Hello

I'm currently looking at what appears to be the canonical setup for home-testing of filters. This setup i outlined here in the "Measuring Double-Tuned Circuit Performance"-box:

http://www.robkalmeijer.nl/techniek/electronica/radiotechniek/hambladen/qst/1991/12/page29/index.html

and here in the "A Method to Measure Insertion Loss or Gain" section:

http://www.qrp.pops.net/RF-workbench-2.asp

and seems to be very common. Basically a 50 Ohms signal generator is connected to the filter which is then connected to a 50 Ohms terminated scope. Now the first link shows 6 dB attenuators both between the signal generator and the filter and between the filter and the scope. The second link only shows the first of the two and says:

I show an attenuator pad in this diagram to stress that the signal generator output port must have a return loss >= 20 dB.

Edit: The first link says that the attenuators present "clean" 50 Ohm impedances to the filter. I do not understand the difference between a "clean" and a "dirty" 50 Ohm termination.

I cannot really make sense out of this. I could imagine that an impedance mismatch could lead to reflections at the filter input that could mess up things. But I have no conception of what "messing up things" constitutes of.

With my limited knowledge, this is my process of thoughts which must be wrong: An attenuator would dissipate the reflection, but how can the presence of the attenuator influence the final result? In the end, there is stuff going into the filter and stuff coming out of the filter on the other end, and if you measure whatever comes out of the filter and compare it to what you put in, how is that not quantifying the loss of the filter properly? Also, when the filter is put to use, there is no such attenuators, so whatever is measured in the test setup is not how the filter eventually behaves when used...

On a related note, why is the second attenuator, between the filter and the scope, missing from the second web page?

Best regards and thank you!

 

dirty impedance has inductive or capscitive reflections. a perfect attenuator or load would be 50 ohms resistive with no reflections. the attenuators are there to attenuate any reflected rf back to the signal generator if you have a good buffered generator, they arent needed. remember if you are testing filters, the tuning might not be perfect, and poor tuning means impedance mismatch that might affect the signal source.

 

alfacliff, thank you for replying to my questions. This cleared up things a little, but to be honest I never felt like I fully grasped the concept of complex impedance. I tried the textbook approach which I can understand mathematically but I sort of lack the connection to the real-world scenarios (i.e. "how bad is it if I mismatch the impedance"), and I also tried some simpler documentation that was more connected to experiments, but never really hit the core of the concept. I have yet to find a nice introduction that fits the way I think about these things (or helps me to alter that)

All that said, I am now wondering about the following things: First of all, that double-tuned circuit in the first link has input and output transformers. I guess these are to have the filter in a higher impedance "environment" than 50 Ohms. I see how that affects the filter when I simulate it. The question is whether these impedance matching transformers contribute reactance. Naively, I would assume that this is the case (it's an inductor after all), but when I read about impedance matching transformers, little is said about correcting for that reactance introduced by the transformer.

Second, related question: If I think of the windings of the transformers as shunt inductors, I could resonate these with a parallel capacitor. In fact there is a parallel capacitor, but only on the secondary winding. Assuming that the above is correct in the first place, is this compensating for the transformer's inductance even though it's on the secondary side? And if not, why does the transformer not contribute reactance to the system?

Lastly, I always had the feeling that there must be some kind of connection between filters and impedance matching. I'm looking at a filter right now, but I am thinking about impedance matching and in fact, I change the impedance of the system if I change the filter response by changing the component values. Is there more to that thought than it just being some kind of large system of equations -- some satisfying the impedance requirements and some others satisfying the desired filter response -- that has to be solved when designing the filter? Your reply somehow suggests that by "tuning the filter", I automatically also set the desired impedance. If I understand your reply correctly, having reactance in the source and/or load and going through the tuning procedure would then result in a filter that is compensating the reactive aspects of the test rig, and when the filter is put to use in an environment that lacks these reactive aspects, the filter itself has complex impedance. Does that make sense in any way?

Thanks again for your reply and best regards!

 

The attenuators are to provide a proper termination to the filter, as filters are designed for a specific impedance. The pads also tend to swamp out inconsistend impedances in the test generator or other devices themselves. This becomes extremely crucial at microwave frequencies.

 

First -Another way to look at attenuators between test equipment and UUTs is that they isolate the effects of cable, connectors, sources, and loads. All of these components inject ripple in amplitude and phase. They are modeled as variations in impedance and length. Putting a 6 db pad at the input and on the output of the UUT attenuates these interactive variations by greater that 12 db.

Second - Filters are often specified with a "not to exceed" return loss at the input and output. An attenuator is the industry accepted way of achieving/guaranteeing it.

Mark