SETI LEAGUE UK - Technical scrapbook
Frequency translation- a brief introduction
Within the two windows, the frequency relationship between any of the signals remains the same in either case - only their absolute values have changed.
This is in sharp contrast with the red shift or doppler effect which will be more familiar to the optical astronomer. With red shift, the whole spectrum has been compressed towards the (zero) origin, so in our example, a 3% red shift would result in the original 30KHz window reducing to 29KHz, as it now centred on 1378.655MHz. Compressing right down to the origin in attempt to copy the earlier translation example would result in a window that had become infinitely narrow.
How frequency translation is achieved
It is worthwhile considering how translation is achieved, if only because some of the secondary effects have a bearing on how it is actually carried out in practice.
In radio applications, frequency translation uses a circuit block known as a 'mixer'. The mixing circuit is made to operate in a non-linear (but defined) manner. Take as an example the single diode commutating, or switching, mixer (because it is notionally the easiest to analyse). The incoming signal is either allowed to pass unhindered, or not at all, through the circuit (ie, the ultimate in changed or non-linear operation), depending on the state of a control input. This control can be considered to be either fully 'on' or fully 'off', with no intermediate state. Now if the control input is toggled at a steady rate, an interesting effect (amonst others) is noticed at the mixer output. The spectrum not only contains the original signal, and the much larger, non linear inducing, toggling frequency, but some new products also, as the following drawing shows. Here, to put some values to the example, the input signal frequency is 1MHz and the toggling frequency 10MHz:
The important products are the 1MHz sidebands arround the 10MHz toggle signal (ie, at 9 and 11MHz)
[because the toggle signal is not a pure sine wave but a square wave, it contains an infinite number of harmonics, and a number of these - together with their 1MHz sidebands are also present and shown above. They are unwanted products, but can usually be filtered out easily enough. Although its another story (but an interesting one, so worth briefly telling(!)), there are no effects shown above due to harmonics of the 1MHz input signal, which can usually accurately be regarded as a pure sine wave. If a mixer is overloaded by an input signal however, harmonics of the 1MHz signal will be generated within the mixer itself, and these will appear as additional sidebands arround the toggle frequency and its harmonics. This is clearly not a very nice situation - pretty soon you end up with components at every MHz, all the way across the spectrum. It is best not to dwell on this situation, which perhaps is akin to a loaf of bread being baked at 700 degrees C!]
Returning to the frequency translation example right at the start, it can be seen that the right hand picture is exactly what would happen if a mixer were used with a toggle frequency (usually refered to as the Local Oscillator) of 1420.000MHz. There is a slight problem, however, with this direct translation of the 1420MHz signal down to the origin, and it is this. An adjacent window extending downwards from 1420.000MHz to 1419.970MHz will also translate to the origin. Intuition suggests this from the mixer output plot above - if you reverse the input and output frequencies it is fairly clear that a signal at either 9 OR 11MHz causes a signal to be generated at 1MHz.
It could be argued that this 'image' effect doesn't really matter - after all, the PC soundcard being fed from the mixer will now respond to frequencies from 1419.970 to 1420.030MHz or double the intended window width. There is a problem however. Even if there is no intelligent signal on the image side, there will always be thermal noise - and it is thermal noise that ultimately limits the sensitivity of a receiver. So now, whatever bin width the PC software is looking at, the receivers thermal noise will be double (ie, 3dB worse) what it need be. There are several ways round this problem:
a) Ignore it! - and accept the 3dB degradation.
b) Use a very sharp filter to select 1420.000-1420.030 and reject 1419.970-1420.00MHz - but no such filter exists for such an example
c) Use a phasing arrangement that nulls out the image whilst enhancing the wanted side - this is what the 'R2' receiver does
d) Use more than one mixing process so that the filtering method of b) IS practical
The latter solution is the one favored in this series of article.
Multiple mixer solution
Most radio deficiencies are ultimately resolved by adding filter circuits to attenuate unwanted responses. This may sound like a 'bodge' solution, but then this is exactly what it is.
Frequency filtering can only work if the wanted and unwanted frequencies have some reasonable degree of frequency separation. This is why b) above was not practical - the wanted window butted up right against the unwanted one, giving a filter nothing to work with. The answer to this problem is simply not to translate all the way back to the zero origin - at least, not all in one go. Infact, the main article involves three translation stages, the first of which is the 1420/144MHz converter. As the following two drawings show, the image response moves from being adjacent the wanted response to being a huge 288MHz away - easy for uncomplicated filtering to deal with
Now the question arises "Doesn't the same problem exist at 144MHz as it did at 1420MHz?". Whilst there are some Surface Acoustic Filters (SAWs) that could be used directly at 144MHz, the answer is still really "Yes!". In any case, such a filter would be fixed at one frequency, and as the main article goes on to explain, it is useful to have the receiver capable of changing channel to increase the effective window width being monitored. Another more obvious point, is that the article intended to show how surplus radio equipment could be re-used for SETI work, and this equipment will certainly down-convert at least one more time.
Eventually, the window does have to be moved to the origin, so that the frequencies are low enough for the PCs soundcard to deal with, so a filter with a sharp adjacent window rejection is needed. A crystal filter can do this - most surplus receivers will have one of these operating at 10.7MHz or 21.4MHz ( these are discussed later in the main article) so the 144MHz signal will be converted to that. Again, the image response (at 101.2MHz, for a 21.4MHz filter) is filtered out.
The final conversion (to the origin) is described in the main article
Return to main article