EMRFD Message Archive 11319

Message Date From Subject
11319 2015-07-16 13:32:21 jmlynesjr Reverse engineering a BPF?
Greetings:

This is my first post.

Question: Given the Ls and Cs of a bandpass filter, is there a method to back into an approximation of the original design goals?
          Also, same question as regards a Tapped-C Transformer.

Background: The July 2015 issue of Nuts and Volts magazine has an article "An Ultra Modern Shortwave Radio" that describes using a cheap DVB-T USB tuner with a 24 MHz upconverter as an HF SDR. While I'm waiting for delivery, I have been studying the design of the upconverter. The optional tuned input filters I understand. The 1.5-30 MHz input bandpass filter and the Tapped-C output transformer I'm having trouble with.

The upconverter uses an SA612. The datasheet shows the Rin and Rout of the SA612 to be 1.5K ohms. The BPF input impedance was not stated(assume 50 ohms?). So, the 1.5-30 MHz(stated operating BW, not specifically stated as the 3db design BW) input filter should match 50(?) to 1500 ohms. The Tapped-C transformer should match 1500 to 75 ohms(Rin of the DVB-T module) from 25.5 to 54 MHz. I can't duplicate the values given in the article:

        BPF: 150 pF
11320 2015-07-16 18:07:10 Tayloe, Dan (Noki... Re: Reverse engineering a BPF?
You can model this in the free Linear Technology Spice circuit simulation tool. You can easily model the 1500 ohm 602 load and sweep the input with a voltage source with 50 ohms in series.

Simulation allows you to tweak some of the values like the Capacitive divider or the coupling capacitor between tuned circuits if there is more than one.

Neat stuff to play with!

- Dan, N7VE

Sent from my Windows Phone

-----Original Message-----
11321 2015-07-16 18:09:46 Dana Myers Re: Reverse engineering a BPF?
11322 2015-07-17 03:23:33 kb1ckt Re: Reverse engineering a BPF?
The 602, I forget its input capacitance; want to say I read either a reflector post, or maybe it was in the documentation, for the SST20 about how the filtering for it was designed. Its slight capacitance had to be taken into account.

shawn kb1ckt

Sent from my iPad

11323 2015-07-17 15:42:07 jmlynesjr Re: Reverse engineering a BPF?
Thanks for the good input.

Is the free LT Spice package available for Linux? I run Ubuntu 14.04LTS.

I found a posting from UCSB that used a Tapped-C transformer to match a Rin of 50 ohms to the input of an SA602. They used 1500 ohms plus 3 pF as the SA602 input impedance. For easy math they used f0 as 159 MHz and a BW of 10 Mhz. the result was 41 nH
11324 2015-07-17 18:03:09 dwines Re: Reverse engineering a BPF?
James,

No, but I run the Windoz version under Wine in Ubuntu 14.04.2 and it runs just fine!

Don
K5DW
Sent via the Samsung GALAXY S® 5, an AT&T 4G LTE smartphone


-------- Original message --------
11325 2015-07-17 23:21:04 vasilyivanenko Re: Reverse engineering a BPF?

I might misunderstand -- but it sounds like you’re trying to analyze a 1.5-30 MHz band-pass filter that matches the resonators to their source/load with a tap on 1 end and a capacitor divider on the other .


The Z matching would really be a compromise affair across such a bandwidth (that’s 1 reason why multi band receivers switch in the ‘correct’ BP filter for each band) – even a 1 MHz frequency change would  require a significantly different ratio of capacitor tapping if you seek strong  impedance matching  into the input Z of that Gilbert cell mixer. 


While sweeping 200-300 KHz 3 dB wide BP filters into a NE612 to look at impedance matching ---  for best results I had to make one of the tapped caps variable [or just run a single series tuned cap out of the resonator] and tweak it to get the best input match at the center frequency.  Mismatch occurred presumably due to stray C, parts tolerances  - and also it’s not a resistance, but a complex impedance. The skirt pass-band and insertion loss were affected by impedance mismatch.


Another problem with a huge bandwidth FL will be over-coupling at some frequencies – the dreaded double-humped response.


I’ve made HF input filters that cover 0.5-1 MHz by analysis and tweaking of the W7ZOI high-pass/peaked low-pass filter found in the Progressive Receiver.

[Reference: QST for November 1981, the Progressive Communications Receiver (PR) by Wes, W7ZOI and John, K5IRK]



I’ve seen builders combine appropriate 50 ohm high-pass and low-pass filters in series to filter the input on dongle SDR receiver converters. 


Since there is no image frequency – I think that might be a good approach to minimize losses and then only a single 5.5 :1 turns ratio broadband ferrite transformer to match the NE612 input is needed

 

Best to you

Todd/Vasily

11326 2015-07-17 23:34:34 vasilyivanenko Re: Reverse engineering a BPF?
...sorry didn't mean to say there is no image frequency for converter mixer...
11327 2015-07-18 10:45:11 mvs_sarma@ymail.c... Re: Reverse engineering a BPF?
James,
 of late there is a software patch for SDRsharp to make RTL SDR work from almost 4MHz to 30MHz
 
 Originally these SDRs used to start responding from 24MHz upwards.
  Once you setytle to use sdrsharp software, you can downloafd the sdrsharp patch files.
 here is the link.  Direct Sampling Mode with No Hardware Modifcations - rtl-sdr.com

 

11342 2015-07-19 13:56:16 jmlynesjr Re: Reverse engineering a BPF?
Dan & Don:

Update: I have relatively painlessly installed wine and LTspiceIV on Ubuntu 14.04LTS. Seems to run fine. Opened and ran several of the supplied demo circuits. Now I will try to enter the circuits in my original question and see how it goes.

Thanks for the suggestions. A new tool to play with.

James
11343 2015-07-19 13:57:05 jmlynesjr Re: Reverse engineering a BPF?
Vasilyi:

Yes, i saw a recommendation in the ARRL Handbook that such a wide filter might be better designed as coupled LP and HP filters. Unfortunately it wasn't one of their worked out examples! and I can't find the referenced page number at the moment.

James
11344 2015-07-19 13:57:24 jmlynesjr Re: Reverse engineering a BPF?
Thanks for the reference.

I'll take a look even though I think I still want to play with the upconverter.

James
11346 2015-07-19 17:10:36 vasilyivanenko Re: Reverse engineering a BPF?
Hi James:

If you go the converter route: 

You may wish to just view some designed Chebyshev low-pass and high-pass filters from the tables in older
Handbooks (not sure about the new ARRL Handbooks).

Pick a high-pass cutoff, a low-pass cutoff, the number of elements and you're away. I tend to use 5-7 elements, but I live in "the city" with ++  BCI and such.

You then may 'build' your combined LP - HP filter in Wes' EMRFD software: ladbuild08 and then simulate it in GPLA. In GPLA, I'll swap nearest standard cap values in while watching the filter skirt and return loss.
Else software to make these filters may be found online. 1 example = Elsie from the  Tonne Software Homepage

Best to you!
T/V
11347 2015-07-20 21:53:17 John Maxwell Re: Reverse engineering a BPF?
11394 2015-08-04 23:40:14 jmlynesjr Re: Reverse engineering a BPF?
Another Update:

I found a short LTspice tutorial on the U of Minn site and went through it entering and simulating the given examples. I then entered the BPF circuit from the NV article. The response was pretty flat across the HF range with the pertinent data:

    -3db points    2.87 MHz    26.91 MHz    (data points picked off of the response graph)
   -32db points   1.23 MHz    62.22 MHz

    BW-3db        24.04 MHz
    BW-32db      60.99 MHz

    Bw/BWc = f/fc = BW-32db/BW-3db = 60.99/24.04 = 2.54

From fig 3-9(Bowick), Look up 2.54 @ -32db, 3+ or 4 elements required, the article used 3.

The geometric center frequency f0 = sqrt((fa)(fb)) = sqrt((2.87)(26.91)) = sqrt(77.23) = 8.78 MHz.

This is in the ball park of the resonance of the article component values.

The RS/RL = 50/1500 = .034. which is not within the table range in  Bowick.

Next is to try and back into the article values. Where was this tool in 1972? LOL.

James
11479 2015-08-10 08:58:59 jmlynesjr Re: Reverse engineering a BPF?
I seem to have come full circle...AKA I'm stuck!

Is there a formula that would let me calculate the 3 element Butterworth lowpass prototype values for a RS/RL of .034? i.e. 50 ohms to 1500 ohms transformation.

I proved to myself using the LTspice model that the BPF values given in the NV article do the job. Would still like to recreate the design calcs.

Thanks,
James
11481 2015-08-10 10:31:57 iq_rx Low-Pass Impedance Transformer
The following is not a Butterworth response, but since you have LTspice running it is easy to check what it does.  It is particularly useful when you want low loss at the design frequency, good attenuation above, and aren't too concerned with perfect flatness below.

This is an old, classic method that we use all the time for low-pass transformation between two impedances:

Step one: calculate the geometric mean of the two impedances, for example sqrt(50x1500) = 274

Step two: calculate a capacitor and an inductor with that impedance at your design frequency.  For example, at 9 MHz, a 64.5 pF capacitor has 274 ohm reactance, and a 4.85 uH inductor has 274 ohm reactance.

Step three: arrange two 64.5 pF shunt capacitors and the 4.85 uH inductor in a low-pass pi network.  If you connect 1500 ohms on one end, you will see 50 ohms on the other at 9 MHz.

That might be all, back in 1950 or so.  But now that you have LTspice, the next step is to do a simulation and see what happens when you use 68 pF off-the-shelf capacitors instead of those odd values, and whether the response is flat enough for your application.  This particular approach is fairly broad, which is why it is used so often.

Next: If it isn't flat enough below the design frequency, you can broadband it by doing the transformation in two steps.  For example, you could transform from 1500 to 300 and then 300 to 50.  At 9 MHz that would have these intermediate impedances and component values:

Z1 = 123 ohms  C = 144 pF  L = 2.18 uH
Z2 = 671 ohms  C = 26.4 pF L = 11.87 uH

The 50 ohm to 1500 ohm transformation network would then be:

shunt 144 pF series 2.18 uH shunt 144 pF -- shunt 26.4 pF series 11.7 uH shunt 26.4 pF

The parallel 144 pF and 26.4 pF capacitors in the middle of the network could be combined into a single 170 pF cap.

You'll want to check those values in LTspice and note the response.

Now that you have ideal 3 and 5 element networks in the simulator, it is easy to play with the values and observe the effects.

At my bench, I'd use fixed capacitors and wind toroids for the inductors.  That would allow me to adjust the inductors to any value I want.  In LTspice it's easy to see what happens when you use the nearest standard value caps and then tweak the inductors for an acceptable response.

For example, in the second example above with two cascaded pi networks, I would try:

150pF   L1 ~ 2.2 uH   180pF   L2 ~ 12 uH   27pF

in LTspice and see what I get.  I expect I'd have to tweak the inductor values down a bit.

Bear in mind that broadband 50.0 ohm and 1500.0 ohm terminations only exist in your imagination, in textbook examples, and in LTspice.  So what you want is not a perfect Butterworth response, but something that is fairly broadband and relatively insensitive to component variations.

If you enjoy doing network problems on the Smith Chart, that is an alternative (or supplement) to LTspice.  Folks who did the problem that way in 1950 could instantly see the effect of component or frequency variations, and laboratory Network Analyzers still offer that option.

If you'd like to learn how to use the Smith Chart, this is an ideal example problem, as a generation of recent RF students can attest.

Best Regards,

Rick


11482 2015-08-10 10:33:46 Thomas S. Knutsen Re: Reverse engineering a BPF?
Simulations only get you so far, and in most filter work, the simulation are the greater experiment!

What you want is a impedance transforming filter. This is usualy done either by calculating 2 filters, one for each impedance, with the same bandwith and center frequency, then the filters are combined. It sounds more advanced than it is. Just remember that since the terminatiopn impedance is different from 50 ohms, the adjustment of the filter may require some patience. 

The information is found in A. Zverev: " Handbook of filter synthesis". 

Almost any question regarding filters can be found in this book. Its from a age where carefull analyzis were important, and there are tables of filter coeffecients and prototype filters for almost any filter type you may want. 

73 de Thomas LA3PNA.


2015-08-07 20:23 GMT+02:00 jmlynesjr@gmail.com [emrfd] <emrfd@yahoogroups.com>:
 

I seem to have come full circle...AKA I'm stuck!

Is there a formula that would let me calculate the 3 element Butterworth lowpass prototype values for a RS/RL of .034? i.e. 50 ohms to 1500 ohms transformation.

I proved to myself using the LTspice model that the BPF values given in the NV article do the job. Would still like to recreate the design calcs.

Thanks,
James

11483 2015-08-10 10:47:33 Ashhar Farhan Re: Reverse engineering a BPF?
why don't you just use the TTC.EXE and the DTC.EXE that comes along with the EMRFD CD? (that's so many acronyms in a sentence!)

- f