Sound wave with squashed valleys after recording a call with appropriate level

Report · Mar 20, 2024

Dear community,

My issue is most probably related to hardware issues when recording calls, therefore not directly related to AU. Sometimes I get a sound wawe with flattened or squashed valleys. I cannot figure out why it is happening even though I can sort of correct it with AU. So I come to you with two questions:

1. Why this might be happening?

2. Am I solving this situation properly with AU using the "graphic phase shifter" effect? If so, how can I properly determine the points and its values in the graph?

3. Is the quality of my recording affected if I get such flattenned valleys (that do not sound like clipping_?

Let me detail everything

How we record and what is happening

Me and my team produce a news podcast and we usually contact our sources via WhatsApp (a phone to phone call). The phone making the call is connected to a ZoomH6 recorderMic Inputs with a P2 / P10 cable to one of its , so the call can be recorded (at 44,100 KHz / 16 bit depth). Not the ideal platform, but it's usually the most agile way to contact people.

We always monitor levels properly, so clipping is not an issue.

We edit the interview with AU CC 2019, on a Mac with iOS Sierra 10.12.6

The problem is that we sometimes end up with a recording that has flattened valleys on the sound wave. We don't know what causes it or what it is.

The other day for example, we recorded with an iPhone 13 connected to the H6 ang got the squashed wave. We changed the phone, called our source again and the issue didn't happen. However, the problem also occurs with this second phone when the reporter records from her home with another H6.

How we are solving it

We found out that in the Waveform panel, the effect "Graphic Phase Shifter" (Effects > Special > Graphic Phase Shifter) can correct this notable assymetry. We set two points on the graph, one on 0 Hz and the other on 20050 Hz, and set the same negative phase degree (since the flattening happens on the negative phase of the wave). This brings back the peaks as if the wave was normally recorded.

_____

I appreciate your help and attention

Raphael

Report · Mar 20, 2024

That's a perfectly acceptable way of fixing this issue - it's been done like that for years. Usually it's the result of sound being carried on breath that's moving forwards, and therefore moving the mic diaphragm forward - which results in larger positive peaks. You can often minimise it by using a pop filter, which will impede the air flow without impeding the speech, although that's not going to be so easy with a phone... and yes a phone could easily exacerbate this, as you're much closer to the mic in it, as a rule. Usually it's less of a problem with headsets.

Does it matter? The one thing that it does do is restrict your dynamic range slightly, as you haven't used up all the available 'space' in your recording. And yes, it looks a bit weird. You get a similar thing happening if you record some brass instruments - much larger positive peaks than negative ones. If it's only a few dB you generally live with it. How large is the difference between positive and negative peak sizes you are experiencing?

Report · Mar 20, 2024

Many thanks for your reply, Steve. (sorry everybody for the typos in the original message). Let me bring two more questions.

I'll take the example of image 01 attached to this reply. This excerpt shows an uncompressed recording normalized to -24 LUFS made with an iPhone 13 connected to a Zoom H6. The flattened valleys reach -12 db on the negative phase of the wave, while the positive phase has most of its peaks close to -9db, sometimes reaching -7db.

Between thethe dotted lines of the marker is the voice of the interviewer speaking to her phone. Here we do not have the flattened peaks.

In this example, is there an appropriate way to calculate which values to apply in the Graphic Phase Shifter to reduce the assymetry?

But there is another thing that bugs us. When possible, we do a backup recording - that is, someone else on the call records it using another Zoom H6 connected to his/her phone. Oftentimes one recording shows the flattened valley, but the other does not.

Image 02 here shows it. It is the exact same excerpt of image 01, but recorded by another person on the same WhatsApp call, with his phone (a Samsung J6 Plus) connected to another Zoom H6.

They are both recordings of the same WhatsApp call. The one with the flattened valleys is from an iPhone 13 connected to a Zoom H6. The other one is from a Samsung J6 Plus connected to another Zoom H6.

If I understood correctly, the problem starts with the person receiving the call and speaking to his/her phone. If so, why we have different outcomes for the same recordings being made by other devices?

Report · Mar 21, 2024

The first screengrab is the interesting one. It's pretty obvious (to me, anyway) that the negative excursions have been limited in some way - you simply can't get negative excursions to cut off that accurately when the positive ones don't. Not only that, but some of these level-limited blocks are happening at different levels, which is quite a big clue... Is it the iPhone or the H6? And very significantly, how are they connected? (Here's a clue - I'm pretty sure it's not the H6...)

Report · Mar 21, 2024

Yes, that's well noted, the limiting vairies. it seems that this issue happens more often with iPhones. I cannot remember if we have ever experienced it with another type of phone. Four of our team members who have experienced this problem use iPhones. They are connected to the Zoom H6 through a cable with a 1/4'' TS male connected to one of Zoom's mic inputs / 1/8'' male connected to an adapter (1/8'' female - Thunderbolt male) which is then connected to the iPhone thunderbolt port. However, it is not always that we have the wave limitation when using this setup.

Report · Mar 21, 2024

Sorry, where I wrote thunderbolt port it should actually be lightning port