simple absorptive lowpass filter

Jeroen Belleman has designed some spiffy absorptive lowpass filters that look resistive at one port.

formatting link

Mini-circuits has some new filters that are resistive on both ports:

formatting link

I have some samples coming to test. What are the input and output shunt resistors for? What is their value? We'll see.

So I thought I'd take a whack at it myself. I wanted something simpler than Jeroen's and cheaper than Minicircuits, and something I could make from parts in stock. The idea to to remove ringing and junk from a very fast but pretty ugly pulse generator output, giving up a little speed for beauty.

Version 4 SHEET 1 960 680 WIRE 32 128 -16 128 WIRE 96 128 32 128 WIRE 224 128 176 128 WIRE 272 128 224 128 WIRE 432 128 352 128 WIRE 480 128 432 128 WIRE 592 128 480 128 WIRE 816 128 688 128 WIRE 864 128 816 128 WIRE 224 160 224 128 WIRE 592 160 560 160 WIRE 720 160 688 160 WIRE -16 176 -16 128 WIRE 432 192 432 128 WIRE 864 192 864 128 WIRE 560 208 560 160 WIRE 720 208 720 160 WIRE -16 304 -16 256 WIRE 224 304 224 224 WIRE 432 320 432 272 WIRE 864 336 864 272 WIRE 432 416 432 384 FLAG -16 304 0 FLAG 224 304 0 FLAG 432 416 0 FLAG 560 208 0 FLAG 720 208 0 FLAG 864 336 0 FLAG 32 128 GEN FLAG 480 128 OUT FLAG 816 128 EOL SYMBOL res 80 144 R270 WINDOW 0 79 60 VTop 2 WINDOW 3 69 59 VBottom 2 SYMATTR InstName R1 SYMATTR Value 50 SYMBOL res 416 176 R0 WINDOW 0 -56 31 Left 2 WINDOW 3 -56 71 Left 2 SYMATTR InstName R2 SYMATTR Value 50 SYMBOL cap 208 160 R0 WINDOW 0 -52 44 Left 2 WINDOW 3 -58 79 Left 2 SYMATTR InstName C1 SYMATTR Value 6.8p SYMBOL cap 416 320 R0 WINDOW 0 -61 17 Left 2 WINDOW 3 -68 50 Left 2 SYMATTR InstName C2 SYMATTR Value 6.8p SYMBOL ind 256 144 R270 WINDOW 0 80 56 VTop 2 WINDOW 3 71 56 VBottom 2 SYMATTR InstName L1 SYMATTR Value 33n SYMBOL voltage -16 160 R0 WINDOW 0 19 117 Left 2 WINDOW 3 -50 195 Left 2 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V1 SYMATTR Value PULSE(0 1 1n 800p 800p 8n) SYMBOL tline 640 144 R0 WINDOW 0 -2 -72 Bottom 2 WINDOW 3 -3 -62 Top 2 SYMATTR InstName T1 SYMATTR Value Td=10n Z0=50 SYMBOL res 848 176 R0 WINDOW 0 57 40 Left 2 WINDOW 3 56 75 Left 2 SYMATTR InstName R3 SYMATTR Value 1K TEXT 592 384 Left 2 !.tran 50n TEXT 464 296 Left 2 ;Simple Absorptive Lowpass Filter TEXT 552 336 Left 2 ;JL May 24, 2016

The test case is to set the terminator R3 high, above 50 ohms, to let the txline reflect a pulse back into the filter, and see how much it bounces around. This is pretty good, just a few percent echo.

This filter was designed in about a half hour purely by instinct and fiddling, with parts values selected from what we have in stock; no explicit math involved. That works for simple stuff like this but blows up for not-much-more complex networks where the fiddling tends to diverge. That becomes the "lost in space" syndrome.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin
Loading thread data ...

IOW a diplexer with one arm terminated.

Way back in the archives, Rick Karlquist recommended the following paper on this:

Wenzel, R. J., "Wideband High-Selectivity Diplexers Utilizing Digital Elliptic Filters" , IEEE Trans. MTT, Dec 1967. I think that's probably "digital" meaning "interdigitated", but I haven't read it.

formatting link

For the normal LC case, there's a pretty readable thesis on non-reflecting multiplexers at

formatting link
.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

IIRC, it's possible to design the perfect diplexer for passive Butterworth ladder filters. So for O(2), which is what you're apparently after, the whole thing would be two caps, two coils and one resistor. It's been a while though, and I'm a little busy...

Jeroen

Reply to
Jeroen Belleman

One advantage that I have is that my filter only has to absorb the reflection of the pulse that I allowed through in the first place. That's a lot easier than trying to make the port absorb over an infinite bandwidth.

Single-port absorption makes sense here; there's no point in wasting pulse generator energy on the left side. We just want to kill reflections in the case that a customer drives a high-Z load at the end of a cable. Which brings up the interesting case of the customer driving a capacitive load, which I just tried. Looks pretty good.

The design methodology is to conjecture a topology and then plug in permutations of the relatively small choice of the parts we have in stock, and hope to get lucky.

There are some filter design programs (NuHertz, Filtercad) that let you specify E12/E24/E96 value constraints, and they seem to work very well. I wonder how they work.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

The classic RCLC wideband termination can be flipped into an absorptive lowpass filter...

formatting link

My little thing just adds one more cap to snap things up a bit.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

Hmm OK. You sorta split the cap. L/C = R^2 which gives C = ~14 pF.

George H.

Reply to
George Herold

Huh ?>:-} C=L/R^2 gives 20pF ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
 Click to see the full signature
Reply to
Jim Thompson

Equal C values isn't optimum in the filter I first posted, but it works with the parts we have in stock. We're running out of room for bins, so we're reluctant to add parts to inventory. I suppose I could parallel caps to get more choices.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

I punched quickly... 33nH/2500 = 13.2 pF.

Reply to
George Herold

JL's drawing is 50nH and 50 Ohms => 20pF

This scheme dates back to (at least) GenRad bridges. I've used it to keep amplifiers happy driving inductive loads... as in my ancient muff modem of the '70's. ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
 Click to see the full signature
Reply to
Jim Thompson

Not a bad approach to start with--if you get lucky, you get lucky, and if you don't, it's no more work than starting out with the full method.

The basic filter calculations are so quick on a modern processor that they could just take the nearest 3 values for each part and see which combination came out best. I doubt that they'd be that crude, but if I were slapping one together for my own use, that's probably what I'd do.

As the estimable Forman Acton said in 1970: "We grudgingly concede that there are times when it is better to use the computer inefficiently than to saddle a professor with a laborious search for a better algorithm; nevertheless enough identifiable nonsense goes on in the computer room to justify a brief abut hopefully cautionary exhibition."

He goes on to tell the tale of one time in 1949 when he and his colleagues were asked to invert a 16x16 matrix of 10-digit floating point numbers with a 1-flops relay computer (CPC model 1, number 1), spending three days on it, shuffling card decks full of intermediate results and so on, and then finding that, to 8 digits, the the inverse was the transpose of the original matrix. That is, once they knew how the matrix was generated, it took them about 10 minutes to prove that it was orthogonal. Seems the customer hadn't done his homework. ;)

"Numerical methods that work"--still a really great read on the lore of numerical computing. Highly recommended.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

Like in many situations, I have no hard criterion for "right". Risetime, rise/fall symmetry, ringing, pulse slope, reflection shape and amplitude, are all beauty contest issues. And different people here have different preferred tradeoffs.

And when I build it, it's likely going to have to be fiddled, given caps around 7 pF and a 33nH inductor and however the real pulse driver works. Fiddling on a screen is bad enough but fiddling with a soldering iron is a lot worse.

--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

Jim Thompson is living (so far) proof that the universe is expanding. It is not large enough to contain his ego.

Reply to
John S

That's the usual rule for Zobel networks. Alas, these always have O(1) slopes. Of course, you could cascade a few, but the poles would still all be real, so you can't make the usual 'interesting' filter responses by cascading Zobel networks. Still, I use them a lot. They are great for muffling out-of-band nasties, or for polishing amplifier and coax cable frequency responses for example.

Jeroen Belleman

Reply to
Jeroen Belleman

Adding the extra cap lets it overshoot a little.

It seems to work; 6.5 volt pulse into 50 ohms:

formatting link

That little blip at the end isn't the filter's fault. I can probably fix that.

Smaller pulses look better:

formatting link

I have spent some fraction of my adult life pursuing the perfect pulse generator output stage. This is another increment.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.