EMRFD Message Archive 13851

Message Date From Subject
13851 2017-05-10 11:34:48 swift_glen low-loss extreme impedance matching

I see some recent measurements where a 50 ohm source is impedance-matched to an ambiguous load of an end-fed half-wave dipole. Ambiguous perhaps because the ground connection is part of the load that is somewhat ill-defined. With that caveat, an equivalent load of about 4Kohm is assumed. OK for an attempt at comparing methods of impedance matching. Gotta start somewhere....

A low-loss ferrite transformer is compared with the good 'ol L-match. Both seem low-loss. But I'm here to suggest other low-loss scheme, and I'm open to see any flaws:
A single-stage L-network has to have a fairly high loaded Q to take 50 ohm source up to 4K load. If you spread the transformation into 2-step (two inductors, 2 capacitors), each step-up section has lower Q than the single-step-up case. Is the addition of two extra (lossy) components worth the trouble? Certainly, the bandwidth over which a match is maintained is wider, but it'd be a pain to tune.

Perhaps one section could be fixed while the other section is a conventional L-tuner. Then there's the other option: choose a quarter-wave length of 450-ohm ladder line and do a distributed impedance match with that alone. But then do you need to drive with a single-ended to balanced "balun"? Perhaps not.
So you never get away without losing power behind in one-or-more components. It sure is worthwhile to see some carefully-done measurements to help choose your poison. Thanks, Wes.

13852 2017-05-10 20:24:20 w7zoi Re: low-loss extreme impedance matching
Hi Glen,

I think that's the right name.   Can't really tell with the info on the web and without a call, so we will just assume it's Glen.   If my fading memory serves me, you are a VE3, but I'm never sure these days, hi.

If some of the readers are wondering what we are talking about here, see my web site ( http://w7zoi.net/ ) in the technical notes section, the very bottom item.

Anyway, yea, it's all in the numbers.   We can use EZnec or similar program to estimate the impedance of the End Fed Half Wave antenna.   You really do have to be careful to model it carefully.   If you are going to hang a bunch of coax from the network and you have no common mode choke to decouple the feedline, you need to model it that way.  Roy (w7el) has covered this with his software manuals.

First, let me briefly comment that the thing that got me started with this subject was the rash of "information" on line where folks were building these conventional transformers with very high impedances, but were using -43 material cores.    Experience in an earlier lifetime with some wideband transformers to match crystal filters had shown high loss with some ferrite cores, but low loss with others.   I was really suspicious about the -43 material.   Measurements did indeed show that it was pretty lousy.    -61 cores were fine, but not so much so with the -43.   

The L-network was reasonable.   It had about the same performance as a -61 core transformer.  You ask about using a double L-network and what that might do to the insertion loss.  That turned out to be a really interesting calculation.

A single L-network was designed to go from 50 to 4000 Ohms.   I assumed an unloaded inductor Q of 250 and a capacitor Q of 800.  I designed it for 14 MHz.   The components for the network (low pass form) were 5.052 uH and 25.26 pF.    The simulated insertion loss was 0.20 dB.   This is in the exact area that I measured.

Now for the double L network:   I decided to design for an intermediate resistance that was the geometric mean of the two end resistance values of 50 and 4000.  This was 447 Ohms.   The first L-net from 50 to 447 was 1.6017 uH and 71.66 pF.   The second L-net from 447 to 4000 was 14.3266 uH and 8.01 pF.   I again used component Q values of 250 for inductors and 800 for the capacitors.    The simulated IL for this was 0.13 dB.  That's close to half the loss of the single L.    That second inductor, however, is not a very comfortable value for 14 MHz.     This is reminiscent of the pi-L networks that folks used to use with high power vacuum tube amplifiers (I'm told.) 

What might be a practical topology would be a transmatch (with adjustments) to get from 50 to 447 or so.   Then use a wideband transformer to go from the 447 to 4000.   The automatic transmatch in the appliances of the day (KX2 or KX3) will probably do the job in going from 50 to 447 or so.   I would guess that the ferrite transformer would have about the same loss as those that I measured, but the builder will certainly want to do his or her own measurements.

The difference between 0.2 dB and 0.13 dB is really a minor thing.   I measured similar low loss values around 0.2 dB for the transformers wound on -61 ferrite.   These loss values are all super low and should never be an issue.    On the other hand, the loss I measured with an otherwise identical transformer on a -43 core was 1.3 dB.   That's getting to be significant.    That much loss says that 26 % of the applied power will go up as heat.   

Anyway, thanks for the feedback.   

73, Wes