Choke input filter power supply "Critical value of Inductance" questions

[KJ4OLL]

Hi,
Using my last few days of vacation to make some progress on my first attempt on a homebrew amp.
N2ZAB educated me (I'm not an engineer!!) about the flaw in the power supply design, so I am fixing it.

It is fun to learn this technology, even though at my age, there are only about 3 functioning synapses remaining, and two of them are always looking for my glasses.

The power supply uses a choke input filter.
The power supply is a single phase, full-wave rectifier, sine wave source voltage design.

Filter circuit:
- not resonant, no luck finding the components at auction sites and hamfests that would form a resonant circuit, can't afford to have them manufactured.
- not a swinging choke.
- Just using what I could find: Fair Radio had some 10H fixed chokes from some kind of Gates.



The "Critical Value of Inductance" is what I missed.

Searching the forums and the web and found lots of good information.

The 2013 handbook has a couple of paragraphs on choke input filters.

Found many different formulas to calculate the Critical value of Inductance for this application.

I have used two of them, (the one from the handbook, and one from a paper by Henry Pasternack) just to see if they agree,
and if I am getting close.

PLATE VOLTAGE
There are three output taps on the plate transformer.
Using the "2600" to start with.
4000 and 4500 are also available if I feel lucky.



 

If the filter choke can be made to work as it should, I could use the "4000" tap and obtain a B+ of about 3600 VDC, which is probably OK for the two 6MFD filter caps, as they are rated at 5kv.

The PA is a 3CX3000A7, which the EIMAC spec sheet says is OK up to 5kv.
 

ARRL HANDBOOK INDUCTANCE CALCULATION
7.8.6 in the 2013 handbook has a chart for finding a constant, based on line frequency.
So I used 60Hz, which gives a "choke inductor constant" of 1100.

The formula in paragraph 7.8.6 is L =R/A

The existing 225 watt bleeder resistors yield R=25k ohms.

So if I am doing this right, L= 25000/1100

L=23H

I have two 10H chokes I can connect in series, so this is pretty close.



ALTERNATIVE INDUCTANCE CALCULATION
From a paper by Henry Pasternack in 2009:

L = vdc/ima
L = minimum choke inductance
vdc = dc voltage output of rectifier
ima = load current in milliamp

ima: I am still experimenting with the Zener circuit that sets the idle current for the PA.
Currently a 5.1v Zener yields 100ma.
It may be that I can just ground the CT of the filament transformer, and delete the Zener, to obtain the correct PA idle current.
The EIMAC 3CX3000A7 specs for a "Radio frequency linear amplifier cathode driven class AB2" call for 250 mills "zero signal plate current" @ 4kv.

So my assumption is for about 200-300ma, (bleeder plus PA idle current), for purposes of this calculation.

2500VDC is what the volt meter actually reads from the rectifier output, using the 2600vac output tap from the plate transformer. I think the higher reading (should be about 2300) is due to the utility presenting 245vac to the plate transformer primary.

2500/200 = 12.5H choke
2500/300 = 8H choke

If I dare to use the 4000VAC tap on the plate transformer x .9 for a choke-input design:

3600/200 = 18H
3600/300 = 12H


- Seems like if I use the two 10H chokes, I will have enough inductance for either of the two plate transformer taps (2600 or 4000)
- Planning to keep experimenting w/ the PA idle current to get close to 250ma @ zero signal.
- Planning to experiment with the 225 watt bleeders to get a bit more current.

Thoughts? Ideas? Am I on the right path, or is it time to drag out the fire extinguisher?

73
Frank
KJ4OLL

[N2DTS]

I have built many power supplies and never did any sort of calculation.
I use chokes that are good for the current, and caps that are good for the voltage and whatever bleeders I can get that put up with the wattage.
I do not like burning up a lot of power in bleeders in big supplies.

I thought many amps did not even have chokes in the power supply.

I would use one choke and as much capacitance as you can, with the voltage ratings on everything higher then the supply could get to.

Are you going to leave the supply on or key it with the amp?

[KA2DZT]

Don't over think the power supply.  That choke will handle 4500 VDC probably more.  Don't worry about critical inductance, a 10Hy choke is good for the input choke.  What happens if the input choke does not have enough inductance, is the  filter capacitors will charge up closer to the peak voltage.  The more bleeder current you use and in addition the xmtr is drawing current the supply won't charge up close to the peak voltage.


The thing you need to use is filter capacitors that have a high enough voltage rating.  If you are going to key the HV supply on and off with xmit and receive you don't to need to have caps with a voltage rating much higher than the DC voltage.  OTOH if you are going to leave the HV supply on during receive and key off the xmtr some other way then the HV may rise closer to the peak voltage at which point the only current pull from the HV supply would be the bleeder current.  Critical inductance depends on the minimum amount of current load on the supply.

Fred

I just re-read you post.  You have two 10Hy chokes.  Two 10Hys in series is more than enough inductance.  You can use a LLC filter or a LCLC filter.  You'll get better filtering with a LCLC supply.  The first cap needs to have the higher voltage rating and it doesn't have to be a lot of capacitance 6-10mfd is good.  The second cap in the LCLC filter can have much more capacitance.

I would suggest keying the HV supply on and off with xmit and receive, it solves a lot of issues and is much safer.

[gerry_w1id]

For a full wave rectifier the frequency is 120 Hz, not 60 Hz. Also keep in mind the DC resistance of the choke as this will result in a voltage drop under load. However with the choke you  show I don't think it will be very much but just keep it in mind.

[John K5PRO]

IN the olden days, critical inductance was important factor to keep regulation from no load to full load good. We had to pay attention to it in commercial gear as FCC regs in part 73 suggested keeping the final stage voltmeter operating in the upper third of scale. If you ran it this way, and removed the excitation of a class C amplifier, voltage would soar to the point of pegging the meter offscale. I've seen this is some old Gates rigs. Its always the first choke between the rectifier output and the capacitor that cause this. As long as your bleeder and the combined effect of modulator bias or RF PA idling current are presenting the minimum load for the calcs (the 1100 x) at 120 Hz, then it would only soar hundreds of volts when RF is removed. If you completely ignore this, be prepared to design some overhead in the capacitor voltage ratings. I wouldn't worry about the 3CX3000A7 handling the voltage of sloppy power supply voltage. By removing the first choke entirely and using a big big capacitor as some do (40-80 uF) you can still get effective ripple reduction, and the voltage stays closer to the peak of the rectifier output. In this case, the variation is due to rectifier Vdrop, transformer impedance, and value of cap/maximum load current. The disadvantage of doing this is that now you have a lot of stored energy (joules=1/2 x C x V^2) in that capacitor. Like 250 Joules with 40 uF and 3600 VDC. Eimac recommends limiting any tube fault current to 5-10 joules, to prevent destroying the grid inside the tube. You have to account for this with either am active crowbar (don't go there) or a dissipative series resistor like 10-20 ohms. This is good practice even with an LC filter and smaller capacitor.

Something else to watch for (power supplies are not trivial, despite what some say!) is the LC resonance in the supply. This is seen in two places. When you switch on the HV first time, you may get a very large transient voltage spike, hard to see on a meter but on a scope there it is. This ringing can be 50% of your DCV or more. The other time is when keying the rig with CW, at some rates you find the voltage hunting up and down. It also was a concern in old TV transmitters with tubes, the visual signal was AM and had a lot of regular modulation repetition in it (sync and dark part of picture for example). You can read about this is some of the papers on this site, old copies of GE Ham News that describe it.

So yes, you should at least calculate the critical inductance and improvement from lower bleeder R and idling current, and compare to your ratings for capacitor voltage, meter scales, voltage divider ratings, bleeder voltage rating. To ignore this and tempt fate would be foolish and poor engineering. Going to a dual section filter LCLC isn't recommended with intermittent loads like AM, just put all the L up front like you planned, 2 x 10 Hy in series.

[John K5PRO]

Link to GE Ham News articles on Dynamic power supply regulation

http://www.amwindow.org/tech/htm/tutor.htm

[Opcom]

On 'critical inductance", or the relationship between current and inductance,

Thus wrote Terman:

L1 = Reff / 1130

L1: input inductance.
Reff: effective load resistance.
1130: constant based on 120Hz ripple from a single phase 60Hz full wave rectifier.

With a given L1 ... the equation ... is not satisfied when the load resistance exceeds a critical value.  (meaning when the load current is less than a certain value)
{Radio Engineering, 2nd ed. 16th imp.,  pp.489 eq.198a, 198b}

This is like the OP's original statement, which seems right. A graphical idea of this can be made by running different currents or load resistances in PSUD and copying the voltage and current data to a spreadsheet and graphing it. There is a definite knee type function where the critical area is.

About transients and oscillations, I wonder when it was that engineers started paying attention to them.

[KA2DZT]

All this info is good to know.  I read some of what John posted from the GE papers.

Bottom line,  most hams are lucky to even have a few pieces of iron to work with.  The OP has pictured better iron than a lot of hams wish they had.

Build the supply with what you have.

I looked at the graphs in the GE papers, seems a LC filter is better than a LCLC.  But it seems to become really good with a lot of capacitance, they show some 90mfd.  Who has 90mfd at 5KV?  The OP has a total of 12mfd to work with.  With only 12mfd, I agree with John and put both chokes in series LLC and it should work OK.  He can always add more if can find the caps.  The xfmr is rated for an amp so he can increase the bleeder current to whatever he can do with the power resistors he can get.  He can run 100ma of bleeder current if he has big enough resistors.  At 5KV with 100ma current you're looking at 500 watts of bleeder power.  He would need a minimum of at least 1KW of resistors maybe more and fans to cool them.

So, you can see that trying to maintain critical inductance with high bleeder current can become a real headache.

Fred

My power supplies use LCLC filters for the PA and the modulator uses an LC filter, seems to work OK and I key my supplies on with a step start.

[N2DTS]

I myself would use one choke and as much cap as I had on hand and do step start.
Nice iron, nice chokes!

[Opcom]

For me the bleeders are economical and a technical necessity. For others, it depends on what the rest of the set looks like.

I use two 100mA bleeders, one on each of two 3500V supplies in the COL Tucker transmitter. The only disadvantage is the space it takes up but I am very pleased with it. Current/power being wasted does not matter because the plate transformers are each rated more than twice what is required. - like the nice stuff the OP has, but older.

Each bleeder is made of two 1" diameter by 14" long, 17K/225W tubular resistors, and all 4 are mounted via 1" ceramic standoffs to an old 4U rack panel. The whole thing is mounted vertically / end-wise in the bottom of the TX for convection through the resistor bodies. The cabinet is relatively spacious like the OP's, and it has generous screened louvers in the lower areas of the back doors through which air is entrained to cool the transmitter. The air is moved out of the transmitter by the blower of the 4-1000 final amp so it does not take any extra effort.

It was necessary to use a decent bleeder in order to maintain critical RL relationship for good regulation.
The B+ does not move around as much under varying modulation and this keeps distortion lower and reduces oscillations in the supply when syllables start and end.

[KA2DZT]

I took another look at the xfmr, I stated earlier that it was rated for an amp.  It is the chokes that are rated for an amp.  Can't see what the current rating of the xfmr is,  But it looks like it could be an amp or something close to it.  Anyway, it's more than enough to run the amp.

Fred

[W7TFO]

A 2-pole mercury plunger relay is a cool way to start and stop the primary in a supply like that using PTT.

Lasts forever, doesn't shake the cabinet either.

73DG

[K1JJ]

I've always liked the additional advantage of a power supply choke: The ability to short it in or out of the circuit by using a vacuum relay to select another voltage level.

Without using a Variac and using a transformer tap and the choke relay, I can get 2KV, 2.5 KV, 3KV and 4KV from my HV supply.   The 2KV is very useful for testing.

Yes, adding 10-20 ohms in series with the HV supply output is a very good idea. I do it on all my HV supplies. In addition, I use a HV fuse in series with the diode rectifier stack. This has saved the diode stack many times.  I use a pair of ceramic insulator standoffs and bridge them with one or two strands of thin copper wire from RG-213 coax shield. Works like a charm. Be sure to space the insulators at least 2" apart.

BTW, I also use 5 ohms of step-start in the 240VAC primary of the HV transformer. This allows me to get away with 140 uf of filter capacitance without a choke, when selected.

T

[KJ4OLL]

Hi,
Thanks very much for all the information and suggestions!

The amp is currently designed with B+ on at all times.
I'm going to leave it that way until I get the rest of the thing done.
Can always go back and do mods later.
I have been told that a homebrew of this type is never actually "done".

My goal is to have the amp operational in some manner before I retire in about 18-24 months.

I was actually pretty far along, working on the cathode drive, hoping to get to the point that an exciter could be connected and some output might appear.

But after learning about the power supply inductance value issue, seemed like a good time to go back over the power supply again and get it right.

The plate transformer is rated at 2A, so there is more than enough energy to work with.



There is a step start circuit, because KLUNK is necessary here.



If I use the suggested 120Hz (instead of 60Hz, as I originally used) for the "line frequency" calculations  from the ARRL 2013 handbook, the inductance value really improves, confirming that there is all the iron needed. This is good as there is not much more room.

I once thought "This cabinet is too big, never fill it up"



And then:


There is just enough room to put an antenna tuner in the top!

Other things that have been suggested, and I am trying to include in the design:

FILAMENT CHOKE
- replace the existing "high voltage drop" homebrew filament choke with a commercial unit.
Plan "A" homebrew:


Plan "B" Replacement:


If this does not work, plan "C" is to float the filament transformer and put chokes in the primary side.

SWITCH TO A GROUNDED GRID
- Originally floated the grid so as to make it easy to measure grid current.



- Now grounding the grid instead.

TUNED INPUT CIRCUIT
Originally just going to drive the cathode thusly:



Now adding variable caps and inductor.

GLITCH FUSE
- adding to rectifier output a piece of #28 AWG solid on ceramic stand-offs.

TUNER
- Once all the basics are operational, using these to build an antenna tuner into the top of the cabinet:



The "job" thing is really interfering with my hobby, can't wait to retire and do this full time!

73
Frank
KJ4OLL

[KA2DZT]

WOW!  2 amp xfmr.  I can power my xmtr with that also, if I had a long enough HV cable to reach NJ

Whole rig as pictured looks great.

Good luck with it and keep us posted on your progress.

Fred

[John K5PRO]

Nice looking build, Frank. For those two paralleled mica capacitors, did you measure them with that dandy HP 4262 LCR meter or better yet, with a vector Z meter up at a few MHz? They get inductive pretty fast, so account for that. Don't try to run them above the first resonance that they have.  Glad the power supply components figured out to your satisfaction.

[KJ4OLL]

John,
I tested them on the HP 4262, but only @ the default frequency.
The HP only goes to 10khz, I can re-test @ that frequency.

No vector Z meter, just old HP stuff (another hobby).

Would doorknob caps be better than the Mica for this application?
I have a pile of doorknob caps left.
73
Frank

[KD6VXI]

Doorknobs rule,  as long as they aren't TV style.

Got a MFJ259, if so,  you have a vector impednallance meter.   

--Shane
KD6VXI

[KA2DZT]

John,
I tested them on the HP 4262, but only @ the default frequency.
The HP only goes to 10khz, I can re-test @ that frequency.

No vector Z meter, just old HP stuff (another hobby).

Would doorknob caps be better than the Mica for this application?
I have a pile of doorknob caps left.
73
Frank

Only thing I didn't like on your project, those two old mica caps.  I have draws full of them, probably not a one is any good.  I would go with doorknobs but like Shane mentioned, not the TV type.  Although I did use some TV types in my HB 6146 rig.  Low power, they've been working FB.  At the power level you're building go with some good RF doorknobs.

Fred

[Opcom]

The plate transformer for my 3CX3000 project is 4430-0-4430 but is only rated 1.5A. I feel so inadequate!  

Ought be Ok in reality. I like yours better though, new not 300LBs and a selection of voltages meaning no need for a variac. I'm glad to see this project moving along, been a while since so many good pictures!



Fred,
I have a bunch of mixed old and NOS micas more or less like that. They could be bad.. who knows. How much current has to pass through them, in the filament coupling circuit from the driving source? 3CX3000 looks like 50 ohms and needs 400W peak for max. output, that is maybe 2.8A and 140V at 400W.

I'm not experienced in what might make them go bad except for over-current and excess voltages or maybe QRO at self resonance. How do they fail? Do they rot from age? It's very interesting.

Also, what is the adequate or proper value for this use?

Thanks!

[John K5PRO]

The micas get stressed from both current and voltage. Find a Hp 4815A vector Z meter, WITH A GOOD PROBE. I got mine years ago, a friend just found one on ebay last month for <$200. All it had was one bad fast recovery diode in the sampler board, not the probe. You can see that the old tank micas of values like 0.01 uF are very inductive and have ESR above a few MHz. This eats their lunch when running current through them. Forget trying them at 10 meters. Depending on the Z, you can either stress them from RF voltage, or from I^2xR heating. I didn't know about them being often dead in the drawer ('DID') as well. That sheds a new light on what some sell as 'NOS'. I have a lot of them, both the stand up type pictured and the flatter ones that look like my VW's key fob.

I use old Jennings, CRL and High Energy 5 kV ceramic doorknobs for some amplifier jobs, the little ones like HT-58 series and 850 series. They aren't stable in tuned circuits, unless you use < 100 pF values, due to the type of ceramic used. If they are just DC blocking capacitors then they are fine. If you have too much reactance (too low a value of C for the lowest frequency) then they can overheat. In bigger industrial/scientific amps, I use 7.5 kV rated ones that are couple inches long, for things like screen bypasses (100 kW tubes).

[KA2DZT]

Most all the old micas I have are recycled ones that I don't know the history of.  They've been sitting around in draws for decades.  I have some of those big tank style ones also.  A lot I pulled from scrapped GR equipment, they're in different draws (the maybe they're useable draw).  A lot of those postage stamp silver micas I also wouldn't trust.  I have hundreds of more modern silver micas (the little brown ones) that I do use.

As for the OP's two micas, I couldn't tell what circuit they're in.  I thought they were possibly some loading caps, but, I guess it may be the input grid circuit.

Fred

[KJ4OLL]

John,
The HP 4815A looks really nice!
I sure would like to add one to the pile'o HP test equipment.
But after some searching, these seem to suffer from the same syndrome as the LCR and power meters - the probes went one way, the meter went the other way!

Lots of HP 4815A's for sale in lots of places, not one probe to be found anywhere.

I paid more $$$ for the fixture for the HP 4262A and the power sensor for the HP 436A than for the meters themselves, due to this issue.  

Frank

[KJ4OLL]

John,
Thanks for the tip about how an HP Vector Impedance meter can help with the amp circuit design and component testing.

Completely addicted to old HP test equipment anyway, and thus did not need any other justification to acquire
one of these wonderful instruments (Thanks to Paul, K0UYA for selling me his HP 4815A)



Would appreciate some guidance on practical usage tips.
I read the manual and even found an HP app note for the 4815A.

Of course, this part was completely obvious:



A more serious example is the cathode input circuit, where there are currently two Micas installed:



These test fine on the LCR meter, and so the next test is to see where they are resonant and/or become inductive at the frequencies the amp will initially be used at, 75 and 160M.

When I sweep these caps w/ the 4815A, from 1.8Mhz to 3.85Mhz, this is the result:

www.youtube.com/watch?v=uUWh_8qOmN4

Not an engineer here, have NO IDEA what the "Phase Angle" means about the suitability of these capacitors.

Google tells me that "Phase Angle is the phase difference between the voltage applied to the impedance and the current driven through it"

Google also found this explanation, which gives me a dim and opaque level of comprehension of the principles involved, but not how to determine if the meter readings indicate capacitor suitability for this application:

"The phases angle (measured as cos phi) is the degree that current lags behind voltage in the sine wave (voltage lead).
So, if you consider that a sine wave passes through 360 degrees and the voltage (v) starts at point zero rising through 90 following to 180 (and similar on the negative side of the sine wave) the phase angel would be the point at which the current (A) starts to follow the voltage lagging behind it.
Lets say by 10 degrees then the phase angle would be cos 10 = 0.97.
This "lagging phenomena" effect is due to the the reactive load on the circuit and has an effect on the power available,
For instance in an ac circuit power = VxA cos Phi so for a load of 10 amp from a 10 volt supply you would have a true power output of 97 watts (as opposed to 100 watts in a dc equivalent circuit).
Most supply load are inductive reactant which results in this lag, however capacitive reactant will have a lead effect - this is why capacitors are used as power factor correction to bring the angle back to as near as unity (angle of 0) as economically feasible* "


Going to receive some doorknob caps in a couple weeks @ the Orlando Ham fest, will test them in the same manner and see if there is any difference.
73
Frank
K4ILL

[KA2DZT]

Current leads the voltage in a capacitive circuit.  Current lags the voltage in an inductive circuit.  When both reactances are equal and opposite the circuit is at resonance.

Fred