Monthly Archives: May 2019

47 Concertina Again – A Gift that keeps on giving

On 22-07-2014, I posted a post about some favourite questions that I had used when interviewing electronic engineers and technicians for a job. One of the favourites that I described was a single transistor circuit with outputs from both the collector and the emitter. Later, a friend identified this as a “Concertina Circuit”

That 2014 Blog Post can be seen at
http://richardschurmann.com.au/Other/Electronic_Burrow/?p=293

Over the years, I had been amazed by how, on the one hand, the circuit was so simple, and on the other, it was so powerful in providing for a candidate to expose his merit (or lack of).

This is the circuit that I used.

You see I used an npn transistor, but just about any active device except an SCR or a TRIAC could be used. Old fogies will remember it being used as a phase splitter in valve amplifiers. I don’t know whether it is popular in latter day valve amplifiers as are used in electric guitar service.

I recently saw it in a final year project report that was written by a friend in 1962.

The “Concertina” is the second triode. Note the twin triode to the right acting as a buffer on the concertina stage outputs. Maybe this is significant. See below. Note that in those days, circuits were simpler, and the argument that one might make these days that it is important to show active devices the right way up for clarity (an argument that I hold to), had not really come into effect.

My title for this post included the words “A gift that keeps on giving”. The reason for this is that there are two extra aspects of complexity that never arose when I was using this circuit for job interviewing. In all its simplicity, it is really more complex than I realized.

Gift 1.

I had actually been aware of this years before I ever conducted a job interview, but I hadn’t thought of it in the context. The argument goes that one possible problem with the circuit is that the two outputs have different impedances, so that they will suffer to different extents if capacitively loaded. For discussion of this, I will stick to the npn transistor case, but the argument holds for the 12AU7 as well. The output on the collector will have an impedance which will be the parallel connection of the 1k load resistor and the resistance of the collector itself. For practical purposes we can take this to be the load resistor alone. That is 1k.

The output impedance at the emitter on the other hand will be the load resistor in parallel with Re

Re = 25E-3/Ic

= (approx) 25E-3/3.15E-3

= 8 ohms

In this case, it is the load resistor that becomes insignificant, and we can say that the output impedance at the emitter is eight ohms.

Let us imagine that we load each output with a 100nF capacitor. We might expect that the Collector output will droop at high frequencies with a pole constructed of the 1k output impedance and the capacitor.

RC = 1E3 * 100E-9

= 100 us

Break freq. = 1.59 kHz

Similarly, at the emitter, R = 8 ohms

RC = 8 * 100E-9

= 800 ns

Break freq. = 199 kHz

This is not how it works out, however. The capacitor on the emitter provides emitter bypassing, so that as the frequency at which the capacitor breaks with the emitter load resistor is passed, the gain of the transistor as a common emitter stage starts to rise. Over the next decade, this rise cancels exactly the fall in output at the collector due to the shunting by the capacitor there.

Here are three cases:

Case 1. 100nF Capacitor on Collector.

Here are the Bode plots for the outputs.

At the low frequency end, the gain is limited by the coupling capacitor on the input.

The Green line represents the Collector. The pole at 1.59 kHz is evident.

The Blue line represents the Emitter. No high frequency attenuation in the frequency range of interest.

Case 2. 100nF Capacitor on Emitter

Here are the Bode plots for the outputs.

The Blue line represents the Emitter. Notice that the bandwidth at the Emitter is limited by the pole at 199 kHz as predicted in the above sums.

The Green line shows the voltage at the Collector. Note the zero at 1.59 kHz which takes effect as the capacitance on the Emitter breaks with the 1k Emitter load resistor.

Above 1.59kHz, the gain rises as Xc on the emitter falls with increasing frequency. This continues up to 199kHz where Xc has fallen to the same magnitude as Re. Above that, Re dominates and the gain flattens out.

Case 3. 100nF Capacitor on Collector and Emitter

Here is the bode plot for the two outputs:

The simulator has drawn the Collector response (Green) first. Then the Emitter response in Blue. The Emitter response overwrites the Collector response for the whole plot, except for up near 100 MHz where the Green line peeps out from under the blue.

The Emitter response (Blue) is the same as in Case 2.

For the Collector response, the pole that was evident in Case 1. is exactly cancelled by the zero that we saw in Case 2.

This result which is, at first blush unexpected, provides a rewarding little bit of sparkle to this circuit. I have to admit that I did not initially think this out for myself. I read an explanation similar to the above in the Wireless World magazine in the 1970s. I found that it fell to me to explain it to a friend recently, so having thought out the explanation properly, I share it with you here. My friend, by the way, suffered from no disadvantage with respect to me except that he is not old enough to have read Wireless World in the 1970s.

Gift 2.

Here is another gem that emerges from the apparent simplicity of this circuit. Unlike Gift 1. I have thought this all out for myself.

One possible application for a phase splitter is to provide a “balanced” output for a connection between items of equipment where there might be some common mode signal between the different local grounds. A “balanced” signal path might be used even where the two grounds are solid, but where magnetically induced or capacitatively coupled interference might advantageously be cancelled.

The concertina circuit is NOT suitable for use as the phase splitter to provide the balanced signal in such an application. I have seen it so used (many years ago), but at that time, I didn’t realize what the problem was.

There are two distinct factors at play here. First the two outputs have different output impedances. Any common mode current to both outputs will give rise to a much higher noise voltage on the Collector where the output impedance is about 1k (in our example) than at the Emitter where the output impedance in our example is about 8 ohms.

Earlier in this post, I made a snide remark about drawing active devices lying on their side. One problem with this is that some circuit configuration might inadvertantly arise, and not be immediately recognized as it does not show the active device in the way we are used to. My habit is to strictly constrain the way I draw these things. Although it is common to draw a transistor lying on its front with its nose in the mud when wired in the common base configuration, it need not be. I drew it the right way up when I designed a common base circuit in an earlier Blog post.

Post number 29 “I did the test question – and failed!” 23-07-2014
http://richardschurmann.com.au/Other/Electronic_Burrow/?p=308

This was how I drew my common base circuit on that occasion:

The similarity to our concertina circuit is immediately apparent. To apply a noise current to the Emitter, is to apply it to the input to a common base stage. Whatever current we feed into the Emitter will come out the Collector (ignore base current a.c. component). If the current is a “common mode” current, applies to both outputs, then the Collector will get a double dose. One from the noise source, and one from the collector itself. All this is clearly seen in simulation.

I used a sinusoid of amplitude 1 millivolt and frequency of 10 kHz to provide the signal “Common”.
Here is what we get:

Don’t you worry about absolute magnitudes here. What we are looking at is how the noise voltages on the two outputs compare. The Green trace is the noise voltage on the Emitter.

The Blue trace is the noise voltage on the collector with noise current applied to the emitter only. That is, the signal we see on the collector is there as the transistor is acting as a common base amplifier.

The Red trace is the voltage on the Collector when identical noise currents are applied to both outputs.

I found it interesting that if I placed a 2k resistor in the connection between C3 and the current source G1, then the two outputs have matched noise voltages. This means that if the receiver at the other end of the “balanced” line, has good common mode rejection, then the noise will not appear. This would be a very unsatisfactory circuit though. The balancing of the two noise voltages would be precarious, and anyway, the introduction of the 2k resistor would spoil the matching of frequency responses that I discussed under the heading of “Gift 1.”.

The best way to utilize the Concertina in a circuit with any interference or capacitative loading on the outputs is to isolate it with a buffer stage as shown in the vacuum tube example above. This is almost exactly what Williamson did in his celebrated audio amplifier https://en.wikipedia.org/wiki/Williamson_amplifier design in 1947.

Sorry about those old 12AU7s, Cyril!

In my earlier discussion of the Concertina circuit, I referred the reader to an article on the Concertina circuit at
http://www.r-type.org/articles/art-010p.htm.
Looking at it again now, I am concerned that it might contain piffle. Go into it with care.

46 Even More Oliver Heaviside

The purpose of this post is to mop up a few bits and bobs that turned up whilst I was researching the previous two posts 44 and 45.

Just as electrical concepts had not fully revealed themselves in the 1800s, the language for discussing these things had not evolved. I read in the Kelvin biography that at the time that the first Atlantic cable was laid, the names of the electrical units had not been agreed on.

Heaviside invented the word “impedance” in 1886. This word seems so normal to us today, but to learned latinists it seemed barbaric at the time.

Even in the 1920s when Henry Fowler wrote “Modern English Usage”, “Impedance still had the power to raise the philologist’s ire. In that work, Fowler had three entries that mentioned “impedance”, and they were all negative. Here is an example:


Although Heaviside was not the sort of bloke to pander to the sensibilities of latinists, other were more inclined to take care. Michael Faraday, for instance.

In the Wren Library at Trinity College Cambridge there is a collection of the correspondence of William Whewell. Librarian Nicholas Bell read from a letter from Whewell to Michael Faraday on a recent radio program. There is an MP3 at:
http://radio.abc.net.au/programitem/pgPGNJBLyG?play=true

It ran (in part) like this…

“My Dear Sir,

I still think anode and cathode the best terms beyond comparison for the two electrodes. The terms which you mention … show that you have come to the conviction that the essential thing is to express a difference and nothing more. This conviction is nearly correct but I think one may say that it is very desirable in this case to express an opposition: a contrariety, as well as a difference. The terms you suggest are objectional in not doing this.”

Whewell was responsible for the coining of “anode”, “cathode” and “electrode”.

It interests me that Fowler should get so hot under the collar about “impedance”, yet ignore other electrical terms that seem (on the face of it) to be worse. He does not mention “voltage” for instance. OED2 (The second edition of the Oxford Dictionary) records the first use of the word “voltage” in 1890 in Pall Mall magazine.

Wikipedia says: The Pall Mall Magazine was a monthly British literary magazine published between 1893 and 1914. Started by William Waldorf Astor as an offshoot of the Pall Mall Gazette, the magazine included poetry, short stories, serialized fiction, and general commentaries, along with extensive artwork.

In other words, this was a source for a new technical term coining that was about as technologically sophisticated as Women’s Weekly. It seems that at the time, the word would have been about as acceptable amongst those who took an interest in electrical matters as “amperage” is today. Yet “voltage” somehow has become widely acceptable. Why would this be?

In the 1800s two distinct concepts crystallized which had the same unit: the volt. The first was “electromotive force”, and the second was “potential”. A year or so ago, I wrote a regular magazine column that had a tutorial aspect for some who needed to strengthen their understanding of electrical matters. I invented a circuit to help make the distinction between electromotive force and potential clear. It had four 1.5 volt primary cells and four incandescent lamps in series.

The emf in this circuit is six volts, and yet if you poke around with a volt meter, you will not find a potential difference anywhere that exceed 1.5 volts.

As conceptualization advanced, this distinction between emf and potential came to be seen differently. Now, we speak of a Thevenan equivalent circuit. When we do this, we mean (of course) “Linear Circuit Model”. In this context, we speak of the circuit’s “Open Circuit Voltage”, or the voltage at the terminals”. We do somehow need this more general concept “voltage”, and then use other words to set the details and show what we really mean by voltage. “Amperage” has no corresponding utility.

As I indicated in recent posts, in the early days of telegraphy, the speed were so low that transmission lines could be modelled with (distributed) resistance and (distributed) capacitance, and that inductance could be ignored entirely. It was Heaviside who first worked out the significance of inductance where it did have to be taken into account, and how to address it. He worked out the criterion for distortionless transmission.

I find it interesting that in the early days of transmission line research, the inductance was completely ignored. It wasn’t until telephony made its demands that inductance became important. Here is a note about the properties of a transmission line in which inductance can be ignored. This is from “Life of Lord Kelvin by Silvanus Thompson P329:


You don’t hear much about this “square law” these days!

Back to the Present.

(My mate Cyril complains that this Blog dwells too much on the “very old”.)

The modelling of a transmission line in which inductance does not appear at all, is not common. There are circumstances in which it is still completely appropriate. One example is in the probes used to measure the electrical potential inside a living cell. This interesting electrical measurement problem is mentioned in ADALPAD (P845), where it is stated that “high impedance is essential in these applications, since living cells are destroyed by the passage of quite minute currents”, but that is only a part of the story. The electrical activity in living cells involves the movement of ionized molecules. It does matter exactly what these molecules are, as every species of molecule will have its own characteristic preponderance to take up or dispose of charge. The introduction of a metallic probe, would involve the doping of the cell interior with metal ions, which would invalidate the investigation.

For this reason, probes are made of very fine glass tubes which are filled with an aqueous liquid charged with ionized molecules to match or mimic the liquid in the cell where the potential is to be measured. A metal electrode is installed in the other end of the tube, but the tube life is limited to the time before metal contaminants reach the active end.

This construction gives a probe with a very high series resistance, which is in itself a reason for a very high input impedance on the attached equipment. As well as that the shunt capacitance in the glass tube wall is distributed along the probe resistance. Here in the most up-to-date biological research work, we find an analogue for the submarine cables of 160 years ago.

I first read about design solutions to the problem of such a high impedance probe in the lamented Wireless World magazine. (This was very different from its successors in that design details and the design process were discussed.) The idea that I read about there was to apply a negative capacitor to the probe to partially cancel the probe capacitance. This was done with a non-inverting amplifier (I think the voltage gain was 3) with a capacitor to apply capacitative positive feedback.

When I try this now, I find evidence that the old idea of modelling the line by assuming lumped capacitance all in one spot doesn’t look that good.

Years ago, before I had either circuit modelling or filter design software available, I did some work on ladder networks in which R(series, C(shunt) networks are strung together in cascade. Of course, for any particular RC, the subsequent members of the cascade provide loading and spoil the simple determination of a pole frequency. A usual trick here is to make the impedance of each stage, higher than the preceding. If a stage is ten times the impedance of its predecessor, it will not make a significant impact on the pass response of the earlier one.

In the following picture, I show the amplitude and phase responses for three networks. The green lines represent a network with three stages or RC low pass filter with RC = 100us. Corner frequency = 1591 Hz. The stages are isolated from each other with unity gain buffers. That is, each stage suffers no loading from following stages.


The Blue lines represent a “ladder network” with three stages of RC low pass filter. These are directly coupled, but the second and third stages have an impedance that is 10 times the previous stage. The Blue line is a little less sharper in the knee than the green one, but the difference is not great.

The red line represents a “ladder network” with three identical RC low pass filters. The second and third stages impose a load on the previous stage. For the red line, the three poles are “spread out”, and the result is a much less clearly defined knee in the response.

I have taken an interest in extending this to the situation where there are a very large number of identical RC stages. Such a circuit might serve as a model for a transmission line with distributed resistance, and distributed capacitance, such as a glass biological probe discussed above.

Maybe I will go into this a little more in a later post.