Topics

Airspy HF+ BDR 110dB


adam9a4qv
 

Can somebody explain how this number is calculated or measured?
BDR 110dB?
The method used ?

thanks
Adam


prog
 

On Mon, Nov 13, 2017 at 12:29 am, adam9a4qv wrote:
Can somebody explain how this number is calculated or measured?
BDR 110dB?
The method used ?

thanks
Adam
Measured with a weak signal at -115 dBm and a blocker 20kHz away. The blocker is increased until the weak signal decreased by 1 dB - and we are still conservative here. The datasheet says 113 dBm typ.
Also, given all the AGC loops involved, the whole gain distribution is optimized for best sensitivity while all the stages are continuously operating in a comfortable region.


jdow
 

On 2017-11-13 00:52, prog wrote:
On Mon, Nov 13, 2017 at 12:29 am, adam9a4qv wrote:
Can somebody explain how this number is calculated or measured?
BDR 110dB?
The method used ?
thanks
Adam
Measured with a weak signal at -115 dBm and a blocker 20kHz away. The blocker is increased until the weak signal decreased by 1 dB - and we are still conservative here. The datasheet says 113 dBm typ.
Also, given all the AGC loops involved, the whole gain distribution is optimized for best sensitivity while all the stages are continuously operating in a comfortable region.
Is their feedback to the program running it for the current AGC setting so that metering can be accurate in a spectrum analyzer application?

{^_^}


prog
 

On Mon, Nov 13, 2017 at 01:12 am, jdow wrote:
Is their feedback to the program running it for the current AGC setting so that metering can be accurate in a spectrum analyzer application?

{^_^}
The AGC's are disabled for the BDR tests. You don't want to disable them in real world use.

accurate
These stunts were done by trained professionals, don't try this at home ;-)


jdow
 

On 2017-11-13 01:19, prog wrote:
On Mon, Nov 13, 2017 at 01:12 am, jdow wrote:
Is their feedback to the program running it for the current AGC setting so
that metering can be accurate in a spectrum analyzer application?
{^_^}
The AGC's are disabled for the BDR tests. You don't want to disable them in real world use.
accurate
These stunts were done by trained professionals, don't try this at home ;-)
I have my trained professional hat and aged tattered and torn tee-shirt among the clothing I am metaphorically wearing at the moment.
{^_-}


adam9a4qv
 

How will then the trained professionals explain this:

BDR= P1dB - Noise floor

P1dB=IP3-10db
P1dB=15-10 = +5dBm

BDR=+5- (-140)
BDR= 145dB

This is a big difference, 110db or 113dB over to 145dB ?

Adam


Mark Jessop
 

What kind of response time (time constant?) is the AGC running?

I will be using my AirSpy HF+ in the receive loop of a Icom IC-7610, and when I transmit there is likely to be a bit of RF ending up in the AirSpy. (not 100% sure how much yet, will be a function of the TX/RX relay isolation.
Would be good if it didn't take a long time to recover... 

73
Mark VK5QI

On Mon, Nov 13, 2017 at 9:32 AM, adam9a4qv via Groups.Io <adam9a4qv@...> wrote:
How will then the trained professionals explain this:

BDR= P1dB - Noise floor

P1dB=IP3-10db
P1dB=15-10 = +5dBm

BDR=+5- (-140)
BDR= 145dB

This is a big difference, 110db or 113dB over to 145dB ?

Adam



prog
 

On Mon, Nov 13, 2017 at 01:32 am, adam9a4qv wrote:
BDR= P1dB - Noise floor

P1dB=IP3-10db
P1dB=15-10 = +5dBm

BDR=+5- (-140)
BDR= 145dB

This is a big difference, 110db or 113dB over to 145dB ?

Adam
Read my reply to Johanne. There's also the action of the AGC which was completely disabled during the tests. The gain was fixed so that the SNR of the weaker tone 3dB.


prog
 

On Mon, Nov 13, 2017 at 02:17 am, Mark Jessop wrote:
What kind of response time (time constant?) is the AGC running?
 
I will be using my AirSpy HF+ in the receive loop of a Icom IC-7610, and when I transmit there is likely to be a bit of RF ending up in the AirSpy. (not 100% sure how much yet, will be a function of the TX/RX relay isolation.
Would be good if it didn't take a long time to recover... 
 
73
Mark VK5QI

Good question. The AGC has 4 ms attack and 40 ms decay. Note that the AGC attenuation is compensated digitally to maintain the same signal level.


adam9a4qv
 

There is nothing wrong in that procedure (with AGC off) and measuring the weak signl 1dB drop, just the math and published figures are not matching to the measured BDR.
Maybe, even better should be to measure the BDR with AGC ingaged and then measure the S/n drop for a 1dB.

I agree, the BDR should be close to 110db but then there is something odd with the P1dB or most probably with the MDS figures?

Adam


Mark Jessop
 

Well, once I have my AirSpy HF+, i'll post some examples of it in this situation :-)

When I was doing something similar with an SDR-IQ on another rig, I'd occasionally see the overload light come on... not sure what level that occurs at though.

73
Mark VK5QI

On Mon, Nov 13, 2017 at 10:26 AM, prog <info@...> wrote:
On Mon, Nov 13, 2017 at 02:17 am, Mark Jessop wrote:
What kind of response time (time constant?) is the AGC running?
 
I will be using my AirSpy HF+ in the receive loop of a Icom IC-7610, and when I transmit there is likely to be a bit of RF ending up in the AirSpy. (not 100% sure how much yet, will be a function of the TX/RX relay isolation.
Would be good if it didn't take a long time to recover... 
 
73
Mark VK5QI

Good question. The AGC has 4 ms attack and 40 ms decay. Note that the AGC attenuation is compensated digitally to maintain the same signal level.



prog
 

On Mon, Nov 13, 2017 at 02:33 am, Mark Jessop wrote:
Well, once I have my AirSpy HF+, i'll post some examples of it in this situation :-)
 
When I was doing something similar with an SDR-IQ on another rig, I'd occasionally see the overload light come on... not sure what level that occurs at though.
 
73
Mark VK5QI
You can see the AGC reaction to strong signals here (actually two of them) :

https://www.rtl-sdr.com/wp-content/uploads/2017/08/hf_pagers.jpg

Note that the pagers themselves are not very clean and have terrible phase noise, hence the rise of the noise floor. Otherwise, the action of the AGC is compensated to avoid any pumping.


prog
 

On Mon, Nov 13, 2017 at 02:33 am, adam9a4qv wrote:
There is nothing wrong in that procedure (with AGC off) and measuring the weak signl 1dB drop, just the math and published figures are not matching to the measured BDR.
Maybe, even better should be to measure the BDR with AGC ingaged and then measure the S/n drop for a 1dB.

I agree, the BDR should be close to 110db but then there is something odd with the P1dB or most probably with the MDS figures?

Adam
Well, I agree. Maybe more tests in different scenarios can be published later. The current testing methodology heavily depends on the underlying architecture because the components that "break" first in a more traditional architecture operate at much relaxed constraints here, and most of the amplification (and hence the biggest contribution to the linearity) happen at narrow baseband.

More background:
The ADC has 113+ dB of dynamic range (with 105 dB SFDR) and there is a spur avoidance mechanism to improve its dynamic behavior (by changing both the sample rate and the RF/IF frequencies). Also, there are tunable N-path filters at RF/IF, plus the passive polyphase mixer. The RF gain is very low - just enough to compensate for the attenuation of the N-path filters. Then the 6 dB stepped AGC controls the operating points for all the stages.


jdow
 

On 2017-11-13 01:32, adam9a4qv via Groups.Io wrote:
How will then the trained professionals explain this:
BDR= P1dB - Noise floor
P1dB=IP3-10db
P1dB=15-10 = +5dBm
BDR=+5- (-140)
BDR= 145dB
This is a big difference, 110db or 113dB over to 145dB ?
Adam
Grasshopper, your calculations are incomplete. They need more consideration of bandwidth and a gentleman named Boltzmann.

{^_^}


jdow
 

On 2017-11-13 02:20, prog wrote:
On Mon, Nov 13, 2017 at 01:32 am, adam9a4qv wrote:
BDR= P1dB - Noise floor
P1dB=IP3-10db
P1dB=15-10 = +5dBm
BDR=+5- (-140)
BDR= 145dB
This is a big difference, 110db or 113dB over to 145dB ?
Adam
Read my reply to Johanne. There's also the action of the AGC which was completely disabled during the tests. The gain was fixed so that the SNR of the weaker tone 3dB.
In what bandwidth? (1Hz is meaningful but a bit of a cheat when dealing with non-professionals.)

{^_-}


prog
 
Edited

On Mon, Nov 13, 2017 at 03:14 am, jdow wrote:
On 2017-11-13 01:32, adam9a4qv via Groups.Io wrote:
How will then the trained professionals explain this:
BDR= P1dB - Noise floor
P1dB=IP3-10db
P1dB=15-10 = +5dBm
BDR=+5- (-140)
BDR= 145dB
This is a big difference, 110db or 113dB over to 145dB ?
Adam
Grasshopper, your calculations are incomplete. They need more consideration of bandwidth and a gentleman named Boltzmann.

{^_^}
RF bandwidth? IF bandwidth? Sigma Delta bandwidth? It's not as simple as it seems.

OK, the last one is pure zeal... Now let's just wait for Leif to show up.


prog
 

On Mon, Nov 13, 2017 at 03:17 am, jdow wrote:
On 2017-11-13 02:20, prog wrote:
On Mon, Nov 13, 2017 at 01:32 am, adam9a4qv wrote:
BDR= P1dB - Noise floor
P1dB=IP3-10db
P1dB=15-10 = +5dBm
BDR=+5- (-140)
BDR= 145dB
This is a big difference, 110db or 113dB over to 145dB ?
Adam
Read my reply to Johanne. There's also the action of the AGC which was completely disabled during the tests. The gain was fixed so that the SNR of the weaker tone 3dB.
In what bandwidth? (1Hz is meaningful but a bit of a cheat when dealing with non-professionals.)

{^_-}
I know you want the datasheet. ;-)


Simon Brown
 

{^"^}

A BDR of 110dB should be enough - if anyone has 3rd-party signals strong enough to cause a problem then the first indication will probably be their underwear catching fire due to the radiation rather than funnies in the waterfall 😊 .

Simon Brown, G4ELI

www.dxgalaxy.com
www.sdr-radio.com
www.sdr-satellites.com

-----Original Message-----
From: main@airspy.groups.io [mailto:main@airspy.groups.io] On Behalf Of jdow
Sent: 13 November 2017 11:18
To: main@airspy.groups.io
Subject: Re: [airspy] Airspy HF+ BDR 110dB

On 2017-11-13 02:20, prog wrote:
On Mon, Nov 13, 2017 at 01:32 am, adam9a4qv wrote:

BDR= P1dB - Noise floor

P1dB=IP3-10db
P1dB=15-10 = +5dBm

BDR=+5- (-140)
BDR= 145dB

This is a big difference, 110db or 113dB over to 145dB ?

Adam

Read my reply to Johanne. There's also the action of the AGC which was
completely disabled during the tests. The gain was fixed so that the
SNR of the weaker tone 3dB.
In what bandwidth? (1Hz is meaningful but a bit of a cheat when dealing with
non-professionals.)

{^_-}


Dana Myers
 

On 11/13/2017 12:52 AM, prog wrote:
On Mon, Nov 13, 2017 at 12:29 am, adam9a4qv wrote:
Can somebody explain how this number is calculated or measured?
BDR 110dB?
The method used ?

thanks
Adam
Measured with a weak signal at -115 dBm and a blocker 20kHz away. The blocker is increased until the weak signal decreased by 1 dB - and we are still conservative here. The datasheet says 113 dBm typ.

Measured at 20KHz spacing. Perhaps naively, I speculate that the DDC performance is
relatively the same at a closer spacing and BDR is noise-limited by the LO.

Is this correct? I am trying to get some idea of how the BDR deteriorates with
respect to spacing.

Cheers,
Dana  K6JQ


jdow
 

On 2017-11-13 03:25, prog wrote:
On Mon, Nov 13, 2017 at 03:17 am, jdow wrote:
On 2017-11-13 02:20, prog wrote:
On Mon, Nov 13, 2017 at 01:32 am, adam9a4qv wrote:
BDR= P1dB - Noise floor
P1dB=IP3-10db
P1dB=15-10 = +5dBm
BDR=+5- (-140)
BDR= 145dB
This is a big difference, 110db or 113dB over to 145dB ?
Adam
Read my reply to Johanne. There's also the action of the AGC which was
completely disabled during the tests. The gain was fixed so that the SNR
of the weaker tone 3dB.
In what bandwidth? (1Hz is meaningful but a bit of a cheat when dealing with
non-professionals.)
{^_-}
I know you want the datasheet. ;-)
That's cheating, and leads to the question, "Where did he get 145 dB BDR?" It is not in the official specifications. With 18 bits of quantization at 768 ksps (sortakinda) and 500 Hz bandwidth 145 dB sort of makes the BDR work out sanely. if we ignore component imperfections. (I mean silly things such as connectors that themselves generate distortion due to their magnetic properties, filters with nonlinear elements which generate distortion, and so forth.)

Now, where did HIS figures come from? And what assumptions did he make that might be wrong?

18 bits of sampling at what level of linearity per sample comes to mind as an issue. I also notice he fixated on the IP3 figures, which are not necessarily what he wants to work with here. (And as you've noted some of these figures have AGC on and others have it off.)

{^_^}