Thursday, 31 August 2017

Fine tuning noise floor testing methodology

A dongle's noise floor represents a value in decibels for a particular frequency and gain setting.
Measuring the noise floor of any RTL-SDR dongle is beneficial for the following:
- establishing whether any modding or attempted improvement resulted in the expected outcome,
- comparing dongles as long as the same testing equipment and commands used.
Comparisons are NOT recommended by the software developer, kmkeen on rtl_power's webpage, yet it's commonly done by reviewers, including me, as 1) numbers are easy to comprehend, and 2) resulting graphs look cool in a post.
Not being content with the accepted methods, and looking for a reliable way to test 30+ dongles, went ahead with following experiments, using the same blog v.3 as test subject - full size dongle with thermal pads and metal case, all other dongles are smaller / lighter. Nooelec's Mini 2 / 2+ Al donges are the two exceptions, but they're far from widespread and do not ship with thermal pads, takes ages to warm up.
Equipment was two digital thermometers taped to metal case to monitor temperature changes, and the same pigtail / resistor throughout the test.
Values in Celsius.

Hot vs Cold

Assumption: a colder dongle has a lower noise floor. Fully warmed up dongle after one hour at full gain at 38 degrees, cold dongle at 24 degrees when finishing scan, 5 minute pass:

Result: depends on frequency. I listen to airband and shortwave with an upconverter, chew numbers below:

Higher up, situation is slightly reversed with a hot dongle having a slight advantage.

Reaching thermal equilibrium

All dongles need to warm up to operating temperature to be able to compare apples to apples, question is, for how long?
To find out, full gain whilst videotaping thermometers, then compiled values against time:

25 minutes is a safe bet, 30 minutes is more than enough - smaller dongles e.g. SMArt and Nano 3 reach operating temperature much faster, Nano 3 is on the ball in no time.

Time required for a scan

Keenerd recommends "It is important to let the scan run for at least 15 minutes. Any less and repeatability suffers". Fair enough, went by this advice for years, but a full noise profile (at 0-5-10-15-20-25-30-35-40-45-50 db gain) takes 11 measurements, that's at least three hours with a warm-up run. Absolutely fine for an individual dongle test, would take ages with multiple dongles.
Assumption: results do not significantly change as long as dongle is warmed up properly.
To test, 30 min at full gain, then 1-2-5-10-15-20-30 minute passes, data below (click or tap for full screen):

Full data set on demand, apart from a few frequencies, I did not see any major difference which would justify spending hours for a noise test.

Reliability and repeatability

Noise floor testing is fun, as long as testing environment and equipment follows the golden rule of comparisons: only one variable, which is the dongle used. This is impossible with multiple connector standards, then environmental and circumstantial factors will change results:
- Dongle position: even a small push on a crowded table with multiple RFI sources around skews by 0.2 - 0.5 dB. Ran multiple tests, LCD screens and especially power supplies for smartphones / tablets wreck results, the closer, the worse.
- Connectors, starting from USB port. If you replace one variable, it must be noted.
- Ambient temperature. Premium gear (SMArt, v.3, XTRs, etc) are very sensitive. Just because you can have a receiver for less than $30 in a week on the floor, do not underestimate what contents can do.

What I'll go with

Warm-up run at 50 gain for 15 minutes, v.3 was within 2 degrees of operational temperature by this stage.
Secondary at 45 gain for 5 minutes.
Tertiary at 40 gain for 5 minutes. 25 minutes elapsed, more than enough, apart from ADS-B nobody uses dongles at 40 dB on the slider anyways.
Measurements at 35, 30, 25, 20, 15, 10, 5, 0 dB gain for three minutes each.
Total time: 49 minutes for one dongle.

I know nothing

Agree? Disagree? You know a better way? Know more than I do? Help me, and many others, contribute, comment, share your wisdom.
I want to hear from you. Let me, and countless readers know how to do better.
No man is an island.

No comments:

Post a Comment