[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: [linrad] OP amps and antialiasing was...a simple front end

Hi All,

> Here are Leif's comments on the Delta44 anti-aliasing filter, from:  
> http://www.antennspecialisten.com/~sm5bsz/linuxdsp/filter.htm
> ----start quote
> The builtin filter of Delta44
> The -3dB point of the builtin filter is at 46.3kHz and the 
> attenuation is only 10dB
> at the Nyquist frequency 48kHz. Adequate attenuation, -60dB is 
> obtained at 52kHz
> making the useful range 0 to 44kHz. 
> Low pass filter with notch
> Since the builtin filter is perfectly ok above 53 kHz the 
> additional filter does not
> have to attenuate above this frequency. To get a very steep 
> filter around 48kHz
> notches are used, see fig.1. Having the components at hand, the 
> easiest way of
> producing notches is to use LC series links. Filter design in the 
> audio range is
> conventional engineering and many other ways are possible to 
> realise a filter with
> notches slightly above 48kHz. The useful frequency range is 
> increased by about 2kHz
> by this filter and it is certainly questionable if it is worth 
> the trouble of adding
> it to increase the total bandwidth from 88 to 92kHz (4.5%) 
> -----end quote
> So you can do reasonably well with a Delta44 without additional 
> filtering, but do get
> a bit of extra bandwidth by adding external antialias filtering.
This is "true" with a mediocre mixer. It is not quite true though 
because one has to make assumptions about how many strong signals 
there are outside the desired passband.

With a really good mixer the above is not true at all, look at
the details of the RX2500 filter chain where a lot of money
is spent on the filters.
The AD797 is sensitive to HF. Very low levels of LO, LO overtones
or RF signals causes degradation of the excellent linearity.
A first filter is needed to protect the AD797, otherwise the
intermodulation characteristics of the op-amp will suffer.

The Delta44 has a simple analog lowpass filter at the input. 
It does not attenuate much below 1 MHz so if the signals within 
+/- 1 MHz add upp in voltage to saturate the A/D, there will be 
strong intermodulation due to saturation but it will not be visible
as saturation in the 96kHz output data stream because sampling is 
actually done at 6.144 MHz and the chip contains a digital 
anti-alias filter that removes anything above 50kHz or so. 

> So, again, start simple and add on as you go.  This makes getting 
> everything working
> easier and less intimidating, and you get to see the benefits of 
> the added complexity
> as you create it.  And when a problem crops up, you have a pretty 
> good idea of where
> it is.  I find that fact particularly helpful!
I agree very strongly with this. You are likely to find that the 
weakest link is the mixer. Then with better low noise amplifiers
and mixers you will reach the level where the Delta44 is limiting
your system. When you want to go beyond that you will find that
complexity grows rapidly....

If you make a simple X-tal filter at RF or IF, all these 
problems disappear but you loose bandwidth;)

If there were not so many more interesting things to do I should
rewrite a lot of elderly things on the Linrad site. The anti-alias
filter above is designed to work with a poor mixer because the high
noise floor of the low cost op-amps forces the mixer to run at a 
rather high level. It is not clear enough that conclusions made on the
page are for this amplifier/filter when used with the 74HC4052
mixers described together with it.

Another way to say it: The amplifier is the weak link. Improving
the noise floor of the amplifier is more important than anything 
else but the notches give a few kHz extra also if the poor amplifiers 
are used.


Leif  /  SM5BSZ