Thanks again Joe! Yep that would work!! :-) I came in at Zo in/out ~= 75 ohms for what looked best to me. Somewhat subjective are these real world adjustments. Dishal software was predicting 94.5 ohms. Matching network built as best I could for 75 ohms. Attached are the resulting VNA plots. Extremely happy with the VNA transmission loss plot. Less than 0.5dB ripple now. The return loss plot has evened itself up somewhat but it is still not real flash. Best I could manage today. I think it is probably usable. probably even pretty good for a homebrew filter. But I'm thinking I might try order in some quality crystals with a brand name you can trust and see what difference that makes. These crystals are just incredibly cheap eBay ones. I pretty sure the VNA software is telling me that the crystal Q is about 14K (14,428). Is this good or bad for a 20Mhz crystal? I read about crystal Q in the 100,000 range which is an order a magnitude greater than what I have here. Also another thank-you for pointing out the other day that if your crystals cases are grounded then this reduces your Cp value. Turned out to be critical. In this process I discovered that the Dishal software can be trusted to predict correct values for QER filters including the bandwidth outcome. (Which is good because so far as I know it is the *only* software with a built in QER calculator). But the crystal holder capacitance (Cp) is critical. I found I had to measure the capacitance value in pF of he crystal between the leads, record that value *AND* then short both the leads together and measure capacitance between the shorted leads and the case of the crystal. Deduct the second value from the first and that's what you plug into Dishal. Then my actual bandwidth was very, very close to that predicted. 73, Steve. > Steve > How much ripple is acceptable is a a judgement call, but we can generalize > that higher ripple in the passband will equate to higher distortion in the > recovered audio. This is not only because the different audio frequencies > across the passband have different attenuation, but also because these > ripples are caused by phase shifts that vary up and down across the > passband. In general, we want phase response across the passband to vary > linearly in one direction, not up and down sharply, because the latter > causes ringing in the impulse response. This is visible more clearly in > the return loss response than in the transmission response. In the QER > filter topology, all the caps are equal value, so the only remaining > variable to adjust for good return loss (match) is the termination > impedances. > The termination impedance of the QER xtal filter depends on center > frequency, filter bandwidth and motional parameters of the crystals. At > this frequency (20 mhz), it typically comes out around 40-70 ohms, but it > will be higher at lower frequencies. Since you have a spectrum analyzer > with tracking generator, you're in a good position to optimize this easily > by looking at the return loss while adjusting the source/load impedances, > which should be equal. Build a TEE resistor network that looks something > like 40 ohms series out of the generator, 10 ohms shunt to ground and 100 > ohm pot to the filter. Put a similar network from the filter to the > analyzer input, pot-side to the filter. Adjust both pots for best return > loss in the passband. This will equate to minimum ripple in the > transmission response. Once you get that adjusted, remove the analyzer, > terminate the network in 50 ohms and measure the resistance to ground on > the filter side. This is the optimum termination resistance. To transform > it to 50 ohms, you might need a transformer or an LC network, but I > suspect it will come out fairly close in this case. If you can't adjust to > at least -15db return loss across the passband, I'd look for differences > in the crystals. When all is said and done, you should be able to get well > under 1db ripple. > > Joe > -------- Original Message -------- > Subject: [minima] Xtal Filters > From: "Steve VK2SJA" > Date: Sat, October 11, 2014 7:16 am > To: minima@xxxxxxxxxxxxx > > Hi All, > > Tried asking this one on the emrfd Yahoo Group but nothing seemed to pop > out at the other end. So sorry if this becomes a cross forum post. > > Question: What is the typical passband ripple for a 2.7kHz SSB filter > both > commercial and home brew? Or to ask the question another way. What level > of passband ripple is Good, Passable and Bad? > > If I'm reading the VNA plots (see attached) correctly I seem to have > about > ~2dB to 3dB of passband ripple on my 2.7kHz filter. I just don't know > what > is acceptable here. The 1.9kHz filter only has about 0.5dB ripple (these > widths are the 3dB point). > > Next question. What is the correct procedure to experimentally confirm > the > filters input and output impedance. Two resistive pads on input and > output. Parallel 50 ohm resistor to ground at both input and output. > Series variable resistor between input and output and crystal filter both > ends. Adjust variable resistors for best passband ripple. Then measure > resistance? Should I expect resistance values to be same at both ends? > > 73, Steve. >
Attachment:
Filter2_20Mhz_2k7BW_matched.png
Description: PNG image
Attachment:
Filter2_20Mhz_2k7BW_matched_closeup.png
Description: PNG image
Attachment:
IMG_0641.JPG
Description: JPEG image
