Output Filter Issue?

[Zazzle]

Hey,

Jep, it's about the Morse Beacon project. I like to discuss this issue in a different Topic an hope that's okay. If no, please merge the thread with the Beacon topic.

I wonder whether I might face some problems with the Output Filter. It's a 5-Pole PI-Filter. A Low-Pas (cut-off is 11,5Mhz) that also manages the impedance matching between PA Transistor an Output (which is 50R).

I'd build the Filter by using some basics formula and a lookup-table (for the impedance transformation). The Filter didn't work (almost no output signal) well and I started modifying values. The PA stage had been calculated for approx 1,5W. After some modification of the component values I was soon able to obtain 1,8W. I tested other values and eventually made it to about 4W with the PA becomming quite hot. Still, the signal looked fine on the Oscilloscope.

But I became sceptical about the 4W and looped a Antenna Tuner in between Transmitter and Dummy. The signal became heavily distorted after introducing just the slightest amount of reactive load (C and L). I undid the latest modifcations and the filter was now able to withstand reactive loads to a certain degree before the signal became distorted again.

I ran a test with my HAM Transceiver. At a certain point it also started to show a distorted signal when too much reactive load was introduced.

My basic question is: how must a output filter respond to mismatching? I attached some screenshots from the Beacon Transmitter output in different situations.

I wonder whether the Filtert causes problems and doesn't feed the Antenna properly... which would be odd, because the SWR Meter says: 1,5W forward, 0,3W reflected.

Thanks and Greetings,
~Zazzle

[Zazzle]

Oh, forgot to attach a picture of the new Filter. Don't ask. Soldered too many times on the board.

[ff]

Hi Zazzle - good to see that you're still fighting the good fight!  Admittedly my understanding of filters is limited, but to my knowledge, if yours is configured for 50 ohm input - 50 ohm output, any large variation from those ideal values will distort the filter's passband - the filter won't be blocking what you want it to block.  I don't recall the specifics of your project but if your PA is developing about 4 watts and the stage is powered from a 12 volt supply, its output impedance is approximately 18 ohms - a long ways from 50.  An easy match at those power levels can be made with an impedance transformer.   If those stated figures are correct, a 6 turn primary (PA side) and a 10 turn secondary (filter side) on an FT50-43 toroid should give a decent match. 

You probably won't want to do this - BUT - if I were doing it, I would remove the existing filter and build a completely new one, AFTER getting a decent match.  If your twaddling is anything like mine, by now the existing filter is completely at sea, and rebuilding it will be the easiest course of action.

Get your match right, which will get the amp running cooler.  Then re-do your filter and tweak for a good-looking waveform.  This is an online LPF calculator I have used many times to good effect:

http://www.calculatoredge.com/electronics/ch%20pi%20low%20pass.htm

Best of luck with your project and 73!

[redhat]

Keep in mind too that the online calculators for filter design rely on ideal component values.  With toroids, different winding techniques, ect will affect their value, and thus the performance of the filter.  Their proximity to other objects or components in the circuit will also affect their performance.  The ones I've built over the years vary quite a bit from the ideal designs, for lack of the right capacitor values, or an inductance value that falls between one turn on the toroids, ect.  Fortunately, a spectrum analyzer and a tracking generator can help verify and tune such networks.  Just because the waveform looks good on a scope, don't trust it.  I seen many transmitters that have fine shaped waveforms, and don't meet specifications.

Good luck!

+-RH

[Zazzle]

Hi ff!

Admittedly my understanding of filters is limited, but to my knowledge, if yours is configured for 50 ohm input - 50 ohm output, any large variation from those ideal values will distort the filter's passband - the filter won't be blocking what you want it to block.  I don't recall the specifics of your project but if your PA is developing about 4 watts and the stage is powered from a 12 volt supply, its output impedance is approximately 18 ohms - a long ways from 50.  An easy match at those power levels can be made with an impedance transformer.   If those stated figures are correct, a 6 turn primary (PA side) and a 10 turn secondary (filter side) on an FT50-43 toroid should give a decent match. 

Nono, like written, the lowpass ALSO takes care about the impedance matching. It's possible to combine both. There isn't much to match anyway. I'm aiming for 1,5W, which means an PA impedance of approx 52R.

(VCC+Uce)² / 2xPo

(13+0,5)² / 2x1,5 = 52,083R

The 4 Watts was what I was able to obtain into a perfect load with a lot of tweeking (just for testing. I went back to ~1,5W)

Maybe I should loop in a 1:1 transformer before the filter. Just to see whether anything changes

You probably won't want to do this - BUT - if I were doing it, I would remove the existing filter and build a completely new one, AFTER getting a decent match.  If your twaddling is anything like mine, by now the existing filter is completely at sea, and rebuilding it will be the easiest course of action.

Well, I've build flters before, primary for the 2m Band. That worked well all teh time and that's why I don't get the problem with this thing. I don't know whether the filter is the problem at all. Maybe it's something else. Or a mix. Maybe I should dump this whole concept and start angew from scratch with a completely new circuit whan I'm back from my vacation.

Thanks
~Zazzle

[Zazzle]

Hey redhat,

Keep in mind too that the online calculators for filter design rely on ideal component values.

Jep. The C's I've caculated (the old school way with formulas from books and with a cualtaor) ended up with C-values way off from the standard values. But well, paralleling and adding a variable C helped here. I verified my values by using online calcs. Teh values are about the same. +/+ some percent. So I think I got that right.

Sadly, I don't call such equipment my own.

Greetings,
~Zazzle

[ff]

Nono, like written, the lowpass ALSO takes care about the impedance matching. It's possible to combine both. There isn't much to match anyway. I'm aiming for 1,5W, which means an PA impedance of approx 52R.

We all have our own ways of slaying the RF dragons, Zazzle!     Over the years I have found that my highest chances of success are had when I'm not getting too ambitious.  Absolutely you can roll an LPF that will do the impedance match but I have found it much less frustrating to use a separate transformer.  As Redhat said, different ways of winding coils and capacitance variations within their tolerances will skew a whole lot of good math (is there any such thing as GOOD math?)  Yes, if you're shooting for 1.5 watts then you should need no matching - IF - the feedpoint of your "fuchskreis" is approximately 50 ohms.  I know nothing about that antenna design so I leave that to you.  Other things I would check:

RF choke - do you have shorted windings?  Also bad solder joints here will make the choke look like a resistor and will cause heating of the PA.  A compromised RF choke will also mess up the stage loading.

Do you have a series coupling cap in the PA output circuit?  A lot of low power designs have one.  If you have one that is starting to go bad it can cause similar problems as yours.

Since your operating frequency is around 10 MHz and your LPF is calculated well above that frequency, my guess is that is not really your problem.  I have had troubles when I try to calculate the filter cutoff too close to the operating frequency (see Redhat's admonishment above).

One more thing - I don't trust SWR meters to give me more than a RELATIVE reading of the situation unless the meter is connected right to the antenna feedpoint.  But that's just my opinion of course.  Okay Zazzle, good hunting!