Help me understand this spectrogram - band at 5500hz

[oester]

Most of my spectrograms look like this. I'm using the audio feed from my Unifi Cameras. What's the band at about 5500 in every one the sound files?

[birdnet-go-bot]

Automated message from birdnet-go support assistant.

A constant, sharp horizontal band across the spectrogram at a fixed frequency (like 5.5 kHz) is electronic noise coming from the camera audio stream rather than something created by BirdNET-Go. On IP cameras like UniFi Protect cameras, this is usually coil whine or switching noise from the internal DC-DC power converter or PoE circuitry leaking into the camera microphone preamp.

You can often reduce how prominent it is by lowering the camera microphone sensitivity in UniFi Protect settings (for example, reducing it from 100% down to around 50% to 70%), which drops the electronic noise floor.

I would avoid trying to filter it out with a notch or band-reject filter in BirdNET-Go. The 5.5 kHz range sits right in the core frequency band for many songbirds, so carving out that frequency can degrade detection accuracy. The BirdNET classifier typically handles narrow stationary tones without issue if the overall audio is clear.

I'm an automated assistant so take this as a starting point. If it doesn't help, reply here and the maintainer will take a look.

[tphakala]

Most of my spectrograms look like this. I'm using the audio feed from my Unifi Cameras. What's the band at about 5500 in every one the sound files?

That's odd. There was one other report of same exact issue, I dismissed at environmental noise as there was no evidence of anything other.

• What file format are you using for clip exports? If you switch to any other format than you are currently using, does this still appear?

• If that clip does not contain anything privacy sensitive, could you please attach that clip to this thread.

• Do you have any systems with microphone input, does they have this or is this occurring only in RTSP streams?

I need more details of your setup as I don't see this myself so I don't know where to start looking at to debug this.

[oester]

• File format is the default (m4a). I'm attaching a mp3 format of the file since github won't allow me to attach an m4a file.
• I dod try switching to mp3 for a few recordings, same banding on the Spectrogram
• No Microphone input devices (tho I have an AudioMoth on the way)
• I uploaded a support dump (fec06b44-f3a2-497c-bd2c-53ce98555f85) so you can see my config.
  Blue_Jay_2026-09-01_10-18-32_0.0-15.0s.mp3

[tphakala]

Thanks, got the clip, could it be cicadas or some other insects?

[birdnet-go-bot]

Automated message from birdnet-go support assistant.

The dump config shows Dog-day Cicada (Neotibicen canicularis) in your excluded species list. Dog-day cicada calls are continuous drones centered right at 5.5 kHz to 6 kHz. During late summer, active cicadas create a persistent horizontal band across spectrograms on outdoor audio feeds.

If the band is strongest during warm daylight hours and fades at night or during cooler weather, that points to cicadas or similar insect noise rather than camera hardware or audio encoding.

I'm an automated assistant so take this as a starting point. If it doesn't help, reply here and the maintainer will take a look.

[oester]

Thanks, got the clip, could it be cicadas or some other insects?

None were active at that time (10AM local) but they have been active in the past month. If I listen to the camera feed it seems clean.

There are some insect noises present,but I don't think it could account for the continuous band that I see.

[tphakala]

I applied high pass filter at 5000kHz and I can hear a sound which resembles cicadas or something similar, we don't have them here. But in any case I am 99% confident that it is real sound at your location, not an bug caused artifact.

Edit, I increased gain and now I am 100% sure it is insects of some sorts.

[oester]

I applied high pass filter at 5000kHz and I can hear a sound which resembles cicadas or something similar, we don't have them here. But in any case I am 99% confident that it is real sound at your location, not an bug caused artifact.

Ah, OK. Let me investigate further. The camera could be more sensitive than I realize. Going to get a recording from my iphone right under the camera.

[oester]

@tphakala I did some more investigation, doing recordings with Merlin ID on my phone. I'm now convinced your conclusion is correct, it's a cicada or something similar. Does make it a challenge to pick out actual bird calls, but I'm surprised that BirdNET-go finds them. I would love to find a way to filter out the insect noises present but I don't think there is a way to do that. Thanks for investigating.