Has anyone performed TIA SPD/ACPR testing?

[xmo]

The Telecommunications Industry Association has published documents to assist system designers, engineers, spectrum planners, and frequency coordinators.

Included are SPD (Spectral Power Density) plots and ACPR (Adjacent Channel Power Ratio) data for the common land-mobile modulation formats.  Instructions on how to perform the testing are included.

An amateur researcher could perform the tests:
1) As a learning exercise.
2) To add results for amateur formats such as Fusion and D-Star.
3) To evaluate the impact of changes in repeater transmit audio processing made in the interest of better sounding repeater audio.

D-Star is especially interesting in view of I-com's claim that D-Star only needs 6.25 kHz channels.

An example of SPD/ACPR data is attached.  It shows the difference between ACPR for +/-5 kHz analog FM at 15 kHz channel spacing vs. 20 kHz.

[rf-messkopf]

Has anyone performed TIA SPD/ACPR testing?

Many modern analog spectrum analyzers have builtin channel power and ACLR measuring functions, even low-budget kit like Rigol and Siglent, so no problem for a well equipped amateur. Attached is an ACLR measurement for a C4FM signal according to ANSI/TIA-102, with the the specifications mentioned here, directly from a signal generator. The symbol rate is 4800 Baud, with 12.5 kHz channel bandwidth, and no channel spacing. Further, the pulse shaping filter was Gaussian with shape a factor of 0.3. See the attachment. The slight imbalance between the upper and lower adjacent channels is because the spectrum analyzer and signal generators were not running from a common reference frequency, and one is slightly off.

[xmo]

rf-messkopf,

Thanks for taking the time do do some testing and cotribute to this discussion.

Yes, many spectrum analyzers can perform ACPR/ACLR testing.  The results may be useful in some context such as a learning exercise or for comparison purposes when constructing or modifying equipment.  Producing results comparable to the TIA spectrum management data is quite another matter.

Regarding your test results - your spectrum analyzer appears to show an average ACPR/ACLR of aabout -53 dBc.  This is based on integrtion of the power in an entire adjacent channel - in this case 12.5 kHz

ACPR measurement in accordance with a standard such as TIA-102 is performed in accordance with a specified procedure and in a defined bandwidth.  For example, TIA-102-CAAB-C states: "The adjacent channel power ratio shall meet or exceed 67 dB using an adjacentchannel power measurement bandwidth of 6 kHz and a resolution bandwidth of 100
Hz.

When measured in the entire adjacent channel, the result will be on the order of -40 dBc as shown in the attached Agilent test result.

One difference between the Agilent test and yours is that the Agilent test was measuring an actual P25 radio generating the standard P25 test signal whereas  you measured a signal generator which was set up incorrectly based on that wikipedia information which is wrong.

That information is apparently derived from System Fusion C4FM, not TIA-102 P25 C4FM (see attached excerpts from TIA-102-BAAA-A and Aeroflex)

Thanks again for participating.

[rf-messkopf]

Yes, many spectrum analyzers can perform ACPR/ACLR testing.  The results may be useful in some context such as a learning exercise or for comparison purposes when constructing or modifying equipment.  Producing results comparable to the TIA spectrum management data is quite another matter.

True. I'm not claiming that my casual measurement complies with the standard. That would require a lot more care (and a thorough reading of the relevant standards). And thank you for pointing out my incorrect setup and the error in the Wikipedia page.

Regarding your test results - your spectrum analyzer appears to show an average ACPR/ACLR of aabout -53 dBc.  This is based on integrtion of the power in an entire adjacent channel - in this case 12.5 kHz

[...]

When measured in the entire adjacent channel, the result will be on the order of -40 dBc as shown in the attached Agilent test result.

I've re-run the same measurement but with the 4-FSK/C4FM settings that you posted. Once with a Gaussian pulse filter with a BT of 0.3 as before, and for comparison with a root raised cosine filter with a roll-off factor of 0.4. This has a significant impact on the relative ACPR (-65 dB vs. -44 dB). I didn't take the trouble to check if the standard actually makes any requirements about pulse shaping, or if that is up to the user, as long as the spectrum requirements are satisfied.

Also, for comparison, I've attached the measurement with QPSK modulation, which requires less bandwidth.

One difference between the Agilent test and yours is that the Agilent test was measuring an actual P25 radio generating the standard P25 test signal

I wonder how big the difference would be compared to a pseudo random bit sequence as in the case of the signal generator. Most channel encoding techniques end up with bit sequences that are close to a uniformly distributed random sequence, but I'm not sure about this case.

[rf-messkopf]

I didn't take the trouble to check if the standard actually makes any requirements about pulse shaping, or if that is up to the user, as long as the spectrum requirements are satisfied.

According to this webpage TIA-102 P25 C4FM calls for a root raised cosine filter with a roll-off factor of 0.2, followed by an inverse sinc filter "to compensate for the sinc response of a P25 receiver integrate and dump filter". Not sure what this means exactly, and if we end up with an effective root raised cosine response in the end. It seems one has to thoroughly read the standards in order to do this properly. Also, the setup of the signal generator becomes complicated when a user defined filter (among other things) is needed.

[rf-messkopf]

I didn't take the trouble to check if the standard actually makes any requirements about pulse shaping, or if that is up to the user, as long as the spectrum requirements are satisfied.

According to this webpage TIA-102 P25 C4FM calls for a root raised cosine filter with a roll-off factor of 0.2, followed by an inverse sinc filter "to compensate for the sinc response of a P25 receiver integrate and dump filter". Not sure what this means exactly, and if we end up with an effective root raised cosine response in the end. It seems one has to thoroughly read the standards in order to do this properly. Also, the setup of the signal generator becomes complicated when a user defined filter (among other things) is needed.

I've been fiddling with the darned signal generator for more than three hours now because its firmware kept crashing. I recovered the embedded Windows, reinstalled the firmware, twiddled endlessly to get all instrument drivers behaving properly, and ran it through its internal calibrations. Windows based instruments are a pure joy.

After doing this I noticed that the sig gen actually supports APCO Phase 1 C4FM signals in its firmware, including the requisite pulse filtering. With that I get the exact same ACPR as in the Agilent test result which you posted, i.e., -40 dB relative ACPR in a 12.5 kHz wide neighboring channel.

Sorry, I totally was unaware that the generator actually supports TIA-102 P25 C4FM. Will post pictures tomorrow.

[xmo]

Here is some information from the TIA documents regarding the P25 C4FM modulation filtering and the standard test signal.

I was thinking that something like R&S WinIQSim2 has everything built in.

Looking forward to your pictures tomorrow.

[xmo]

File didn't attach

[rf-messkopf]

Here is some information from the TIA documents regarding the P25 C4FM modulation filtering and the standard test signal.

Thank you, very interesting. However, since I realized that the sig gen knows that standard, it is now only a button press. It then selects the correct pulse shaping filter for C4FM, i.e., a raised cosine filter (not a root raised cosine as I erroneously said yesterday), followed by a sinc filter. So we don't have to provide the impulse response as a custom filter. The data used in the attached measurement is a pseudo random bit sequence from a maximal-length linear feedback shift register of length 9 (PRBS9), i.e., with a period of 511 bits, as required by the TIA-102.CAAC-C excerpt you posted. This should now be a compliant test signal, and the result is the same as that by Agilent.

Also attached is, for the C4FM case, a frequency deviation pattern over 50 symbols measured with the signal analyzer (it also has a APCO25 compliant reconstruction filter implemented in its firmware), and an eye diagram over 400 symbols.

For comparison also did the same measurement for the APCO25 DQPSK signal.

I was thinking that something like R&S WinIQSim2 has everything built in.

Yes, but WinIQSim generates encrypted waveform files, and for the more complex digital standards the generator won't play them back unless you have the right option license installed. If you don't have the option, you would have to build the waveforms yourself if you want the correct frame structure and synchronization patterns, or the correct subcarriers in multicarrier standards. Since C4FM is only standard 4-FSK with a nonstandard filter which the generator knows, and the data is just a pseudorandom bit sequence, no license is needed.