[linux-dvb] [RFC] SNR units in tuners

[linux-dvb] [RFC] SNR units in tuners

[Devin Heitmueller]

Hello,

In response to Steven Toth's suggestion regarding figuring out what
the various units are across demodulators, I took a quick inventory
and came up with the following list.  Note that this is just a first
pass by taking a quick look at the source for each demodulator (I
haven't looked for the datasheets for any of them yet or done sample
captures to see what the reported ranges are).

Could everybody who is responsible for a demod please take a look at
the list and see if you can fill in the holes?

Having a definitive list of the current state is important to being
able to provide unified reporting of SNR.

Thank you,

Devin

===
af9013.c        dB
at76c651.c      unknown
au8522.c        0.1 dB
bcm3510.c       unknown (vals > 1000)
cx22700.c       unknown
cx22702.c       unknown
cx24110.c       ESN0
cx24116.c       percent scaled to 0-0xffff, support for ESN0
cx24123.c       Inverted ESN0
dib3000mb.c     unknown
dib3000mc.c     always zero
dib7000m.c      always zero
dib7000p.c      always zero
drx397xD.c      always zero
dvb_dummy_fe.c  always zero
l64781.c        unknown
lgdt330x.c      dB scaled to 0-0xffff
lgs8gl5.c       unknown
mt312.c         unknown
mt352.c         unknown
nxt200x.c       dB
nxt6000.c       unknown
or51132.c       dB
or51211.c       dB
s5h1409.c       0.1 dB
s5h1411.c       0.1 dB
s5h1420.c       unsupported
si21xx.c        unknown (scaled to 0-0xffff)
sp8870.c        unsupported
sp887x.c        unknown
stv0288.c       unknown
stv0297.c       unknown
stv0299.c       unknown
tda10021.c      unknown
tda10023.c      unknown
tda10048.c      unknown (looks like 0.1dB)
tda1004x.c      unknown
tda10086.c      unknown
tda8083.c       unknown
tda80xx.c       unknown
ves1820.c       unknown
ves1x93.c       unknown
zl10353.c       unknown


-- 
Devin J. Heitmueller
http://www.devinheitmueller.com
AIM: devinheitmueller

_______________________________________________
linux-dvb mailing list
linux-dvb@linuxtv.org
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb

Re: [linux-dvb] [RFC] SNR units in tuners

[Steven Toth]

Devin Heitmueller wrote:
> Hello,
> 
> In response to Steven Toth's suggestion regarding figuring out what
> the various units are across demodulators, I took a quick inventory

Cool.

> and came up with the following list.  Note that this is just a first
> pass by taking a quick look at the source for each demodulator (I
> haven't looked for the datasheets for any of them yet or done sample
> captures to see what the reported ranges are).
> 
> Could everybody who is responsible for a demod please take a look at
> the list and see if you can fill in the holes?

Some minor comments in line, basically confirming your assessment. When 
I can get some time I'll go back and do the cx22702.

> 
> Having a definitive list of the current state is important to being
> able to provide unified reporting of SNR.

Indeed, good work, thanks.

> 
> Thank you,
> 
> Devin
> 
> ===
> af9013.c        dB
> at76c651.c      unknown

Old code, being removed I think.

> au8522.c        0.1 dB

Correct.

> bcm3510.c       unknown (vals > 1000)
> cx22700.c       unknown
> cx22702.c       unknown

I need to add dB support for this.

> cx24110.c       ESN0
> cx24116.c       percent scaled to 0-0xffff, support for ESN0
> cx24123.c       Inverted ESN0
> dib3000mb.c     unknown
> dib3000mc.c     always zero
> dib7000m.c      always zero
> dib7000p.c      always zero
> drx397xD.c      always zero
> dvb_dummy_fe.c  always zero
> l64781.c        unknown
> lgdt330x.c      dB scaled to 0-0xffff
> lgs8gl5.c       unknown
> mt312.c         unknown
> mt352.c         unknown
> nxt200x.c       dB
> nxt6000.c       unknown
> or51132.c       dB
> or51211.c       dB
> s5h1409.c       0.1 dB
> s5h1411.c       0.1 dB

Correct.

> s5h1420.c       unsupported
> si21xx.c        unknown (scaled to 0-0xffff)
> sp8870.c        unsupported
> sp887x.c        unknown
> stv0288.c       unknown
> stv0297.c       unknown
> stv0299.c       unknown
> tda10021.c      unknown
> tda10023.c      unknown
> tda10048.c      unknown (looks like 0.1dB)

Correct.

> tda1004x.c      unknown
> tda10086.c      unknown
> tda8083.c       unknown
> tda80xx.c       unknown

Old code, being removed I think.

> ves1820.c       unknown
> ves1x93.c       unknown
> zl10353.c       unknown
> 
> 


_______________________________________________
linux-dvb mailing list
linux-dvb@linuxtv.org
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb

Re: [linux-dvb] [RFC] SNR units in tuners

[Darron Broad]

In message <412bdbff0810171104ob627994me2876504b43c18d8@mail.gmail.com>, "Devin Heitmueller" wrote:
>Hello,

hiya :-)

>In response to Steven Toth's suggestion regarding figuring out what
>the various units are across demodulators, I took a quick inventory
>and came up with the following list.  Note that this is just a first
>pass by taking a quick look at the source for each demodulator (I
>haven't looked for the datasheets for any of them yet or done sample
>captures to see what the reported ranges are).
>
>Could everybody who is responsible for a demod please take a look at
>the list and see if you can fill in the holes?
>
>Having a definitive list of the current state is important to being
>able to provide unified reporting of SNR.
>
>Thank you,
>
>Devin
>
>===
<SNIP>
>cx24116.c       percent scaled to 0-0xffff, support for ESN0
<SNIP>

There is no hole here but I thought I would pass you by some
history with this.

The scaled value was calibrated against two domestic satellite
receivers. The first being a nokia 9600s with dvb2000 and
the other being a Fortec star beta. At the time there was
no knowledge of what the cx24116 value represented and no
great idea of what the domestic box values represented.
However, the scaling function matches very closely to those
two machines. What this means in essence is not much but
may be useful to you.

cya!

--

 // /
{:)==={ Darron Broad <darron@kewl.org>
 \\ \ 


_______________________________________________
linux-dvb mailing list
linux-dvb@linuxtv.org
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb

Re: [linux-dvb] [RFC] SNR units in tuners

[Devin Heitmueller]

On Fri, Oct 17, 2008 at 3:55 PM, Darron Broad <darron@kewl.org> wrote:
>>===
> <SNIP>
>>cx24116.c       percent scaled to 0-0xffff, support for ESN0
> <SNIP>
>
> There is no hole here but I thought I would pass you by some
> history with this.
>
> The scaled value was calibrated against two domestic satellite
> receivers. The first being a nokia 9600s with dvb2000 and
> the other being a Fortec star beta. At the time there was
> no knowledge of what the cx24116 value represented and no
> great idea of what the domestic box values represented.
> However, the scaling function matches very closely to those
> two machines. What this means in essence is not much but
> may be useful to you.

By all means, if you have information to share about how the
calculation was arrived at, please do.

At this point the goal is to understand what the value means for
different demods.  For the simple cases where the answer is "it's the
SNR in 0.1db as provided by register X", then it's easy.  If it's "I
don't really know and I just guessed based on empirical testing, then
that is useful information too.

Once people have reported in with the information, I will see about
submitting a patch reflecting this information as a comment in the
driver source for the various demods.

-- 
Devin J. Heitmueller
http://www.devinheitmueller.com
AIM: devinheitmueller

_______________________________________________
linux-dvb mailing list
linux-dvb@linuxtv.org
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb

Re: [linux-dvb] [RFC] SNR units in tuners

[Darron Broad]

In message <412bdbff0810171306n5f8768a2g48255db266d16aa8@mail.gmail.com>, "Devin Heitmueller" wrote:

hi

>On Fri, Oct 17, 2008 at 3:55 PM, Darron Broad <darron@kewl.org> wrote:
>>>===
>> <SNIP>
>>>cx24116.c       percent scaled to 0-0xffff, support for ESN0
>> <SNIP>
>>
>> There is no hole here but I thought I would pass you by some
>> history with this.
>>
>> The scaled value was calibrated against two domestic satellite
>> receivers. The first being a nokia 9600s with dvb2000 and
>> the other being a Fortec star beta. At the time there was
>> no knowledge of what the cx24116 value represented and no
>> great idea of what the domestic box values represented.
>> However, the scaling function matches very closely to those
>> two machines. What this means in essence is not much but
>> may be useful to you.
>
>By all means, if you have information to share about how the
>calculation was arrived at, please do.
>
>At this point the goal is to understand what the value means for
>different demods.  For the simple cases where the answer is "it's the
>SNR in 0.1db as provided by register X", then it's easy.  If it's "I
>don't really know and I just guessed based on empirical testing, then
>that is useful information too.
>
>Once people have reported in with the information, I will see about
>submitting a patch reflecting this information as a comment in the
>driver source for the various demods.

The trouble there is that the scaling for the cx24116 already works
from an end-user perspective. The value derived in the code is
a possible maximum of 160 from the chip. REELBOX decided on 176
which may be more accurate.

A quick glance here:
http://www.mathworks.com/matlabcentral/files/19717/ExactBER.jpg
Would suggest that if that 160 equates to around 10 esn0 (QPSK)
then the register on that chip may equal -5 when 0. I have no real
idea of course as I have no access to any confidential information.

Also, if you refer to that graph, we can see that to scale esn0
for the end user it also needs to take into account that it's
maximum requirement varies per modulation scheme.

I am no expert on this but it doesn't seem as simple as it
may do on first sight.

cya

--

 // /
{:)==={ Darron Broad <darron@kewl.org>
 \\ \ 


_______________________________________________
linux-dvb mailing list
linux-dvb@linuxtv.org
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb

Re: [linux-dvb] [RFC] SNR units in tuners

[Georg Acher]

On Sat, Oct 18, 2008 at 06:42:08AM +0100, Darron Broad wrote:
<cx24116> 
> The trouble there is that the scaling for the cx24116 already works
> from an end-user perspective. The value derived in the code is
> a possible maximum of 160 from the chip. REELBOX decided on 176
> which may be more accurate.

The reelbox code was just a heuristic approach to scale the value so that
less than 30-40% is where the trouble starts... I've more or less matched it
to femon's colors. There was no intention to indicate dBs, as end users
don't understand dBs anyway ;-)

The docs for the 24116 say that the snr is measured in 0.1dB steps. The
absolute range of registers a3:d5 is 0 to 300, so full scale is 30dB. I
doubt we will see the 30dB in a real-world setup...

The signal strength in 9e/9d is the value of the AGC voltage. Register da
seems to contain the estimated power level (-25 to -70dBm), but there's no
further information about that (step size etc). I guess the firmware derives
it from the AGC settings.
-- 
         Georg Acher, acher@in.tum.de         
         http://www.lrr.in.tum.de/~acher
         "Oh no, not again !" The bowl of petunias          

_______________________________________________
linux-dvb mailing list
linux-dvb@linuxtv.org
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb

Re: [linux-dvb] [RFC] SNR units in tuners

[Darron Broad]

In message <20081019195409.GW6792@braindead1.acher>, Georg Acher wrote:

hi

>On Sat, Oct 18, 2008 at 06:42:08AM +0100, Darron Broad wrote:
><cx24116> 
>> The trouble there is that the scaling for the cx24116 already works
>> from an end-user perspective. The value derived in the code is
>> a possible maximum of 160 from the chip. REELBOX decided on 176
>> which may be more accurate.
>
>The reelbox code was just a heuristic approach to scale the value so that
>less than 30-40% is where the trouble starts... I've more or less matched it
>to femon's colors. There was no intention to indicate dBs, as end users
>don't understand dBs anyway ;-)

All sat receivers I see just scale to fit. We can read lots of forum
posts where people exchange % readings for sats :-)

>The docs for the 24116 say that the snr is measured in 0.1dB steps. The
>absolute range of registers a3:d5 is 0 to 300, so full scale is 30dB. I
>doubt we will see the 30dB in a real-world setup...

Okay, so we know the step size of 0.1 per bit and that's measured
within a range of 0 to 300 but that doesn't actually say what it's
value is? Ie, is 50=5dB or something else?

All the graphs I see for QPSK and 8PSK in use in the real-world
suggest the theoretical limit of esn0 is a lot less than that available
range. I don't know what is the accepted error rate to set this limit.
Perhaps someone who has authority on this subject can chime in?

On the cx24116 testing observed that a register max of 160 from QPSK
gave good approximation to that given by regular sat-kit sitting
around 100%. If that really means 16dB then it doesn't look right
compared to the graphs I see, what's wrong here?

>The signal strength in 9e/9d is the value of the AGC voltage. Register da
>seems to contain the estimated power level (-25 to -70dBm), but there's no
>further information about that (step size etc). I guess the firmware derives
>it from the AGC settings.

Okay. All that's done in cx24116.c is take that val and invert
it, which is enough I suppose.

cya!

--

 // /
{:)==={ Darron Broad <darron@kewl.org>
 \\ \ 


_______________________________________________
linux-dvb mailing list
linux-dvb@linuxtv.org
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb

Re: [linux-dvb] [RFC] SNR units in tuners

[Georg Acher]

On Sun, Oct 19, 2008 at 09:43:53PM +0100, Darron Broad wrote:
 
> >The docs for the 24116 say that the snr is measured in 0.1dB steps. The
> >absolute range of registers a3:d5 is 0 to 300, so full scale is 30dB. I
> >doubt we will see the 30dB in a real-world setup...
> 
> Okay, so we know the step size of 0.1 per bit and that's measured
> within a range of 0 to 300 but that doesn't actually say what it's
> value is? Ie, is 50=5dB or something else?

I guess so.
 
> All the graphs I see for QPSK and 8PSK in use in the real-world
> suggest the theoretical limit of esn0 is a lot less than that available
> range. I don't know what is the accepted error rate to set this limit.
> Perhaps someone who has authority on this subject can chime in?
> 
> On the cx24116 testing observed that a register max of 160 from QPSK
> gave good approximation to that given by regular sat-kit sitting
> around 100%. If that really means 16dB then it doesn't look right
> compared to the graphs I see, what's wrong here?

I've looked at the IF output of the frontend (it should be always a CX24118)
with spectrum analyzer. The input was from Astra 19.2 with a 80cm dish. The
slope top above the noise floor indicates also a SNR of 16, maybe 18dB, but
not more. It's really that low...

-- 
         Georg Acher, acher@in.tum.de         
         http://www.lrr.in.tum.de/~acher
         "Oh no, not again !" The bowl of petunias          

_______________________________________________
linux-dvb mailing list
linux-dvb@linuxtv.org
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb

Re: [linux-dvb] [RFC] SNR units in tuners

[Darron Broad]

In message <20081019215225.GA11678@braindead1.acher>, Georg Acher wrote:

hiya.

>On Sun, Oct 19, 2008 at 09:43:53PM +0100, Darron Broad wrote:
> 
>> >The docs for the 24116 say that the snr is measured in 0.1dB steps. The
>> >absolute range of registers a3:d5 is 0 to 300, so full scale is 30dB. I
>> >doubt we will see the 30dB in a real-world setup...
>> 
>> Okay, so we know the step size of 0.1 per bit and that's measured
>> within a range of 0 to 300 but that doesn't actually say what it's
>> value is? Ie, is 50=5dB or something else?
>
>I guess so.

Okay, thanks.

>> All the graphs I see for QPSK and 8PSK in use in the real-world
>> suggest the theoretical limit of esn0 is a lot less than that available
>> range. I don't know what is the accepted error rate to set this limit.
>> Perhaps someone who has authority on this subject can chime in?
>> 
>> On the cx24116 testing observed that a register max of 160 from QPSK
>> gave good approximation to that given by regular sat-kit sitting
>> around 100%. If that really means 16dB then it doesn't look right
>> compared to the graphs I see, what's wrong here?
>
>I've looked at the IF output of the frontend (it should be always a CX24118)
>with spectrum analyzer. The input was from Astra 19.2 with a 80cm dish. The
>slope top above the noise floor indicates also a SNR of 16, maybe 18dB, but
>not more. It's really that low...

Thank you for this information Georg. I am now going to leave this
case closed mostly because I have looked at too many graphs, mistrusted
registers, failed to see the wood for the trees and finally you give
me absolute evidence to trust what were simple observations as
being partially true at least.

Thanks again

darron

--

 // /
{:)==={ Darron Broad <darron@kewl.org>
 \\ \ 


_______________________________________________
linux-dvb mailing list
linux-dvb@linuxtv.org
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb

Re: [linux-dvb] [RFC] SNR units in tuners

[Darron Broad]

In message <5905.1224308528@kewl.org>, Darron Broad wrote:
>In message <412bdbff0810171306n5f8768a2g48255db266d16aa8@mail.gmail.com>, "Devin Heitmueller" wrote:

hi.

>
>hi
>
>>On Fri, Oct 17, 2008 at 3:55 PM, Darron Broad <darron@kewl.org> wrote:
>>>>===
>>> <SNIP>
>>>>cx24116.c       percent scaled to 0-0xffff, support for ESN0
>>> <SNIP>
>>>
>>> There is no hole here but I thought I would pass you by some
>>> history with this.
>>>
>>> The scaled value was calibrated against two domestic satellite
>>> receivers. The first being a nokia 9600s with dvb2000 and
>>> the other being a Fortec star beta. At the time there was
>>> no knowledge of what the cx24116 value represented and no
>>> great idea of what the domestic box values represented.
>>> However, the scaling function matches very closely to those
>>> two machines. What this means in essence is not much but
>>> may be useful to you.
>>
>>By all means, if you have information to share about how the
>>calculation was arrived at, please do.
>>
>>At this point the goal is to understand what the value means for
>>different demods.  For the simple cases where the answer is "it's the
>>SNR in 0.1db as provided by register X", then it's easy.  If it's "I
>>don't really know and I just guessed based on empirical testing, then
>>that is useful information too.
>>
>>Once people have reported in with the information, I will see about
>>submitting a patch reflecting this information as a comment in the
>>driver source for the various demods.
>
>The trouble there is that the scaling for the cx24116 already works
>from an end-user perspective. The value derived in the code is
>a possible maximum of 160 from the chip. REELBOX decided on 176
>which may be more accurate.
>
>A quick glance here:
>http://www.mathworks.com/matlabcentral/files/19717/ExactBER.jpg
>Would suggest that if that 160 equates to around 10 esn0 (QPSK)
>then the register on that chip may equal -5 when 0. I have no real
>idea of course as I have no access to any confidential information.
>
>Also, if you refer to that graph, we can see that to scale esn0
>for the end user it also needs to take into account that it's
>maximum requirement varies per modulation scheme.
>
>I am no expert on this but it doesn't seem as simple as it
>may do on first sight.

LOL. not only am I no expert on this I am also stupid. Please
forgive the ramblings of an old man who totally misread this.
Please refer to other post where Georg has clarified all of
this simply without pointing out the obvious flaws in my
pursuit of signal nirvana.

I wish you well, sorry for the rambling.

darron

--

 // /
{:)==={ Darron Broad <darron@kewl.org>
 \\ \ 


_______________________________________________
linux-dvb mailing list
linux-dvb@linuxtv.org
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb

Re: [linux-dvb] [RFC] SNR units in tuners

[Richard Scobie]

Devin Heitmueller wrote:

> At this point the goal is to understand what the value means for
> different demods.  For the simple cases where the answer is "it's the
> SNR in 0.1db as provided by register X", then it's easy.  If it's "I
> don't really know and I just guessed based on empirical testing, then
> that is useful information too.


According to the datasheet, the carrier to noise ratio for the BSRU6 
front end, (which contains an stv0299), can be calculated in dB using 
the following information:

Register ADD 24 NIRH 8 bits (MSB of noise indicator)

Register ADD 25 NIRL 8 bits (LSB of noise indicator)

C/N (dB) = -0.0017 * NIR + 19.02

Regards,

Richard

_______________________________________________
linux-dvb mailing list
linux-dvb@linuxtv.org
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb

Re: [linux-dvb] [RFC] SNR units in tuners

[Patrick Boettcher]

Hi Devin,

On Fri, 17 Oct 2008, Devin Heitmueller wrote:

> dib3000mb.c     unknown
> dib3000mc.c     always zero
> dib7000m.c      always zero
> dib7000p.c      always zero

All of them have the ability to provide SNR in dB 0.1 scale . To have it, 
I think I should use dvb-math.

best regards,
Patrick.

--
   Mail: patrick.boettcher@desy.de
   WWW:  http://www.wi-bw.tfh-wildau.de/~pboettch/

_______________________________________________
linux-dvb mailing list
linux-dvb@linuxtv.org
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb

Re: [linux-dvb] [RFC] SNR units in tuners

[Trent Piepho]

On Fri, 17 Oct 2008, Devin Heitmueller wrote:
> lgdt330x.c      dB scaled to 0-0xffff

It's dB * 256

> or51132.c       dB
> or51211.c       dB

These are also dB * 256

_______________________________________________
linux-dvb mailing list
linux-dvb@linuxtv.org
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb

Re: [linux-dvb] [RFC] SNR units in tuners

[Matthias Schwarzott]

On Freitag, 17. Oktober 2008, Devin Heitmueller wrote:
> Hello,
>
> In response to Steven Toth's suggestion regarding figuring out what
> the various units are across demodulators, I took a quick inventory
> and came up with the following list.  Note that this is just a first
> pass by taking a quick look at the source for each demodulator (I
> haven't looked for the datasheets for any of them yet or done sample
> captures to see what the reported ranges are).
>
> Could everybody who is responsible for a demod please take a look at
> the list and see if you can fill in the holes?
>
> Having a definitive list of the current state is important to being
> able to provide unified reporting of SNR.
>
> Thank you,
>
> Devin
>

> mt312.c         unknown

The hardware provides two values:
* AGC: For now the AGC feedback value (14 bit) is returned unchanged.
* ERR_db: Also available is a 10-bit value representing the signal level 
difference between AGC-Reference and received signal level (most likely in 
1dB steps).

If an absolute value is needed: I don't get it how to calc absolute signal 
levels from AGC-Reference.

so code for now:
mt312.c: agc-feedback 0-0x3FFF

@Obi:
Any additions?

Regards
Matthias

_______________________________________________
linux-dvb mailing list
linux-dvb@linuxtv.org
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb

Re: [linux-dvb] [RFC] SNR units in tuners

[Andreas Oberritter]

Matthias Schwarzott wrote:
> On Freitag, 17. Oktober 2008, Devin Heitmueller wrote:
>> Hello,
>>
>> In response to Steven Toth's suggestion regarding figuring out what
>> the various units are across demodulators, I took a quick inventory
>> and came up with the following list.  Note that this is just a first
>> pass by taking a quick look at the source for each demodulator (I
>> haven't looked for the datasheets for any of them yet or done sample
>> captures to see what the reported ranges are).
>>
>> Could everybody who is responsible for a demod please take a look at
>> the list and see if you can fill in the holes?
>>
>> Having a definitive list of the current state is important to being
>> able to provide unified reporting of SNR.
>>
>> Thank you,
>>
>> Devin
>>
> 
>> mt312.c         unknown
> 
> The hardware provides two values:
> * AGC: For now the AGC feedback value (14 bit) is returned unchanged.
> * ERR_db: Also available is a 10-bit value representing the signal level 
> difference between AGC-Reference and received signal level (most likely in 
> 1dB steps).
> 
> If an absolute value is needed: I don't get it how to calc absolute signal 
> levels from AGC-Reference.
> 
> so code for now:
> mt312.c: agc-feedback 0-0x3FFF
> 
> @Obi:
> Any additions?

>From the manual:

7.2.3 Measured Signal to Noise Ratio. Registers 9 - 10 (R)

M SNR[14:0]: These two registers provide a indication of the signal to noise
ratio of the channel being received by the MT312. It should not be taken as
the absolute value of the SNR.

Eb/N0 = ~ (13312 - M_SNR[14:0] / 683) dB.

The equation given only holds for Es/No values in the range 3 to 15 dB, i.e.
Eb/No values in the range 0 to 12 dB.

Regards,
Andreas


_______________________________________________
linux-dvb mailing list
linux-dvb@linuxtv.org
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb

Re: [linux-dvb] [RFC] SNR units in tuners

[wk]

Devin Heitmueller schrieb:
> Hello,
>
> In response to Steven Toth's suggestion regarding figuring out what
> the various units are across demodulators, I took a quick inventory
> and came up with the following list.  Note that this is just a first
> pass by taking a quick look at the source for each demodulator (I
> haven't looked for the datasheets for any of them yet or done sample
> captures to see what the reported ranges are).
>
> Could everybody who is responsible for a demod please take a look at
> the list and see if you can fill in the holes?
>
> Having a definitive list of the current state is important to being
> able to provide unified reporting of SNR.
>
> Thank you,
>
> Devin
>
> ===
> af9013.c        dB
> at76c651.c      unknown
> au8522.c        0.1 dB
> bcm3510.c       unknown (vals > 1000)
> cx22700.c       unknown
> cx22702.c       unknown
> cx24110.c       ESN0
> cx24116.c       percent scaled to 0-0xffff, support for ESN0
> cx24123.c       Inverted ESN0
> dib3000mb.c     unknown
> dib3000mc.c     always zero
> dib7000m.c      always zero
> dib7000p.c      always zero
> drx397xD.c      always zero
> dvb_dummy_fe.c  always zero
> l64781.c        unknown
> lgdt330x.c      dB scaled to 0-0xffff
> lgs8gl5.c       unknown
> mt312.c         unknown
> mt352.c         unknown
> nxt200x.c       dB
> nxt6000.c       unknown
> or51132.c       dB
> or51211.c       dB
> s5h1409.c       0.1 dB
> s5h1411.c       0.1 dB
> s5h1420.c       unsupported
> si21xx.c        unknown (scaled to 0-0xffff)
> sp8870.c        unsupported
> sp887x.c        unknown
> stv0288.c       unknown
> stv0297.c       unknown
> stv0299.c       unknown
> tda10021.c      unknown
> tda10023.c      unknown
> tda10048.c      unknown (looks like 0.1dB)
> tda1004x.c      unknown
> tda10086.c      unknown
> tda8083.c       unknown
> tda80xx.c       unknown
> ves1820.c       unknown
> ves1x93.c       unknown
> zl10353.c       unknown
>
>
>   
What about cinergyT2? Seems to be missing in list.

Regards, Winfried

_______________________________________________
linux-dvb mailing list
linux-dvb@linuxtv.org
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb

Re: [linux-dvb] [RFC] SNR units in tuners

[Morgan Tørvolt]

> In response to Steven Toth's suggestion regarding figuring out what
> the various units are across demodulators, I took a quick inventory
> and came up with the following list.  Note that this is just a first
> pass by taking a quick look at the source for each demodulator (I
> haven't looked for the datasheets for any of them yet or done sample
> captures to see what the reported ranges are).
>
> Could everybody who is responsible for a demod please take a look at
> the list and see if you can fill in the holes?
>
> Having a definitive list of the current state is important to being
> able to provide unified reporting of SNR.

As I used to work with satellite signals on an earth station, and was
responsible for the development of measurement techniques, I thought I
should join in here for some hopefully revelaing info.

I am guessing here of course, but I believe that there is no real SNR
measurement in any of the tuners available for computers. It is quite
easy to calculate SNR from BER, and even some professional satellite
modems in the 10kUSD price range and up uses the received BER to
calculate SNR. The received BER is calculated from how many bit errors
the FEC fixes. On a QPSK, a SNR of 16 would generally mean that the
time between biterrors is so long that it is near impossible to
calculate a SNR from the ber unless you run an average of multiple
days, or even weeks. I remember doing a test for a week on a satellite
mux without getting a single biterror. This means that you might never
get a reading above 16 on the SNR. Looking at the curves on wikipedia
(http://en.wikipedia.org/wiki/Image:PSK_BER_curves.svg), you can
assume that with an EbNo of 12 db, on a 27500MS 3/4 QPSK mux, the
average raw biterror will arrive every ( 27500 * 2 * (3/4) * (188/204)
= 38015*10^6 Mb/s ) 10^8 / 38015*10^6 = 0.0026 s.
If we then use EbNo at 13dB, the snr is 10^-10, at EbNo at 14 it is
10^-13. At this point, you have one bit-error each 260s on the same
mux. Suddenly the internal SNR calculation algorithm has a zero
element, and it gets impossible to calculate SNR.

Conclusion is, that no matter what you do on most tuners, you will not
get more than roughly 16-17dB SNR on QPSK signals. If they say that
they will give a range og 0-30dB, then the higher numbers will most
probably never happen. On other modulations this highest number is
different of course, and on lower bandwidth signals, this number is
reduced somewhat as a ber of 10^-9 will happen less frequently with
fewer bits per second coming trough the receiver.

Hope this made any sense (and was correct).

-Morgan-

_______________________________________________
linux-dvb mailing list
linux-dvb@linuxtv.org
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb

Re: [linux-dvb] [RFC] SNR units in tuners

[Georg Acher]

On Mon, Oct 27, 2008 at 10:37:52AM +0000, Morgan Tørvolt wrote:
 
> As I used to work with satellite signals on an earth station, and was
> responsible for the development of measurement techniques, I thought I
> should join in here for some hopefully revelaing info.
> 
> I am guessing here of course, but I believe that there is no real SNR
> measurement in any of the tuners available for computers. 

My guess is that it is possible. Actually, it is quite easy for QPSK ;-) You
only need to calculate the distance of the IQ-value from the ideal symbol
center ( (sqrt(0.5),sqrt(0.5)) or whatever) after the
rotator/retiming-block.

-- 
         Georg Acher, acher@in.tum.de
         http://www.lrr.in.tum.de/~acher
         "Oh no, not again !" The bowl of petunias

_______________________________________________
linux-dvb mailing list
linux-dvb@linuxtv.org
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb

Re: [linux-dvb] [RFC] SNR units in tuners

[Andy Walls]

On Mon, 2008-10-27 at 15:03 +0100, Georg Acher wrote:
> On Mon, Oct 27, 2008 at 10:37:52AM +0000, Morgan Tørvolt wrote:
>  
> > As I used to work with satellite signals on an earth station, and was
> > responsible for the development of measurement techniques, I thought I
> > should join in here for some hopefully revelaing info.
> > 
> > I am guessing here of course, but I believe that there is no real SNR
> > measurement in any of the tuners available for computers. 
> 
> My guess is that it is possible. Actually, it is quite easy for QPSK ;-) You
> only need to calculate the distance of the IQ-value from the ideal symbol
> center ( (sqrt(0.5),sqrt(0.5)) or whatever) after the
> rotator/retiming-block.
> 

Isn't that just Error Vector Magnitude (EVM)?

<musing>
I suppose over a number of samples, EVM gives you an idea of the noise
power + source transmitter deformation, if the received signal remains
at a constant power.  I suppose you could estimate signal strength by
examining amplifier gain settings.
</musing>

Regards,
Andy


_______________________________________________
linux-dvb mailing list
linux-dvb@linuxtv.org
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb

Re: [linux-dvb] [RFC] SNR units in tuners

[Morgan Tørvolt]

>> My guess is that it is possible. Actually, it is quite easy for QPSK ;-) You
>> only need to calculate the distance of the IQ-value from the ideal symbol
>> center ( (sqrt(0.5),sqrt(0.5)) or whatever) after the
>> rotator/retiming-block.
>>
>
> Isn't that just Error Vector Magnitude (EVM)?
>
> <musing>
> I suppose over a number of samples, EVM gives you an idea of the noise
> power + source transmitter deformation, if the received signal remains
> at a constant power.  I suppose you could estimate signal strength by
> examining amplifier gain settings.
> </musing>

Easy and possible as it might be, my experience, this is not done.
Since one can _very_ easily get the actual bit-error rate using FEC
error corrections, there is no need to implement anything but the very
simplest demodulator. The formula converting from BER to SNR is quite
accurate and not very calculation intensive. You could just have a
small look-up table for it if you wanted.

As for the signal level, then yes, it is usually "measured" by the
needed preamp gain. Since this is highly temperature dependent, most
professional equipment do not give you a signal level measurement.

I can list a few equipment names that does the one or the other if
anyone is interested. NEC actually measure the SNR properly in some of
their equipment, which makes it possible to compare the SNR and BER to
figure out if something is wrong (like a interference carrier hidden
inside the data carrier).

I must say that if any low-end equipment acutally measures the SNR
properly, I will be very surprised. To get this feature usually costs
thousands of dollars. If anyone could point me to a low cost DVB
receiver that actually does this right, I would very much like to
know.


Regards
Morgan

_______________________________________________
linux-dvb mailing list
linux-dvb@linuxtv.org
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb

Re: [linux-dvb] [RFC] SNR units in tuners

[Georg Acher]

On Mon, Oct 27, 2008 at 04:16:01PM +0000, Morgan Tørvolt wrote:
 
> I must say that if any low-end equipment acutally measures the SNR
> properly, I will be very surprised. To get this feature usually costs
> thousands of dollars. If anyone could point me to a low cost DVB
> receiver that actually does this right, I would very much like to
> know.

The datasheet of the "good old" STV0299 shows the C/N indicator before the
decoding chain (Viterbi, deinterleaver, ...).

-- 
         Georg Acher, acher@in.tum.de
         http://www.lrr.in.tum.de/~acher
         "Oh no, not again !" The bowl of petunias

_______________________________________________
linux-dvb mailing list
linux-dvb@linuxtv.org
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb

Re: [linux-dvb] [RFC] SNR units in tuners

[Manu Abraham]

Andy Walls wrote:
> On Mon, 2008-10-27 at 15:03 +0100, Georg Acher wrote:
>> On Mon, Oct 27, 2008 at 10:37:52AM +0000, Morgan Tørvolt wrote:
>>  
>>> As I used to work with satellite signals on an earth station, and was
>>> responsible for the development of measurement techniques, I thought I
>>> should join in here for some hopefully revelaing info.
>>>
>>> I am guessing here of course, but I believe that there is no real SNR
>>> measurement in any of the tuners available for computers.


True, most demodulators do provide a CNR from which SNR can be
calculated out. CNR is measured from the early stages, while SNR is
measured out from the final stages.


>> My guess is that it is possible. Actually, it is quite easy for QPSK ;-) You
>> only need to calculate the distance of the IQ-value from the ideal symbol
>> center ( (sqrt(0.5),sqrt(0.5)) or whatever) after the
>> rotator/retiming-block.
>>
> 
> Isn't that just Error Vector Magnitude (EVM)?

Not only that, by the time a driver reads that through a slow bus such
as I2C (of course not to be forgotten about other delays: inherent to
demodulator internals) though it can be used to evaluate SNR on the
demodulator IP core, the events being at 2 distinct points of time as it
is not RT wrt the driver, i guess it makes no sense.


Regards,
Manu

_______________________________________________
linux-dvb mailing list
linux-dvb@linuxtv.org
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb

Re: [linux-dvb] [RFC] SNR units in tuners

[VDR User]

Is this topic going to be address at the upcoming Plumbers conference?
 Several months have passed with seemingly no resolution or even any
progress made even though getting this sorted out would provide a very
useful tool.

-Derek

_______________________________________________
linux-dvb users mailing list
For V4L/DVB development, please use instead linux-media@vger.kernel.org
linux-dvb@linuxtv.org
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb