Take a piece of graph paper and put a cross on it every ten millimeters. Initially, this only results in a sequence of numbers. If you have a vibration and always take just one sample, a point on this vibration, then you cannot know what has happened before or after it. In other words, the more vibration points I take, the more precise my representation of this analog signal. I can make a perfect representation with an infinite number of points. This is exactly what the analog vinyl record does. It has an endless number of points. But, the analog copies of them are individually subject to interference – every record crackles differently. Digital copies, on the other hand, are perfect replications because one only as to copy numbers – a comparatively easy technical process. One does signal processing digitally because it has become so unbelievably cheap – and because I can precisely reproduce the most complicated computing operations with digital representations of the analog world. The processing of digital signals is called computing. And affordable computing capacity is becoming increasingly more powerful, although the space requirement is getting smaller and smaller. So that now even digital representations of complex analog, natural, learning processes are reasonably priced and compact – in our sensors too. This opens up a new world full of possibilities. The number of possibilities we implement in our sensors ultimately simply depends on the price and on whether the customer actually wants to have these possibilities implemented in the sensor. But there is always an analog component right at the start as an interface to the real world, for example a photodiode or a pressure sensor. The beginning is always an analog component that detects analog variables. And the question is how much I have to ‘fatten up’ this little signal in analog form so that I will then be able to digitally process it further very cheaply.
AJ What I find, unfortunately, is that the analog interface is not always understood in depth, or carefully enough processed. One very quickly distances oneself from the original. One can see this with mobile phone cameras, for example. Their pictures are highly computed and one has to ask oneself what has been reinterpreted in the picture and what is actually real. These possibilities simply do not exist in this form with analog photography. My point is that everything that is lost at this interface can never be recovered. I increasingly distance myself from the original. So I would wish that people better perceived the special aspects of the analog/digital interface. If I want to take the digital path I should try to stay as close to the original signal as is sensibly possible. Then I should not compute out any errors that are actually only caused by the way in which I obtained the digitalized signal. There are always intermediate values that I do not know. If the deviation from the original is too great I can always increase the number of measurement values. After a certain point, however, I would then drown in the quantity of data and return to the analog signal. It’s important to find a good balance here. So one really must get a feeling for the analog signal and see what part of this signal is really needed to transfer it to this other world as exactly and true to the original as necessary.