The Sonocent guide to
recording devices
Introduction
What is the best way of recording lectures?
In the past, a specialised recorder – rst of all cassette, then minidisk, then digital recorder – was the
only way to take a reasonable recording. But in recent years PCs, phones and, latterly, tablets have
been equipped with quality microphones and now oer a viable alternative.
Many DSA assessors and Sonocent Audio Notetaker users ask us whether, in light of these advances,
there is still signicant dierence in the standard of recordings taken with dierent recording devices.
So, as audio experts with decades of experience in this area, we thought it time we did some proper
tests to nd out exactly how much variation in sound quality there is between devices, and what can
be done to take the best recording possible.
This guide is the result of our investigations. Its purpose is to equip you with all the information you
will need when advising students how best to take recordings of their lectures.
If you have any questions about the guide or our ndings, please don’t hesitate to get in touch by
visiting www.sonocent.com/dsa or contacting [email protected].
Best wishes,
Roger Tucker
Founder and CEO, Sonocent
The Sonocent guide to recording devices
Contents
Section 1 – Summary
1. Conclusion
2. The devices we tested
3. Our tests
4. Using Sonocent software
5. Audio post-processing
Section 2 – What makes for a good recording?
1. Sound level
2. Intelligibility
3. Sound colour
4. Interference
Section 3 – What makes for a bad recording?
1. Mobile phone interference
2. Sound from the recording surface
3. Extraneous sounds – coughs, talking, movement
Section 4 – Recommendations
1. Tips
2. Noise cancellation
3. University recordings
4. Android
Section 5 – Device guide
Appendix 1 – Recording devices tested
Appendix 2 – Spectrogram analysis
The Sonocent guide to recording devices
Section 1 – Summary
Our conclusion
The conclusion may seem an unusual place to start this guide, but the results of our tests surprised us,
and are worth sharing rst.
In a nutshell, we found there to be little dierence in raw audio quality between the devices tested.
It didn’t matter if it was an iPhone or a top-of-the-range digital voice recorder (DVR) equipped with
an external microphone; intelligibility was comparable between all recording options (listen to our
summary examples and judge for yourself).
On the other hand, there was a dierence in the level of interference from device to device. That
is, external sounds which can make it dicult to understand recorded speech. Each device had
its own susceptibilities, and some can be negated by taking some simple measures. We give our
recommendations for achieving optimum recording quality, and minimising the eects of interference,
throughout this guide and have also included a ‘Recording Tips’ section at the end for students.
We also briey surveyed the native recording quality of a range of Android devices. As with recordings
taken with laptops, quality varied enormously. The better ones were as eective as our test devices,
the poorest ones were very bad.
See Appendix 1 for full details of the devices tested.
Our main test devices were:
• iPhone 4 using ‘Sonocent Recorder’
• iPad/iPad mini using ‘Sonocent Recorder’
• Olympus DM670
• Philips DVT6000
• Olympus DM670 with Olympus ME32 external analogue microphone
For the second tests (see next section), we also tested:
• Samson Go external (digital) microphone, plugged into a laptop
• Kindle Fire HDX using ‘Simple Recording App’
Prior to releasing this report, we also performed informal tests on the very latest Olympus DVRs,
the DM901 and DM7, and a range of Android devices. The Olympus DVRs performed as well as the
DM670 tested, and were better in the respect that they do not have the same susceptibility to mobile
interference – see section 3.1. The Android results were a dierent story – see section 4.4.
The Sonocent guide to recording devices
Our tests
In the rst of our tests, we made live recordings in average-sized lecture theatres (approx 200 seats) of
two lecturers who had both chosen not to use the PA system. We sat in the back row each time, to give
a worst-case scenario, and recorded on all the main devices. Both lectures were well attended with well
over 50% occupancy of the seating. While one lecturer projected reasonably well, the other spoke very
quietly. This was the very worst-case scenario for recording a lecture – a quiet speaker, no PA, sitting at
the back with lots of other people – and a very tough test for any recording device.
In the second tests, we played a lecture that had previously been recorded using a headset
microphone through the PA system of two lecture rooms, one large lecture room (seating around 400)
with good acoustics and one small room with poor acoustics (seating around 40). The repeatability of
the environment meant that we could record as many dierent conditions as we liked, and we added
a laptop, with an external Samson Go digital microphone, and a Kindle Fire HDX with a recording app
to the array of devices tested. There were no people in the two rooms during the recording, except for
the two people making the recordings.
In the informal tests of new DVRs and Android phones, we played a recording quietly from a laptop at
2m and 5m, and recorded it using the test devices alongside one or more of our main test devices. We
were then able to judge how well the new devices performed compared to our main test devices, both
by listening and by carefully observing spectrograms of the recordings.
Using Sonocent software with audio devices
Although this guide has been written to answer the question of how audio quality varies between
devices, it is important to remember that the decision on which device to use will primarily depend on
how the student wants to take their notes, their needs and the available budget.
In general, simply having access to an audio recording is only one step along the road to eective
note taking, either for study or as a form of accommodation for a disability. Listening back to a long
recording without actively engaging with the content is time consuming and, studies indicate, of little
benet. Using recordings as part of a study strategy which involves active listening, summarising and
review is likely to be of far greater value.
Sonocent’s note-taking software supports recording on PC, Mac, phone, tablet or digital recorder
through our products Audio Notetaker (PC/Mac) and Sonocent recorder (iPhone/iPad). The software
oers a wide choice of annotation features - sectioning, colouring, typing and pictures - as well as a
visual display of the audio recording.
Our approach to note taking allows students to colour-code parts of their lecture recording before
revisiting this material to take notes in their own time. This process is simplest and most ecient when
the student can annotate as they go, for which they will need a PC or mobile/tablet.
The Sonocent guide to recording devices
Audio post-processing
It is also important to note that the audio recording quality delivered by most devices can be improved
by playing the recording through audio software such as Audio Notetaker itself.
Audio Notetaker features various post-processing playback options: noise cancellation (from v3),
automatic volume control, key-click suppression, rumble removal. This helps make all recordings
usable, whatever recording device was initially deployed.
However, for the purpose of our comparison, the only feature that we used when testing Audio
Notetaker was volume adjustment, after which the speech level from all the devices was approximately
the same.
The Sonocent guide to recording devices
Section 2- What makes for a good recording?
Sound Level
If you listen back to a recording and the speech is quiet, it is natural to assume that the recording
quality was poor. However, sound volume and audio quality are not the same. Speech from some
recording devices sounds louder because these devices include Automatic Gain Control which
dynamically adjusts the volume of the recording. However, volume dierences are easily corrected and
Audio Notetaker (v4) will do this automatically when recordings are imported.
During our tests we found a large variation in the sound level of the recordings taken. All the
devices oered automatic recording level setting, but this seemed to work to dierent degrees of
eectiveness, with the Apple devices producing the quietest audio and the Olympus devices the
loudest. However, once the recordings were imported into Audio Notetaker (v4) these dierences
disappeared.
Conclusion:
There is a large variation in sound levels between recordings taken with dierent devices, but these
variations can be simply and eectively removed with audio software such as Audio Notetaker (sound
example 1 & sound example 2).
Intelligibility
The key criterion for assessing recording quality is whether, when listening back to a recording, you
understand what was said.
Overall, the tests showed a surprising consistency in quality between all the devices in terms of their
raw recording intelligibility, although other problems did emerge, mostly due to interference (see
the next section). Almost all the recorded speech was intelligible on all the devices, even when the
lecturer spoke very quietly. So it is only in a very few places of the recordings that it is possible to rank
intelligibility.
When recording the audio played through a PA system (i.e. through speakers in the lecture hall),
recordings from all devices were equally intelligible.
In the quieter of the two live lecture recordings there were a few places where it was possible to rank
devices on the basis of intelligibility (excluding those instances cause by mechanical or electrical
issues – more of which below). In these places, recordings taken with the iPhone/iPad were slightly less
intelligible than the two DVRs, which in turn were slightly less intelligible than the DVR with an external
microphone, but the dierences were marginal. We have provided you with two examples of these
instances so you can decide for yourself (sound example 3 & sound example 4).
There is a more technical way in which we can assess the relative intelligibility of the recordings, and
this can be applied to new recordings without needing to record an actual lecture. Please see Appendix
2 if you’re interested. This method corroborates the above results.
The Sonocent guide to recording devices
Conclusion:
Although there was minor variation between devices, almost all the recorded speech was intelligible
on all the devices. The biggest factor aecting intelligibility is whether there is interference. Some
interference is preventable, which means that it is possible to optimise recording quality regardless of
which device you use.
Sound colour
Although all the recordings oered very similar intelligibility, the timbre or “colour” of the sound (i.e. its
spectral characteristics) produced by each device did dier (sound example 2 & sound example 5).
While, with one recorder, the sound changed depending whether you selected “dictation” or “lecture”
mode (sound example 6).
When listening to our audio examples, you will notice that it is the ambient noise that is most
noticeably dierent for each device, rather than the speech. It varies from hissy (Kindle Fire HDX)
right down to rumbly (some DVRs). But this does not aect the intelligibility of the speech, since in the
important frequency range for speech (300-3400Hz), the noise is actually very similar between devices.
If listening to music, this variation could be signicant. With speech, however, you quickly adjust to
the sound colour and after a short while you no longer notice it. (This eect is comparable to the
experience of attending an amateur lm showing where, at rst, the sound is very mued, but after
a while you adjust and it becomes perfectly clear. Or when you talk on the telephone and at rst the
person’s voice sounds dierent to when you communicate face to face, but after a while it sounds
quite normal.)
How you rank the various sound colours depends very much on personal taste. In any case, the
sound can usually be adjusted using the playback program – For instance, Audio Notetaker allows the
removal of low-frequency “rumble” sounds, while you can use the software’s noise cancellation feature
to reduce all the noise, which makes for a more enjoyable listening experience.
Conclusion:
Sound colour varies from device to device and the type that you prefer is largely matter of personal
preference. Moreover, listeners quickly adapt to sound colours to a point where they no longer notice
them.
The Sonocent guide to recording devices
Audio format and compression rate
All Sonocent products record by default at 32 kbits per second, which equates to 14.4MB per hour. For
the types of recordings made for the purpose of note taking, this is of more than sucient quality and
has the advantage that the disk space on the recording device will not ll up too quickly.
How have we reached this conclusion?
Well, the technical explanation is that audio can either be recorded uncompressed (PCM), in which
case it is usually found in a WAV format le, or compressed using one of a number of common audio
compression standards – MP3, M4A/MP4, Opus - or proprietary formats such as WMA or DSS. All these
formats reduce the le size. Most have dierent settings so you can choose the level of compression
you want, which is usually measured in kbits per second or Mbytes per hour. PCM is around 1411 kbps
(635 Mbytes/hour), and most compression schemes oer rates from 8 to 192 kbits/sec (3.6 to 86.4
Mbytes/hour).
Using audio compression makes the audio very much smaller, but does it aect the quality? It
depends what you are listening to and why you are listening. For example, if you are recording an
unaccompanied choir, you need the best quality possible, since not only is the sound itself complex
and subtle, but your listeners are listening for the pleasure of the beautiful sound produced. But if you
are recording a lecture, any subtlety in the voice has already been lost because of the distance you
are recording at, and in any case you are listening to what is being said, not the beautiful sound of the
lecturer’s voice.
Don’t assume that because the recording quality is already degraded by ambient noise you will need
the best possible quality setting on your recording device. If anything, the opposite is true – the
ambient noise will tend to mask any subtle artifacts of the compression scheme.
However, as the compression setting gets more and more severe, the artifacts do become noticeable,
and you have to trade o le size with audio quality. For lecture recordings, this tradeo will occur in
the 16-64 kbits/sec region, depending on the recording format and the quality you want.
Audio Notetaker and Sonocent Recorder both record in the Opus format, the most recent of all audio
standards and the one which oers best quality compression. (Listen to the sound le examples
provided - recoded into high-rate m4a so you don’t need an Opus decoder to hear them - and see if you
notice any dierence above 32kbits/sec).
The Sonocent guide to recording devices
Section 3 - What makes for a bad recording?
There were a number of types of interference which, when they occurred, aected the intelligibility of
the recordings signicantly:
Mobile phone interference
We are all familiar with the Morse-code-like interference that you sometimes experience when
your phone is too near a microphone, cable or PA system. Whether or not your recording will suer
depends on how far you are from the base-station (your phone adjusts its transmit power accordingly)
and, of course, how far your phone is from the recorder. But it also depends on the susceptibility of
the recording device itself.
The iPhone that we tested occasionally experienced a minor level of interference from its own
transmission. Although audible as a quiet ticking sound, the speaker could still be heard clearly and
so it did not aect the intelligibility of the recording (sound example 7). By contrast, the Samson Go
digital microphone and the Olympus DM670 both picked up the iPhone’s transmission very severely
(the iPhone was 30cm distant), rendering the recording unusable in places (sound example 7), but,
fortunately, not very often:
Mobile interference (in red) over 55 minutes. Each line represents 3 minutes.
However, with forethought, it is possible to mitigate the impact of mobile interference. For example,
with the Olympus DM-670 we found that there was no mobile interference when the ME32 external
mic was plugged in to the recorder. No problems with mobile interference were encountered with the
other brand of recorder (Philips), the iPad, or the new Olympus DVRs (DM7 and DM901).
The Sonocent guide to recording devices
Bangs from the recording surface
We found a susceptibility to mechanically-transmitted interference through the desk that the device
was placed on, which in our lecture rooms was a desk shared by the whole row (for example someone
kicking the desk or a seat in the row in front banging the desk). The iPhone/iPad seemed particularly
susceptible, digital recorders less so and an external mic the least. These eects were reduced by
putting the device on something which dampened the vibrations through the desk, such as a soft case
designed to angle the device. One of the recorders had a ip-down stand which seemed to help a lot
(sound example 8). Most of the adverse impact on the iPhone/iPad was due to the automatic recording
level system dropping the level when a “bang” occurs, and then taking a short while to ramp it up again
so that the speaker was very quiet for a few seconds (sound example 9) but, fortunately, this did not
occur often.
Extraneous sounds – coughs, talking, movement
The eect of extraneious sounds would be minimised by sitting as close as possible to the lecturer or
the PA speakers (if the PA is on), but in our tests we maximised them by sitting on the back row. We
expected the “zoom” directionality of the digital recorders (particularly with an external microphone) to
help reduce these eects compared to the phone/tablet, but in practice there were so many reective
surfaces around us (particularly the desk itself) that the zoom settings seemed to make little dierence
(sound example 10 & sound example 11)
Although recording near the front minimised these sounds, a word of warning is needed: we did try
placing the recorder on the podium at the front of the seminar room and found that the projector
(which came down from the ceiling in the same place) added a little bit of fan noise to the recording.
No problem in our case, but it is always worth being aware of noise sources near the recorder, even at
the front.
Key Clicks
A particular problem with recording on a laptop without an external microphone is the distraction
caused by the background sound of key clicks from typing (sound example 12a). In our experience,
key/mouse clicks are particularly pronounced when the internal laptop microphone is positioned
on the lower half of the laptop (as it was in the sound example above), but are minimised when the
internal microphone is at the top of the screen.
With an external microphone, however, the problem of key clicks is much reduced (sound example
12a). In our tests, we had the Samson Go microphone clipped to the screen (which is how it is designed
to be used), but the key clicks would have been further reduced if it had been placed on the desk.
You can also reduce the prominence of the clicks by typing slowly and gently - it is typing at touch-type
speed that really does the damage!
Audio Notetaker v4 oers key-click suppression on playback. This makes the key clicks much less
intrusive (sound example 12b).
The Sonocent guide to recording devices
Section 4 - Recommendations
Tips
If you pay £5 for a recorder on the internet, you should not expect a good quality recording. However,
the recording devices that we used, each of which were manufactured by reputable brands, all oered
a reasonable recording. Any dierences in the quality of sound paled into insignicance compared to
interference problems which can be minimised by following the simple tips below:
• Put your phone on “airplane” mode, particularly if you are recording with it. This will also help with
battery life! or with a digital recorder or external digital microphone that is susceptible to mobile
noise. If you are using a digital recorder which is susceptible (Olympus DM670 or Samson Go),
consider using an external Olympus ME32 microphone if the noise turns out to be a problem.
• Don’t place your recorder directly on the desk, place it on something soft which will absorb
vibrations and reected sounds from the desk.
• If using an iPhone or iPad, using a case which can prop it up at an angle will help limit mechanical
noise.
• Sit reasonably near the lecturer or PA system.
• If the recorder is too near a projector, you may lose some of the advantage of having the
microphone close to the lecturer due to the fan noise.
Noise Cancellation
Some devices and recording programs oer noise cancellation during recording. A PC’s internal
microphone usually has this on by default and it takes quite a lot of burrowing in settings menus to
nd where to turn it o ¹. It is better to record with it o as it can be applied afterwards (in Audio
Notetaker for example), but if you record with it on you have no option to listen without it. While noise
cancellation makes for easier listening, if the voice is borderline unintelligible, you have the most hope
of hearing what was said with noise cancellation o.
University Recordings
If you are able to persuade them, an almost ideal option is for the lecturer to wear a tie-clip or headset
microphone. This will give the best quality possible for the lecturer’s speech and is the recommended
solution when quality really matters – for instance, if the recording is for a general audience and not
just for you.
There are some drawbacks though. Questions/comments from the lecture theatre itself do not come
through well, if at all. Depending on the class, this can be a signicant disadvantage of centralised
recording, which can only be overcome by a second recording device and some careful audio editing.
¹ Playback Devices > Speakers > Properties > Enhancements
The Sonocent guide to recording devices
Centralised recording also makes it harder for students to tie the notes they take during the lecture
with the recording they get afterwards. If they rely totally on the recording, listening becomes a passive
process, both during the lecture and after. Research shows that active participation in a lecture (even
just pressing a button when you think its an important point) helps students recall the information at a
later date.
In order for students to use centralized recordings and annotate their lectures as they listen, we are
adding an ‘Audio Replace’ feature to Audio Notetaker v4 so that users can simply replace the audio
recording made on their laptop or mobile device with one provided by the university. Audio Notetaker
will automatically transfer all annotations and notes, along with any extra audio, from the old
recording to the new recording.
Android
The Sonocent Recorder app will be available for Android early 2015, so the quality of audio captured
with Android devices is a pertinent issue.
Over the years, we have discovered a wide variation in the quality of the internal microphones in
laptops, and we expected to nd a similar variety amongst Android devices. We were not disappointed.
We informally tested 3 Android devices – the Kindle Fire HDX, the Sony Xperia SP and the Google Nexus
5. The Kindle Fire HDX, although a little bit hissy (i.e. noisy at high frequencies), is similar in quality to
the other of our main test devices. The Sony Xperia SP is noticeably (about 3dB) noisier and would not
have performed well at the back of the quiet lecture, but would have been ne for the other lectures. It
was however very much better than the Nexus 5, which gave terrible audio quality, and could only be
used for recording close-up (sound example 13).
With laptops, the advice we have always given is to have an external USB microphone like the Samson
Go available. Not all laptops in all recording situations will need it, but it ensures a reasonable quality
of recording is always possible. With Android, this has not been an option until recently with the
release of 5.0 (Lollipop), where it is now possible to plug in an external USB microphone, providing
that the phone’s hardware supports USB audio input. We tried this on our Nexus 5 with a Samson Go
and cheap OTG connector cable, and it worked very well. However, it did double battery life during
recording – from 6% to 13% for a 55 minute recording.
An alternative to a USB microphone is to use Audio Notetaker 4’s ‘Audio Replace’ facility, which allows
users to replace a bad recording with a good one. The good one could either be the university’s own
recording if one is available, or one taken by a student with a separate DVR, which could be an entry
level device placed at the front of the lecture.
Note: we do not recommend a Bluetooth microphone for Android as Sonocent Recorder will not
support recording from Bluetooth, and in any case we have not yet found a Bluetooth microphone of
sucient quality to record a lecture.
The Sonocent guide to recording devices
Section 5 - Device guide
Device Positives Negatives
Laptop Good if you want to type lots of textual
notes to go with the recording
Recording quality varies between brands
and even within brands, so you can never
be quite sure how good it will be if you can’t
try before you buy
No extra purchase needed (assuming you
already have one)
Quite large to carry around
Need to be careful with key clicks aecting
audio quality
Battery may not last if you want to record all
day away from a power plug
...with
external
microphone
Reliable recording quality One more thing to setup and to lose/forget
Reduces key click issue Samson Go susceptible to mobile noise
iPad/iPad Mini Good battery life Onscreen keyboard not as easy to type
on as a laptop (although many external
keyboards are available)
Portable, though you still have to think
ahead to bring it
Reliable recording quality
iPhone Don’t need to plan in advance – already
have it
Battery may run low if recording all day,
especially for older phone (use airplane
mode to minimize battery drain)
Don’t need to buy an extra piece of
hardware
You need to ensure you have enough space
on your phone for your recordings
Reliable recording quality
Android
devices
These have the same advantages as iPad/iPhone but the drawback of the laptop –
recording quality is an unknown until you try it. New devices may support an external
USB Microphone.
Digital voice
recorder
Excellent battery life Can’t take live notes on the device
Reliable recording quality Another device to remember to take with
you
Additional cost when you probably already
own a device that can record audio
...with
external
microphone
Best possible recording quality One more thing to carry/keep/setup
Avoids mobile interference if recorder is
susceptible to it
The Sonocent guide to recording devices
Appendix 1 - Recording devices tested
Device Model External
Microphone
Microphone
setting
App Live test PA test
Phone Apple iPhone Sonocent
Recorder
X
Tablet Apple iPad/
iPad Mini
Sonocent
Recorder
X X
DVR Olympus
DM670 Maximum
zoom; low
cut (live test
only)
X X
DVR Olympus
DM670
Olympus
ME32
X X
DVR Philips
DVT6000
Long
distance
X X
DVR Philips
DVT6000
Olympus
ME32
X
Laptop Toshiba
Satellite
Z930-10X
Default Audio
Notetaker
X
Laptop Toshiba
Satellite
Z930-10X
Samson Go
(digital)
Directional Audio
Notetaker
X
Tablet Kindle Fire Simple Voice
Recorder
X
Phone Sony Xperia
SP
Sonocent
Recorder
(prototype)
Phone Google
Nexus 5
Sonocent
Recorder
(prototype)
DVR Olympus
DM901
Maximum
zoom
DVR Olympus
DM7
The Sonocent guide to recording devices
Appendix 2 – Spectrogram Analysis
Intelligibility is determined primarily by the amount of background noise obscuring the speech signal
you want to hear, so a very good guide to intelligibility is the amount of noise compared to the speech
signal. When listening to the background noise in the recordings (listen to sound sample 2 again),
you are at rst aware that some have low frequency noise (rumble) and others have high frequency
noise (hiss). But neither of these necessarily make the speech less intelligible. What matters most is
the amount of noise in the main speech frequency band of 300-3400 Hz. This is the frequency band
that is used for telephony, and it contains most of the speech information. When the recordings are
ltered to this frequency range (sound example 14), you can hear that the variations between devices
becomes very much less.
A useful alternative to listening is to look at the spectrogram of the speech. This is a plot of signal
strength (indicated by colour) with frequency on the vertical axis and time on the horizontal axis. Here
is the spectrogram for example 2:
And here is just the 300-3400 part of the spectrogram:
You can see that whereas there is a lot of dierence in the amount of noise over the whole spectrum
(the blue background), in the crucial range 300-3400 there is very little dierence.
However, having said all that, the frequency range above 3400HZ does add some intelligibility – think
how hard it so to distinguish between “f” and “s” on the telephone and how many times you end up
saying “foxtrot” and “sierra”. And so from the top plot you can see that for the DVRs the lower noise
level will make these kinds of sounds more distinguishable, and it is this that gives them the slight
edge on intelligibility. Although the DM670 has a better high-frequency response than the DVT6000,
we did not think this made any dierence to intelligibility in any of our recordings, but it did make for a
“brighter” sound.
The Sonocent guide to recording devices
