COMPARING THE CIRCUITS [Crystal Oscillator Circuits]

Home | Audio Magazine | Stereo Review magazine | Good Sound | Troubleshooting



This section describes the relative performance of all the oscillator circuits and lists their strong and weak points. To obtain the performance data, a test circuit of each type was built and its performance measured. The circuits are discussed individually, and their performance is summarized in four tables: discrete transistor circuits, ICs, bridge circuits, and harmonic circuits. At the end of this section, the circuits are rated on a relative scale of outstanding to poor.

12.1. PERFORMANCE CRITERIA

There are several criteria for rating the performance of oscillator circuits.

One important factor is whether the crystal likes the circuit, as indicated by voltage waveforms in and out of the crystal. Another important factor is the in-circuit Q, which depends on the ratio of the circuit's external resistance as seen by the crystal across its terminals to the crystal's internal series resistance Rs. The frequency should be reasonably independent of power supply and temperature changes, and power dissipation in the crystal should be low. Circuit complexity and parts count are factors to be considered.

Spurious oscillations when the crystal is removed from the circuit are un desirable, as are parasitics.


TABLE 12.1 Performance of Discrete Transistor Oscillator Circuits

12.2. OSCILLATORS USING DISCRETE TRANSISTORS

Performance of the different discrete transistor circuits varies from outstanding to poor, and is listed in Table 12.1. The data were obtained from the circuits in Figs. 10.1-10.13, and their performance is summarized in the following paragraphs.

Miller circuits (Figs. 10. la and 10.2a) give poor performance. The crystal waveform is awful in the transistor circuit, and only fair in the FET circuit.

The frequency is 30-139 ppm above series resonance. The frequency shifts considerably with power supply and temperature changes, and with the gain of the transistor used. In addition, the frequency is very sensitive to the feedback capacitance used.

Colpitts circuits (Figs. 10.5a-d and 10.4a-d) give average performance.

The crystal waveform varies from poor to very good. The frequency is 94 257ppm above series resonance, and is very sensitive to the value of capacitor C2. A FET works much better in this circuit than a transistor. Colpitts circuits have very few parts, a practical advantage.

The low capacitance load circuit (Fig. 10.5a) gives average performance.

The crystal waveforms are only fair, but the circuit is very stable in the face of power supply and temperature changes. It has a good short-term frequency stability of 0.1 ppm. Its oscillation frequency is very high, however, at 301 ppm above series resonance. This is due to the low capacitance load on the crystal acting in series with the crystal's internal motional capacitance, reducing the total net series capacitance and raising the oscillation frequency.

The high resistance load circuit (Fig. 10.6a) turns out to be essentially the same as the low capacitance load circuit, but with poorer performance.

The former oscillates at an even higher frequency than the low capacitance load circuit, at 671 ppm above series resonance. This is due to the very small capacitance load on the crystal.

Common base amplifier circuits (Figs. 10.7a and b and 10.8) give above average performance. Crystal waveforms and the short-term frequency stability are both very good. The crystal has very low source and load resistances, which give it its very good short-term frequency stability. The frequency is somewhat sensitive to temperature changes. The circuit oscillates exactly at series resonance.

Pierce circuits (Figs. 10.9a, d, and g and 10.10a) give outstanding performance. Crystal waveforms are good to very good. The short-term stability is 0.1 ppm, as good as any circuit tested with the exception of the bridge circuits. The oscillation frequency is normally a little above (5-40 ppm) series resonance but can be set exactly at series resonance or even below it. The frequency sensitivity to power supply and temperature changes is very low, as good as any circuit tested. The circuits dissipate very little power in the crystal and provide a high output voltage at the same time. The increased gain advantage of operating the Pierce without a series resistor (Fig. 10. 11a) does not offer much in the discrete transistor circuit, but it does permit making a Pierce oscillator out of a CMOS inverter, which does not have enough gain for the standard Pierce circuit. Operating the Pierce at exact series resonance requires more parts in the circuit and does not seem to _ offer much advantage in exchange for the higher parts count.

The Emitter coupled circuit (Fig. 10.12) gives average performance. Its measured performance is above average, but the sensitivity of its waveforms to stray capacitance effects brings its overall performance down to average.

Its sensitivity to power supply and temperature changes is very low, and its short-term frequency stability is very good. One disadvantage of the circuit is its relatively high parts count.

12.3. SERIES-RESONANT OSCILLATORS USING ICs

The performance of series-resonant IC oscillators varies from above average to poor. The measured performance of a test circuit of each type is listed in Table 12.2; the data were obtained from the circuits shown in Figs. 11.1 11.7. The performance of these series-resonant IC oscillators is summarized in the following paragraphs.

The CMOS two-inverters circuit (4009 in Fig. 11.1) gives poor performance. It has good waveforms at the crystal, but its frequency is very sensitive to power supply and temperature changes. Its upper frequency limit is 300 kHz. The 74C04 (Fig. 11.2) also gives poor performance. It has good waveforms at the crystal, but its frequency is very sensitive to power supply changes. The 74C04's performance is slightly better than that of the 4009, and it has a higher upper frequency limit of 500 kHz.

The TTL two-inverters circuit (7404 in Fig. 11.3~) gives average performance. Waveforms at the crystal are good, and the frequency is reasonably independent of power supply and temperature changes. The 74804 offers no advantages over the 7404, but the 74LS04 does. The 74LS04 can be used at lower frequencies (down to 100 kHz), but its frequency is a little more sensitive to power supply changes. All three TTL inverters oscillate at a spurious frequency when the crystal is removed from the circuit, a practical disadvantage. The upper frequency limit for all three TTL inverters is about 3 MHz. This is controlled by the crystal's power dissipation limit rather than by the inverters themselves.

TABLE 12.2 Performance of IC Oscillator Circuits Measured Performance

The TTL voltage comparator circuit (LM319 in Fig. 11.4) gives below average performance. Waveforms at the crystal are only fair, and the frequency is rather sensitive to power supply voltage changes. The crystal's source and load resistances are a little high, reducing the short-term frequency stability. The circuit's upper frequency limit is 1 MHz.

As to the two TTL receiver circuits (9615 and 8820 in Figs. 11.5 and 11.6), the performance of the 9615 is above average. Waveforms at the crystal are very good, and its frequency is relatively insensitive to power supply and temperature changes. The 9615 has a low parts count and an upper frequency limit of 2 MHz. The performance of the 8820 is below average. Crystal waveforms are only fair, its frequency is sensitive to temperature changes, and the crystal's load resistance has to be relatively high, reducing its short term frequency stability. Its upper frequency limit is 1 MHz.

All three of the ECL receivers (10114, 10116, and 10216 in Figs. 11.7~ and c) give above average performance as series-resonant circuits. Crystal waveforms are very good, and the frequency is relatively independent of power supply and temperature changes. The crystal's source and load resistances are low, giving good short-term frequency stability. The upper frequency limit is at least 20 MHz. One drawback for TTL use is that they need a buffer to convert the oscillators' ECL output to the TTL format.

12.4. PIERCE OSCILLATORS USING ICs

Pierce-IC oscillators give above average to average performance. The measured performance of a test circuit of each type is listed in Table 12.2. The data were obtained from the circuits in Figs. 11.8-11.10. The performance of Pierce-IC oscillators is summarized in the following paragraphs.

The Pierce-CMOS inverter circuit (74C04 in Fig. 11.8) gives average performance. Crystal waveforms are good, and the oscillation frequency is reasonably independent of power supply and temperature changes. Maximum frequency for the Pierce-CMOS circuit is about 1 MHz for most manufacturers' CMOS (B series), 2 MHz for National Semiconductor's 74C series, and 5 MHz for Solid State Scientific's 4000A/B series CMOS.

The Pierce-TTL circuit (Fig. 11.9) uses an LM319 voltage comparator and gives above average performance. Crystal waveforms are good, and the oscillation frequency is reasonably independent of power supply and temperature changes. A voltage comparator is the only TTL circuit that has enough gain to make a Pierce oscillator. Maximum frequency for the voltage comparator circuit is about 2 MHz. Although Pierce circuits normally operate at 5-40 ppm above series resonance, the oscillation frequency in the test circuit was only 1 ppm above series resonance, due to the large 60-nsec time delay in the voltage comparator.

Pierce-ECL circuits (10114, 10116, and 10216 in Figs. 11. 10a and c) give above average performance at 1 MHz and average performance at 20 MHz.

Crystal waveforms are good at both frequencies, and at 1 MHz, the frequency is almost completely independent of power supply and temperature changes.

The three receivers in a DIP are connected in series to get enough gain for the Pierce circuit. The 10114 is used at 1 MHz and the 10116 or 10216 at 20 MHz. When used to drive TTL circuitry, the parts count is relatively high. This is due to the large number of emitter pull-down resistors used in the circuit and the ECL-to-TTL interface circuit that is needed at the output.

12.5. SPECIAL OSCILLATOR KS

As a group, ICs that are specifically designed for use as crystal oscillator circuits do not perform very well when compared with discrete transistor circuits. The special oscillator ICs generally try to cover a wide range of frequencies with a fixed circuit, and this is very difficult to do. As a rule, 'the ICs work best at the low end of their frequency range and poorest at the high end. Many of them drive the crystal with a badly distorted wave form. Their general performance ranges from average to poor. Special oscillator ICs can usually provide reasonable source and load resistances for the crystal at the low-frequency end of their range, but source and load resistances are invariably too high for high-frequency operation, and, hence, their short-term stability is not too good at the higher frequencies. To maintain a perspective, poor does not mean unacceptable. The label poor performance here is with respect to what can be done with discrete transistor circuits. Even the poorest crystal oscillator circuit will provide a more stable frequency source than an LC-type oscillator circuit.

The measured performance of a test circuit of each type is listed in Table 12.2. The data were taken from the circuits shown in Figs. 11.12-11.17.

The performance of the special oscillator ICs is summarized in the following paragraphs.

The 7209 (CMOS) uses the Pierce circuit and operates from at least 4 kHz up to 10 MHz. It gives good performance at 1 MHz and 10 MHz. Crystal waveforms are very good, and the frequency is reasonably independent of power supply and temperature changes. And as with any Pierce circuit, capacitors Cr and C2 in the circuit should be changed when the frequency is changed.

The 74LS321 (TTL) uses a variation of the common base amplifier circuit and operates from l-20 MHz. Performance is poor at 1 MHz and 20 MHz.

The circuit has two main drawbacks. The waveform driving the crystal is poor, and an inductor is used in the circuit when it is not necessary (from a design viewpoint). The frequency is reasonably independent of power supply and temperature changes at 1 MHz. At 20 MHz, the frequency drifts 1 ppm short term and changes a large IO2 ppm when the power-supply voltage changes from 4 to 6 V.

The SP705B (TTL) uses the common base .amplifier circuit and operates from l-10 MHz. It gives fair performance at 1 MHz but is poor at 10 MHz.

12.6. BRIDGE OSCILLATORS

All of the discrete transistor bridge circuits give outstanding performance, as shown in Table 12.3. They all operate at or near series resonance, de pending on the tuning of their LC tanks. The oscillation frequencies are all relatively insensitive to power supply changes.

The short term stability of all the bridges is equal to or better than the 0.1 ppm limit of measurement. The electronic improvement in in-circuit Q is experimentally shown by temporarily decreasing the in-circuit Q of a Feedback bridge oscillator (Appendix C) by 10X, so that its electronic improvement (3X) in in-circuit Q could be seen with the measurement equipment available. The temporary 10X reduction in in-circuit Q was obtained by adding nine times more resistance in series with the crystal.

In the Modified Meacham circuit (Figs. 10.13a, c, & d), the crystal wave forms are fair to good. The maximum gain obtained in a stable two- or three stage transistor amplifier varied from 28--100X, considerably less than the 300-422X obtained by Clapp and Meacham in their original circuits using a pentode vacuum tube amplifier.

The Modified Meacham circuit gives bridge arm resistance ratios (2R,lR,) of 1.4 to 2.4, depending on frequency. A resistance ratio of 1.0 indicates a crystal with all of its internal losses cancelled out, equivalent to an unobtainable infinite Q. These resistance ratios are converted to in-circuit Q improvement ratios by (ratio)/(ratio-1), giving electronic Q improvement factors of 1.7X to 3.5X.

The in-circuit Q can be calculated from the crystal's source resistance R, and load resistance R_load, that add to the crystal's internal resistance R, and decrease the crystal's internal Q by R,I (R,s + R, + R_load. This in-circuit Q (without electronic Q improvement) is then multiplied by the electronic Q improvement factor. For the Modified Meacham at 1 MHz (Fig. l0.13a), the source resistance R, driving the crystal is 8 ohms and the load resistance R_load is approximately 290 + 51/2 or 315 ohms. This gives an in-circuit Q without electronic improvement of 240/(240 + 8 + 315) or .43 times the crystal's internal Q. The bridge resistance ratio is 2 R2/RS, = 2(290)/ 240 = 2.4. The electronic Q improvement factor is ratio/(ratio - 1) = 2.4/ (2.4 - 1) or 1.7. Multiplying the unimproved in-circuit Q by the electronic Q improvement factor gives the actual in-circuit Q of .43 x 1.7 or 0.7 times the crystal's internal Q (at 1 MHz).

Repeating the above calculation procedure, the Modified Meacham's in circuit Q is 0.9 times the crystal's internal Q at 100 kHz, and 1.5 times at 10 MHz. The in-circuit Q at 100 kHz and 1 MHz is low for a bridge circuit, and is due to the crystal's load resistor (R,) being too large. The in-circuit Q can be raised about 50% by increasing the bridge's voltage excitation ratio from the existing 2-to-1 up to 6-to-1 or so. This is done by changing the ratio of the emitter to collector resistances of transistor Qr, which provides 180 degr. phase inversion and drives the bridge. The crystal's load resistor (R,) will decrease proportionately with the bridge's voltage excitation ratio, thereby increasing the in-circuit Q and the short term stability.

In the RLC half-bridge circuit (Figs. A.3, A.4, & A.5) the crystal is driven with a square wave, as compared to the sine wave drive in the Modified Meacham, but that had no apparent effect on performance, as both circuits give the same very good performance. The RLC half-bridge gives resistance ratios (2Rr/R,J of 1.4 to 2.6, depending on frequency. These are converted to in-circuit Q improvement ratios by (ratio)/(ratio - 1), giving an electronic Q improvement factor of 1.6X to 3.5X. Crystal waveforms are good to very good.

Using the same calculation procedure as used previously for the Modified Meacham, the in-circuit Q of the RLC half-bridge is 1.5 times the crystal's internal Q at 100 kHz, 1.9 times at 1 MHz, and 0.6 times at 10 MHz. The high in-circuit Qs at 100 kHz and 1 MHz should provide extremely good short term stability. The in-circuit Q at 10 MHz is low for a bridge, because of low amplifier gain (17X). The low in-circuit Q can be increased about 50% by using the same technique proposed above for the Modified Meacham, which is to increase the bridge's voltage excitation ratio to 6-to-1 or so. This is done by increasing the ratio of the emitter resistance to the collector resistance in the transistor phase inverter driving the bride. Increasing the voltage excitation ratio will reduce the crystal's load resistor (II,), and raise the in-circuit Q and the short term stability.

In the Feedback bridge circuit (Fig. C.2) the crystal waveforms are good.

The negative feedback through bridge arms R1 and R2 gives an electronic Q improvement factor of 2.8X. The in-circuit Q is high, at 1.9 times the crystal's internal Q. The Feedback bridge has only three fourths the number of component parts that the Modified Meacham and RLC half-bridge have, a practical advantage. On the negative side, its amplifier (21X) gain is less than that of the modified Meacham (28X) or the RLC half-bridge (50X) at 1 MHz, so its short term frequency stability should also be less.

The Meacham-IC bridge is discussed in the next section.

12.7.

---- BRIDGE OSCILLATOR USING AN IC Measured Performance

BRIDGE OSCILLATOR USING AN IC

The Meacham-ECL circuit (Fig. 11. 11a) is a full-bridge circuit giving above average performance. The ECL output stage has a difficult task driving the low resistance R, (5 ohms) of the crystal at 10 MHz, and this results in only fair waveforms at the crystal and increases the frequency sensitivity to power supply changes. Both crystal waveform and power supply sensitivity would improve at frequencies lower than 10 MHz, where the crystal's internal series resistance A, would be higher and easier to drive.

The crystal has a 5 ohm source resistance R, and a load resistance R_load of (8.2 + 5) or 13.2 ohms. The crystal's in-circuit Q without improvement by the bridge's negative feedback is &I(& + R,, + Rload) = 5/(5 + 5 + 13.2) or .22 of the crystal's internal Q. The bridge's negative feedback reduces resistive losses in the crystal's bridge arms, and improves the in-circuit Q by means of the bridge's resistance ratio R,/R,y (= 1.64) by (ratio)/(ratio - 1) = 1.64/(1.64 - 1) = 2.6.

The in-circuit Q with negative feedback is then .22 x 2.6 or 0.6 times the crystal's internal Q. This is a relatively low in-circuit Q for a bridge circuit, due to the source and load resistances being too high for the crystal's low internal series resistance R, (5 ohms) at [...] The Colpitts harmonic circuit (Fig. 10.14), using an inductor for the emitter load, gives average performance at 20 MHz. Crystal waveform is good, but the frequency is somewhat sensitive to power supply changes. The circuit has a low parts count, a practical advantage.

The Emitter coupled harmonic circuit (Figs. B.2, B.6 & B.7) gives out standing performance at 20 MHz to 100 MHz. Its short term stability of 0.1 ppm is better than that of any other harmonic circuit. The crystal waveform is good to very good, and the frequency is relatively insensitive to power supply voltage changes. 100 MHz is the circuit's upper frequency limit, because the transistor's collector load resistance has to increase with frequency to maintain a fixed stage gain. This conflicts with the need to reduce the same load resistance as the frequency increases, because of stray shunt capacitance effects.

12.9. WHICH IS THE BEST CIRCUIT?

Which is the best circuit? Among the fundamental circuits, there are two that stand out above the rest: the discrete transistor bridge circuits (as a group), and the Pierce circuit. The discrete transistor bridge circuits appear to have the best short term stability of any circuit type, due to their ability to e electronically increase the crystal's in-circuit Q by subtracting out part of the resistive losses in the crystal's bridge arm (or arms).

The Modified Meacham's in-circuit Q is 0.7 to 1.5 times the crystal's internal Q, depending on frequency. The RLC half-bridge's in-circuit Q is 0.6 to 1.9 times the crystal's internal Q, depending on frequency. The in circuit Q S of these two bridge types can be increased about 50% by raising the bridge voltage excitation ratio, as described in section 12.6. And the Feedback bridge's in-circuit Q is 1.9 times the crystal's internal Q.

Contrast these in-circuit Qs with those of non-bridge circuits, which are only 0.1 to 0.8 times the crystal's internal Q. The Meacham-ECL bridge, which uses an IC amplifier at 10 MHz, has a relatively low in-circuit Q of 0.6 times the crystal's internal Q. It's in-circuit Q will improve at lower frequencies, where the crystal's internal resistance R, will be higher than the 5 ohm output resistance of the ECL amplifier driving the crystal. Below 5 MHz the Meacham-ECL's in-circuit Q will double to about 1.2 times the crystal's internal Q, a much better value for a bridge circuit.

The electronic Q improvement factor in the various bridge circuits varies from 1.6X to 3.5X. If desired, this Q improvement factor can be made larger by increasing the circuits' amplifier gain.

TABLE 12.4 Performance of Harmonic Oscillator Circuits

Experimentally, the bridges' short term stability measured 0.1 ppm, at the limit of the available measuring equipment. The Feedback bridge's stability measured 0.1 ppm without its electronic Q improvement factor, indicating its stability would be even better with it connected into the circuit.

On the negative side, bridge circuits of any type are much more complex than other types of oscillator circuits, are more difficult to design, and have high parts counts.

Among the various bridge circuits, the RLC half-bridge should be the best circuit with the best performance. Its untuned amplifier and simpler half-bridge configuration make it easier to design than the Meacham. And its amplifier gain is higher than the Feedback bridge's, which should give it better short term stability. The Feedback bridge performs almost as well as the RLC half-bridge. It has less amplifier gain, so its short term stability should also be less. Its parts count is only three fourths that of the RLC half bridge, a practical advantage. The word "should" rather than "will" is used here because such stability performance is beyond the measuring equipment available, and hasn't been verified.

Among the non-bridge fundamental circuits, the Pierce stands out on almost every count. It has good waveforms at the crystal, a frequency that is relatively independent of power supply and temperature changes, very good short-term frequency stability, high output voltage, low crystal dissipation, low cost, and it is usable at any frequency from the highest to the lowest. The only disadvantage of the Pierce is that its frequency stability appears to be less than that of the bridge circuits.

Above 20 MHz, where harmonic circuits are used, the Emitter coupled harmonic is the best circuit up to 100 MHz, and the Butler emitter follower is the best one above 100 MHz. The Emitter coupled harmonic is relatively independent of power supply changes, its crystal waveform is good, and it has the best short term stability of any of the harmonic circuits. Its upper frequency limit is 100 MHz. The Pierce harmonic circuit also works well as a harmonic oscillator, but it has a relatively high parts count.

Above 100 MHz, the very broad and flat frequency response of an emitter follower, due to its built-in feedback, makes the Butler emitter follower the best circuit at these frequencies. It is a simple low cost circuit, with good waveforms at the crystal. It has no parasitics, and its frequency is relatively immune to power supply changes.

Table 12.5 lists the relative performance of various circuit types. All of the bridge circuits, the Pierce, and the Emitter coupled harmonic are in the outstanding group, for the reasons just given.

In the above average group are the Butler emitter follower (harmonic), Pierce harmonic, common base amplifier (discrete transistors), and series resonant-ECL circuits. All of these circuits perform very well, with good frequency stability and very good waveforms at the crystal.

In the average group are the emitter coupled, Colpitts, Colpitts harmonic, TTL two-inverters, and low capacitance load circuits.

TABLE 12.5 Relative Performance of Circuit Types

In the below average group are the high resistance load, Butler common t base (harmonic), and CMOS two-inverters circuits. The high resistance load circuit has the disadvantage of an oscillation frequency way above series resonance. The Butler common base (harmonic) circuit suffers from bad parasitics. And the CMOS two-inverters circuit has the disadvantage that its frequency is too sensitive to power supply changes.

In the poor performance group are the special oscillator ICs and the Miller circuit. The special oscillator ICs actually fall into more than one group, but the majority of them fall into the poor performance group because of poor crystal waveforms and a frequency that is too sensitive to power supply and temperature changes. The Miller circuit has bad crystal waveforms and an unstable frequency. To maintain perspective, labeling the performance of one circuit as poor with respect to another circuit is a relative matter. In this case, it means poor with respect to what can be done with a discrete transistor circuit. It is worth repeating that even the poorest crystal oscillator circuit will provide a more stable frequency than an LC-type oscillator circuit.


Prev. | Next | Index



Home
This page was last updated: Friday, 2007-07-20 17:18,Thursday, 2023-07-27 14:11 PST