[Amps] Filter-C

[R.Measures]

On Aug 18, 2004, at 12:15 AM, Ian White, G3SEK wrote:

> R. Measures wrote:
>> RE:  Filter-C, how much is enough?  A friend of mine built a 
>> 4-1000A/8166 g-g amplifier that ran c. 6500v on the anode from a FWB 
>> rectifier/C-filter.  The only filter-C he had on hand that would 
>> handle this much V was a 2uF unit.  To get the project going, he 
>> installed the 2uF C, but he left extra space on the chassis so that 
>> he could add more C when it became available - at the right price, of 
>> course.  To initially check out the amplifier, he fired it up with 
>> the 2uF filter C.  He fired up the amplifier on 80m.  I could hear no 
>> ripple in the SSB signal.
>
> It isn't likely that you would hear any 120Hz ripple on SSB, in among 
> all the other voice-frequency components.

   However, with 50Hz mains I might have noticed something.
>
>> Is this plausible?  Hmmmmm.  During a SSB voice transmission, 
>> indicated anode-I is roughly 1/3 of the max anode-I rating, so in 
>> this case, the average load would be 750mA / 3 = 250mA.  Thus, the 
>> equivalent RL on the PS was 6500V / 0.25a = 26,000-ohms.  The rule of 
>> thumb for the filter-C for a FWB is 70k / RL, or 70,000 / 26,000 = 
>> 2.7uF.  This led me to conclude that this rule of thumb might be a 
>> bit conservative. Eventually, a 4uF unit was added to the 2uF, 
>> bringing the total to 6uF. No subsequent change in audio quality - or 
>> change the HV meter indication  - was observed.
>
> 6500V @ 0.25A is an unusual combination of high voltage and low 
> current, where you can get away with a very low filter C.

In Hamdom, 4-1000As running at this potential on SSB are quite common.

> Running the numbers through 'PSU Designer' (and estimating the 
> transformer voltage and winding resistances) it seems like 2u would 
> give about 7% ripple at 120Hz. Increasing the capacitance by a factor 
> of 3 will decrease the % ripple by the same factor.
>
> As Angel has pointed out off-list, tetrodes are quite insensitive to 
> anode voltage, so HV supply ripple tends not to get through to the RF 
> output (remember how tetrodes needed to be plate *and* screen 
> modulated?). That isn't true of triodes, so I'd guess that if Rich's 
> friend had been using a triode-connected tube, Rich might just have 
> been able to hear 7% of 120Hz ripple on a CW carrier.

It was triode-connected, g-g.
>
> But that really was a far-out case with a very high value of RL. 
> Almost any other amplifier is going to run at both lower voltage and 
> higher current, so RL will be much lower. That means almost everybody 
> is going to need *much* more filter C than Rich's friend did.

My current project is using 40uF at 9900WV, but that's a FWD.  The 
original "plywood box" amplifier used only 15uF at 9000WV, and no one 
noticed any ripple.  This time I decided to use more C because 
Matsushita's new line of 105ºC-rated electrolytics are so compact that 
I had more room to fill up   --  i. e., it would have been just too 
embarrassing to have empty space.

>
> Yer typical 8877 is operating at around 3kV and 0.8A peak, so RL is 
> only 3750 ohms.

A typical 8877 draws c. 350mA average from the PS on voice-SSB - - even 
though the current peaks are c. 1000mA.  In other words, PS RL for NØN 
(was AØ)  is not the RL for SSB voice service

> ... ... ...
cheers, Ian
>
Richard L. Measures, AG6K, 805.386.3734.  www.somis.org