RF transitting capacitors - current ratings at different frequencies

[Opcom]

I made up a spreadsheet to help me figure out how much to derate a "RF transmitting" type doorknob capacitor that gives a rating of 15A @ 30Mhz, to use on lower bands.

Does this make any sense or are there factors I missed?
I don't know so much at all, but believe they have to be derated for lower frequencies.

The output of the sheet gave this chart.

The chart has two caps.
470pF, 15A @ 30MHz
1000pF, 25A @ 5Mhz.
The green column shows the derated current at a given frequency.
The blue column is just the spec of the cap in question.

Lines 1-4 show different currents allowable to lower frequencies for the first cap.

Lines 5-9 show different currents allowable to lower frequencies for the second cap.

(ignore the lines 10-12 - the current was calculated to be higher at frequencies above th4 cap's specified rating. I don't think that's appropriate.

[Steve - K4HX]

What was the derating formula or factor?

Good stuff.

[W1ITT]

I am guessing that you are talking about doorknob capacitors of the Centralab type.  I see that Comet vacuum capacitors have a similar derating curve and it appears that heat generated by losses is the limiting factor.  They go so far as to make water cooled vacuum capacitors!
This leads me to believe that if we, in common amateur fashion, wish to push the ratings a bit, some sort of cooling might be advantageous.  While we probably wouldn't use a blower in an antenna matching unit, it might help to arrange that any doorknob capacitors in a transmitter be placed where the cooling air could get on them.  If nothing else, this would at least provide a bit more safety factor.

https://www.comet-pct.com/getmedia/47fcbd2f-1729-4ffe-800e-14b62c58561d/SB-44_Current_Limits_for_COMET_Vacuum_Capacitors.aspx

73 de Norm W1ITT

[K8DI]

Interesting stuff. I get that as frequency decreases, heating loss goes up due to the reactance. The Comet data show a knee at varying frequency where the current rating also drops as frequency increases, which I’m finding counterintuitive. Anyone know why it has a break over and what influences it?.

Ed

[WD5JKO]

Interesting stuff. I get that as frequency decreases, heating loss goes up due to the reactance. The Comet data show a knee at varying frequency where the current rating also drops as frequency increases, which I’m finding counter intuitive. Anyone know why it has a break over and what influences it?.
Ed

   Take a .01uf ceramic disk, and wrap the leads around a Grid Dip Oscillator coil and tying the ends together. You will see a resonance around 7 Mhz. This is a resonance where Xc = Xl of equal amount and opposite sign. That leaves just the ESR, effective series resistance. The ESR is what gets hot from I^2*R loss. The charts in the Comet Word Doc seem to show the series resonance point at the slope change up top, or close to it depending on lead length. Above series resonance, a capacitor is inductive. Perhaps the Max current decline from there is due to skin effect losses increasing with frequency? Edit: Has more to do with the inductive reactance as frequency rises above resonance.

Jim
Wd5JKO

[Tom WA3KLR]

(This is what happens when I have time on my hands.) Attached is a photo of a ceramic transmitting cap I have.  It has 3 current versus frequency ratings on the nameplate and I have wondered about this situation too.  This is bigger than a ‘doorknob’ type, the round ceramic body about 3 ¼ “ diameter.

I think this is the only cap I have like this and it has been lying under the workbench for many years.  Apparently at one time probably more than 10 years ago I did some measurements, and some are calculated, I don’t recall how I came up with the series resistance. I think I was starting to investigate the seemingly counter-intuitive ratings.

Most of the name plate may not be legible in the photo unfortunately. Here is what it says:

Aerovox Type 1960-304 mica capacitor
.003 Mfd. 7000 Volts 60 C ambient
20 Amps at 3000 kcs
16 Amps at 1000 kcs
10 Amps at 300 kcs.

Also showing is the paper tag I made stating I measured 2950 pf at 1 kHz. Somehow I measured the self-resonance at 22 MHz. This calculates to 18 nH series inductance at 22 MHz. Also stated is the series resistance of less than 1.2 milliOhms at 3. 9 MHz. Again, I don’t recall how I did this measurement.

I guess we would need a ceramic capacitor engineer to explain everything but here is my take.  Basically the part is made of thin metal layers and ceramic dielectric.  No rating can be exceeded.  There is a voltage breakdown rating, maximum current determined by the cross-section of the metal layers & number of layers, maximum temperature reached based on the acceptable maximum of each of the materials used. Temperature rise from internal dissipation plus ambient also gives a limit.  Just an intuitive guess, I think this big flanged part could handle 50 – 100 Watts at 60 C. I’m just throwing together some numbers that may be needed to figure what is going on. Margins placed on ratings.

I would model the part as, between the two end flanges, the series resistance which I presume in this case is not frequency dependent, next in series - the series inductance - not frequency dependent, then the  capacitance – perhaps slightly frequency dependent, and the dielectric loss which is frequency dependent as a resistor in parallel with the capacitor.  

Reactance in itself is lossless. Reactance goes up as the frequency goes down. Dielectric losses increase as the frequency increases. 20 Amps with 1.2 milliOhms series resistance gives 0.48 Watts.  The dielectric loss I don’t know, I don’t think it would be too many times this? AC voltage developed across 3000 pf at 3000 kc and 20 Amps = 354 V rms, 500 V peak. AC voltage developed across 3000 pf at 300 kc and 10 Amps = 1768 V rms, 2500 V peak.  My only guess is that the limiting factor is actually the voltage across the capacitor and they are adding dc voltage and margin? My conclusion, current-limit at 3000 kcs, voltage limit at 300 kcs. Just my guess. Perhaps someone like broadcast engineers Tim WA1HLR or Jamie N2VJ has an idea on this.

[Tim WA1HnyLR]

Hi all I got an E mail from Tom,KLR Re: RF  handling of capacitors in RF service. In a way comparing mica to ceramic capacitors is like comparing apples to oranges. In the case of a mica condensher. , the ratings plate indicates an increase of current at progressively higher frequencies. as you can see the curve flattens out about the 20 A rating on the .003 Mfd cap. Micas are not used in many HF operations above 7 Mhz except for the much smaller types such as the molded postage stamp type and dipped silver mica's which have a low ESR compared to some ceramic designs. The physically large construction of the ceramic cased transmitting capacitors. will exhibit resonances that my wind up in the area of interest. It would be best to do the grid dip osc with a metal strap across the cap to see if there is a potential problem. It will be noticed that the smaller mica transmitting capacitors are rated to higher frequencies. It is also dependent on what purpose you are looking for. The large ceramic cased mica's will work well in 160 & 80/ 75 meter applications in the tank circuit such as padding the PA tuning,and addition capacitance needed for loading. A much larger 300Pf cap in a bakelite type case perhaps @5Kv makes a good permanent replacement in a Viking Valiant in place of that troublesome cluster"F" of small mica caps used for the 160/80 meter tank padder. Ceramics on the other hand behave differently.  A typical 500Pf 20 Kv TV door knob capacitor does exhibit a high ESR and will not tolerate much RF current as I found out. Case in point: early experience with using 2 500 pf 20 Kv door knobs is series
as the replacement in a Valiant. It appeared to work at first but I noticed when making an old buzzard transmission the PA plate current kept going up as the the capacitance value kept going down. I had to re-dip the plate frequently. I felt the caps after obviously going into standby. The were quite hot. . These caps are ok for medium powered transmitters like a single 813. I have had large ceramic caps that were like TV door knobs on steroids , 2000 Pf @ 30Kv blow up and catch fire in my 4-1000 transmitter years ago. Only a very large RF rated 500Pf 15 Kv was used did the problem end. I am not sure how well a large ceramic cased mica would have handled 10 meters. There are now a number of good RF rated ceramic caps that are out the such as the Coment brand that Norm W1ITT mentioned. The construction is like that of a shopping cart wheel. Once again it depends on your application. when in doubt,always overkill.
Tim WA1HnyLR

[Opcom]

What was the derating formula or factor?

Good stuff.

I missed getting back to answer the question. the chart was based on the VAR calculated from the capacitor's current rating @ MHz as stated on the capacitor.

It turned out to be the same as the very simple concept of derating the current by % decrease in frequency.

The VAR is what I believe heats the capacitor due to current, and the caveat is that as the frequency goes down, the voltage across the cap will go up, so be careful of the voltage rating of the capacitor.

However, I can see from other replies that the chart is not necessarily to be trusted.

It is not for "DC filter" doorknobs as from old TV sets and laser power supplies, but was intended to be for 'solid' RF-rated caps like the micas and the RF doorknobs.

I can see from the good information in the other posts that the chart is not reliable and shouldn't be used. That is a good answer.

[VE7RF]

You can obtain the RF current ratings  for any  TX doorknob cap, from  HEC  (high energy corp)  website...for any frequency.
I  typ will  use the HT-50/58  series for  some applications....and the larger  HT-57's  for  plate block caps.   The...'go to' cap in the
HT-57 series is the 200 pf  unit..  rated at  15A.  Typ  4-8  are used in parallel for plate block caps on the metal tubes.  Sometimes the even larger  HT-59 caps are  used for plate block service.  The ..'go to'  cap in the HT-59 series is the 250 pf unit,  rated for 35.5A.

The current  flowing through any plate block cap is substantial  at higher  freqs,  like 18-60 mhz.  I just finished designing a YC-179  6M  amp for a fellow, and  with a modest 4.5 kv  under load, on  50.125 mhz,  a whopping  44A flows through the plate block cap assy.   4 x HT-57's in parallel  are used,  all are  200 pf  units. 

Download the HEC  catalog here.   http://www.highenergycorp.com/High%20Energy%20Corp.%20Ceramic%20Capacitors.pdf
It  will  list max  current  vs freq (1  khz  up to  100  mhz)  for every cap in their catalog..in graph  form.   Starting on page 36, the theory is  in great detail.

Knowing the cap value, the  XC  for a given freq is  easily calculated.  Knowing the  RF current  flowing through the cap, the  AC  V drop across  the  cap  is easily calculated..(  typ  barely minimal for a plate block cap at  upper HF).  V  drop = XC  X current.   Peak Ac V will be  41% higher of course.    What HEC  calls....'DC bias',  hams  call.. "B+".   B+  and the peak  AC  Vdrop  across the cap,  can not exceed the caps  V  rating.

 Calculating  VAR  power at any freq is also a simple matter. ( RMS V drop  across  cap  X  current through  the cap)

All of this is already done for you on  most,  but not all caps  they list in their catalog.   The graphs for each cap value  will  graph   current,  voltage, and  VAR.

Be careful,  selecting a cap is  going to be for a specific application.   IE:  if  say padding a cap for a tuned input..or padding a load cap  on the lower bands,  typ current  ratings of the  HEC  series caps  will  be lower.   Padding caps are  critical, since you don't want them to drift, since it's part of a tuned circuit.   Plate block caps are  coupling caps,  drift is a non issue....ditto  with bypass caps.

Typ plate block cap assy  won't handle as much current on  160 +  80m..vs  10m....  but this is a  non issue.   There is  very little  current flowing through a plate  block cap on the lower freqs.

Centralab caps have  virtually identical  specs  to  HEC caps.  HEC HT-57  series  caps  are the same as  centralab  57  series  etc.

Jim  VE7RF

[VE7RF]

As a supplement to  my previous post,  calculating the  RF current through a plate block cap at a  given freq is as follows.
On  upper  HF freqs, like 15-12-10-6m,  the  tube C  on  a  GG triode  makes up most of the  required C1  tune cap  value  for a PI, or  PI-L  output tank.   The  tube C is the  C between the  anode and the  grid (grid bonded to chassis).   The  tube C  is  directly in parallel  with the C1  tune cap.  Problem is.... the plate block cap  resides  between the anode..and the hot side of the  C1  tune cap.  On  10m  band,  think of it as the  tube C... plus a tiny variable padder(C1).. in parallel.   All the RF current that flows  from anode to grid..now also has to flow through the  plate block cap.   On the lower bands, like 160-40m, the  tube C  is  only a very small  portion of the total tune C required.

Xc  of the tube C is higher on lower bands.   Xc  of the tube  is lower, on  upper HF bands.

To calculate the RF current flowing through the tube C, (and also the plate block cap) is  just RF  voltage / XC.   (rms)  RF  voltage is .6 X loaded  B+  voltage. 

As  you can see, on  upper hf bands,  tube  XC  is much lower..and more current flows.

Beware, on metal tubes like my  3CX-3000A7 etc, the tube C (24 pf for the  3CX-3000A7)  will  increase to 33 pf, when  tube is inserted into the  grid ring /socket.   The  increase in  C is  because of the proximity of the lower anode fins.... to the chassis below the  fins.  This is easily measured   with any  digital  LCR meter (I use both a B+K  875A..and also a 875B).

Tubes like the  YC-179 and  YC-156  have  35 pf of tube C  (anode to grid)..which increases to 50-52 pf..when  integral  grid flange of tube, is bolted to  chassis.   Increased tube C with other metal  GG  triodes... like 3x6,  3x10 /15/20.

Just something to be aware  of when  doing calcs.

The  4500 vdc of loaded B+... and the 52 pf of  tube C  of the YC-179...and a freq of  50.125 mhz... results in a whopping  44A  flowing through the  tube  C...and also the plate block cap  assy.  And it all  gets  worse, if  loaded B+ is further increased.      The same tube, same loaded B+  is a non issue on lower freqs.   Plate block cap  value is also a non issue  for the most part.  W8JI  replaced the  1000 pf plate block cap in an  Ameritron  AL-80..with a  170 pf  cap..... and on 160M,  zero difference in power output..and only a tiny tweak on the
 tune + load  caps.

Sri  for the  diatribe.

Jim  VE7RF