App Note 106

"Instrumentation circuitry using RMS-to-DC converters: RMS converters rectify average results." 20 pages.

This app note is a sad postscript to Jim's love of RMS-to-DC converters. One of Jim's favorite chips was the LT1088, which he originally discussed in App Note 22, and then used in many, many app notes, including App Notes 47, 49, 55, 61, 65, 75, and 83. Jim has heaped high praise upon these thermal RMS techniques. Appendix A in App Note 61 is particularly noteworthy, where he includes pictures and schematics of the innards of an HP3400A RMS voltmeter, and then compares it to the elegant simplicity of an LT1088 circuit.

However, the LT1088 is now obsolete. And with the present app note, it seems that it has been replaced by the LT1966 family, discussed here. There is no mention of thermal RMS techniques here, despite all of his previous rants and raves. (For example, see Figure C5 in App Note 83: "Four thermally based types agree within 1%. Other instruments show relative error as large as 48%.") If these parts were better than the LT1088 (and not just "newer"), wouldn't there be a direct comparison?

It gets worse... The only mention of thermally based RMS measurements appears in Appendix B, where he discusses laboratory meters by HP and Fluke. The footnote on page 17 says, "See references 1 and 2 for details on thermally based RMS-to-DC conversion" which are applications notes from Hewlett-Packard and Analog Devices. Why wouldn't he reference any of his own (voluminous) writing on thermal RMS-to-DC converters? Was he told to pretend that the LT1088 never existed?

Sad.

That all said, some of the applications circuits are quite interesting. Figure 8 is effectively a linear regulator that regulates an AC waveform (the MOSFET Q2 in the diode bridge is the series-pass element). Figures 9, 10, and 12 show high-fidelity preamplifiers. Figure 15 is an oscillator with frequency stabilization (using a quartz crystal) and amplitude stabilization (using an RMS measurement loop). Figure 17 is a RMS-leveled noise generator. Figure 20 is an AGC loop that uses the RMS measurement in the feedback path (see also Figure 139 in App Note 47).

The app note ends with a cartoon. "The Good, the Bad, and the Ugly."

I suppose that the LT1088 would have taken the crown as "the beautiful one".

App Note 83

"Performance verification of low noise, low dropout regulators: Silence of the amps." 20 pages.

This app note discusses testing LDO regulators, primarily the measurement of output noise. In effect, this app note is the linear-regulator version of App Note 70 (which discussed the measurement of output noise for low-noise switching regulators).

The schematic of the measurement chain is shown in Figure 2. Note that single-pole highpass filters on use on the input and output to remove any DC components from the signals. This arrangement leads to a non-optimal low-frequency cutoff. (See the frequency response in Figure 3, which shows a nice high-frequency roll-off, but a considerably "rounder" low-frequency roll-off... Butterworth is better. He could have done better by moving the second-order Butterworth poles to the third-order locations, since he already had a pole (or two) on the axis.)

Nevertheless, the plots on pages 8 and 9 are very nice, and are clearly the product of painstaking work.

The best quote (from page 3, with footnote): "The metal can encloses the regulator under test and its internal battery power supply (the cookies were excellent, particularly the thin ones with sugar on top)."

Appendix A discusses the internal architecture of low-dropout regulators. This verbiage is an updated and expanded version of Appendix A from App Note 32. Note that Figures A3 and A4 show the transient response for a load step (from 10mA to 100mA); he will return to this topic in a later app note.

Appendix B discusses capacitor selection, and warns of the voltage-dependent and temperature-dependent horrors of some ceramic types.

Appendix C is an in-depth discussion of RMS voltmeters. He has discusses RMS voltmeters before (most notably in Appendix C of App Note 65), but this discussion is the most detailed, and includes comparisons of voltmeter types (Figures C1, C2, and C3), as well as comparisons of specific models (Figure C5). Of course, the HP3400A and HP3403C work very well, as does his special baby, the LT1088-based circuit in Figure C6. Thermal great, logarithmic good, rectify and average bad.

The app note ends with a cartoon. "Speak softly and carry a big PNP."

App Note 75 part 1

"Circuitry for Signal Conditioning and Power Conversion: Designs From a Once Lazy Sabbatical." 32 pages.

This app note is another collection of circuits, like App Note 45 (June 1991) and App Note 61 (August 1994). He seems to do one of these collections every four or five years. This one was inspired by his sabbatical and the acquisition of a HP 215A pulse generator. "I took it home, repaired it, and used it to characterize a fast coincidence detector... I had previously abandoned. This exercise proved fatally catalytic."

The first two circuits are improved voltage-to-frequency converters, loosely based on "The Zoo Circuit" from App Note 23. Figure 1 shares much in common with the original Zoo Circuit, exploiting some component improvements (using an LTC1441 in place of the original LT1017) and consuming only 20 microamps at full scale. Figure 4, using a reworked reference chain, is even lower power (less than 9 uA at full scale). These are impressive results. As he says, "these voltage-to-frequency circuits are the beneficiaries of considerable attention over a protracted period of time."

Figures 6 and 10 show low-power single-slope analog-to-digital converters. One interesting feature of Figure 6 is his use of diode-connected transistors instead of diodes, because "Q2, lacking gold doping, temperature tracks the LM334 more closely than a small signal diode would." (The 1N914 is doped with gold to increase its speed, but I had never thought about the effect on temperature.)

Figure 11 shows another RMS-to-DC converter using the LT1088. This circuit is very similar to the circuit in App Note 61 Figure 22, with the addition of a differential front-end, using the LT1207 dual power op amps.

The best circuit, the inspiration for this app note (as explained on page 1), is the coincidence detector in Figure 14. The circuit is relatively simple, but the construction and instrumentation are certainly not. "Evaluating circuit performance requires a sub-nanosecond rise-time pulse generator and a very fast oscilloscope."

I'll cover the rest of the circuits next time.

The best quote is the footnote on page 3: "Okay all you SPICE types out there, start your computers and model the charge pump drift and the reference compensation mechanism." Had someone been needling him about the supposed superiority of simulation?

---

Related:

• App Note 75 part 2

App Note 65 part 4

Sixty-four total pages of appendices!

Appendices A and B are reprinted directly from App Note 55 (and App Note 49). Appendices C and D both have significant updates.

Appendix C discusses instrumentation for electrical measurements. In App Note 55, this discussion was six pages long, but here it has been expanded to twenty-four pages. His instrumentation suggestion are much more detailed. Figures C1 and C2 illustrate the problem: the necessary measurements need to be made with wide-band tools. For example, a new 10-MHz current-probe amplifier is detailed in Figures C3 through C7, and Figures C8 and C9 show a Wien-bridge oscillator used for its calibration and demonstrate its use.

Voltage probes and RMS voltmeters are discussed starting with Figure C10, with a calibration circuit shown in Figure C13. With the new emphasis on floating-lamp circuits in the main text, a new section on "Voltage Probes for Floating Lamp Circuits" starts on page 72. "Measuring voltage of floating lamp circuits requires a nearly heroic effort." Figure C14 shows a wide-band differential amplifier for use with high-voltage probes and a single-ended RMS voltmeter. Calibration of this fully differential RMS voltmeter apparatus is a significant challenge. Jim explains the difficulties, starting with first principles and a detailed schematic of a high-voltage probe in Figure C15. "No less than seven user adjustments are required to compensate the probe to any individual instrument input." And the warning in footnote 10, "Allow at least six hours for the entire session. You’ll need it."

A fully floating, differential, high-voltage calibrator is shown in Figure C16 (schematic in Figure C19). Photos of the details and internals are shown in Figures C17 through C23. The calibrator produces a 500-VRMS waveform to check the calibration of the probes. "Those who construct and trim the differential probe and calibrator will experience the unmitigated joy that breaks loose when they agree within 1%."

RMS voltmeters are discussed starting on page 82. Much of this discussion is copied from App Note 55, with the addition of Figures C25 and C26 (which are mostly borrowed from App Note 61 Figures 22 and 23). The appendix ends with the discussion of calorimetric measurements from App Note 55.

Appendix D discusses instrumentation for photometric measurements. In App Note 55, this discussion was less than two pages long, but here it has been expanded to six pages. He again starts with a picture of his "glometer" tube in Figure D1, which has been modified to allow variable-frequency drive to test individual lamps. The drive schematic is shown in Figure D2, and the measurement schematic is shown in Figure D5.

Finally, Figure D7 shows his complete "CCFL test set", including current probe, differential RMS voltmeter, calibrators, and photometer. Compare to the CCFL test set in App Note 55 (Figure C6). He has also upgraded his computer to a slide rule!

I'll discuss the rest of the appendices next time.

---

Related:

• App Note 65 part 1
• App Note 65 part 2
• App Note 65 part 3
• App Note 65 part 5

App Note 61 part 2

Continuing on, Figure 26 shows an improved high-speed pulse generator. Compared to Appendix D in App Note 47, this pulse generator has a distinct advantage in that it is triggered. "This feature permits synchronization to a clock or other event." The 2N2369 is biased just below avalanche breakdown, and the incoming trigger input causes breakdown (due to the difference between Vceo and Vces). Note the comment about the Tektronix 661 with the 4S2 sampling plug-in ("I'm sorry, but 3.9GHz is the fastest 'scope in my house"). Nice upgrade from the 1-GHz 1S2 sampler that he has used previously. "Ground plane type construction with high speed layout, connection and termination techniques are essential for good results from this circuit." Yep. (I think this circuit is the best circuit of the app note.)

Figure 29 shows a special voltage regulator for flash-memory programming. Unfortunately, he does not compare and contrast this design with the designs in App Note 31. Is this circuit just as good? (It probably is, but why?). What are the features of the LT1109 that make it a good fit for this (touchy and sensitive) application?

Figure 31 shows a low-voltage voltage-to-frequency converter (operating from a 3.3-volt supply). The next two circuits are broadband noise sources. Figure 33 uses a Noisecom NC201 noise diode with a selectable filter to produce the noise, and a RMS AGC loop to set the amplitude. Figure 37 uses a standard zener diode as the noise source, but requires a trim to set the initial noise level. Figure 38 is a switchable-output crystal oscillator, which uses diodes to select which crystal is active in the feedback path of the comparator. Cute.

Appendix A includes significant reprints from the manual of an HP3400A True RMS Voltmeter. Jim compares the approach used here to his approach in Figure 22. The instrument he discusses here is a classic, with an impressively creative solutions to several design challenges. The input buffer in Figure A1 uses Nuvistors (because JFETs weren't good enough in 1965). Figure A2 shows a photograph of the input-buffer circuit board. The "video amplifier" in Figures A3 and A4 is an impressive design with DC and AC feedback loops and clever bootstrapping. The chopper amplifier in Figures A5 and A6 uses neon bulbs and photocells for the chopping action! (As Jim says, "Hewlett-Packard has a long and successful history of using lamps for unintended purposes.") Figure A7 shows the circuit board for Jim's RMS-to-DC converter from Figure 22.

The topic of RMS-to-DC conversion was near and dear to Jim's heart. He covered these circuits in detail in App Note 22, and he used them in his CCFL explorations. He helped to design the LT1088 IC and was its main evangelist. In the footnote on page 28, Jim tries to explain the context: "We are all constantly harangued about the advances made in computers since the days of the IBM360. This section gives analog aficionados a stage for their own bragging rights. Of course, an HP3400A was much more interesting than an IBM360 in 1965. Similarly, Figure 22's capabilities are more impressive than any contemporary computer I'm aware of."

Do I detect some frustration in Jim's voice here? Compare all of the circuitry of the HP3400A with the circuit in Figure 22. In effect, I think Jim is saying, "Look how easy I'm making it for you! Why aren't you buying more LT1088 chips?"

Best quote (page AN61-38): "Incidentally, what were you doing in 1965?"

The app note concludes with a great cartoon (perhaps the best cartoon so far).

---

Related:

• App Note 61 part 1

App Note 61 part 1

Practical circuitry for measurement and control problems: Circuits designed for a cruel and unyielding world. 40 pages

Another collection of random circuits, like App Note 45 (from June 1991). In fact, the introduction says this circuit collection includes circuits from June 1991 to July 1994. Unfortunately, there is no baby-bottle-rating system, however, there are plenty of great circuits.

Figure 1 shows a switching regulator where the switching frequency is synchronized to the system clock. (Synchronization was also discussed in App Note 55.) This trick is a great one. I once designed a "noise-less" power supply for a sensitive RF receiver with this method... I simply "hid" all of the EMI from the switching regulator underneath the EMI from the digital hardware (I admit that this approach is cheating, but it works).

Some of the circuits were supplied by Steve Pietkiewicz, and many of them were "inspired" by Jim's recent work on CCFL power supplies in App Note 55. Two methods for 1.5V-to-5V conversion are shown in Figure 3 and 6 (high power and low power, respectively). Figure 9 is a low-power CCFL supply (an alternative to App Note 55 Figure 11). Figure 10 is a LCD contrast supply (compare to App Note 55 Figure 14). Figure 11 is a power supply for a HeNe laser (the same as App Note 55 Figure H2), and Figure 12 is an electroluminescent-panel supply.

Jim's interest in instrumentation also appears. Several barometer circuits are shown in Figures 13, 14, 15, and 16, and a quartz-based thermometer appears in Figure 17. A FET-input instrumentation amplifier is shown in Figure 18.

There are also some improvements on circuits from App Note 47. The high-speed adaptive trigger makes another appearance in Figure 21 (compare this circuit to Figure 131 in App Note 47). Another wideband thermally based RMS-to-DC converter is shown in Figure 22 (compare to Figure 137 in App Note 47 and Figure 8 in App Note 22).

The best circuit (so far) is Figure 24, not because it's a great circuit in isolation, but because it is true to his "fixer mentality." Although he doesn't admit it here, I imagine that he was elbow deep into fixing a Tektronix P-6042 current probe when he realized that he couldn't get the replacement parts that he needed (see Figure 25: is M18 still available?). So, he designed his own replacement, and Figure 24 was born.

I'll discuss the rest of the circuits and the appendices next time.

Best quote (page 18, discussing figure 22): "It is worth considering that this circuit performs the same function as instruments costing thousands of dollars."

---

Related:

• App Note 61 part 2

App Note 55 part 2

As always, the appendices (up to K!) are great.

Appendices A, B, and G are copied from App Note 49.

The first half of Appendix C, "Achieving meaningful efficiency measurements", is also borrowed from App Note 49, but the second half (starting with the calibration source in Figure C3) shows that considerable effort has gone into making the measurements more meaningful. Jim always devotes great effort to instrumentation, and it shows here. Again, precision RMS-to-DC conversion plays a large role. (Although Figure C6 is captioned "Typical efficiency measurement instrumentation", there's nothing typical about it.) The appendix concludes with a discussion of using calorimetric measures as a efficiency double-check (Figures C7, C8, and C9). As the man says, "Calorimetric measurements are not recommended for readers who are short on time or sanity."

Appendix D discusses more instrumentation, now focusing on photometric measurement. How much light is the CCFL producing? This question raises a key issue, as he explains in the footnote, "It is possible to build highly electrically efficient circuits that emit less light than "less efficient" designs." (See Figure J3.)

The next four appendices are short. Appendix E discusses protection circuitry (important!) for broken lamps. Appendix F discusses shutdown control, and a calibration source (Figure F2) for intensity control. Appendix G is copied from App Note 49, and Appendix H (HeNe laser power supplies) is mostly copied from App Note 49.

Appendix I is a brief (too brief!) discussion of the history and operation of the Royer topology.

Appendix J is my favorite. Titled "A lot of cut-off ears and no Van Goghs", it discusses some not-so-great ideas. On the whole, we engineers don't spend enough time talking about engineering failures, and there is often a lot to learn. Jim relates, "Backlight circuits are one of the deadliest places for theoretically interesting circuits the author has ever encountered." Figures J1, J2, and J3 attempt to increase efficiency by removing the losses in the LT1172. Unfortunately, the resulting lamp drive waveforms are undesirable. Figures J5 through J8 show suboptimal sensing schemes for measuring the output current. It is very instructive to consider why these circuits didn't work: I wish Jim had written more appendices like this one!

Appendix K is a brief discussion of the various sources of inefficiency in a backlight application, including the electrical-to-electrical step, the electrical-to-light step, and the light-to-light step (see Figure K1). Importantly, the electrical-to-electrical efficiency stays high under a variety of operating conditions (see Figure K2), which extends battery runtime.

The app note concludes with a cartoon of his son Michael sitting in front of a 556.

---

Related:

• App Note 55 part 1

App Note 49

"Illumination circuitry for liquid crystal displays: Tripping the light fantastic..." 16 pages.

This app note is Part 2 in the grand saga of cold-cathode fluorescent lamps (CCFL). Part 1 was a single schematic (Figure 36 back in App Note 45, remember the 48 baby bottles?), reproduced here with minor changes as Figure 2. This problem is an important one in portable computers, since "The CCFL and its power supply are responsible for almost 50% of the battery drain."

The converters here use the Royer switching topology. Back in Appendix A of App Note 29, Jim discussed the disadvantages of the Royer topology at length. Mostly, he mocked it for being noisy. Here, he uses it as a high-voltage step-up circuit in a current-control loop. Footnote 3 explains, "Controlling a non-linear load's current, instead of its voltage, permits applying this circuit technique to a wide variety of nominally evil loads." Evil loads.

Of course, instrumentation plays a large role in this effort. High-voltage oscilloscope probes are recommended and discussed ("Don't say we didn't warn you!"). The dual-beam dual-time-base Tektronix 556 get considerable praise: Figures 3 and 6 would be nearly impossible without it ("Most oscilloscopes, whether analog or digital, will have trouble reproducing this display."). True RMS voltmeters ("the meter must employ a thermal type RMS converter") are also discussed on page 4 and in Appendix C. Of course, proper instrumentation is critical, as the footnote on page 11 explains, "It is worth considering that various constructors of Figure 2 have reported efficiency ranging from 8% to 115%."

Other design considerations are mentioned. Parasitic capacitance from mechanical layout is discussed. Transformer leakage due to silkscreen ink is considered. System designs employing two tubes are discussed on page 4 (this advice will be retracted in App Note 65). LCD-contrast-bias sources are shown in Figures 7 and 8.

Best quote (footnote 3 on page AN49-12): "The translation of this statement is to hide the probe when you are not using it. If anyone wants to borrow it, look straight at them, shrug your shoulders and say you don't know where it is. This is decidedly dishonest, but eminently practical. Those finding this morally questionable may wish to re-examine their attitude after producing a day's worth of worthless data with a probe that was unknowingly readjusted."

The next app note significantly expands on this topic.

App Note 47 part 4

The final two application sections (pages AN47-51 to AN47-66) cover data-conversion circuits and some miscellaneous things. As footnote 15 on page AN47-51 says, "Seasoned readers of LTC literature, a hardened corps, may recognize this and other circuits in this publication as updated versions of previous LTC applications. The partial repetition is justified based on improved specifications and/or simplification of the original circuit." Jim undersells the application circuits here; some of the improvements (utilizing these high-speed op amps) are a big deal.

Figure 118 shows a sine-wave VCO, again using a AD639 as a triangle-wave-to-sine-wave converter (see Figure 21 in App Note 13). As I've said before, the AD639 was a Barrie Gilbert's brilliant Universal Trigonometric Function Converter, and using it as a triangle-to-sinusoid converter is like using a Lamborghini to go get your groceries.

Figure 121 is a 1Hz-to-10MHz V-to-F converter. Ten megahertz is pretty fast, but remember that App Note 14 includes the 100-MHz "King Kong" V-to-F converter. Of course, the advantage here is that with the high-speed op amps, there's no more need for the scope-trigger circuits and exotic 10H ECL parts. I especially like the caption of Figure 123 "(Whoosh!)".

Figure 124 is a high-speed (100 ns!) sample and hold, using a four-diode gate with transformer drive. This topology is an elegant solution for a high-speed S&H. Compare to the discrete 200-ns sample-and-hold circuit in Figure 23 of App Note 13.

Figure 131 is a trigger circuit with adaptive threshold, basically the adaptive subcircuit from Figure 97 (I wonder why he didn't present these two circuits in the opposite order?).

Figure 132 is a simple pulse-width-to-voltage converter, using a current source charging up a capacitor, like the single-slope converter in Figure 33 of App Note 13. Despite (or, perhaps, due to) being simple, the performance is very fast: able to resolve 1% accuracy on 250-ns pulses. The drive circuitry for Q3, including the Baker clamp and the speed-up capacitor, is especially instructive.

Figure 137 shows another application circuit for his LT1088 RMS-to-DC converter. This circuit is the same as Figure 8 in App Note 22, except the LT1223 is used instead of the discrete buffer suggested in App Note 22. Figure 139A shows a RF-leveling loop (from App Note 22 Figure 27), using the RMS-to-DC converter from Figure 137. (Figure 139B show a much simpler RF-leveling loop.)

Figure 140 shows a voltage-controlled current source (basically, it's Figure 8 from App Note 45 with much faster op amps). Figure 142 shows a higher-power version of the current source, using a discrete output stage.

Figure 144 shows a high-speed (18 ns!) circuit breaker (compare to the 12-ns version in Figure 40 of App Note 13, which required a floating load).

The best circuit in these sections is a toss up between the elegant sample-and-hold circuit in Figure 124 or the high-speed pulse measurement in Figure 132. (I still think the best circuit in the whole app note is the AM radio station in Figure 116.)

Best quote (page AN47-58): "Digital methods of achieving similar results dictate clock speeds of 1GHz, which is cumbersome." Understatement?

---

Related:

• App Note 47 part 1
• App Note 47 part 2
• App Note 47 part 3
• App Note 47 part 5
• App Note 47 part 6

App Note 22

"A monolithic IC for 100MHz RMS-DC conversion." 16 pages.

App Note 22 is a first chapter in the sad story of an underappreciated design. This app note discusses the LT1088, an IC building block for RMS-to-DC conversion. Jim advocated, designed, and was the primary evangelist of this product. In addition to this app note, Jim also co-authored the 1986 ISSCC paper that described its function and fabrication. For someone with a instrumentation background, an RMS-to-DC converter is an important functional block. If you look in future app notes, you see that he often used this part in his circuits, and he referenced this app note (and the ISSCC paper) quite a few times. It should have been a popular part (and I think that Jim was proud of it), but unfortunately the LT1088 was a poor seller and has been discontinued.

The first part of the app note discusses challenges for RMS-to-DC conversion, and explains how the feedback approach (Figures 2 and 3) is the preferred solution. The text on page AN22-4 (and Figures 5 and 6) discusses the problems and proprietary solution to the thermal conductivity of the die attach (the complexity of which may be a contributing factor to the part's discontinuance). (The box section on page AN22-15 details a method for measuring the thermal resistance of the die attachment.) Figure 7 is a die photo, showing the extreme symmetry necessary for good performance. (Also, there's an inside joke here about "Counts' Theorem" (that 9 does not equal 10) that I don't fully appreciate. Someone is going to have to explain it to me. Anyone?)

Figure 8 shows a complete application circuit for the part, and its performance is shown in Figures 11 through 16. Figure 16 shows the stereotypical response for a thermal-computation system such as this one, where the positive-going and negative-going transient responses are different (because of the differences in heating and cooling time constants). Figures 20 and 21 show protection circuitry to prevent damage to the LT1088.

As Jim says, "some applications may require buffering the LT1088's relatively low input impedance." Several buffers are shown in Figures 23, 24, and 25. Figures 24 and 25 are from App Note 21's collection of composite amplifiers.

I think the best (most interesting) circuit is Figure 27, the RF leveling loop, using a AD539 wideband multiplier, a wideband discrete amplifier (from App Note 21 Figure 6), and a LT1088 RMS-to-DC converter.

Best quote (from page AN22-1): "Applications such as wideband RMS voltmeters, RF leveling loops, wideband AGC, high crest factor measurements, SCR power monitoring, and high frequency noise measurements require the advantages of thermally based conversion." There are a large number of applications; it's too bad it wasn't a more popular part.

App Note 5

"Thermal techniques in measurement and control circuitry." 8 pages.

This app note has two best circuits: first, Figure 6 shows a wideband thermal RMS-to-DC converter. This circuit uses the old servo-multiplier (or light-bulb multiplier) technique of using a feedback loop to force two matched parameters (in this case, the heat produced in the two composite thermistors) to have equal value. We'll see this circuit again. The flowmeter (Figure 8) and anemometer (Figure 11) are related tricks.

The second celebrity circuit is on the final page: Jim's obsession with the HP200 makes an early appearance. Figure 12 is an op-amp recreation of the Hewlett-Packard HP200 audio oscillator circuit (and the references include William Hewlett's thesis). While the flowmeter and anemometer are useful applications, I wonder if they weren't just a set up so that he could justify spending time at work playing with Hewlett's light-bulb circuit.

The other applications are also interesting and instructive. Figure 1 shows an interesting temperature controller (but the LT3525A is now discontinued). The explanation includes a simplified discussion of the various delays and time constants in a heat flow problem, but the full story is much worse: it's not just a multiple-time-constant system, it's a diffusion-equation system! Also, his discussion of insulation needs a caveat about the trade-off: you want to keep the losses small to save power, but large enough to allow the system to quickly recover from an over-temperature condition. Regardless, using the 50-ohm resistor and switch to check the loop response is a great idea, but the loop response needs to be checked for positive steps (when the heater heats) and negative steps (when the losses cool). He only shows the former. (Also, I don't think he needs the 100M resistor in the integrator, but that's another discussion.)

The circuit in Figure 4 uses a CA3096 array to thermally stabilize the feedback transistor in a log amp. This neat trick is also used to thermally stabilize a VCO in National Semiconductor's App Note 286. I'll comment here: Yikes! That LM301A has a huge compensation capacitor! A 33nF capacitor gives a unity-gain frequency of 1 kHz and a slew rate of 0.5 V/ms (yes, volts per millisecond). Replace the LM301A and 33nF with an integrator, and I think you'll get better (and more repeatable) performance. Looking back at Figure 1, is Jim afraid of integrators? I'm guessing Bob Widlar didn't consult on this circuit.

Best quote (page AN5-1): "The close relationship between temperature and electronic devices is the source of more design headaches than any other consideration." Word.