Tools measurement and tolerances - Electronics

Created; 05/02/2015, Changed; 04/03/2024, 11/01/2024

Previous page; Control Loops

Harmsworth Encyclopedia about 1901 

The Electromagnetic section describes the cohere as the best wireless detector.  It was not until later in 1904 that the diode valve had been developed, then in 1906 the cat's whisker semiconductor diode was developed.  These were all susceptible to vibration.  Valves were also susceptible and could behave like a microphone, and the term microphoney is used for the issue. 

The cohere wireless detector is formed of a tube with iron filings and a magnet to stick them together and so it conducts electricity an electro magnet was wound around the tube supplied with the alternating current radio signal received.  I believe by looking at examples that the alternating current tends to weaken the strength of the magnetism to stick the iron filings together and so reduces the direct current flowing through the iron filings, thereby demodulating the wireless signal.

The electrical and electromagnetic sections of the encyclopedia have quite long articles, but there is no electronics section, the term was not in wide use at this time, I read.

When you assemble things from components, sometimes the completed assembly will not fit together, and other times the components will go together so well that the assembled item is superior to normal.  Electronics components are now made with very good tolerance control so that in practice the extreme (worst and best) cases of the situation do not apply, defective parts almost never enter a legitimate supply line.  It is usually the easiest and cheapest way to put a bigger margin in the design so that assembly is consistently fault free due to the accumulation of tolerances.  So instead of using a sum of squares, use the sum of worst case errors or some factor between the two if that is not practical.

If you want a good camera now pick a make and model and the next box off the shelf will be fine.  When my Granny bought her Rolliflex in the 1940s, mum tells me she made a lot of fuss to get the one demonstrated to her rather than another off the shelf.  The point must have been significant for granny to tell mum and for mum to tell me.  But the story is generally very true and well understood until about 1980 then anything purchased had become consistent that is anyone off the shelf is the same as any other off the shelf.  The first Zero Defects programme was launched by the big semiconductor manufactures such as Motorola about 1980.

St. Marys, The Castle in the Rocks, Hastings, Easter Bank Holiday, April 2015. 

This is as good as it gets - the cars are there, they do not need to be seen as spoiling the picture - This is as it is.  I held the camera above my head even so to reduce the impact of the foreground, and in that way the photograph is as I perceive rather that what I see it.  That is, the cars are seen, but you look past them, therefore perceive the scene differently.  Camera is a Cannon IXUS 60.  The camera would have been the next one off the shelf, and it would perform as well any other of the same model.  This camera is better than a very good film camera.  This picture is particularly sharp, I put down to experience and luck, because it was a windy day and given that I was holding the camera above my head.

Best tools for the job

If you are working on a higher performance instrument design, there may be no accurate test instrument as good as the instrument you are designing.  In any case the required test equipment for the particular job may not be available or may change the way an instrument under tests functions such as when considering using an oscilloscope even with x10 attenuator probes severely loads a high frequency and high impendence circuits by its capacitance, so it is important to consider what parts of a circuit you place the oscilloscope probe.  Either way, spilt the system up into parts that you can test in different ways so that you can figure out logically the likely performance in the parameter you are interested in.  In this case, build up a pattern of cross-checks that give you a subjective confidence.  Continue to build up the system you are developing cross-checking in the way I have described, and reviewing what you had done previously until the work is complete, and you can say what you have achieved and how much confidence you have in its performance.

In conclusion, your best tool is your reasoning, and to develop an ability to look at things in alternative ways. Nothing is absolute, even the accuracy of a measurement, but by building up a pattern of measurement with tolerances you will come to a high degree of certainty.

Worst Case Design and understanding tolerances

This is about understanding shades of gray.  To start with, Sum the squares of the errors, to give you a figure of merit. You should, but it is not essential, square the sum of rooted numbers to give a meaningful value.

If you try calculating the values of resistor that you require in a network such as a Wheatstone bridge using Super-position and Maxwell's circulating currents, the outcome is a set of non-preferred value resistors.  So what they taught you at school was no use then, but we learnt superposition, which was useful.  So what you do is create a model on paper or office spreadsheet which you can repeatedly try preferred values of resistors and use the values that best give the lowest number using the least squares of the errors' method.  {squaring the errors and sum them - but don't bother with square rooting those sums that you would have been taught}

AL-0050-01A OrCAD 17.2, Maxwells circulating currents.  Wheatstone Bridge.  - The curves, lines and wavy line drawing tool is unique to OrCAD, Cadence and is fiddly to use and possibly using arks and lines rather than the curve tool would be easier? 

Super-position, I did not use this for solving resistor networks but to find the constants from a number of test points for a polynomial equation in the form y = a + b.x + c.x^2 + d.x^3 ... This way an equation could be found to fit a data set forming an X-Y graph.   Depending on the number of terms resolved using 2, 3, 4 or other by the number of points term determinants.  The problem was rationalised a little by summing the data, the squares of the data, the cubes of the data, but the number of points required grows exponentially so it is necessary to limit the number of terms and pick the most useful terms.  I subsequently learnt this is quantum maths that is used to solve multiple, variable problems like this.  But this was 40 years ago, and I have now forgotten much of what I developed then. 

The resistor network calculation was a lot of work, and I developed programs written in BASIC that by trial and error found the best preferred value set of resistors to best meet the requirement.  Later solution was to breakup and solve smaller networks and do any arithmetic in the microcontroller.  So what I needed to do could be done using mental arithmetic, a calculator, pen and paper, entering preferred values. 

So to sum up; nearly all circuits with resistor networks can be split down to two resistors networks which come down to a voltage or a current source with an impedance, which you then load with the next resistor in the network.  In fact, if you can avoid resistor networks that is even better, resistors age and have temperature coefficients.  Spreadsheets are very useful for comparing and changing a simulation. 

My point is that the best lesson in tolerating and getting a feel of circuit function and actual behaviour is to do the maths, until you feel it, without having to go through everything to prove it.  Once you are, there, use a simulator rather than spreadsheet if you want. 

Often working out the solving equation is difficult, but there are often short-cuts that suffice for now and may be satisfactory.  

For example, 4 = 2 x 2 but if multiplication is difficult 2 + 2 addition may suffice.   I can not think of an example where this case has been useful.

A small angle triangle which has two long sides can be treated as a right angle or an equilateral triangle, and you can use sine instead of tangent, to solve your problem. 

An attenuator attenuating by 10% so pass 90% of the signal, but if the signal passes through two attenuators then roughly 80% using addition of the signal remains is a good approximation, although 0.9 x 0.9 is 0.81 (81%) is the correct answer. 

Look for shortcuts and use them, but also check their validity.   Shortcut approximation can be applied in computing where simple arithmetic is much faster than carrying out more complex maths, such as inside an interrupt routine. 

During World War two fast calculation for gun ranging was carried out using tables but synchros and servos systems were developed which only used addition to calculate the ranging of big guns. 

If you do start with a simulator, have a guess and do some arithmetic to support your guess first, use the simulator to test your guess.  If there is a disparity, prototype it and see which is correct and analyse it to understand it.  I have never used a simulation tool for anything I have designed and had put into production. 

I recommend that you download and use any spreadsheet or other models that an integrated circuit or other component the maker provides, such as for a switch mode power supply coil maker and controller manufacturing, provides they are well tested and good.  But give yourself some margin, the IC makers tend to specify to just the limit.

I find thermal modelling from data sheets difficult.  An IC won't dissipate much without an area of copper.  Hardly anything can be dissipated if the IC is faced down, face up is better, and with the board stood vertically is best.  I have found exposed tinned or silvered copper better than with solder resist covering a copper area.  Tinned or silvered copper does go mat white quickly, which could mean the board tends to improve with age slightly.  I have heard others contradict my observation on removing solder resist, so try it for yourself. 

A mat black anodised heat sink is 50% more efficient than an untreated aluminium heat sink.


Magnetic near field probe -  the inner cable is pulled out of a coaxial cable.  A thick piece of enabled copper wire is pushed instead.  The new cable is looped around, and the inner wire soldered to the outer braid.  Attach coaxial cable.

Electrostatic near field probe - is simply a disk of metal is attached to the screened enamelled copper wire.

This works with a wideband voltmeter, rather than say an oscilloscope.  Telling you what parts of a circuit are hot.  

You can find components on a PCB that are emitting high frequency radio interference.  I found as I expected that the linear regulator was oscillating.  xx78xx regulators oscillator at low amplitude 55MHz, but I doubt that those are made now?  A solution I was told of was to fit bead chokes in the input and output leads.


Cannon AE-1

This camera uses a very low power Microcontroller made by Texas Instruments, with Ferranti's IIL (I squared L) low voltage transistor Intellectual Property. 

Cannon AE-1

My father bought his first AE-1 in 1976 with the 50mm lens when the camera was first launched.  The camera was particularly popular with amateur photographers for at least a decade.  The A-1 was launched later and was the professional version.

I find this camera has everything to hand and only features that you need, and it does all that you expect of it well.  The design was not compromised in order to persuade you to purchase the next model but to keep the commercial pressure, on any remaining German or Nikon camera supremacy.  A classic camera, surely?

This camera made photography consistent.  To achieve the correct exposed pictures.  The AE-1 cost about £250 and by comparison a teenage electronics technician would be earning £19 to £24 weeks before deductions.  The camera has a low power Texas Instruments microprocessor in which one of the significant Intellectual Property parts was that it used Ferranti I2L (I squared L) low voltage transistor integrated circuits.  These I2L was developed in 1973 and probably continued being used in watch IC's for the next 20 years?

The right-hand camera; Clockwise from the top.

Left Camera;

Significance difference between these two AE-1 cameras is that there is no difference between them.  They were different to British, European, American products in the respect that any two of the same product would be the same by comparison. 

I took the photo with a digital Cannon IXUS 60 - that camera takes comparable quality photos.  The camera's flash has done well, I had zoomed to x3 and photo edited to crop and turn the picture black and white.


Important points on testing;

Useful Links;

I have variously been recommended and received criticism of the following modelling tools (which are free or limited demo's); The schematic and circuit board layout is the most significant factor and CAD that you can work with is important.  I use power supply component maker's modelling tools, particularly for wound components they are very good.  I use spreadsheet or perhaps basic program based tools provided by manufactures of inductive or integrated circuits such as for switch mode power supply design, these are very good.

Conversely, the discrete circuits on the previous pages could be modelled rather than prototyped, and I can see a benefit with the high voltage circuit, if you wish to see how fast it can operate accurately.  That was not a requirement, though.

To discuss electronics, see; Blog page Electronics 

Next Page; Software and Digital