Date   

Re: Transmission lines, Input match and Connectors #bestpractice #experiment

David Eckhardt
 

Most of the newer professional signal generators bottom out at -130 dBm.  The 608s were microphonic as the dickens and drifted like crazy.  They were the best we had at the time.  I use good (HP) attenuators if I require levels less than -130 dBm.  Professional grade Attenuators are necessary and required equipment for any RF lab.  I really do not understand the bias against using them other than adding another error bar to a measurement where they are used.

Dave - WØLEV

Dave - WØLEV


On Fri, Oct 18, 2019 at 11:22 PM jdow <jdow@...> wrote:
Si. Of course, if I did not mind a lot of up front attenuation a waveguide
beyond cutoff attenuator as in the venerable HP-608 series would be REALLY
interesting. They work surprisingly well. (I still own a 608F. It's fun when you
develop a transceiver that you can feed a -109 dBm signal and not be able to jam
it with the 608F fed through a 20 dB coupler. That's more than 1mW of jammer at
the radio input. I reversed the coupler and made the transceiver collapse when
the input got high enough to cause gain compression. Hedy Lamar had a really
good idea. It works.)

{^_-}

On 20191018 15:46:55, Dana Myers wrote:
> On 10/18/2019 3:38 PM, jdow wrote:
>> I haven't quantized it. 20 or 30 dB would probably be close. (Airgaps can work
>> as attenuators if you are REALLY delicate creating the tiny gap. Just don't
>> ask for precision. {^_-})
>
> "Really small" air-gaps are *always* capacitors, right?
>
> 73,
> Dana  K6JQ
>
>
>
>





--
Dave - WØLEV
Just Let Darwin Work
Just Think


Re: Transmission lines, Input match and Connectors #bestpractice #experiment

David Eckhardt
 

We are operating n a high RF environment with a 'tower from hell' cell installation.  Ask why there?  We desire the installation including the optical side (18 and 24" optics) available to  the students at the HS and the community of Berthoud, Co. and beyond.   At the time we were trying to answer where and if in the receiver system, gain compression was occurring.   

My own installation is in a valley and is relatively RF quiet (except for my own computers and laptops and new appliances......ggggrrrrr).  My strongest FM brtadcast station measures roughly -40 dBm on an FM antenna in the radio room window and I still need an FM stopband filter on any of the SDRs. 

Dave - WØLEV


On Fri, Oct 18, 2019 at 10:03 PM Marcus D. Leech <patchvonbraun@...> wrote:
On 10/18/2019 03:55 PM, David Eckhardt wrote:
We have fed the antenna port with broadband noise well below the 12-bit limit and well over the 12-bit limit for saturation.  Even with the low input we see raw I/Q values in excess of what should be spit out with a 12-bit system.  With the well into saturation, we see many more raw I/Q values well in excess of saturation for a 12-bit system.  These are raw I/Q values grabbed on the USB output.  We discovered this some 1.5 years ago when we were attempting to put the 'nominal' output from our H1 telescope no more than 3 to 5 bits above the bottom of the bit stack.  We asked more questions that we were able to answer, as a result, and are still scratching our heads as to how this can occur.  We were seeing values upwards of roughly 2X that of the +/- 33k values (for a 16-bit system).  If its a true 12-bit system without attempting to internally adjust the position of the signal within the bit stack in the airspy FW, how can this occur?   

Dave - WØLEV
I use a pair of AirSpy R2 for our hydrogen spectrometer, and I've never seen any hint of unexpectedly-large values coming out of it, although
  I don't use anything close to a "raw" interface, but use gr-osmosdr and Gnu Radio.

Do your results show a gain dependency?



On Fri, Oct 18, 2019 at 7:23 PM prog <info@...> wrote:
On Fri, Oct 18, 2019 at 09:14 PM, David Eckhardt wrote:
It appears it is not a true 12-bit system as we are getting raw I/Q values well in excess of that permitted by a strictly 12-bit system.
The output is scaled to fill a 0 dBFS unit. It can be either float (-1 .. +1) or a 16bit signed integer (-32768 .. +32767). You also have some DSP running in the library to convert the RAW samples into IQ.
Speaking of the RAW samples, the 12bit data is stuffed in the LSB bits of the unsigned 16bit samples and no further processing is done. You can read the code of the library for the details.

Not sure what you are doing, but I recommend you check your math. You can't get more resolution than what the device is giving.


--
Dave - WØLEV
Just Let Darwin Work
Just Think



--
Dave - WØLEV
Just Let Darwin Work
Just Think


Re: Transmission lines, Input match and Connectors #bestpractice #experiment

jdow
 

Si. Of course, if I did not mind a lot of up front attenuation a waveguide beyond cutoff attenuator as in the venerable HP-608 series would be REALLY interesting. They work surprisingly well. (I still own a 608F. It's fun when you develop a transceiver that you can feed a -109 dBm signal and not be able to jam it with the 608F fed through a 20 dB coupler. That's more than 1mW of jammer at the radio input. I reversed the coupler and made the transceiver collapse when the input got high enough to cause gain compression. Hedy Lamar had a really good idea. It works.)

{^_-}

On 20191018 15:46:55, Dana Myers wrote:
On 10/18/2019 3:38 PM, jdow wrote:
I haven't quantized it. 20 or 30 dB would probably be close. (Airgaps can work as attenuators if you are REALLY delicate creating the tiny gap. Just don't ask for precision. {^_-})
"Really small" air-gaps are *always* capacitors, right?
73,
Dana  K6JQ


Re: Transmission lines, Input match and Connectors #bestpractice #experiment

Dana Myers
 

On 10/18/2019 3:38 PM, jdow wrote:
I haven't quantized it. 20 or 30 dB would probably be close. (Airgaps can work as attenuators if you are REALLY delicate creating the tiny gap. Just don't ask for precision. {^_-})
"Really small" air-gaps are *always* capacitors, right?

73,
Dana  K6JQ


Re: Transmission lines, Input match and Connectors #bestpractice #experiment

prog
 

On Sat, Oct 19, 2019 at 12:40 AM, jdow wrote:
I've not taken the time to nail it down.
Please do. This may help future developments.


Re: Transmission lines, Input match and Connectors #bestpractice #experiment

ebrombaugh1@...
 

"This is the main reason I do not take anyone's tests on their face value. Understanding the phenomenons involved in the "bad" results and being able to reproduce them is proper engineering. The rest is, well... just noise."

Yes - when I'm taking feedback from beta testers on my product development projects I don't classify something as a real bug until I can reproduce it under conditions that I control. Before that it's merely an interesting anomaly that needs further study.


Re: Transmission lines, Input match and Connectors #bestpractice #experiment

jdow
 

I haven't quantized it. 20 or 30 dB would probably be close. (Airgaps can work as attenuators if you are REALLY delicate creating the tiny gap. Just don't ask for precision. {^_-}) I've never bought myself a suitable RF attenuator. The ones I've seen I've not been inclined to purchase. It's WAY too easy to destroy their accuracy with a little too much power put into them. I don't do 160 listening much these days so I have not obsessed about it. (My ProII's display backlight is out. SIGH! And if I can't TX why RX on that band?)

At the moment I am running with a 6 dB SMA attenuator in line. With that attenuation I've also noticed some apparent IMD artifacts as high as 6 MHz with two broadcast stations beating with each other. Since it's near 4th harmonic of the really big signal I figure it could be transmitted harmonics, too. But it sounds like I hear audio from two stations mixed. I've not taken the time to nail it down. It was quite weak. 75 meters seems to be free of the artifacts, at least with no tuner on the long wire (1/2 wave end fed Zepp for mid 75 meters.)

{^_^}

On 20191018 14:44:32, prog wrote:
On Fri, Oct 18, 2019 at 11:40 PM, jdow wrote:
While you are on the topic - is the AGC correction developed before or after
any tuneable front end filtering Discovery may have? Symptoms I see here
suggest that it is taken after the filtering. (160 meters with a very strong
AM station at about 1.5 MHz out of the receiver band.) Adding attenuation
mitigates the problem for the most part.
{^_^}
The AGC was always there. Now running with slightly different parameters.
The power detector for the RF AGC loop is located after the filters. How much attenuation is needed in your case?


Re: Transmission lines, Input match and Connectors #bestpractice #experiment

Marcus D. Leech
 

On 10/18/2019 03:55 PM, David Eckhardt wrote:
We have fed the antenna port with broadband noise well below the 12-bit limit and well over the 12-bit limit for saturation.  Even with the low input we see raw I/Q values in excess of what should be spit out with a 12-bit system.  With the well into saturation, we see many more raw I/Q values well in excess of saturation for a 12-bit system.  These are raw I/Q values grabbed on the USB output.  We discovered this some 1.5 years ago when we were attempting to put the 'nominal' output from our H1 telescope no more than 3 to 5 bits above the bottom of the bit stack.  We asked more questions that we were able to answer, as a result, and are still scratching our heads as to how this can occur.  We were seeing values upwards of roughly 2X that of the +/- 33k values (for a 16-bit system).  If its a true 12-bit system without attempting to internally adjust the position of the signal within the bit stack in the airspy FW, how can this occur?   

Dave - WØLEV
I use a pair of AirSpy R2 for our hydrogen spectrometer, and I've never seen any hint of unexpectedly-large values coming out of it, although
  I don't use anything close to a "raw" interface, but use gr-osmosdr and Gnu Radio.

Do your results show a gain dependency?



On Fri, Oct 18, 2019 at 7:23 PM prog <info@...> wrote:
On Fri, Oct 18, 2019 at 09:14 PM, David Eckhardt wrote:
It appears it is not a true 12-bit system as we are getting raw I/Q values well in excess of that permitted by a strictly 12-bit system.
The output is scaled to fill a 0 dBFS unit. It can be either float (-1 .. +1) or a 16bit signed integer (-32768 .. +32767). You also have some DSP running in the library to convert the RAW samples into IQ.
Speaking of the RAW samples, the 12bit data is stuffed in the LSB bits of the unsigned 16bit samples and no further processing is done. You can read the code of the library for the details.

Not sure what you are doing, but I recommend you check your math. You can't get more resolution than what the device is giving.


--
Dave - WØLEV
Just Let Darwin Work
Just Think


Re: Transmission lines, Input match and Connectors #bestpractice #experiment

jdow
 

Can you rephrase that? It left me a little unclear over what you are seeing. With AGC on I'd expect values up to full scale. But, I would expect those values I see to have steps of about 8 on the 16 bit integer values transmitted if no internal decimation is used on board the HF+. (I'd expect the equivalent in floating point mode.) So it might be helpful to describe what mode you are in and what the problem is. If it's in 16 bit integer mode (you cannot see more than -32768 to +32767 so your description REALLY confuses me o that basis) you should not see values that vary the LSBs. BTW - where are you seeing those values? My natural conclusion was that you were monitoring the digital stream coming out of libusb directly. If you are reading data out of SDRSharp that is another story. You have a lot of digital processing which can push the dynamic range beyond the bare 12 bits.

{^_^}

On 20191018 12:55:10, David Eckhardt wrote:
We have fed the antenna port with broadband noise well below the 12-bit limit and well over the 12-bit limit for saturation.  Even with the low input we see raw I/Q values in excess of what should be spit out with a 12-bit system.  With the well into saturation, we see many more raw I/Q values well in excess of saturation for a 12-bit system.  These are raw I/Q values grabbed on the USB output.  We discovered this some 1.5 years ago when we were attempting to put the 'nominal' output from our H1 telescope no more than 3 to 5 bits above the bottom of the bit stack.  We asked more questions that we were able to answer, as a result, and are still scratching our heads as to how this can occur.  We were seeing values upwards of roughly 2X that of the +/- 33k values (for a 16-bit system).  If its a true 12-bit system without attempting to internally adjust the position of the signal within the bit stack in the airspy FW, how can this occur?
Dave - WØLEV
On Fri, Oct 18, 2019 at 7:23 PM prog <@prog <mailto:@prog>> wrote:
On Fri, Oct 18, 2019 at 09:14 PM, David Eckhardt wrote:
It appears it is not a true 12-bit system as we are getting raw I/Q
values well in excess of that permitted by a strictly 12-bit system.
The output is scaled to fill a 0 dBFS unit. It can be either float (-1 ..
+1) or a 16bit signed integer (-32768 .. +32767). You also have some DSP
running in the library to convert the RAW samples into IQ.
Speaking of the RAW samples, the 12bit data is stuffed in the LSB bits of
the unsigned 16bit samples and no further processing is done. You can read
the code of the library for the details.
Not sure what you are doing, but I recommend you check your math. You can't
get more resolution than what the device is giving.
--
*Dave - WØLEV
*
*/Just Let Darwin Work/*
*/Just Think/*


Re: Transmission lines, Input match and Connectors #bestpractice #experiment

prog
 

On Fri, Oct 18, 2019 at 11:40 PM, jdow wrote:
While you are on the topic - is the AGC correction developed before or after any tuneable front end filtering Discovery may have? Symptoms I see here suggest that it is taken after the filtering. (160 meters with a very strong AM station at about 1.5 MHz out of the receiver band.) Adding attenuation mitigates the problem for the most part.

{^_^}
The AGC was always there. Now running with slightly different parameters.
The power detector for the RF AGC loop is located after the filters. How much attenuation is needed in your case?


Re: Transmission lines, Input match and Connectors #bestpractice #experiment

jdow
 

While you are on the topic - is the AGC correction developed before or after any tuneable front end filtering Discovery may have? Symptoms I see here suggest that it is taken after the filtering. (160 meters with a very strong AM station at about 1.5 MHz out of the receiver band.) Adding attenuation mitigates the problem for the most part.

{^_^}

On 20191018 10:52:01, prog wrote:
The internal IF AGC of the ADC will not allow any digital saturation to happen. It will silently adjust the level at the input of the ADC and keep the signals below 0 dBFS. The same applies to the RF AGC chain plus or minus the threshold level. This excludes many of the assumptions stated in the previous emails.
What could happen, however, is all these non-linearity problems that could be introduced by bad contacts (mostly in connectors), bad cables, and even rusty metallic structures surrounding the antenna that could mix signals and translate them into unexpected frequencies. Here's an example from the recent memory: https://www.pressreader.com/canada/windsor-star/20191009/281517932872802
I have seen this happen too many times to ignore it. Many experienced ops overlook these details. Even Leif found some noise humps (/similar to those we see in these screenshots which are unaffected by the attenuation level/) when testing the original HF+ and he traced them back to a faulty (leaky) attenuator. Once he replaced the attenuator, things went back to normal. At the same time, I am being realistic by not expecting everyone to have Leif's experience. Someone else would have just trusted his apparatus and produced a bogus appreciation.
This is the main reason I do not take anyone's tests on their face value. Understanding the phenomenons involved in the "bad" results and being able to reproduce them is proper engineering. The rest is, well... just noise.
</rant>


Re: Transmission lines, Input match and Connectors #bestpractice #experiment

jdow
 

(warning - smartass reply) Luke was told to use the force. You are being told to use the AGC. (I believe it has a readback which can be used to provide at least roughly calibrated displays on software which uses AGC.) The AGC will keep the strongest signal near but below the top of the display giving you the best usable instantaneous dynamic range the device can deliver. (This is still not enough dynamic range to recover a week signal when your buddy fires up 100 Watts on the voice portion of the band when you are in the CW portion. Heck, the AGC might not even handle that case gracefully at all. But it will try. Suggest your buddy go for QRP points and it'll probably work out nicely.)

{^_-}

On 20191018 10:27:38, David Eckhardt wrote:
For the results of the tests presented:  The tests and data have nothing to do with input filtering or very little with input match.  Everything within each data set is within the same band.   Of course, any input filtering will pass all signals close to the target signal.
Another point seen especially in data sets Airspy7_0dB.jpg and to a lesser extent in data set Airspy 9_0dB.jpg:  At issue may be digital saturation.  In an SDR when all available bits are utilized as with possible reception of a strong signal, the output becomes garbage and unpredictable.  This is highly evident in the Airspy7_0dB.jpg.  Attenuation between the antenna feedline and the input to the receiver, in this case, is mandatory.  I wrote an article some 4 or five years ago in the Journal of the Society of Amateur Radio Astronomers (SARA) addressing this issue for reception around 20 MHz (when solar conditions were pretty much the best of Cycle 24).
The issue of 'bogus' signals in the FM broadcast band:  This is common in nearly all receivers be they the older technology or the newer digital receivers like the Airspy and others.  The problem is that within the 88 to 108 MHz FM broadcast band, there are many truly strong signals, especially true in or near large cities.  This is required to minimize the noise and maximize the level of quieting in detection of a wideband signal.  The input stage(s) (front end) acts as a mixer as its driven into unintended non-linearities by all the strong stations within that 20 MHz of spectrum.  The strong stations mix with each other to produce bogus or false signals.  Again, too much signal requires either a stopband filter (if you're not interested in receiving FM), or insertion of an appropriate amount of attenuation between the antenna feedline and the input of the receiver.  Any input passband filter present for the FM band, specifically, will aim at passing the whole band on to the mixer or ADC.  Again, an attenuator is required due to too much in-band strong signal levels.
The suggestion has been made to understand how receivers are tested and the basics of their internal design.  The information is on the Internet, so you don't have to visit a library.  For proper evaluation of the newer SDR technology, to properly design tests and evaluations of the item, a knowledge far deeper than just the basics of the design will be required.
In conclusion, your measurements and data do not address input filtering and/or input match.  To properly conduct the tests you desire, you must first understand what you are evaluating.
Dave - WØLEV
<https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail&utm_term=icon> Virus-free. www.avast.com <https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail&utm_term=link> <#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
On Fri, Oct 18, 2019 at 5:07 AM jdow <jdow@... <mailto:jdow@...>> wrote:
Please look at "https://en.wikipedia.org/wiki/Radio_noise". It is a simple
discussion. Then explain to me why you insist on NOT using an attenuator at MW
as a matter of course? If you had a SMALL antenna like one of those telescoping
whips the low noise figure of the HF+ would be really good for you. But with an
antenna such as you describe in an environment with a "typical" level of man
made noise a 30 dB attenuator would do you no ill and probably a 40 dB
attenuator would do you no ill. Look at Wikipiddle er Wikipedia for a perhaps
over-simplified discussion of noise figure to see how it works.
Now, knowing the input impedance of the HF+ input may help you optimize a
filter
to reduce signals in the MW broadcast band when you are working elsewhere. But
that would probably be for one tuning setting. The input impedance may vary
with
the frequency to which HF+ is tuned.
My purpose in trying to drive this home is to help you optimize your time and
efforts to best effect. I am trying to boost your game not deride you. Too many
years in a very male oriented occupation has left me a tad abrasive. Please
forgive.
{o.o}
On 20191017 10:23:46, Wes Stewart via Groups.Io wrote:
> On Thu, Oct 17, 2019 at 07:45 AM, prog wrote:
>
>     Let's be more pragmatic. What is the real world performance using an
antenna?
>
> Thank you for asking.  Let me preface the following by explaining the
> environment.  The antenna is a 55 foot vertical over ground radials,  For
the
> results that follow, the antenna is connected to the Airspy through a
Telonic
> Industries Model TG-950 stepped attenuator.  Although not a traceable
device I
> have verified its performance with my DG8SAQ VNWA. I have several plots, all
> taken during the same session, which are screen grabs of the SDR#
program.  The
> model, serial number, firmware and software versions are all obvious.
This unit
> is as received a few days ago.
>
> I'm not sure I can intersperse comments between the attachments so I will
> summarize them here.  The image "Airspy_1_40db" is of a local MW station
with
> the inline attenuator set to 40dB.  The image  "Airspy_2_20db" is the
same as
> before except the attenuation has been reduced to 20 dB.  It was
pointless to go
> to 0 dB.  Images "Airspy_3_40dB" and "Airspy_4_20db" are similar but of a
> different station.
>
> The next set, "Airspy_5_20db", "Airspy_6_10db" and "Airspy_7_0db" show
station
> WWV being received on 10 MHz with 20, 10 and 0 dB attenuation respectively.
> Note that with 0 dB attenuation, the Airspy is essentially useless.  The
next
> set,  "Airspy_8_10db" and "Airspy_9_0db" show station WWV being received
on 15
> MHz with 10 and 0 dB attenuation respectively.  Same situation.  Now you
might
> understand my interest in knowing the performance of the input filtering
because
> obviously it isn't up to the task.
>
> But it gets worse.  Image "Airspy_10_Spurious_of_88_1_0db" shows an FM
broadcast
> signal, fully readable so I could identify it, that doesn't exist.  Image
> "Airspy_11_Spurious_of_88_1_20db" shows the same spectrum with 20 dB of
> attenuation inline. The non-existent station disappears.
>
>
> AIrspy_1_40db.jpg
>
>
> AIrspy_2_20db.jpg
>
>
> Airspy_11_Spurious_of_88_1_20db.jpg
>
>
> Airspy_3_40db.jpg
>
>
> Airspy_4_20db.jpg
>
>
> Airspy_5_20db.jpg
>
>
> Airspy_6_10db.jpg
>
>
> Airspy_7_0db.jpg
>
>
> Airspy_8_10db.jpg
>
>
> Airspy_9_0db.jpg
>
>
> Airspy_10_Spurious_of_88_1_0db.jpg
>
--
*Dave - WØLEV
*
*/Just Let Darwin Work/*
*/Just Think/*


Airspy R0/R2/Mini Sample scaling

prog
 
Edited

On Fri, Oct 18, 2019 at 10:01 PM, David Eckhardt wrote:
We have fed the antenna port with broadband noise well below the 12-bit limit and well over the 12-bit limit for saturation.  Even with the low input we see raw I/Q values in excess of what should be spit out with a 12-bit system.  With the well into saturation, we see many more raw I/Q values well in excess of saturation for a 12-bit system.  These are raw I/Q values grabbed on the USB output.  We discovered this some 1.5 years ago when we were attempting to put the 'nominal' output from our H1 telescope no more than 3 to 5 bits above the bottom of the bit stack.  We asked more questions that we were able to answer, as a result, and are still scratching our heads as to how this can occur.  We were seeing values upwards of roughly 2X that of the +/- 33k values (for a 16-bit system).  If its a true 12-bit system without attempting to internally adjust the position of the signal within the bit stack in the airspy FW, how can this occur?   
 
Dave - WØLEV
toggle quoted messageShow quoted text

 


On Fri, Oct 18, 2019 at 7:23 PM prog <info@...> wrote:
On Fri, Oct 18, 2019 at 09:14 PM, David Eckhardt wrote:
It appears it is not a true 12-bit system as we are getting raw I/Q values well in excess of that permitted by a strictly 12-bit system.
The output is scaled to fill a 0 dBFS unit. It can be either float (-1 .. +1) or a 16bit signed integer (-32768 .. +32767). You also have some DSP running in the library to convert the RAW samples into IQ.
Speaking of the RAW samples, the 12bit data is stuffed in the LSB bits of the unsigned 16bit samples and no further processing is done. You can read the code of the library for the details.

Not sure what you are doing, but I recommend you check your math. You can't get more resolution than what the device is giving.

 

 


--
Dave - WØLEV
Just Let Darwin Work
Just Think
I think this all has to do with the sample scaling. Basically, these lines:
https://github.com/airspy/airspyone_host/blob/master/libairspy/src/airspy.c#L300
https://github.com/airspy/airspyone_host/blob/master/libairspy/src/airspy.c#L312
and
https://github.com/airspy/airspyone_host/blob/master/libairspy/src/airspy.c#L57


Re: Transmission lines, Input match and Connectors #bestpractice #experiment

David Eckhardt
 

We have fed the antenna port with broadband noise well below the 12-bit limit and well over the 12-bit limit for saturation.  Even with the low input we see raw I/Q values in excess of what should be spit out with a 12-bit system.  With the well into saturation, we see many more raw I/Q values well in excess of saturation for a 12-bit system.  These are raw I/Q values grabbed on the USB output.  We discovered this some 1.5 years ago when we were attempting to put the 'nominal' output from our H1 telescope no more than 3 to 5 bits above the bottom of the bit stack.  We asked more questions that we were able to answer, as a result, and are still scratching our heads as to how this can occur.  We were seeing values upwards of roughly 2X that of the +/- 33k values (for a 16-bit system).  If its a true 12-bit system without attempting to internally adjust the position of the signal within the bit stack in the airspy FW, how can this occur?   

Dave - WØLEV


On Fri, Oct 18, 2019 at 7:23 PM prog <info@...> wrote:
On Fri, Oct 18, 2019 at 09:14 PM, David Eckhardt wrote:
It appears it is not a true 12-bit system as we are getting raw I/Q values well in excess of that permitted by a strictly 12-bit system.
The output is scaled to fill a 0 dBFS unit. It can be either float (-1 .. +1) or a 16bit signed integer (-32768 .. +32767). You also have some DSP running in the library to convert the RAW samples into IQ.
Speaking of the RAW samples, the 12bit data is stuffed in the LSB bits of the unsigned 16bit samples and no further processing is done. You can read the code of the library for the details.

Not sure what you are doing, but I recommend you check your math. You can't get more resolution than what the device is giving.



--
Dave - WØLEV
Just Let Darwin Work
Just Think


Re: New SDR# release r1730 with High DPI support #software #announcements

prog
 

Quick update: I added a "Snap To Peak" feature that helps clicking more precisely on very narrow signals.
You can now use the Control key when hovering a signal with the mouse to snap to it, then you can click.
This feature is compatible with "Snap To Grid". Which means, if you enable a certain step size, the Control + Hover will snap to multiples of the step size.

Download as usual from: https://airspy.com/download


Re: Transmission lines, Input match and Connectors #bestpractice #experiment

prog
 

On Fri, Oct 18, 2019 at 09:14 PM, David Eckhardt wrote:
It appears it is not a true 12-bit system as we are getting raw I/Q values well in excess of that permitted by a strictly 12-bit system.
The output is scaled to fill a 0 dBFS unit. It can be either float (-1 .. +1) or a 16bit signed integer (-32768 .. +32767). You also have some DSP running in the library to convert the RAW samples into IQ.
Speaking of the RAW samples, the 12bit data is stuffed in the LSB bits of the unsigned 16bit samples and no further processing is done. You can read the code of the library for the details.

Not sure what you are doing, but I recommend you check your math. You can't get more resolution than what the device is giving.


Re: Transmission lines, Input match and Connectors #bestpractice #experiment

David Eckhardt
 

My article in the SARA Journal was based on the Airspy 1.  A lot has changed in the interim.

I should comment that every time Terry and I try to "discover" what makes the Airspy tick inside, we're frustrated - completely.  It appears it is not a true 12-bit system as we are getting raw I/Q values well in excess of that permitted by a strictly 12-bit system.  The Airspy is utilized in our H1 line telescope at the Little Thompson Observatory where both Terry and I volunteer.  I would be "rather" useful in this application to take the lid off the little box!!!!!

Dave - WØLEV


On Fri, Oct 18, 2019 at 5:53 PM prog <info@...> wrote:
The internal IF AGC of the ADC will not allow any digital saturation to happen. It will silently adjust the level at the input of the ADC and keep the signals below 0 dBFS. The same applies to the RF AGC chain plus or minus the threshold level. This excludes many of the assumptions stated in the previous emails.
What could happen, however, is all these non-linearity problems that could be introduced by bad contacts (mostly in connectors), bad cables, and even rusty metallic structures surrounding the antenna that could mix signals and translate them into unexpected frequencies. Here's an example from the recent memory: https://www.pressreader.com/canada/windsor-star/20191009/281517932872802
I have seen this happen too many times to ignore it. Many experienced ops overlook these details. Even Leif found some noise humps (similar to those we see in these screenshots which are unaffected by the attenuation level) when testing the original HF+ and he traced them back to a faulty (leaky) attenuator. Once he replaced the attenuator, things went back to normal. At the same time, I am being realistic by not expecting everyone to have Leif's experience. Someone else would have just trusted his apparatus and produced a bogus appreciation.
This is the main reason I do not take anyone's tests on their face value. Understanding the phenomenons involved in the "bad" results and being able to reproduce them is proper engineering. The rest is, well... just noise.
</rant>



--
Dave - WØLEV
Just Let Darwin Work
Just Think


Re: Transmission lines, Input match and Connectors #bestpractice #experiment

prog
 

The internal IF AGC of the ADC will not allow any digital saturation to happen. It will silently adjust the level at the input of the ADC and keep the signals below 0 dBFS. The same applies to the RF AGC chain plus or minus the threshold level. This excludes many of the assumptions stated in the previous emails.
What could happen, however, is all these non-linearity problems that could be introduced by bad contacts (mostly in connectors), bad cables, and even rusty metallic structures surrounding the antenna that could mix signals and translate them into unexpected frequencies. Here's an example from the recent memory: https://www.pressreader.com/canada/windsor-star/20191009/281517932872802
I have seen this happen too many times to ignore it. Many experienced ops overlook these details. Even Leif found some noise humps (similar to those we see in these screenshots which are unaffected by the attenuation level) when testing the original HF+ and he traced them back to a faulty (leaky) attenuator. Once he replaced the attenuator, things went back to normal. At the same time, I am being realistic by not expecting everyone to have Leif's experience. Someone else would have just trusted his apparatus and produced a bogus appreciation.
This is the main reason I do not take anyone's tests on their face value. Understanding the phenomenons involved in the "bad" results and being able to reproduce them is proper engineering. The rest is, well... just noise.
</rant>


Re: Transmission lines, Input match and Connectors #bestpractice #experiment

David Eckhardt
 

For the results of the tests presented:  The tests and data have nothing to do with input filtering or very little with input match.  Everything within each data set is within the same band.   Of course, any input filtering will pass all signals close to the target signal. 

Another point seen especially in data sets Airspy7_0dB.jpg and to a lesser extent in data set Airspy 9_0dB.jpg:  At issue may be digital saturation.  In an SDR when all available bits are utilized as with possible reception of a strong signal, the output becomes garbage and unpredictable.  This is highly evident in the Airspy7_0dB.jpg.  Attenuation between the antenna feedline and the input to the receiver, in this case, is mandatory.  I wrote an article some 4 or five years ago in the Journal of the Society of Amateur Radio Astronomers (SARA) addressing this issue for reception around 20 MHz (when solar conditions were pretty much the best of Cycle 24).

The issue of 'bogus' signals in the FM broadcast band:  This is common in nearly all receivers be they the older technology or the newer digital receivers like the Airspy and others.  The problem is that within the 88 to 108 MHz FM broadcast band, there are many truly strong signals, especially true in or near large cities.  This is required to minimize the noise and maximize the level of quieting in detection of a wideband signal.  The input stage(s) (front end) acts as a mixer as its driven into unintended non-linearities by all the strong stations within that 20 MHz of spectrum.  The strong stations mix with each other to produce bogus or false signals.  Again, too much signal requires either a stopband filter (if you're not interested in receiving FM), or insertion of an appropriate amount of attenuation between the antenna feedline and the input of the receiver.  Any input passband filter present for the FM band, specifically, will aim at passing the whole band on to the mixer or ADC.  Again, an attenuator is required due to too much in-band strong signal levels.

The suggestion has been made to understand how receivers are tested and the basics of their internal design.  The information is on the Internet, so you don't have to visit a library.  For proper evaluation of the newer SDR technology, to properly design tests and evaluations of the item, a knowledge far deeper than just the basics of the design will be required. 

In conclusion, your measurements and data do not address input filtering and/or input match.  To properly conduct the tests you desire, you must first understand what you are evaluating.  

Dave - WØLEV  
 

Virus-free. www.avast.com


On Fri, Oct 18, 2019 at 5:07 AM jdow <jdow@...> wrote:
Please look at "https://en.wikipedia.org/wiki/Radio_noise". It is a simple
discussion. Then explain to me why you insist on NOT using an attenuator at MW
as a matter of course? If you had a SMALL antenna like one of those telescoping
whips the low noise figure of the HF+ would be really good for you. But with an
antenna such as you describe in an environment with a "typical" level of man
made noise a 30 dB attenuator would do you no ill and probably a 40 dB
attenuator would do you no ill. Look at Wikipiddle er Wikipedia for a perhaps
over-simplified discussion of noise figure to see how it works.

Now, knowing the input impedance of the HF+ input may help you optimize a filter
to reduce signals in the MW broadcast band when you are working elsewhere. But
that would probably be for one tuning setting. The input impedance may vary with
the frequency to which HF+ is tuned.

My purpose in trying to drive this home is to help you optimize your time and
efforts to best effect. I am trying to boost your game not deride you. Too many
years in a very male oriented occupation has left me a tad abrasive. Please forgive.

{o.o}

On 20191017 10:23:46, Wes Stewart via Groups.Io wrote:
> On Thu, Oct 17, 2019 at 07:45 AM, prog wrote:
>
>     Let's be more pragmatic. What is the real world performance using an antenna?
>
> Thank you for asking.  Let me preface the following by explaining the
> environment.  The antenna is a 55 foot vertical over ground radials,  For the
> results that follow, the antenna is connected to the Airspy through a Telonic
> Industries Model TG-950 stepped attenuator.  Although not a traceable device I
> have verified its performance with my DG8SAQ VNWA. I have several plots, all
> taken during the same session, which are screen grabs of the SDR# program.  The
> model, serial number, firmware and software versions are all obvious.  This unit
> is as received a few days ago.
>
> I'm not sure I can intersperse comments between the attachments so I will
> summarize them here.  The image "Airspy_1_40db" is of a local MW station with
> the inline attenuator set to 40dB.  The image  "Airspy_2_20db" is the same as
> before except the attenuation has been reduced to 20 dB.  It was pointless to go
> to 0 dB.  Images "Airspy_3_40dB" and "Airspy_4_20db" are similar but of a
> different station.
>
> The next set, "Airspy_5_20db", "Airspy_6_10db" and "Airspy_7_0db" show station
> WWV being received on 10 MHz with 20, 10 and 0 dB attenuation respectively. 
> Note that with 0 dB attenuation, the Airspy is essentially useless.  The next
> set,  "Airspy_8_10db" and "Airspy_9_0db" show station WWV being received on 15
> MHz with 10 and 0 dB attenuation respectively.  Same situation.  Now you might
> understand my interest in knowing the performance of the input filtering because
> obviously it isn't up to the task.
>
> But it gets worse.  Image "Airspy_10_Spurious_of_88_1_0db" shows an FM broadcast
> signal, fully readable so I could identify it, that doesn't exist.  Image
> "Airspy_11_Spurious_of_88_1_20db" shows the same spectrum with 20 dB of
> attenuation inline. The non-existent station disappears.
>
>
> AIrspy_1_40db.jpg
>
>
> AIrspy_2_20db.jpg
>
>
> Airspy_11_Spurious_of_88_1_20db.jpg
>
>
> Airspy_3_40db.jpg
>
>
> Airspy_4_20db.jpg
>
>
> Airspy_5_20db.jpg
>
>
> Airspy_6_10db.jpg
>
>
> Airspy_7_0db.jpg
>
>
> Airspy_8_10db.jpg
>
>
> Airspy_9_0db.jpg
>
>
> Airspy_10_Spurious_of_88_1_0db.jpg
>





--
Dave - WØLEV
Just Let Darwin Work
Just Think


Re: Transmission lines, Input match and Connectors #bestpractice #experiment

doug
 

On 10/17/2019 07:45 PM, jdow wrote:

/skipped/

Too many years in a very male oriented occupation has left
me a tad abrasive. Please forgive.

{o.o}
You're right. In all my years in that business, I never saw a female engineer. At one place, we took on a young female summer aide who
apparently wanted to be an RF engineer, but that's as close as I ever came to one.
(On the other hand, we had a very competent female software engineer on the projects I worked on at my last place, from which I retired in 2003.
I tried to get her to learn to read schematics, but she wasn't having any.)

--doug, WA2SAY, retired RF engineer