom of the 24"  wide x 24"  deep new hammond rack...with the  smoked plexiglass in the front door.   A  ton of lytics in racks above the choke, + the  air cooled  50 ohm glitch assy  at the top of the  rack.   Key lock  front and rear doors,  +  1/4  turn cam  locks to remove one / both side  panels.   0-10 kv dc   B+ meter inside, can be seen  through the smoked plexiglass.    I also installed 2 x 120 vac neons in series....  and wired that mess  across the very last lytic ( cold end) of one of the strings of lytics.  You can see the pair of red neons lit up from across the room.  I did the same on my last  6700 vdc supply, which had an external  hammond cab with the  20 x 2000 uf lytics in it..and a solid  al  front panel..with the HV meter..... + the series pair of neons  across the last cap.   The neons  are available   in  red /green /amber....and  in  2 x diameters. I use the smaller  diam ones, that only require a 1/4"  hole in the panel, the actual neon is bigger  diam.   A pring tab comes with each neon, and slides over the back end of the tube assy..and locks the neon  solidly in place.   I assume everything in that cab is hot.

I drew this out  on paper, and inverting the entire  filter network is in some of the older arrl handbooks.... putting the choke in the B-  line.  It sorta works  per the handbook, but part of the  choke  gets  shunted..and it's effect is not as pronounced...per the arrl book.    Biggest issue for me is  any  V drop across the  choke in the B- leg is now additional, unwanted bias on the  GG triode tube.    I don't want  any V drops in the B- leg.

...

I searched and searched, and can  not  find  any formulae for a  C-L-C  filter network for  power supplies, only   L-C  filters  ( 12 db per octave) ..... or  R-C filter  networks,  (  6 db per octave).

In one of the glass audio sites  ( some of their members had an  EE  degree), when a C-L-C  was used,   to calculate the  120 hz  attenuation, they  just used the  standard  L-C  formulae.... but  with an added twist.   They  totaled both  C1  and  C2  cap values  1st.... then took that  sum..and stuffed it into the  standard  L-C  filter  formulae...which I believe is incorrect...( but I may well be wrong). 

When I used PSUD-2...and a C-L-C  filter,   (with a big load),  I can see the exact amount of  P-P voltage   across the  C1  cap.....  = 18.674 v  p-p, or -58.51 db.

 I can also see the exact  P-P  voltage  across the  C2  cap (which is  of course... way down in value)....=4.4433  millivolts P-P... = -130.96 db.     I then  deleted the entire  L-C  section..... leaving only the C1  cap intact.  The  P-P  voltage across the  C1  cap remains  identical...whether the following  L-C  section is   deleted, or included.

What it amounted to was the  C1  cap alone attenuated the  120 hz  ripple by a bunch.  Adding the  L-C  to it ( resulting in a C-L-C...or  call it  C1-L-C2)  just increased the  total attenuation  at 120 hz  by a huge amount. (72.45 db more).   If I worked backwards,  at  12 db per octave, the  corner freq  was just   1.875 hz. ( 6 octaves back from 120 hz).    That  1.875 hz  corner freq is  just from the  added  LC section.

C1 is  625 uf
L  is 4 H (9.56 ohms  dc resistance)
C2  is  1250 uf

Loaded  B+  is 5548 vdc
Loaded  current is  1.5 amps.

(Psud-2  does not  depict  harmonic content, nor   resonant  chokes, or negative power  supplies,  3 phase power, and a bunch of other stuff.)

When  John  Lyles designed that single phase  B+  supply for  BE  (FM broadcast  PA)   he used a resonant choke setup..+ C1  filter cap.   Harmonics of  120 hz, like  240,360,480 hz  etc, did not meet spec.   So in went a 2nd  choke ( not resonated)  + a  2nd cap.   2nd  choke and 2nd cap was there to reduce the harmonics  to meet  FCC  specs.   He ends up with a (resonant L)-C-L-C  setup.

Local power supply EE genius is telling me  with lower value chokes, like my  4H, the  required resonating cap  is big enough in value, to pass  harmonics. 0.439762082 µF...which has 1507 ohms of  XC  @  240 hz.    (1005 ohms  @  360 hz )  (754 ohms  @  480 hz)    etc, etc.    My  C-  (resonant choke) - C   setup,  esp the resonant choke + 2nd cap,   no longer forms a standard LC  filter. 

Patrick kindly modelled the above mess for me last year, with and without the resonating cap on my  C-L-C  setup..and   with the resonating cap installed, the  total  attenuation at  120 hz increased by a  meager  4 db, that's it.   Dunno about the harmonics, that was not modeled afaik.

Fwiw,  the glass audio folks  never resonate the  choke on a C-L-C  setup.   But they design the  C-L-C  so the corner freq is  <  20 hz.... like  typ  16 hz.

On a side note, I simulated a C-R-C  setup,  but the  required  R value was  too  high  imo, like  50-52 ohms....  just to get a lousy  6 db  per octave  roll off.   And if a  2nd  R-C  network  was added to the mix,  to make a C-R-C-R-C, (  12 db per octave roll off).... the  2nd resistor  ends up the same value as the 1st resistor.   The 75 vdc  Vdrop  across  each resistor is unacceptable imo,  and the power  dissipated was  wicked,  112 watts  per resistor.  It would work  for a bias  supply, or perhaps a screen supply, that drew very little  screen current.

sri for the diatribe.

Jim   VE7RF

[WBear2GCR]

<snip>
Patrick kindly modelled the above mess for me last year, with and without the resonating cap on my  C-L-C  setup..and   with the resonating cap installed, the  total  attenuation at  120 hz increased by a  meager  4 db, that's it.   Dunno about the harmonics, that was not modeled afaik.

Fwiw,  the glass audio folks  never resonate the  choke on a C-L-C  setup.   But they design the  C-L-C  so the corner freq is  <  20 hz.... like  typ  16 hz.
<snip>

sri for the diatribe.

Jim   VE7RF

Anyhow, filters have Q. One can not disregard Q when looking at attenuation and rolloff?

There's no way a properly designed "resonated" filter will "add only 4 dB of attenuation" at 120Hz. unless
A) the notch is not at 120hz, only the -3dB point (or similar) or B) the Q is thoroughly wrong so that the
filter does not work properly.

So, if, as the glass audio people do, design for very low freq -3dB point, then the Q becomes less of
an immediate worry... but IF you build a 2nd order filter (C-L) with improper values it will look like a 1st order filter.
So, assuming a 15Hz. -3dB target, at 120Hz one would be 36dB + 3dB down. However if it worked like 1st
order one would be only 18dB + 3 dB down.

Guessing at filter values is ok IF one bothers to measure the results empirically. Imho.

                       _-_-bear

[Opcom]

LTspice gets pretty close so far. But measurement is always done once the decision's been made to hook it up. The 40KV TV probe and a scope with vacuum tubes only.