Insync Tech Talk , microsoft teams , | 2021/01/25 at 4:53pm

The No.1 Video Call killer could be a thing of the past

by Tim Presland

Regardless of what cloud-based Video Conferencing (or more now commonly aka “Video calling”) system you currently use or have been exposed to in the past – either as the host or as a guest, you would have had a bad meeting experience due to one common failing point – bad audio speech!

So, put your virtual hand up and keep it up if at least one the following scenarios resonates?

  • Someone says … “can everyone please on mute if you’re not talking?
  • The speaker’s audio drops in and out or clips
  • Robot sounding or even drawn out sounding voices
  • Too much background noise
  • Speakers voice is not loud enough even if everyone else seems fine.

Sure, there might be many more instances you can think of but these would more certainly be the most common experiences I have come across when speaking and dealing with hundreds of people over the years as an integrator of systems.

To rectify bad audio for Video Callings, I’ll broadly categorise into 2 common pillars that will always need addressing in an attempt to ensure that audio quality is of premium quality on any given cloud-based Video Calling system.

Hardware / Software / Configuration – without deep-diving too much, some examples that are notorious of causing bad audio are when connecting to your cloud Video Calling platform via VPN without split tunnelling (can cause robot or drawn out voices),  Windows drivers need updating (audio clipping) or not being supplied certified noise cancelling headsets (too much background noise).

With regards to Adoption Change Management, it might simply be that the end user is not aware of, understands or appreciates the importance of using supplied certified noise cancelling headsets (which strongly migrates background noise).

It’s often the case that certified headsets are left behind, lost or are a hassle to wear, especially when being mobile, so end users will simple self-resolve by using ubiquitous consumer standard grand headsets such as “ear buds” for their video calls.

However, what the end user may not realise is that these devices can pick up and amplify background noise without their knowledge.  These devices simply aren’t designed for cloud-based software codec systems.

The thing is, it’s a double-edged sword!    The wearer isn’t affected, they can usually hear fine but all the other attendees will have a bad experience with amplified background noise and thus the meeting is ruined.   In turn, the natural assumption is that the platform being used is at the heart of the problem and not a singular endpoint device that is the case which can be simply rectified.

So, does that mean you can never use “earbuds” for your Video Calls?  No, it doesn’t.  If earbuds are to be used, ACM simply educates the user on what can happen and advises they should be in a quiet area for their call as this will greatly reduce background noise pick up.  I wish I had a dollar for every time I have explained this.

Real world examples of bad audio within a Video Call using not certified headsets.  

To give weight to this scenario, I’ll give two quick extreme examples that I was directly involved with when someone wasn’t using a certified headset and the meeting was ruined.  The first example wasn’t actually a background noise amplification issue but a constant buzzing sound.   For those familiar with a guitar amplifier earthing buzz, that’s what it was like but in our meeting.   As host, I did the normal thing of getting everyone to mute which didn’t resolve the issue and after that, the most likely step for anyone else would be to restart the meeting.  I on the other hand asked if anyone was wearing earbuds and if so, can they be unplugged and sure enough, this fixed the problem.

The 2nd example only happened as recently as the start of this year when I had a one on one call with a customer that was sitting at their desk in a corporate open air office.  I could hear too many voices and it was loud.  I asked if the customer was wearing earbuds explaining there was an issue and the reply however was that they were wearing Ear Pods.  I explained the importance of wearing certified headsets to which my customer was shocked to know this as their “ear pods were expensive!”

 

So, does Microsoft have a solution for bad audio in Microsoft Teams?

Yeah …. they might well indeed and it’s called Noise Suppression.   But firstly before I cover off testing results, of course, I’ll put in the usual disclaimers of … it’s not a magic bullet, make sure your setup is right (etc), end users are still very much advised and encouraged to wear certified headsets and most of all, be aware of noise pollution in their area.  BUT if all of this has gone out the window then Microsoft may have an AI safety net setting to provide clear speech audio from any device (….potentially).

Straight from Microsoft’s mouth, albeit website, they say this…

Sounds good right!?! (yes pun intended).  So, at this very early stage of its release, our preliminary results achieved from our internal testing it turns out to be very, very promising indeed.   In fact, wearing my old school audio engineering hat and just general tech nerd helmet I’ll go as far as to say …“WHOA MAN!! THIS IS FRICKEN AWESOME!!!”

 

Our test setup.

This wasn’t an over the top test for starters, and I didn’t go low level like running spectrum analysers or picking apart phase cancellation captures but I did have Windows TASK manager open for resource observations.  The test was more of a real end user test but I will also give my 2 cents on a few things as well as considerations on what can and can’t be done at this time of writing.

The test was carried out on a peer-to-peer Teams Meeting.  Endpoints were Surface Book 2 and Surface Laptop 2 on Windows 10, thick client install of Microsoft Teams with all the latest updates, running NBN 100 / 50. 4G/5G was not tested at the time.  Video was on for both endpoints from internal cameras and the headsets were Sennheiser MB660Jabra Evolve 75 and Surface Laptop 2 microphone.

You can see from the test matrix I wanted to test each setting on a certified headset running both on wired connection and Bluetooth.

And then test on a non-certified device in different environments. Originally it was meant to be earbuds but ended up using the Laptop 2 microphone.

Due to COVID-19 social distancing, we had to test at home but pumped up enough noise (streaming music) to gain some truly incredible positive results. Please note that Noise Suppression is only available on the Windows 10 client at the time of this writing.   It is not currently available on Android, iOS, Browser or Microsoft Teams Room (MTR)

 

 

Where is the noise suppression setting in Microsoft Teams?

Here are some screen shots directly ripped from Microsoft 

Is the noise suppression setting dynamic?

The first thing that I asked myself was …“is the setting dynamic?”, meaning if I’m in a meeting and change the setting, is it applied on the fly to the current session?  The short answer is no.   If you change a setting you will need to hop out the meeting and back in for the new setting to be applied.

And to spell out what might seem obvious but just in case – the setting applied is only for that END POINT – that user, that applied the setting and not the other end points that are joined to the current meeting.  This is regardless if you are the host or guest.

 

What were the testing results?

I’ll skip to the chase.  At the time of this writing, regardless of what headset or microphone we used, only the HIGH setting had a positive impact, and it was really quite significant.  The settings of Auto, Low or Off sounded the same especially using the certified Jabra Evolve 75 and the Jabra by design was probably doing the “heavy lifting” from the very start anyway, so with that in mind, we ditched the Jabra testing and focussed on using the laptop microphone.

With the Auto, Low or Off setting for the laptop microphone there was a slight noticeable ambient background noise when my colleague wasn’t talking and when I say “noticeable”, I was intently listening because I was testing.  In a normal meeting, you wouldn’t think twice, it was a somewhat a “normal” meeting experience, which is to say having Auto, Low or Off at the time of this testing was without any discernible impact.

With HIGH turned on though, it was apparent from the get-go the AI had kicked in fully and was doing something.  It was quiet!   And it was wonderful.   My colleague read her repeatable script we were using for the tests and she was clear.  Really clear!   We then decided to introduce some noise in the way of streaming music from her mobile phone and I couldn’t stop grinning.  The music was about 95% filtered out and that remaining 5% you could hear but it was low and certainly wasn’t distracting or impacting my colleague’s voice – she remained clear.    The music was then turned up to full level and shoved right in front of the microphone as we both laughed.  She was laughing because “…it’s like a full-on party happening now” and I was laughing because I could hear every word very, very clearly.

For our testing, the HIGH setting produced something very special.

 

Why was it so special?

Noise suppression technology isn’t new and I’ve been wowed with other products on the market which yield just as excellent results.  But this is a cloud service and doesn’t require additional hardware or software installs. And being Microsoft AI too, it’s only going to improve as it matures.

It’s special though IMO on 2 counts.   Firstly, the vocal that came though sounded natural.  The background noise was removed, and it didn’t it taint the vocal audio I wanted to hear.  It didn’t clip the vocal, it didn’t sound strange, it didn’t sound synthetic, it sounded completely natural.

The second thing is, out of the box, if someone doesn’t use a certified headset or is not aware of noise pollution it has a significate chance to reduce all that stress, anxiety and work arounds for the host, guest and certainly not forgetting the poor support crew to simply make audio in meetings work as intended – to have clear speech.

 

So, did the High setting hog a lot of resources?

Quick answer is, no it didn’t seem to despite Microsoft putting in a caveat.

Having Task Manager open for resource observations showed a little increase but not that much and to be honest, Microsoft Teams is a general resource hog anyway it simply wouldn’t warrant, IMO, not having it on.   My noise suppression setting is always on HIGH.

 

My 2 cents and wrap up.

In our testing we did have my colleague’s wife happen to chat in on the call too in passing (remember testing was done at home and from a laptop microphone) and she was clear as a bell too when the music was playing, which got me to thinking there’s more to the noise suppression setting than just “removing background noise”.  I dare say (without investigation) there would have to be noise gate and noise compression functions at play here too.  So, if that’s the case, at what tolerance does Microsoft Teams noise suppression AI determine what vocal audio should be included from any given distance and then, at what level is that vocal volume boosted?  All very interesting and something I forward look to investigating with another blog at another time.

 

Quick reference note reminders

  1. At the time of writing, Noise Suppression is only on Windows 10 Client – not available on Android, iOS, Browser or MTR.
  2. The Noise Suppression setting is not dynamic, meaning it can’t be changed on the fly, you will need to exit and hop back into the meeting for it to be working.
  3. At the time of writing, our testing revealed only the HIGH seemed to have a positive impact.
  4. The High setting …
  • Suppresses all background sound that isn’t speech.
  • For this option, your computer’s processor must support Advanced Vector Extensions 2 (AVX2).
  • This option is currently not available if the meeting or call is being recorded or live captions is turned on.
  • Enabling this option uses more computer resources.