Bootstrapping

[Synchrodyne]

Here I am talking about the bootstrapping technique used to increase the apparent input impedance of bipolar transistor and IC circuits, particularly for signal-level AF purposes.

Does anyone know when it was first thought of and/or first came into use?

Since input impedance increasing devices were less likely to have been needed in the valve era, it seems more likely to have been a 1950s “invention”.

Certainly one can find 1960s examples of where it was used, so it may be assumed to have been was alive and well by that decade.

Perhaps one of the better known examples is in the RIAA stage of the 1965 Wireless World Dinsdale amplifier, which Doug Self records as the first of the two-transistor series feedback species. Here bootstrapping was used with the input OC44 transistor.

The Quad 33 RIAA stage used what might be seen as a silicon (BC109) version of the same circuit, refined by the inclusion of passive correction for the inherent series feedback curve error at the HF end (actually quite unusual I believe). Presumably the silicon transistors allowed “tighter” bootstrapping, as the input impedance as seen by magnetic cartridges was essentially determined by the input 68k shunt resistor. On the other hand, the Dinsdale circuit was said to achieve 47k net input impedance with a 68k input resistor.

The Quad 33 also had two other bootstraps in the signal path, both in the main section. The input emitter follower was bootstrapped. Thus for example the impedance as seen looking into the tape replay input was determined by the combination of the resistor network and the volume control. The output emitter follower was also bootstrapped. Presumably this was to allow the preceding tone control stage to look into a high enough impedance that its curves were unaffected by its loading. The feedback arm of the inverting tone control stage was also fed from the output emitter follower, tapped down to provide around 7 dB gain whilst still allowing the tone control stage itself to look more-or-less like a unity gain Baxandall stage.

But Quad had earlier used the bootstrap circuit in its 1963-designed FM MPX decoder. The input OC44 was bootstrapped to ensure that the FM tuner looked into a suitably high impedance.

Tripletone used a bootstrapped input to provide a 2M ceramic cartridge input for its 8+8 amplifier (as recorded by Gordon J. King in HFYB 1968-69).

Revox used a bootstrapped input for its A77 tape recorder input amplifier. The bootstrapping must have been very tight, as the auxiliary input impedance of 1M seems to have been defined by the 1M input resistor. Whilst this high impedance input might have had utility for direct input from ceramic cartridges, I suspect that it was provided more in the expectation that the A77 would be used in conjunction with existing valve amplifiers, very few of whose tape record outputs were fitted with cathode followers and thus which required to look into a high impedance, often 500k minimum.

Mullard included a bootstrap circuit, although not named as such, as a high input impedance buffer in its “Transistor Audio and Radio Circuits” book. And Hellyer showed both germanium and silicon bootstrap circuits in “How to choose and Use Tape Recorders”.

The advent of FETs suitable and economic for AF work in the late 1960s, with their inherent high input impedance, seems to have made the bootstrap circuit redundant for the applications considered here.

Indeed, Tripletone moved to a FET input for ceramic cartridge matching (I think using a Mullard IC). And Ferrograph choose a FET input, 2M2 impedance, for its Series 7 tape recorder, extending that to the following 307 amplifier.

But somewhat later as I recall, in one of his WW designs, Self used a bootstrapped input with the NE5534 (bipolar) IC, for lower level applications where its lower noise as compared with the (FET input) TL071 series was desired.

A related question is whether the “bootstrap” name was coined with the circuit invention or whether it came somewhat later. Of the above noted incidences, Hellyer (1970) is the earliest reference I can find in which the “bootstrap” name is actually used.

Cheers,

[Radio Wrangler]

I must admit I'm not so interested in the history of the term as in whether bootstrapping actually works well in a given instance.

It is one mechanism that can behave very badly when input transients drive stages into clipping.

I've just used a bootstrap in an AM transmitter modulator design, but that comes after a DSP based limiter/filter/processor. Ironically, I actually own some real bootstraps, threaded into the tops of my best leather riding boots! It was also a term in computing for the tedious loading by hand of a small programme used to get the paper tape reader going so the big programme could be loaded.

David

[Synchrodyne]

So how do your real bootstraps behave in transient overshoot conditions.....

Insofar as bootstrapping seems to be something for nothing, then I suppose there is likely to be a downside somewhere. Looking at a bootstrapped emitter follower, in transient overload condition, the emitter voltage would momentarily not track the input voltage. Thus the voltage coming back from the emitter to one end of the bootstrap resistor could be a lot different to the signal voltage at the other end, meaning that it would no longer look as if it were a very high impedance. So during the transient, the source would be looking into a much lower impedance than it bargained for, presumably to bad effect.

Cheers,

[Radio Wrangler]

You mean like um, coming off? (No rider EVER uses the phrase 'falling off') That's certainly a transient condition. The principal audio impairment is truncation distortion. I only ever get as far as "Oh shi".

Thinking of a bootstrapped amplifier arrangement where an E-F has its bias supply bootstrapped to increase the input Z, or where an anode/collector load is bootstrapped... if the amplifier is hit with a transient the bootstrap cannot handle, then essentially it's something the output would not have been driven by properly even if the bootstrap had been replaced with an ideal current source.

Often bootstrapping is used to avoid the need for a higher voltage supply for drivers in circuits which could have achieved the necessary gain anyway.

The high input impedance of an opamp in non-inverting circuits is also a form of bootstrap.

David

[G8HQP Dave]

I have always assumed that boostrapping started with cathode followers in the 1930s. The self-biased CF (i.e. not direct-coupled to the previous stage) bootstraps the grid resistor. They might not have called it that but they would have been aware of the effect. See RDH4, Ch 7, section 2iH (page 322) - this was 1953, revised in 1967.

[Herald1360]

Originally Posted by Radio Wrangler

It was also a term in computing for the tedious loading by hand of a small programme used to get the paper tape reader going so the big programme could be loaded.

And lives on today as an abbreviation (or perhaps I'm just showing my age)- I still think of "booting up" the PC when I turn it on.

Winchester drive, anyone? 3030 of course.

[Synchrodyne]

Originally Posted by G8HQP Dave

I have always assumed that bootstrapping started with cathode followers in the 1930s. The self-biased CF (i.e. not direct-coupled to the previous stage) bootstraps the grid resistor. They might not have called it that but they would have been aware of the effect. See RDH4, Ch 7, section 2iH (page 322) - this was 1953, revised in 1967.

Dave, thanks very much for that; quite enlightening. I hadn’t previously seen the connection, so as to speak. But yes, from RDH4, the input impedance multiplying effect of returning the grid resistor to a tap on the cathode resistors was clearly understood. I wonder if it was a serendipitous find when the initial objective was looking for a simple way to set grid bias. And with the capacitor added, as in RDH4 fig. 7.11, one may see the broad similarity to the conventional emitter follower version, which I now imagine was a natural consequence developed early in the transistor era.

Thus one can probably find many examples of valve equipment with bootstrapped cathode followers, but one that comes to mind is the AF output stage of the leak Troughline and Troughline II FM tuners, which I think is of the RDH4 fig. 7.10 form.

Cheers,

[G8HQP Dave]

The earliest use of the name bootstrap I have found is in Terman 4th edition (1955/57). This is for a linear timebase circuit, which charges a capacitor using a resistor whose other end is bootstrapped so the voltage rise is linear rather than exponential. A similar circuit appears in a 1961 book. In both cases it might have appeared earlier, but I don't have earlier editions of the books. Linear timebases would have first appeared in the mid or late 1930s?

[Radio Wrangler]

I'll try to remember to look when I'm home, I've got some earler editions of Terman. (Fred was a nice chap and everyone who worked for HP when Bill and Dave were on the go has an especial liking for their mentor)

Bootstrap is itself a contraction from the phrase "Lifting onself by one's bootstraps". So booting a computer is a contraction squared

Cheers
David

[G8HQP Dave]

I can imagine that in 100 years time arty types will argue with each other about the etymology of 'booting' a computer, or the 'boot block' on a 'bootable' disc. They may think it has something to do with kicking things off!

I am old enough that at one time I knew how to manually enter the boot code for a PDP-11/40 on the front-panel switches.

I have found elsewhere that people are puzzled when I describe something as a bootstrap problem. By this I mean a problem which can only be solved after you have solved the problem.

[Synchrodyne]

In that vein, another bootstrap-like situation is “you don’t know what you don’t know until you know it”. Or more frustratingly in work situations, folks who don’t know enough to know that they don’t know very much.

Back to the original topic, it’s a pity that the various literature references to the bootstrapping of transistor and IC circuits don’t link back to the cathode follower precedent, which seems so obvious now that I know it, but just didn’t occur to me previously. (Even that sounds a bit bootstrappish!)

I don’t have Terman in my library, although it is on the list of S/H books for consideration. It seems though that it would be a worthwhile book to have. Was the mentioned 4th edition (1955-57) the last?

Re future etymologists arguing about the origins of “booting” and “bootstrapping”, one should hope that in today’s e-world, knowledge would not be lost so readily. Who knows, maybe Googling (or whatever has replaced it in 100 years time) “booting” will bring up this forum and thread from the internet dungeons.

Cheers,

[G8HQP Dave]

My guess is that much of today's knowledge will fade more quickly than older stuff, because it is stored on media with a shorter life. I can, with no difficulty at all, read a book published on paper well over a century ago. I have all the necessary interfaces and protocol details, at both the physical and logical level. On the other hand, I can have difficulty reading a document produced by a piece of software which differs by more than a few versions in either direction from the one I have.

There is another valve bootstrap circuit which is more like the bootstrap often seen with transistors. This is the grounded cathode followed by a cathode follower, with the first stage anode load boostrapped from the CF output. It is closely related to the SRPP (and therefore a cousin of the mu follower). Does anyone know when/where this circuit first appeared?

[julie_m]

Not only that, but an alarming amount of today's data is stored in ways that requires the use of proprietary technology in order to be able to read it. As old versions of software are replaced by newer ones, companies go out of business, newer operating systems lose compatibility with older applications and hardware changes occur that were never foreseen by software authors in the past, so documents are gradually becoming unreadable. Would you be able to make sense of a Lotus 1-2-3 spreadsheet, or an AmiPro document?

Known examples of hardware changes breaking old software:

• Some of the example QBasic programmes supplied with MS-DOS died with errors on fast 80486DX2 and Pentium machines. They used empty FOR ..... NEXT loops to create delays. Near the beginning of the program, a loop that should have taken a long time was timed, and this would be used to set a delay. If the loop managed to finish in less than 55 milliseconds (i.e. before the internal timer moved on), it appeared to take zero time; and dividing the maximum count by the time taken to obtain a "counts per second" figure was doomed to failure. Alternatively, the loop might last more than 55 ms., but then the count needed to impose a longer delay would exceed the limit for an integer type.
• An 80x86/64 processor in 64-bit mode cannot understand certain (rare) 32-bit instructions. Modern GCC versions avoid these instructions when compiling for 32-bit architectures. Some software compiled for 32-bits with an older compiler may well be unusable on a 64-bit system.
• The first version of OpenOffice.org is notable for refusing to build on 64-bit architectures, due to schoolkid errors -- fortunately, the Open Source community were able to fix that one.

[dinkydi]

I would like to suggest that the bootstrap principle (though not the name) is considerably older than so far suggested.

Bootstrapping occurs when the apparent size of an impedance is modified by controlling the voltage difference across that impedance. At the risk of labouring the explanation, consider a simple example. If a signal voltage is applied to one end of any resistor in a circuit, its apparent resistance (as far as that end/port is concerned) can be made infinite. To do this a voltage, identical to the signal voltage, is applied to the far end of the resistor. In this state there is no voltage difference across the resistor, so no current flows - the same result as if the resistance was infinite.

For readers who are familiar with the classic bridge circuit, it may now be recognised that bootstrapping is the principle of the bridge circuit. In a bridge circuit, the resistor of the example is a meter, but the principle is identical. Balancing a bridge is the same idea as "tight bootstrapping" of an amplifier input.

For these reasons I think that Samuel Hunter Christie should get the award for the bootstrap principle because in 1833 he published the circuit that we later called the Wheatstone bridge.

Peter

[dinkydi]

Here are some further thoughts to mull over.

Unfortunately, the fundamental importance of the Wheatstone-bridge principle of 1833 to present-day circuit design is not appreciated. At school we probably learnt that a bridge is used to measure impedance very accurately (in advanced classes we may have also learnt its communication applications). However, while a bridge may be an essential laboratory test instrument, its use is cumbersome and the measurement of resistance with a multimeter is good enough for everyday purposes. All that we learnt about bridges was probably pigeonholed. It did not relate to other knowledge.

If you are a circuit designer, what school didn’t teach is far more interesting and relevant: the basis of bridge measurement is a fundamental technique of impedance control with universal application.

The balancing of a bridge involves reducing the voltage across a galvanometer to zero. When the voltage gets smaller the influence of the galvanometer impedance on the measurement becomes less. When the bridge is in balance there is zero galvanometer current and the galvanometer impedance has no effect on the measurement accuracy (this is why a bridge is so accurate) – in essence, the galvanometer impedance apparently increases as balance approaches and at balance is infinite. The importance of this principle of impedance variation has circuit design implications.

The manipulation of impedance at a point in a circuit by voltage control has application in circuit design in both passive and active circuits and all feedback circuits. These applications are widespread but their bridge ancestry remains unrecognised – some examples are “bootstrapping” or buffering (follower circuits) or impedance determination in feedback systems. An understanding of equivalent-circuit models may be required to fully appreciate the bridge behaviour in some applications but nevertheless bridge action is fundamental.

Peter

[julie_m]

Yes, that's a good point. If two points in a circuit are at the same voltage (wrt some reference point), it doesn't matter what resistance you connect between them; no current will flow through it. That is what happens in a balanced bridge. No energy is changing state in the meter, so it has no effect on the circuit. (Ordinarily a voltmeter has current flowing through it, and an ammeter has a potential difference across it, so either instrument will affect the circuit under test. An AVO 8 less so than an AVO 7, and a VVM or DMM even less so than an AVO 8, but it may still influence the circuit conditions.)

For the purposes of circuit analysis you can make various simplifying assumptions by pretending there is a short circuit, open circuit or any convenient resistance value between those points.

This turns up a lot in op-amp circuits. For the output to be anywhere between the supply rails (i.e. stable, linear operation), the difference between the voltages at the (+) and (-) inputs (i.e., output voltage divided by gain) must be tiny. And since the input impedance is huge, meaning hardly any current flows in or out of the inputs, you have two points at the same potential.

[Synchrodyne]

Thanks dinkydi and AJS for the very nice analyses.

Thus for example, in a virtual earth (shunt feedback) valve amplifier stage, the impedance of the grid leak resistor is effectively multiplied many times, and it becomes negligible as its loading effect on the input (or feedback arm) is concerned. Or in other words it is bootstrapped. And bootstrapping is not a discrete circuit technique, but an implicit facet of feedback circuit design. And as such, it was not invented, but simply identified and at some stage called out.

Possibly then the term bootstrapping was coined for those situations where apparent impedance multiplication was the primary objective, rather than a by-product of designing for other objectives. Thus bootstrapping of a cathode follower was probably seldom done to increase input impedance, which was probably naturally high enough for most purposes, but to obtain the desired grid bias. Hence specific naming of the technique might not have been seen as necessary or even thought of. But with an emitter follower, increasing apparent input impedance was often the specific objective of bootstrapping, whence having a convenient “handle” for it had some utility.

Cheers,

[dinkydi]

I'm a bit confused by the first sentence of your argument, Synchrodyne. With shunt feedback the shunt-feedback/grid-leak resistance is divided down (not multiplied), so the input impedance is low (in fact a virtual earth, as you state).

The problem with the term "bootstrapping" is that it has not been formally defined, as far as I am aware. If it is defined along the lines of "voltage control of impedance" as per previous posts then it encompasses a wide range of techniques. I don't see this inclusion as a problem, at least you know the meaning of the term and what it refers to, and can employ it when you desire.

The bootstrap term is handy to explain the relocation of a circuit component or particular additional components added to a design. This is consistent with the above definition and also previous usage.

If you wish to limit the applicability of the bootstrap concept to a subset of the above definition then you need clear rules to delineate the subset. I don't see any reason for exclusion and, in any case, no clear rules are apparent.

I have a problem distinguishing bootstrapping on the subjective basis of whether or not it is the primary objective of a designer. Designers usually have lots of objectives and they aim to satisfy them all. Who knows what the primary objective is, or indeed if a primary objective existed. Sure, you can ask the designer, but that is inconvenient at best.

The bootstrapping of bipolar-junction-transistor (BJT) input stages is common because the base current of a BJT is appreciable; for stability the maximum base bias resistance is limited. Bootstrapping simultaneously achieves high impedance and bias stability. Valves and FETs do not have significant input currents so biasing is stable with high resistance and, in these cases, bootstrapping is unnecessary.

Peter

[Synchrodyne]

Looking first at the more general issue, my commentary was based upon the starting assumption – wrong in hindsight – that bootstrapping was a technique used with bipolar transistors in order to increase the apparent input impedance beyond that which could be achieved by normal circuit techniques. Certainly prior to this thread, the only references that I had seen in respect of bootstrapping were in this regard, and where designers had clearly used it for this purpose.

It was the conditioning of that history, coupled with the much better understanding that I have gleaned from this thread, that led to my thinking that perhaps the bootstrapping name might be applied only to those cases where apparent impedance increase was a primary objective, and not to other cases where the same technique was used for other purposes. However, your further commentary - thanks very much - has indicated the difficulties that exist with this idea, and which I had previously missed.

On the shunt feedback case, my thinking was that let’s say we have R1 on the input arm, R2 on the feedback arm, the junction of R1 and R2 connected to the grid, and R3 as grid leak connected from the junction of R1 and R2 to earth. Quite likely R3 is a lot bigger than R1, but that is not always necessarily so. Anyway, because the junction of R1 and R2 is a virtual earth and so very close to earth potential (in signal AC terms), the sources sees the input impedance as R1 alone. R3 has negligible effect because its bottom end is at earth, and its top end is held very close to earth because of the feedback action. It seems to me that R3 had effectively been “bootstrapped” out of the picture, and that was my intended point; another – and different - illustration of impedance manipulation by control of potentials at each end rather than an illustration of impedance magnification. Unfortunately I was not all that clear about it, particularly in the context of the overall discussion.

Cheers,

[dinkydi]

You are seeing things from a historic/descriptive perspective and your assumption is not wrong but limited. You are asking what historically has been described with the term "bootstrapping". This is only a subset of the circuits that actually use the principle of bootstrapping, as has been noted. I see things from a designer's viewpoint so I am interested in the complete set. How does the big picture fit together - are there any anomalies and why.

Historically the bootstrapping term also found considerable use with BJT drivers for unity-gain amplifier output stages. Bootstrapping enabled increased signal swing and the configuration was popular when the cost of a capacitor was less than a current-source driver load (see David's contribution at post #4).

Thank you for clarifying your circuit example. You are quite right that the input impedance would be R1 alone. In a real design you would remove R3 because it wastes loop gain and R1 & R2 can supply the grid current (or R2 alone if R1 is ac coupled). I'll leave the design considerations for the case of the virtual earth being ac coupled from R3 as an exercise for advanced readers.

Peter