Been testing some RF amplifiers – and came across something interesting …
Is there a ‘normalization’ noise floor??What do I mean by the above term?
Well, when you normalize the signal path when using a TG – you expect any ‘anomalies’ to be eliminated – and the resultant response should be FLAT (or as close to flat as possible)
When using a default SSA and we simply connect the TG output to the RF input AND normalize – we get a relatively FLAT response –
with NO NOISE – something you expect after naomalization.
I have a relatively unusual set-up
– with 50dB attenuation in my signal path (unfortunately required when working with high power RF).
When performing a default TG normalization
– with my 50dB attenuator in the signal path
– I got LOTS OF NOISE in my normalization – not a straight ‘flat’ response as one would expect.
This made me wonder is there is a ‘normalization’ noise floor??I guess when doing a normalization – you need some raw numerical data to perform the normalization math function.
When performing math functions in software
– you need to work with either interges or floating point numbers.
If these ‘numbers’ are extremely small
– because of the massive attenuation in the signal path
– then there might be a problem if extended numerical precision is not used in the math calculations.
Remember I have the following attenuation in the signal path …
TG default setting -20dBm
RF attenuation set at -20 db
My own attenuation - 50db
So I have -
70dB of pure attenuation and a small signal from the TG (-20dBm)
This means that the ‘data’ we are trying to normalize
– may not be possible to fully quantize
– because the SSA simply cannot collect sufficiently accurate (enough precision) information to be successfully used in the normalization math function.
Anyway – I hope you are able to see what I mean by normalization ‘noise floor’Personally I never came across such a term
– and
one would expect the signal to be normalized – no matter what.If the data to be normalized is so ‘small’
(due to the high attenuation in the signal path)
then rather to work with extremely small numbers which the software cannot handle
– why not simply make it ‘ideal’ and zero the result anyway
– since we have such a high attenuation
– instead of trying to calculate and result is massive noise floor??
What do you think?
Is there such a thing as a normalization noise floor?
Or is today simply a start of a bad day for me – and I am doing something stupid
I have attached some screen shots
Screen shot 1 TG and RF both at their default values – with my 50dB attenuator in the signal path.
Here you can clearly see the normalization noise – it is not a flat response!
Screen shot 2 Both TG and RF (built-in) attenuation set at 0 (zero) – and still the 50dB attenuator in the signal path – after normalization – got ‘flat’ response – as expected – with little or no visual noise.
Screen Shot 3 Here I removed the -20dB build-in attenuation and set the TG to -10dBm instead of the default -20dBm. I still have the 50 dB attenuation in the signal path. It can be seen clearly that the ‘noise’ level in the normalized result is still there but much smaller – confirming that we are heading in the correct direction by reducing the overall attenuation in the signal path.