The National Crystal Filter Cut to the Chase We don't need no steenkin math... Tony Casorso ADØVC 8/6/2017 67 Slides 1
Goals Understand this circuit (NC183D Receiver): 2
QST Article This circuit was described in the December 1940 issue of QST by W1BZR. The article was titled Improving Crystal Filter Performance. 3
Circuit Functions The circuit has three functions: Selectivity Control Removing heterodynes Bypass mode 4
Selectivity Use a series resonant crystal at 455Khz. Curve from HRO7 manual, same as QST article 5
Selectivity From Curves we estimate 3db BW for position 1 at about 2Khz for a Q of 227. 3db corresponds to an input/output ratio equal to the square root of 2 (1.41) The idea was to provide a wider range of selectivity with a simple circuit. 6
Crystal Footnote in Nov. 1933 QST article by Lamb: 4 P34, QST Mar 1933. Typical Approximate value for a 525-kc. X-cut crystal with faces approximately 0.4 inch square: Equvalent Inductance 8.6 henrys; resistance 2500ohms, capacitance.01uuf, Q 11400. These values were obtained from the resonance curve of the crystal and indicate considerably higher selectivity than previously published data would indicate. See Terman, Radio Engineering, pp. 264. 7
Crystal The Wiki for crystals: https://en.wikipedia.org/wiki/crystal_oscillator says that Q ranges from 10000 to over 1 million. Statek tech note on crystals: http://www.statek.com/pdf/tn32.pdf says series cap ranges from.0005pf to.01pf typically but inductance can be anywhere from 100,000H to 1mh. So cap is limited range and inductance is whatever needed to achieve resonance. 8
Crystal Since we don't know, pick capacitance as 0.009pf, near the top of the range. Inductance has to be 13.595H based on resonance formula for 455Khz. Pick resistance as 2000 ohms which makes the Q 19425. There is also shunt capacitance of a few pf. 9
Varying the Q To vary the selectivity we need to vary the Q by adding extra resistance in series with the crystal. We estimate the Q by dividing 455Khz by the 3db bandwidth eyeballed off the selectivity curves and then divide it into the reactance of the 13.595H inductor: Q 4550 needs R of 8542 so add 6542 more Q 2275 needs 17083, add 15083 more Q 1138 needs 34153, add 32153 Q 569 needs 68305, add 66305 Q 284 needs 136852, add 134852 10
Varying the Q Simulation Results: 11
But National Circuit Has No Resistors! How do they vary the Q without resistors? A lossy parallel resonant circuit has a resistive component to its impedance that varies immensely as you approach resonance. National exploits this. 12
Nationals Approach Looking at the National schematic from the first slide, the parallel resonant load for the crystal consists of L22 and 100pf and a 35pf trimmer with an assortment of small caps placed in series with the trimmer at different selectivity settings. L22 is nominally 1mh and is adjusted in the calibration procedure. 13
Nationals Approach The circuit shifts the resonance frequency downward as the selectivity switch is advanced to narrower settings by adding more capacitance in parallel with L22. At the narrowest setting, the trimmer is directly added. Note that the tube grid and other stray capacitance must also be considered. Lets look at a curve of the resistance component versus frequency for parallel resonant circuit using a 1mh inductor at 455Khz. The inductor for this curve has a series resistance of 30 ohms based an estimate of the wire resitance, skin effect and other losses. 14
Resistance Curve for Lossy Parallel Resonant Circuit 0 15
Nationals Approach National uses the right side of this curve. For example, lets say we wanted to use the tuned circuit to do the narrow bandwidth for switch setting 5. We said this would require us to add 6542 ohms. If we shift the resistance curve to the left by 15Khz, the resistance at 455Khz will be about 6500 ohms. So retuning down to 440Khz will give us a Q near 4550 for a bandwidth of about 100Hz. Lets look at the shifted curve (next slide) 16
National Approach 40k ohms 6500ohms 17
National Approach Note the 40K ohms of capacitive reactance. That means that the tuned circuit looks like an 8.7pf capacitor at 455Khz. Combined in series with the.009pf capacitance of the crystal, this will shift the crystal peak upward by 235hz. So, lets combine the crystal with the tuned circuit (set to 440Khz) and see if it behaves as expected. Bandwidth about 100Hz and peak shifted upwards by 235hz... 18
Filter at Position 5 (Ftune=440Khz) 455.235khz Amazing! ` 100Hz Tuned to 440Khz 19
Position 1 In position 1, the LC is tuned to its highest frequency. No extra capacitance is put across the LC in this position. The QST article says that the LC is to be tuned to the crystal frequency in position1 so the highest frequency is 455Khz. This puts us at the peak of the resistance curve (273Kohms) and will give the broadest response possible. Lets try it! 20
Position 1 (Ftune=455Khz) Width 0ver 4Khz 21
Position 5 Comments Some things to make note of: The peak is not shifted. This is because the reactance is nearly zero at the LC resonant frequency. The peak is broader than the National curves. Maybe we need to guess a higher Q for the crystal. Doubling the Q will cut the bandwidth in half. 22
Positions 2 through 4 These can be achived by shifting the resistance curve to be somewhere between 440Khz and 455Khz. Evenly spaced would be 443.75khz, 447.5khz and 451.25Khz. The resitance curve isn't a straight line but this should be close enough. Lets try it. We will plot all 5 settings together, for both normal and doubled crystal Q. 23
Positions 1 through 5 Crystal Q=20000 Crystal Q = 40000 24
Thats it for the Selectivity. We will look at the notch next. 25
The Notch Real crystals all have both a series and a parallel resonant mode. This is because the crystal is a slab of quartz with an electrical contact on opposing faces. The quartz is an insulator so the construction we just described sure sounds like a capacitor. It is a capacitor. The capacitance just described is usually called the holder capacitance. It is connected from one end of the series resonant circuit to the other. C3 is the holder capacitance. It is usually several Pf 26
The Notch Up to now we had the holder capacitance set to 3 femtofarads (.003pf) so that the notch wouldn't get in our way. It is really more like 3pf or more even. Let us redraw the crystal circuit to look more like another circuit we have see before. 27
The Notch Since all of the components are in series with each other, we can reorder them. Now the circuit looks a lot like a parallel resonant circuit (next slide). C3 and C4 combine to make a capacitor smaller than either of them and is connected across a lossy inductor. Because the combined capacitor is smaller than C3 alone, the parallel resonant frequency has to be higher than the series resonant frequency which only involves C3. 28
The Notch Here is our crystal with proper holder capacitance. This is the doubled-q version. 3pf and.0045pf in series make 0.00449326pf putting the notch about 340hz above the peak at 455khz. 29
The Notch Remember that C4 is connected across the pins of the crystal package. How then could we move that notch to the right or left in order to remove a heterodyne without opening up the crystal case and messing with it's innards? Adding more capacitance across C4 will push it ever closer to 455Khz. In the limit, we could move it to 455 and that's as far as it goes. But the cap would be so big that we would be shorting out the crystal. To move it to the right we have to reduce C4 so that C3 and C4 combine into an even smaller capacitor. How is that possible? How about pushing to infinity and eliminating it entirely? 30
The Notch The trick to reducing capacitance is inductance. Look what happens when I put 20mh across C4: The notch crossed over to the left side of the peak. 31
The Notch If we start with the holder capacitance of zero and add a very small capacitor across the crystal (where C4 is), we will see a notch appear way above the peak. As we add more capacitance, the notch will move down towards the peak. You won't be able to get it closer than a few hundred Hz to the peak before the capacitor you are adding effectively shorts out the crystal. 32
The Notch If again we start from where the holder capacitance is zero and add a small inductor across the pins of the crystal, a notch will appear way below the crystal peak. As we add more inductance it will move up to with a few hundred Hz of the peak. 33
The Notch But National didn't use an inductor. Or did they? If you put an inductor and a capacitor in parallel across a source, the current in the inductor will be 180 degrees out of phase with the current in the capacitor. This is because current leads voltage in a capacitor by 90 degrees while current lags voltage in an inductor by 90 degrees. 34
The Notch The filter circuit used by National and many others simulated the inductor by using an inverted input signal through a capacitor. If the capacitor is bigger than the holder capacitance it looks inductive and the notch moves to the left of the peak. If the capacitor is smaller than the holder capacitance the the Notch moves to the right of the peak. 35
The Notch Neutral Inductive Capacitive 36
The Notch On the capacitive side, the closest you can get the notch to the peak is the natural parallel resonant frequency of the crystal. This happens if you reduce CB to zero. But real capacitors don't go to zero and we would like to be able to get even closer. To solve this, a dual stator capacitor is used. As you reduce CB you simultaneous add capacitance across C4. 37
Notch Simulation (0 to -90 degrees) 38
Notch Simulation (0 to +90 degrees) 39
Notch Last Words You probably noticed Cmagic in the circuit. This is used to neutralize the holder capacitance when the KNOB is at zero degrees. Dave Wise (www.antiqueradios.com member) has taken internal photos of the crystal unit in his HRO60 and we don't see Cmagic anywhere. Maybe it is hidden in stray capacitance or some asymmtery in the phasing capacitor construction. This is not very important for understanding the circuit though, so it is a mystery for another day. 40
Input Circuit Before we talk about the operation in bypass mode, we should take a look at the input circuit. We have modeled it with a voltage source so far. Maybe it isn't a good voltage source. Maybe it throws some extra reactance or resistance into the circuit that would affect the behavior. Fortunately Dave Wise made a few measurements on the input circuit components. 41
Input Circuit Response Node n002 Component values are thanks to Dave Wise who made the measurements on his receiver. Primary resistance is 5 ohms. Secondary 1.9ohms. I fudged them up some for skin effect. 42
Input Circuit Loading Node n002 By back-driving the circuit we can measure its impedance. Here we plot the reactance (blue) and the resistance (green). The reactance is a straight line that declines with frequency so it is a capacitor. Calculating it out gives (surprise) 152pf which amounts to C1 and C2 in parallel. The resistance is mostly zero except near resonance where it rises to a whopping 125 ohms. I don't think we need to worry about that too much. But the capacitance could matter when we start looking at the bypass mode. The coupling coefficient I chose might be way high (K1 is.2 or 20%). IF transformers are typically more like 1%. 43
Input Loading I was asked about the effect of the transformer. I had to simulate it to be sure that my assessment of the 152pf being due to the sum of the two 75pf caps was correct. Away from resonance, the transformer just looks like a 88uh inductor with a few ohms of series resistance. At resonance, the inductance hides and it becomes a few hundred ohms of resistance. So, I think my assessment of the source of 152pf was OK. A lucky guess. 44
Crystal and Phasing Circuit Loading When we were looking at the selectivity way back in the beginning, we used a crystal model with a nearly zero holder capacitance so that we didn't have to deal with the notch. Without the holder capacitance, the crystal looks like its series capacitance (.0045pf for the double Q case). Basically connecting it to the parallel tuned circuit did not really affect the LC resonant frequency. With the phasing circuit and real holder capacitance, the LC tuning will be affected. By how much? We can backdrive the phasing circuit and measure its impedance. Or you can just look at and see that it will look like something a little under 11pf. The holder capacitance plus the phasing cap plus the Magic capacitance is 11pf but it is in series with 150pf from the input 45 circuit. Backdriving gives a value of 10.5pf.
LC Resonance Looking at the original National circuit again, we must remember the grid capacitance of the tube that the filter is driving. This should be 4 or 5 pf. Combining with the phasing circuit let us estimate a total capacitance across L22 (from the National Schematic) of 115pf. Based on that, the inductor needs to be 1.064mh. Lets remember that number for future reference. 46
Bypass Mode In bypass mode the switch shorts out the crystal. It also opens up the connection between the crystal/phasing circuit and the LC tuned circuit and replaces it with a 3pf capacitor. The reason for this is that shorting the crystal puts the 152pf of input circuit loading directly across the LC. This would make it impossible to resonate at 455Khz in bypass mode. 47
Bypass Mode Why 3pf? It just needed to be small. There is an interaction between bypass mode and positions 2 through 5 because both use trimmer C29 for different purposes. Bypass mode uses C29 to resonate at 455Khz while position 5 uses it to resonate at roughly 440Khz. With 20-20 hindsight, we expect C29 to be set near 8pf in bypass mode so that, when added to the 3pf above we match the capacitance of the crystal/phasing circuitry which is all that is across L22 in position 1 (recalling that we want the LC to be resonant at 455Khz in position 1 also). So lets build bypass mode (a simple circuit) and use 1.064mh for the inductor, 8pf for the trimmer, 4pf for the grid and 100pf for C30, a 3pf coupler to the input circuit and see what happens. 48
Bypass Mode Simulated Node n002 49
Position 1 Simulated Previous we looked at position 1 without the phasing circuits or the holder capacitance. Now lets muddy it up and throw in all of that. In position 1, the trimmer is removed so the capacitor across L22 drops from 112pf to 104pf. The 3pf cap goes away and we hook up the crystal and phasing circuitry. If we did all this right, the response should be very broad with no notches. 50
Position 1 Simulated I had to tweak the cap up by.73pf to get it perfect. 51
Position 5 Simulated In position 5 we add the full trimmer capacitance so we are up to 112.73pf. The bandwidth goes from over 5Khz down to about 50 to 60hz. This is with the doubled Q crystal. Note that we can't really change the bandwidth because adjusting the trimmer will mess up bypass mode. 52
Position 5 Simulated 53
Do you think that's enough already?? Nope. We still need to explain the National Alignment procedure. Whats with setting the generator to 457Khz in position 1 and adjusting for a peak on an output meter?? This gives me an opportunity to go bonkers with LT Spice!!! 54
Doing the Selectivity Switch This is a way to do the selectivity switch in LT Spice. Not beautiful, but effective. Each switch contact is a resistor that we adjust from very small to very large to accomplish the switching.rswxtal and RSWCAP get used for bypass mode (position 0). 55
The Whole Hog 56
The Whole Hog With this I can perform the National Alignment procedure. Lets do it. Start by putting the switch in position 1 and setting the generator to 457 Khz. Then adjust L22 for max output. I will not use the doubled Q crystal this time. Q is about 20000. 57
Alignment Step1 A clear peak at 1.07mh. Whats the frequency response? 58
Alignment Step 1 Frequency Response 59 Dead center on 455Khz with maximum breadth. Its magic.
Response with Various Inductances 1.06mh 1.07mh LC Tank Resonant at 455Khz Resonant at 457Khz 60 Step Inductor 1.006 1.027 1.048 1.06 1.07 1.091 1.113 1.134mh
Alignment Step 1 So, setting the generator to 457 (or 453) Khz and adjusting the slug for maximum output is actually tuning the LC to resonance at 455Khz. Note the 457Khz response with the LC tuned to 457Khz. Output actually falls relative to tuning to 455Khz. This is because the output rises more from the broadening effect than it does from the peaking. Unexpected but true. Important Point!!! National engineers had to find an alignment procedure that tuned the LC to 455Khz without setting the generator to 455Khz. They could not go to 455Khz directly because the crystal peak would cause the best peak to be found with the LC highly detuned from 455Khz (remember how the selectivity circuit works). 61
Align Step1 Comments By experiment with SPICE it can be shown that the proper alignment frequency for step 1 (ie: the generator frequency) is determined primarily by the Q of the crystal. If we double the crystal Q to about 40000, correct alignment occurs with a generator frequency of 456.4Khz instead of 457Khz. Changing the Q of the LC tank only affects the sharpness of the adjustment peak but not the correct amount of inductance for L-22. 62
Align Step 2. Flip to position 0 (off) and adjust the trimmer for a peak at 455Khz. 63
Align Step2 Result A clear Peak at 7.4pf 64
Align Step2 Frequency Response Hmm. Not exactly at 455Khz but very close. 65
Alignment Step2 Comments OK its off a hair. This is only because I used 0.1pf steps. Finer steps would have centered it. 66
Grand Finale Lets run the switch through all 6 positions. Hopefully you will now understand everything you see. 67
Grand Finale Finale 68