EMRFD Message Archive 5347

Message Date From Subject
5347 2010-10-05 07:31:49 Fernando Krouwel A question to Dan Tayloe about audio filters
Hi Mr. Tayloe, good morning:

Please, I ask you a help for an experiment I am testing:

Recently I decide to transform an SDR receiver board (a now classic "Tayloe FST3253 based converter", resulted from a kitting effort/materials group buy of two brazilian hams a few years ago, who probably are reading this - PY2WM and PU2JTE/N1VTN...) to be used as a computer divorced receiver (HDR?!). This is because I have difficult about having computer and radios in the same room, due to space reasons, and is not convenient to have coaxials "travelling along the QTH"...

A very exciting adventure (a mix of some existing circuits joined with others of my own design) with good results until now!

So, I chose to add some circuits to the SDR and transform it into some kind of "analog SDR2GO" (tks to Kees...).

Then, I picked up some phasing network circuit ideas from the NORCAL 2030 transceiver.
After the first test, I added a third opamp stage of phasing network to more easilly cover the entire 2.4 KHz voice channel (the 2030 was intended for CW only), which I built exactly as two original ones, but with smaller caps and it worked OK (I am astonished about the excellent resulting single signal capabilities over all voice channel, only one chinese 40m broadcasting radio over 7200KHz which is really strong can leak a very weak unwanted carrier pitch on the other side of the zero beat!).

As in the "2030", these "I" and "Q" signals are combined at the output of the phasing channels by means of two 2.2Kohm joined to make the phasing math and thus, provide audio output with the desired sideband.

After it, signals go to the audio filter (and after it, my difficult).

The audio filter was designed with the help of the Texas Instruments Utility for active filters, as I saw in your presentation slides and, according to your teachings, I was carefull to choose low Q and low delay, together with the best specs I could achieve using just four Sallen_Key Butterworth stages (2.4KHz passband freq, 5.3KHz stopband, -54dB stopband atten. and -1dB passband ripple (better would require more opamp stages, which I didn´t want to add)). In order to save an audio preamp stage, I tried some filter gain options, but due to the high output of the SDR board being used, the unity gain option worked best.


The filter´s first stage, which is a butterworth config has as its first input component a series resistor (6.8Kohm), and the phasing network has two 2.2Kohm output resistors, each one leaving the "I" and "Q" opamp outputs and connected to it.

"I" opamp output----2.2Kohm-----+
+----6.8Kohm input filter resistor (part of the filter)-----to next filter components - other Rs and Cs etc-----
"Q" opamp output----2.2Kohm-----+

Are these opamp output resistors coming from the phasing network that connect to the 6.8Kohm input filter resistor disturbing anything in the filter response?
If Yes, how much should I deduct from the 6K8ohm input filter resistor value to restore its original specs (I observed that in the "2030" you do not use this resistor, but filter value components are usually lower in your case)?
Could another option be do not to use the input filter resistor (zero ohm instead of 6.8Kohm) and increase the 2.2Kohm phasing opamp output resistors? In this case, what would be the new values?

Thank you very much
Fernando - PY2ETT


No virus found in this outgoing message.
Checked by AVG - www.avg.com
Version: 8.5.445 / Virus Database: 271.1.1/3174 - Release Date: 10/03/10 06:34:00

[Non-text portions of this message have been removed]
5349 2010-10-05 09:16:47 Tayloe Dan-P26412 Re: A question to Dan Tayloe about audio filters
I cheated a bit going into the first stage of the low pass filter strip. This filter stage needed 1K of input resistance. I broke this 1 K filter input resistor into a pair of 2K resistors, one to the I phased output, the other to the Q phased output. On the undesired sideband, these two resistors acted like summing resistors. Thus on the undesired sideband the signal coming out of the two phasing section will be 180 degrees out of phase and thus cancel as they are summed using the input resistor of the filter. That allowed me to leave out a buffer stage that would normally be used to do this summing, I just needed two 2x filter input resistors. For the desired signal, the output of the I and Q are the same signal at the same phase, so my reasoning was that since the two output signals looked identical, both outputs could be viewed as the same output, so the two 2K resistors would "appear" to be in parallel, producing the desired 1K of filter input resistance.

When I was designing this filter using the TI software, I was trying to use the Chebychev configuration with a very, very small ripple (0.05 dB or less as I recall). The latest TI filter software is great, but I find the older TI software a bit more flexible in allowing me to pick specific capacitor values and then having the software compute the updated resistor values. The old software used to allow me to tweak the individual capacitor values until I exceeded a specified C1-C2 limit. It seems like the latest software forces an apparent fixed ratio of C1 and C2 in a given stage, which makes it more difficult to optimize for designs that try to stress minimal unique component values. I have often spent a lot of time tweaking filter cut off frequency, ripple, and capacitor values to be able to maximally reuse existing R and C values elsewhere in the circuit.

- Dan, N7VE


5350 2010-10-05 09:27:00 Tayloe Dan-P26412 Re: A question to Dan Tayloe about audio filters
To answer your specific question, the two 2.2K summing resistors look like 1.1K of resistance for the desired sideband. Thus if you want to keep the 2.2K resistors, you should subtract 1.1K from your current 6.8K filter input resistance. Using a 5.6K resistor will likely be close enough. Alternatively, you could use two summing resistors close to 13.6K (a non standard value) and eliminate the 6.8K resistor altogether.

I use the TI filter output as a starting point. Depending on the Q of the stage, you can change values to something else close (like using two 15K resistors in the example above) and it will not impact the circuit much. That is where I simulate the circuit using one of the freeware simulators (like TI TINA or Linear Technology LTspice) to check the frequency response of the proposed component changes to see if they look ok or not.

- Dan, N7VE