[hpsdr] HPSDR Signal Overload ( In the digital processing???)

[Phil Harman]

I thought that I had adequately covered this in a previous email but it
would seem that I didn't.

The architecture of Mercury is a switchable 20dB attenuator followed by a
20dB RF amplifier that feeds the ADC.

The ideal configuration would be to switch the RF amplifier out of circuit
under software control.  We tried to do this with multiple revisions of
the Mercury PCB (thanks Lyle!) but was unable to economically produce a
design that had an acceptable low level of spurs.

So we had a choice, either

- don't fit an RF amplifier to the board and require the user to fit an
external one. This would have resulted in a Noise Figure of about 30dB
making the receiver useless on the higher bands or where a lossy antenna
(e.g. beverage, flag etc) was used on the lower bands; or
- leave the RF amplifier in circuit at all times and take out the RF
amplifier gain with a switched attenuator under software control.

Rightly or wrongly we choose the latter approach.

Placing an attenuator between the RF amplifier and ADC is the worst of all
worlds since you increase the NF but the amplifier is still subject to
overload due to large signals.

Perhaps we should (have) re-named the Preamp ON/OFF control in HPSDR
version(s) of PowerSDR to be Attenuator 0dB/20dB.

With regard to useful noise figure it really depends on how efficient your
antenna system is.  I find that I don't need the preamp on (err... can
leave the attenuator in circuit) on all bands up to 17m.  If I use a flag
antenna on 80m then I need the preamp on ( attenuator out) and an external
preamp. With a decent antenna a 20/30dB Noise Figure on the lower bands is
quite acceptable.  All switching out the attenuator does is reduce the
dynamic range by 20dB.

A good rule of thumb is that when plug in your antenna if the Bandscope
noise increases by 10dB then your receiver (system) is bandnoise limited
not Noise Figure limited.

A Digital Down Conversion receiver like Mercury has different
characteristics to a conventional analogue receiver. With an analogue
receiver we want to keep strong out of band signals out of the front end
to prevent intermodulation products.

With a DDC receiver strong signals can actually improve the performance.
This is true as long as the instantaneous sum of all the signals don't add
to cause the ADC to overload.  So why are big signals our friends?

The ideal ADC will be completely linear in that what ever signal(s) are
applied to the input are perfectly converted to a digital equivalent.

In practice there will be parts of the ADC curve  that are non-linear,
particularly round the origin where the small signals sit.

If we put a big signal and a small signal into the ADC then, since its a
linear process, the output will be the sum of the two signals.  You can
think of this as the small signal riding piggy back on the large signal.
In which case the small signal spends less time around the origin and any
distortion due to nonlinearity is reduced.

As long as the big signal(s) do not instaneously sum so as to overload the
ADC then they cause no problems.

So does this mean that you don't need any filtering prior to Mercury? 
Well it depends on the level of in and out of bands signals Mercury sees.
In my own case I have a very strong MW broadcast station very close to my
QTH that puts in a 0dBm signal into the front end. On HF the natural
selectivity of my antenna, ATU etc reduce this to a level that does not
cause the ADC to overload.  However, on 80m I need a HPF in circuit to
prevent the ADC overloading.

Hence there is no universal solution. If you don't have an overload
problem then you don't need any external filters. If you still fit
external filters then they are not going to do any harm.

I'll respond to the user who is seeing an occasional increase in his noise
floor in a future email.

73 Phil...VK6APH







 1302139752.0