Ferrite Power Transformers

Welcome and feel free to ask questions about class E, mosfet's, and AM.
You must be registered to post.

Ferrite Power Transformers

Postby WA2WVL » Thu Jan 19, 2006 3:15 pm

When I worked for a living I designed High Power Ferrite Transformers
(2-30 Mhz ) up to 10 KW average. Some of the things I learned may be of interest to Forum members.

1. RF power transformers DO NOT store energy ( except for leakage inductance ) they transfer it.
2. Core loss does increase with RF current thru the transformer but normally current is not used as a design criteria (applied voltage is).
3. Doubling core crossectional area does not double inductance. This is because the flux density drops as one moves away from the center of the core. This means that larger diameter cores is not the answer in reducing core loss. Use more cores. Loss is proportional to the area inside the BH curve so less V/core means a shorter and narrower BH curve.
4. IR scanning of a ferrite power transformer showed that the inner surface of the core heats first and expands (maximum flux density). This
is what cracks cores. Even imersion in oil had little effect on the power handling ability of the transformer.
5. At 160 meters I would recommend using the Amidon FB-43-1020 core run at about 10V/core (carrier). Peak modulation voltage is not much of a problem due to duty cycle and thermal time constant of the ferrite. For full CW power more cores should be used. This same transformer at 80 meters would be coasting (half flux density due to frequency).

I hope my comments help some of the non-engineers out there better understand the in's and out's of ferrite power transformers.
Last edited by WA2WVL on Sun Jan 22, 2006 7:56 pm, edited 5 times in total.
WA2WVL
Class E Enthusiast
 
Posts: 167
Joined: Thu Dec 22, 2005 1:40 pm
Location: Homosassa, Fl.

Postby frank carcia » Thu Jan 19, 2006 9:20 pm

Floyd,
Thank you. I have always wondered about current but never thought I got it right. I went as far as considering it as bad as DC bias so I didn't get myself into trouble. I'm using Chromerics A637s 8 per phase at about 130 volts at modulation peak, about 6 volts per core at carrier. I guess that is why they don't warm up much. A material is about Ui=750

The 160 meter final has 5 per stack at 85 volts peak. Again about 6v per core at carrier. They barely arm up after a long transmission.

Good to have someone around who has lived this stuff to set us hacks right.
frank carcia
Senior Contributer
 
Posts: 765
Joined: Fri Jan 31, 2003 10:48 am
Location: Enfield Ct. / Niantic Ct.

Postby NE4AM » Fri Jan 20, 2006 9:20 am

Floyd, I'm going to try to clarify your statement about loss being related to voltage, at least for my own understanding. Loss is related to voltage - but not directly. The H in the BH curve is in units of Oersteads (ampere turns), not volts. The amperes here are the magnetizing current - the current resulting from the applied RF voltage across the reactance of the open-circuit inductance of the transformer, at the carrier frequency. Further, if I recall my ferrite data curves, the loss in these transformer cores is roughly proportional to frequency and to B (gauss) squared.

73 - Dave
73 - Dave NE4AM
NE4AM
Member
 
Posts: 24
Joined: Fri Mar 11, 2005 1:24 pm
Location: Chatham, Illinois

Postby frank carcia » Fri Jan 20, 2006 9:36 am

Here is a question that came to mind in my crazy dreams. What is the effect of drain current through the core. I have seen many BB amplifier designs with DC current shunt fed to the devices. I found this method does not work well unless the pair of devices are matched so I prefer running DC power through the transformer to force a balance.
This is why I considered the core as one with DC offset to CYA. My 75 meter final will hit almost 50 amps peak current with modulation. I don't have a cross reference to Floyds part number from Amidon but have a ferrite book. Steve and I picked a 1 inch OD core with a 1/2 inch hole to handle the large conductors one night at his place. Everyone has had luck with these cores. I have a pile of the A637 cores so I still use them. Steve found the ran warm so we went to type 43. I just use more cores. I found the core was down to about 20% perm at high DC currents.
Am I missing something guys or was I just lucky. The guys running type 43 use 6 core stacks and I use 8 of the lower perm material. fc
frank carcia
Senior Contributer
 
Posts: 765
Joined: Fri Jan 31, 2003 10:48 am
Location: Enfield Ct. / Niantic Ct.

Postby NE4AM » Fri Jan 20, 2006 10:10 am

DC current through a toroid is OK provided you don't saturate the core. The ferrite material has a specified saturation flux, and the equation for flux is:

B = L * I * 10^8 / N / A

B = Gauss
L = inductance, henries
I = current, amps
N = number of turns
A cross sectional area of core, square cm

You want to keep the peak I (DC + AC magnetization) below the saturation flux of the ferrite.
73 - Dave NE4AM
NE4AM
Member
 
Posts: 24
Joined: Fri Mar 11, 2005 1:24 pm
Location: Chatham, Illinois

Postby frank carcia » Fri Jan 20, 2006 12:50 pm

Dave,
I see now the DC plus AC. In my case the dc was a lot higher. I did the same thing without noticing what I was doing. I looked at dc on the BH curve to see if I could support AC but in reality all you have to do is add them and stay below saturation. A Chromerics guy told me to stay below 150 Gauss when I was checking A material at 4 MHz. I think I was around 125 leaving plenty of room for DC.
So now back to broad band amplifiers. They shunt feed them to get away with smaller transformers. I found this a real problem if the devices are not perfectly matched. You can get some weird waveforms that way.
Good to relearn things. gfz
frank carcia
Senior Contributer
 
Posts: 765
Joined: Fri Jan 31, 2003 10:48 am
Location: Enfield Ct. / Niantic Ct.

Postby WA2WVL » Fri Jan 20, 2006 8:45 pm

Looks like I stired up a lively discussion. Where shall I start.

DAVE: I am not a college professor but as I understand it the RF transformer current consists of two components. First the magnetizing current which we can design to be less than 5% of the total current by designing for an impedance 20 times the load impedance. For a ONE turn winding this has determined how many cores should be used. Since we know the applied voltage that we want to use we can calculate the average flux density in the core. It will generally come out <200G.

We also know what load impedance is expected at the transformer terminals to develop the design power E (sq)/R. Once the R is set the second component of the current (Load current) is known E/R.
Now since it is difficult to measure the RF current but easy to measure the RF voltage (with a scope) we talk Voltage/core as the simplest way to describe the design. Ampere-turns may be driveing the BH but we have defined the design.

My experience is that 200G designs work while 400G is a sure thermal runaway design. This seems to be true for ferrites with perms of 125 to 2500. The higher perm materials have slightly higher loss but reduce magnetizing current.



TESTING at high power is the only sure way to verify that the design is sound.

One very importain point not covered in the textbooks is flux distribution across the core crossection. For an average flux density of 200G the density near the conductor may be 600G while the flux near the outside of the core aproaches zero. In my experience if you double the crosssectional area the inductance may increase by 20%, not double.
This is not true in 60Hz transformers since the material perm is VERY VERY high and all of the flux is confined to the core material.


FRANK: I checked the Collins handbook and they recommend direct DC feed at 100 watts and shunt choke feed at higher powers If you think you may have a problem with direct feed try feeding thru two external RF chokes and see if you can measure any difference.

The Amidon core is 1 in dia, .5 in hole, 1.1 in long in 43 material.
I would rather use shunt chokes than match ON time and V sat of devices. (and critical drive set)

I hope I hit all of the points raised and the average ham should look at what other hams have done and start there.

Floyd WA2WVL
WA2WVL
Class E Enthusiast
 
Posts: 167
Joined: Thu Dec 22, 2005 1:40 pm
Location: Homosassa, Fl.

Postby frank carcia » Fri Jan 20, 2006 9:16 pm

Floyd,
That is exactly the core Steve and I picked one night based on the A637 size. We just use the Fairite part number. It has a large hole to handle large size wire. Yes, the flux distribution is a new one on me and it makes a lot of sense. I was playing with bb amps a while ago and changed to a larger diameter sleeve and nothing changed. I was getting core heating on 160 meters. Then I stacked 2 cores per phase and they cooled down.
2 cores per phase got a bit warm on 160 at 50 volts.
Look at all the motorola high power applications. Above 300 watts the dc is going through the core again. Also they switch to 61 material. I guess the losses are lower and you get similar perm at high flux levels. I found the waveform was pretty odd looking when I did shunt feed with unmatched transistors MRF429s. I switched to dc feed and everything looked even and balanced with the same 2 transistors. I bought a bunch of those square cores Harris uses in their finals and combiners at Hostraders. I also have a nice 4 port with balance resistors. Someday I hope to build a nice big ss linear with all this stuff. I bought a bunch of ENI boards a years ago with a pile of MRF429s and MRF150s.
I have been building bb transformers since the early 70s mostly in receiver circuits. I worked for a German guy who taught me how to build them right. He was a master of the LC delay line. He would adjust the twists per inch and turns could to get the best bb performance. fc
I found the .7 inch sleeves work pretty well also but have a smaller hole.
frank carcia
Senior Contributer
 
Posts: 765
Joined: Fri Jan 31, 2003 10:48 am
Location: Enfield Ct. / Niantic Ct.

Postby WA2WVL » Fri Jan 20, 2006 11:49 pm

Frank:
I sent you a post about 11pm and i think Monkeypuppet crashed about then. No record of my post.
I suggested that you wind your feed chokes on 1/2 inch ferrite rod with magnetic coupling. Then it will act as a "reversing transformer" as is used in some Class AB amplifiers. THe large air gap of a rod will prevent saturation by the DC. What do you think?
Floyd

PS The ct bypass need not be as massive as the current one since it doesn't have to handle the full RF current.
WA2WVL
Class E Enthusiast
 
Posts: 167
Joined: Thu Dec 22, 2005 1:40 pm
Location: Homosassa, Fl.

Postby frank carcia » Sat Jan 21, 2006 8:24 am

Floyd,
I have seen a twisted pair on a rod phased to swing flux both ways. It is in a Raytheon amp I have. I did not get the file. Try: carcia@sbcglobal.net.
When I first tested the 75 meter final I had no bypass. I do have a bit now but nut much compared to the 2 ohm Z. I want to keep RF out of the modulator feed back. fc
frank carcia
Senior Contributer
 
Posts: 765
Joined: Fri Jan 31, 2003 10:48 am
Location: Enfield Ct. / Niantic Ct.

Ferrite Power Transformers

Postby WA2WVL » Sat Jan 21, 2006 10:59 am

Here are some measurements of the inductive reactance of the Amidon FB-43-1020 core.

1.840 Mhz 33 ohms/core ( 264 ohms for 8 cores)
3.800 Mhz 54 ohms/core
7.200 Mhz 65 ohms/core
14.200 Mhz 66 ohms/core (what it measures)
21.300 Mhz 103 ohms/core



Floyd WA2WVL
WA2WVL
Class E Enthusiast
 
Posts: 167
Joined: Thu Dec 22, 2005 1:40 pm
Location: Homosassa, Fl.

Postby frank carcia » Sat Jan 21, 2006 7:15 pm

Floyd,
The A637 core only has Ui=750 and at maximum current the perm is down to about 20%. Sounds like I'm ok on 160 meters for either final.
The main thing for me was the simple fact just add DC to AC and stay under saturation. I did that but my method was more work.
I'm glad you started this post so all of can understand what is going on.
frank carcia
Senior Contributer
 
Posts: 765
Joined: Fri Jan 31, 2003 10:48 am
Location: Enfield Ct. / Niantic Ct.

Ferrite discussion

Postby John K5PRO » Sat Jan 21, 2006 9:27 pm

Very interesting discussion here, with some real meat. A couple of years back I spent a year at work designing an RF system which would sweep frequency in a ferrite-loaded resonator, for a new proton synchrotron. The range was 1-5 MHz, and ran ~200-300 Gauss in the cores. The cores were Ferroxcube 4M2 or 4B3 or 8C12, tried all three. These cores are about an inch tall, and around 19 inches in diameter with a large hole. They cost about $2200 per core, custom made in Eindoven. We would stack enough cores together over the beam pipe (the L) to get the flux density per core that we wanted, and try and develop about 20 kV of peak RF voltage at the end of the L where a parallel vacuum capacitor sets resonance. This has been done this way for years, in the big proton machines at Brookhaven National Lab, Fermilab and others. To sweep frequency, of course, the drive frequency is swept, simulataneously a bias winding around the ferrite moves L rapidly to track. Interesting nonlinear things happen at high power, with rapidly changing bias. One is called High Loss Effect and another is Dynamic Loss Effect. Both have the end result of knocking down the Q and causing loss to dramatically rise during the cycle.

I tested the cores in a resonator having only 2, with a figure 8 bias winding (to cancel the RF at the DC power leads), driven by a single winding off an amplifier at 5 MHz, using an 8877 triode at about a kW CW. My original tester used a Johnson Thunderbolt linear, but it was too hard on that and it was arcing over quite a bit.

Anyhow, keep up the good work passing on the knowledge about ferrite cores, as it ain't so much magic afterall.
John K5PRO
Member
 
Posts: 19
Joined: Tue Oct 05, 2004 6:20 pm
Location: Santa Fe County, NM

Postby frank carcia » Sun Jan 22, 2006 9:48 am

I learned BB transformers from a German guy who was a kid in Germany during WW2, he had some interesting stories. He was a master an holding phase over a wide frequency. Everytime I look at a pulse on a scope and it isn't perfect my mind flashes back to ask myself " what would Gunther do about this pulse" He could look at a pulse on a passive network and know exactly what to do.
I guess we all can agree that it is a bit more than threading wire through the hole. Some only want a transformer and some wonder why it works.
great thread!
frank carcia
Senior Contributer
 
Posts: 765
Joined: Fri Jan 31, 2003 10:48 am
Location: Enfield Ct. / Niantic Ct.

Core Loss of Amidon FB-43-1020 Core

Postby WA2WVL » Tue Jan 24, 2006 12:37 am

Frank C.

Frank, tonight I belive that I measured the core loss of the 1 Inch cores
at 1.8 Mhz.
I measured a 1 turn, 1 core to be 9+j33 so I designed an "L" network to match it to 50 ohms. Then I read the bird wattmeter (50 watt slug) at various voltages across the 1 turn. The results were:

E/core(rms)----P(loss)

5v---------------.5 watts
10v--------------2 watts
15v--------------5 watts
20v--------------20 watts

This is the loss/core. 10v/core still looks like a good number on 160 meters. Core was cool during the measurement but loss may increase
as it heats.

Floyd WA2WVL

PS: The "L" network was 3300pf across xfmr and 5.4 uh to 50 ohm connector.
WA2WVL
Class E Enthusiast
 
Posts: 167
Joined: Thu Dec 22, 2005 1:40 pm
Location: Homosassa, Fl.

Next

Return to Class E Discussion

Who is online

Users browsing this forum: No registered users and 1 guest

cron