[RNLD] Suggestions for simple recording devices for community members in a new MAP pilot project

Thu Sep 6 01:46:40 UTC 2018

>  I'm just wondering, what would be the attributes of "bad" audio be?

This is a good question. Nick answered re microphones. I'd like to touch on
this with particular emphasis on digital recording and phones since that's
my thing.

Aside from microphones, there's two subsystems that affect audio quality.
The first is the audio amp (pre-amp) between the microphone and the ADC
(analogue-to-digital converter). This has been easy to get right for
decades now, so this is rare. Bad implementations will be noisy (hiss when
recording nothing) or have some constant pitch noise (mostly laptops) due
to interference or fans. I've seen this on cheap mobile phones and
particularly on laptops (easily cured by using an external USB microphone).
I've also seen very good audio on cheap mobile phones. Mid to high level
phones are generally excellent. The most significant drawback of mobile
phones is that they are mono but there's solutions to that, see below.

The second subsystem is whatever encodes the raw output of the ADC to the
data format that's saved. The biggest impact is audio compression.
Obviously you can sidestep this by avoiding compression. Audio compression
has a poor reputation stemming from poor experiences with the (ancient) MP3
format. Bottom line, it's easier to just turn off and know it wont be a
problem.

Other sources of 'bad audio' can be somewhat more dynamic. One of the most
common is audio clipping. This is an operator problem that often stems from
outdated practice. In the days of analogue recording, best practice was to
ensure the audio level is at the top end of the range to maximize SNR
(signal-to-noise ratio). 16-bit digital recording has a *massive* SNR,
greater than 90dB in many cases, and unlike tape, if you exceed it, the
effect is catastrophic, with obnoxious clicks and crackles that will not
only destroy the audio for computational acoustic analysis, but will even
make it unpleasant and difficult to comprehend.
The solution is obviously to turn down the gain. You only need to see a
couple of 'pips' or a needle moving a bit, to be sure your audio is
working. You can turn it up with processing later on, with no ill effects.

Another form of bad audio stems from dynamic range compression and/or the
related automatic gain control. TLDR; use manual gain, or turn automatic
gain control off if you can, or at least be aware of when it might be a
problem.

Longer explanation: The pre-amp (hardware) I mentioned earlier is
controlled by software. The software tries to obtain a sufficiently loud
recording based on the conditions. So if you're quiet, it ramps the gain
up, if you're loud, it ramps it down. There's actually a really fast method
to cope with transients (dynamic range compression), and a slower method to
set the general gain level (AGC). Mobile phone AGC is particularly
aggressive since it has to cope with recording everything from quiet
ambient audio, up to yelling into the microphone at close range! AGC can be
an issue with dedicated recorders. Consider this. You're recording a
participant in a noisy village with a mobile phone placed closed to them.
The cultural norm is to wait some time before answering a question. The
problem is that the phone ramps up the gain and all you can hear is the
pots and pans from the nearby kitchen. Then when the participant starts
speaking, the audio clips (is distorted)  for a moment, before clearing up
and the pots and pans recede to the background. That's what AGC might look
like in a typical environment. The solution is to turn AGC off, and do a
quick sound check with the participant speaking. If ambient noise is
remotely an issue then close mic'ing gives you a better result than a
directional microphone. Hence why you see lapels on reporters when they are
standing in front of some unfolding disaster on cable news.

So while I'm here, some common issues/misconceptions of audio recording.
Sorry for those that this is obvious.

In my view the majority of digital recording devices, except for laptops
and cheap mobile phones, is more than up to the challenge, and any noise or
distortion is so small so as to be negligible. The largest source of
unwanted signal (noise) stems from microphone strategy. This is also the
best way to approach evaluating recording devices in general. When you buy
a dedicated recorder, you aren't getting magical capability to produce
better audio. What you do get is a robust unit with good battery life, easy
to reach audio controls, good input options (like maybe XLR) and some kind
of basic functional directional microphone.

These are significant but they are achievable in other ways. *Most* third
party shotgun microphones are better than what you get on a Zoom. Ergo,
plug that one into the device of your choosing and get better audio than a
zoom. That doesn't negate the other reasons for a Zoom, or other, record as
above.

Next, there's a whole load of mythology in this genre around bit depth and
sample rate. Bottom line, but if you are recording voice, you are not doing
yourself any favours by using a sample rate more than you need. It's not
like camera megapixels, you can't scale down and get a better result.
24-bit is pointless. 96/192 or whatever Khz is harmful audiophile snake
oil* used to put bigger numbers on the box for marketing purposes. That's
why Steven said he's recording at 16KHz.

Returning to the parent topic, I suggest a couple of recording options for
either end of the budget.

1. The 'ghetto' lapel. Using a mobile phone in a lanyard or in a top
pocket.

This is basically radio lapels on a budget. For best results, use a
recorder app that allows you to manually set the gain. Cleverly you can
deploy this to several people at the same time and sync them up, giving you
as many audio channels as you want. David Nathan has drawn attention to the
fact that stereo audio ought to be considered the minimum (audio positions
is a kind of metadata).

2. Use of a camera with an external microphone.

Cameras are rugged reliable devices with good tripod mounts. Ideal for
capturing video and audio. A lot of what's good about dedicated audio
recorders (rugged, battery life, big memory capacity) also applies to
cameras.

You  can turn down the video quality to maximize record length. The video
will still be outstanding. Audio input is sadly a feature you have to hunt
for, and you wont be getting XLRs, but audio input can be found on cheaper
cameras, for example the Panasonic  FZ1000 (there's a newer one now)  which
has been a workhorse of mine. It soldiered on while everything else broke
and failed. It has proper gain controls and on-screen audio level
indication. As a far-mic solution, you  need  a directional microphone.
There are many shotgun microphones that will fit the flash shoe at the top
of a camera. I bought a cheap one off eBay and it's outstanding. Camera
audio will be compressed, but it will be 48KHz AAC  usually, more than
sufficient. There's a step required to extract the audio if you need stand
alone audio files, which may be a factor.

Sorry if I strayed the topic a bit :)

* https://people.xiph.org/~xiphmont/demo/neil-young.html

On Tue, 28 Aug 2018 at 14:03, Hugh Paterson <hugh_paterson at sil.org> wrote:

> Nick Williams,
>
> I'm just wondering, what would be the attributes of "bad" audio be? I'm
> never read a technology review of a "bad" piece of equipment in LD&C.
>
> all the best,
> - Hugh
>
> On Mon, Aug 27, 2018 at 7:54 PM, nick williams <n.jay.williams at gmail.com>
> wrote:
>
>> Zoom Q8 with a shotgun microphone (if you can buy fewer units and share
>> or up your budget slightly). you might be able to get a used one for under
>> $300. i know it is a bit over your budget, but they are very good and video
>> is the way to go unless people are for some reason uncomfortable with
>> video. the audio is quite good too. it has xlr inputs as well and
>> possibility of four channels total.
>>
>> On Fri, Aug 24, 2018 at 8:17 PM Steven Bird <stevenbird1 at gmail.com>
>> wrote:
>>
>>> Hi Heather,
>>>
>>> Plenty of voice recorders support uncompressed recording. I've been
>>> using Smart Recorder, Parrot, and Hi-Q, all of which record in RIFF (=wav)
>>> format. The first two also support sample rate selection, and I choose
>>> 16kHz 16 bit mono.
>>>
>>> That's 32kB/s or ~0.1GB/h. For A$40 you can fit out your phone with a
>>> 64GB micro SD card and record for ~600 hours.
>>>
>>> -Steven
>>>
>>> --
>>> Steven Bird  http://stevenbird.net
>>> Professor, College of Indigenous Futures, Arts and Society, Charles
>>> Darwin University
>>> Linguist, Nawarddeken Academy, Kabulwarnamyo Community, Northern
>>> Territory
>>> Research Scientist, International Computer Science Institute, University of
>>> California Berkeley
>>>
>>>
>>> On Tue, 21 Aug 2018 at 21:10, Julia Sallabank <js72 at soas.ac.uk> wrote:
>>>
>>>> Dear Heather
>>>>
>>>> I've used the Zoom H2N with community members, together with very
>>>> simplified instructions, but they still had problems with it. The easiest
>>>> thing would be to use the record or video function on mobile phones. I know
>>>> there are other communities using mobile phones for language documentation,
>>>> and phones also have the advantage of being familiar and thus intruding
>>>> less on the conversation. I know that the recording quality is mp3 not
>>>> .wav, but for language learning and MAP purposes that should not be a
>>>> problem. I would also recommend not recording at all for the first
>>>> meetings, to put people at their ease.
>>>>
>>>> Good luck!
>>>>
>>>> Julia
>>>>
>>>>
>>>>
>>>> On 18 August 2018 at 23:38, Heather souter <hsouter at gmail.com> wrote:
>>>>
>>>>> Taanshi!  Hello!
>>>>>
>>>>> Heather Souter d-ishinikaashon.  En Michif Camperville, Manitoba,
>>>>> Canada oschi niya. Daweeyiteen lii zavii.   Mahti wiichihin...
>>>>>
>>>>> My name is Heather Souter.  I am a Michif from Camperville, Manitoba,
>>>>> Canada.  I need some advice.  Please help...
>>>>>
>>>>> Looking for some collective wisdom here.  I need a range of ideas for
>>>>> the lowest tech to the highest (and a few in between?!) recording devices
>>>>> to use with a group of community-based Indigenous language learners of
>>>>> widely ranging digital skill levels?  The budget is CA$ 160..?
>>>>>
>>>>> Zoom H1 and Zoom H1n have been suggested.  Any other suggestions on
>>>>> the lower  and slightly higher tech sides?  Something my non-tech savy
>>>>> elderly auntie could use and then maybe something with a XLR input?
>>>>>
>>>>> It is for an upcoming MAP pilot project..
>>>>>
>>>>> Kihchi-marsii.  Thank- you!
>>>>>
>>>>> Eekoshi pitamaa. That’s it for now.
>>>>>
>>>>> Heather
>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>> --
>>>> Dr. Julia Sallabank
>>>> Reader in Language Policy and Revitalisation, School of Languages,
>>>> Cultures and Linguistics
>>>> Associate Director for Learning and Teaching Quality (PGT)
>>>> SOAS, University of London,
>>>> London WC1H 0XG, UK
>>>>
>>>> Tel. +44 (0)20 7898 4326
>>>>
>>>
>
>
> --
> *Hugh Paterson III *Innovation Analyst
> *Innovation Development & Experimentation*, *SIL International*
>
> *Web*: Contact & CV <http://hughandbecky.us/Hugh-CV/>
>
>
>

-- 
Regards,

Mat Bettinson
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/resource-network-linguistic-diversity/attachments/20180906/527ded96/attachment.htm>