[Linrad] Re: Progress

[Dave Sublette <arrl.net; k4to@xxxxxxxxxxxxxxxx>]

Leif Asbrink wrote:
> Hi Dave,
> 
>> I have added 10 dB of amplification to the direct input of the SDR-14 
>> for a total of 25 dB now. When I disconnect the 144 mhz signal from the 
>> input to the SDR-14, the main spectrum base line now drops 20 dB, I 
>> believe this tells me that I have enough gain ahead of the SDR-14 to 
>> have proper operation.
> Hmmm, not quite. What happens when you switch off the mast mounted
> preamp? (The 10 dB amplifier could be a noise generator.) You can
> not be sure on the whole system until you verify that the noise you
> hear is dominated by the antenna and first amplifier.
> 
You are correct. When I switch off the mast mounted stuff, the base line 
of the main spectrum display still only drops 5-7 dB, indicating that I 
don't have enough gain at the top.

> There is of course no guarantee that all is ok by this test, but
> as long as you trust the preamp to have a low noise figure, you
> can be syre.

Unless I have a failure I tested the nf as 1.5 dB and the gain is 25 dB 
ftom the intput of the preamp to the output of the transverter at 144 
mhz. I ave an HP8970A automatic noise figure meter.
> 
>> I experimented with RCF shift and found that a value of 4 gives a base 
>> line of +20 dB in the main spectrum display. If I read everything 
>> correctly, I believe this is the desired level.
> On the linrad home page:
> http://www.sm5bsz.com/linuxdsp/linrad.htm
> there is a section: "Dynamic range considerations"
> with a link to
> http://www.sm5bsz.com/linuxdsp/install/dlevel.htm
> 
> Since MAP65-IQ expects 16 bit data you have to run the backwards
> FFT (from fft1 to timf2) with 16 bit arithmetics.
> Rounding happens at the fft1 output and you have to set the
> signal level for the noise floor in each bin of fft1 to be well
> above the quantization noise that the rounding from float to integer
> leads to. Press 'A' and read the value after "Floor" it is given
> in bits at the fft1 output. Never allow floor to go below 3.
> (That corresponds to a level of about 13 dB in the main spectrum.)

I am not using the MAP65-IQ program. I am in the Weak CW mode of LINRAD. 
I don't how the backward FFT relates to WCW mode. Right now, backward 
FFT is not selected.
> 
> Under normal circumstances, place "Floor" between 6 and 9.
> (It is in units of bits in linear scale) The noise floor in the 
> main spectrum will then be around 20 dB.
> 
> Note that the SDR-IQ is a 16 bit device. The output shift it does
> is a rounding and you have to be sure that the associated quantization
> noise is very small.

I am using SDR-14, so I don't know how SDR-IQ relates here.
> 
> When you find 20 dB noise floor with an output shift of
> 4 all is ok on the input side of fft1 (the SDR-IQ output)
> ONLY if the first FFT amplitude is 1000 (or more)
> 
> In case you increase the output shift by 2 units and then
> compensate by making the first FFT gain 4000 you will still
> see the noise floor at 20 dB and a good "Floor" value at
> the output of fft1, but you would have lost 2 bits at the
> input and then the quantization noise caused by the output
> shift in the SDR-IQ would no longer be negligible.
> 
I am understanding more of the concept of optimizing range and 
quantization error, but until I get my gain at the top fixed, I can't 
really worry about it.

> There is one more place where one has to worry about rounding
> errors when using MAP65 and that is at the output of the
> backwards fft1, the timf2 function. The noise floor in timf2 should
> be about 20 dB above 1 bit. That is the red line in the horizontal
> bar in the high resolution graph. There is nothing wrong in setting 
> the noise floor higher as long as timf2 does not saturate.
> (Check by pressing 'A' that there is a margin.)
> 
> The links above give more details.
> 
> 
> 
>> The sensitivity of the LINRAD now appears to be very close to the same 
>> as the K2 receiver that I use to copy by ear. I can see signals on the 
>> waterfall as weak as I can hear them in the K2.
> Yes. I expect you to find that the antenna mounted preamp now
> raises the noise floor by at least 15 dB:-)
> 
>> However, the audio demod from the baseband through the speaker is at 
>> least 10 dB less sensitive than the K2. In other words, I can copy 
>> signals on the K2 by ear and can see them, but they are not audible in 
>> the speaker of the computer, nor is there a visual indication in the 
>> baseband window. I am sure that the BFO tuning is identical on both 
>> because on stronger signals, the CW tone is the same. I use about 450 hz.
>>
>> I think maybe this is due to not having a narrow enough filter setup in 
>> the LINRAD parameters. I am still struggling to understand all of the 
>> various FFT's and multiplying factors.
> OK. You have little vertical yellow bars in the baseband graph to set 
> the filter bandwidth. There is also "First mixer bandwidth reduction"
> that sets the total span of the baseband as a fraction of the input
> bandwidth. 
> 
>> But, progress is progress and I am pleased to have gotten this far.
> Good:-)
> 
> I hope I will read from you soon that you can now copy better with 
> Linrad than with the K3;-)
> (But you will need earphones to use Coh modes.)
> 
> 73
> 
> Leif / SM5BSZ

I am now modeling my system using Agilent's ApCAD and the model doesn't 
show the same problem with gain at the top. I am modeling the SDR-14 
direct input at 33 dB noise figure and 0 gain. Is this reasonable?

Thanks,

Dave
> 
> 
> > 
> 

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups "Linrad" group.
To post to this group, send email to linrad@xxxxxxxxxxxxxxxx
To unsubscribe from this group, send email to linrad+unsubscribe@xxxxxxxxxxxxxxxx
For more options, visit this group at http://groups.google.com/group/linrad?hl=en
-~----------~----~----~----~------~----~------~--~---


LINRADDARNIL