Calibrating SDRs and similar devices--some important facts

[dlwtrunked]

Calibrating SDRs and similar devices
I recommend this be read carefully as I have seen posts
that seem unaware.

(1) Limitations of calibrating an SDR and other devices
with any external source.
The first thing to know is that that you are not going
to get dead on frequency accuracy with any common SDRs
and many other devices simply because the internal device
firmware/hardware frequencies are in steps. These steps
are small enough for frequencies to be "close enough"
to the ones displayed or selected to be "good enough" for
most purposes. So in general,you cannot calibrate
to be dead on accurate. Many have been fooled into think
they have calibrated to being dead on accurate when they
have not. In fact, if you calibrate with a certain
frequency and then look at a signal on that (the one
you calibrated with), you will commonly see a difference
which if frequencies were exact, would not happen);
or if you calibrate using a different frequency,
often frequencies calibrated using the first frequency
will be slightly different.

(2) Calibrating using ATSC 1.0 pilot carrier.
Make sure you first read (1). Now if you want to calibrate
in the U.S. using an over the air signal, generally,
your best choice is to use a digital TV 1.0 pilot carrier.
However, there are two problems with this.
(a) These carrier are *ideally* 0.309411 MHz above
the lower end of the TV channel. There is no specified
tolerance for this frequency. Some stations (approximately
half in my area) use GPS discipliced oscillators to be right
on that frequency. Those that do not have been observed to
to far from that frequency to input that frquency as the
calibration frequency. Sometimes that might be due
to equipment limitations and other times it may be intentional
so that pilot tones do not interfere. Acurately measuring
the actual transmitted pilot frequency is generally beyond
the normal readers equipment (some have spoken to station
engineers to find it out). (Additionally, there may be
more than one ATSC 1.0 pilot received at your location
and the calibration may pick a different one on a slightly
different frequency than the one you intended.)
(b) In many areas, stations are converting from ATSC 1.0
to ATSC 3.0 (Next Gen TV). ATSC 3.0 does not have this pilot.
And even more to note is that the National Association
of Broadcaster has just petitioned the FCC to require
all ATSC 1.0 to convert to ATSC 3.0.

That may or may not happen (many will oppose that ATSC 3.0
would enable encryption and wish over the air TV to remain
freely accessible to the public). But in either case,
there will be diminishing ATSC 1.0 pilot carriers to use
for calibration.

(3) Using other the air signals like FM broadcast,
public safety, or similar to calibrate.
There are few (no?) other sources tranmsitting on VHF/UHF
that would be good for calibration. To be a good calibration
source, the signal needs to be accurate and ideally
a single frequency and with no modulation (AM, FM or digital
on the signal means the frequency is changing or is composed
of various frequencies. Just forget doing this, it is a bad idea.

(4) Calibrating using a signal generator.
If a very good one ($$$)that is GPS disciplined, this is likely
better than (2) but requires the proper equipment.

(5) Calibrating using a GPS disciplined oscillator (a variation
of the last). Though these are usually 10 MHz, they often have
harmonics up into the UHF range (particularly if the have
square-wave output) and those harmonics might be used to calibrate.
However, if your receiver or equipment has a 10 MHz
reference input, that is a better place to use the 10 Mhz GPSDO.
That will give you the result in an easier away. Noting (2) above,
with the majority of equipment, you will *still not have dead
on accuracy+, but you will have extreme stability. Even an
AirSpy R2 is measurable not dead on accurate compared to a GPS disciplined

SignalHound spectrum analyzer (roughly $1000 plus
the GPSDO; I recommend Bodnar for GPSDOs).

(6) Is calibration of the SDR worth it?
One needs to define worth it. If for some reason the SDR or other equipment's
calibration has drifted, "yes". But if you want dead on accuracy, read again (1)
above; many (I once did) spend a lot of time calibrating to obtain something not
really obtainable.

 

[Unitrunker2]

Well said. Allow me to add some details to this:

The first thing to know is that that you are not going to get dead on frequency accuracy with any common SDRs and many other devices simply because the internal device firmware/hardware frequencies are in steps. These steps are small enough for frequencies to be "close enough" to the ones displayed or selected to be "good enough" for most purposes. So in general,you cannot calibrate to be dead on accurate.

Specifically for the Rafael Micro R820T tuner combined with the RTL2832U - the registers allow for a step size of about seven hertz - in the 850 MHz band. The RTL2832U includes a warp correction register that offers about four bits of decimal precision. Oddly enough, some software only allows setting the warp correction value to an integer (no decimal point). The hardware allows parts-per-million corrections in steps of one sixteenth of one part per million (which is quite good for decoding purposes).

However, none of these really matter when the software performs the correction - in software - instead of through a programmable register. Using sine and cosine, a complex signal can be translated to where ever it is needed. You're now in sub-hertz territory.

I think the most important thing is to use an SDR with a temperature compensated clock.

 

[boatbod]

However, none of these really matter when the software performs the correction - in software - instead of through a programmable register. Using sine and cosine, a complex signal can be translated to where ever it is needed. You're now in sub-hertz territory.

In the case of op25, while the application can/does make use of the hardware ppm correction register, the software will be perfectly happy if you get your signal within about 2kHz of where it is supposed to be. An FLL band edge filter is used to track the down-converted IF slice and compensate for drift in the signal, whether caused by the receiving hardware or external/environmental factors. Other tracking loops stabilize phase and clock recovery, so absent of severe multipath effects (which can cause degradation of clock recovery) the result is usually a nice stable demodulated signal.

 

[yugps]

Doppler effect ！

 

[Ubbe]

Doppler effect ！

That's probably one of the reasons why aircrafts still use AM. When you have several transmitters as in a simulcast system and vehicle drives against one tower but away from another, then one towers transmit frequency will increase and the other towers decrease that will create a difference between frequencies that shows as an audible tone that can interfere with a digital signal. So there could be a valid reason why simulcast isn't used in most other countries than US.

/Ubbe

 

[boatbod]

That's probably one of the reasons why aircrafts still use AM. When you have several transmitters as in a simulcast system and vehicle drives against one tower but away from another, then one towers transmit frequency will increase and the other towers decrease that will create a difference between frequencies that shows as an audible tone that can interfere with a digital signal. So there could be a valid reason why simulcast isn't used in most other countries than US.

/Ubbe

My desk isn't moving much, but some days multipath distortion affect the receiver more than others. On those days you can move the desk antenna around a few inches in any direction and eventually you'll find a spot where the signal is less affected and the constellation plot is tighter. Other days the signal is rock solid regardless of where you put the antenna.

On one of my feeds I recently changed from a 5dBi vertical omni to a 7dBi yagi to better pull in a distant tower in a neighboring county. Not only did this fix the crappy signal from the neighbor, but it totally resolved all the multipath problems previously seen on the local tower which is strong enough to receive using a wet noodle.