Topic: PCM-M10 Achilles heels

[earmonger]

I recorded a folk jam session a few days ago just to test it out, and the recording resulted with slight clipping. I used manual recording, low sensitivity, no limiter nor low-cut, and the level dial on 5, using built in mics. I suppose I should have used a lower setting, but this was a small group of people playing acoustic instruments in a pub, not a loud rock concert! I'll try on 2.5 next time.

I remember that with my Sharp MT-88 (+ Sony ECM-DS70P mic) I could record just about anything without caring about settings, be it folk or heavy metal. It could even record with no problem some inhumanly loud concert a few years back

I would suggest using the limiter. That's different from auto level control--it will only kick in if you're about to overload. It's like an airbag, there for emergencies.  Other people will say oh no no, but I just have to disagree. It will do no harm whatsoever and could save you.

Meanwhile, when recording, if you can see the recorder you should see green lights for optimum levels and red if you're overloading. You can even see them if  you have the recorder stashed in a pouch at your waist--glance down and adjust the volume accordingly. I like mine to be where the green light is blinking occasionally,  so the green-light range is at the peak of the music's volume.

You're comparing apples and oranges to compare the built-ins with the DSP-70. The DSP-70 is a very, VERY low-sensitivity mic with a built-in low-cut--the lower end of its frequency response is 100 Hz. The built-ins also taper off the bass but not that much, and they are probably more sensitive. You could always just plug the DSP-70 into Mic-In and see what you get. Or...sorry about your wallet...upgrade your mics and get a battery box...

[F.O.Bean]

I recorded a folk jam session a few days ago just to test it out, and the recording resulted with slight clipping. I used manual recording, low sensitivity, no limiter nor low-cut, and the level dial on 5, using built in mics. I suppose I should have used a lower setting, but this was a small group of people playing acoustic instruments in a pub, not a loud rock concert! I'll try on 2.5 next time.

I remember that with my Sharp MT-88 (+ Sony ECM-DS70P mic) I could record just about anything without caring about settings, be it folk or heavy metal. It could even record with no problem some inhumanly loud concert a few years back

I would suggest using the limiter. That's different from auto level control--it will only kick in if you're about to overload. It's like an airbag, there for emergencies.  Other people will say oh no no, but I just have to disagree. It will do no harm whatsoever and could save you.

Meanwhile, when recording, if you can see the recorder you should see green lights for optimum levels and red if you're overloading. You can even see them if  you have the recorder stashed in a pouch at your waist--glance down and adjust the volume accordingly. I like mine to be where the green light is blinking occasionally,  so the green-light range is at the peak of the music's volume.

You're comparing apples and oranges to compare the built-ins with the DSP-70. The DSP-70 is a very, VERY low-sensitivity mic with a built-in low-cut--the lower end of its frequency response is 100 Hz. The built-ins also taper off the bass but not that much, and they are probably more sensitive. You could always just plug the DSP-70 into Mic-In and see what you get. Or...sorry about your wallet...upgrade your mics and get a battery box...

I agree it could save your ass, but with 24bit recording, why not just run lower levels

[fguidry]

The one possibly adverse thing that I've found about the M10 is that its 24-bit recording mode doesn't seem to offer any wider dynamic range than its 16-bit mode, or at most just a dB or two. Somewhere in another thread I posted spectrum analyzer photos of the unit in both settings, using the line inputs, with the recording level knobs set to low-ish settings.

If anyone here has determined any differently, I would certainly like to know about it (different settings or approaches might yield different results for all I know), but otherwise, it seems to me that we could all spare ourselves 1/3 of the storage-space hassles and just use the recorder in its 16-bit mode.

--best regards

I remember reading this as well. What exactly do you think the 24 bit mode does, then?

Do you understand what 24 bit recording does? Folks talk about "resolution" and "headroom" but the real difference is the level of quantization noise - the noise that is generated as part of the sampling process. Quantization noise in 16 bit is at -96 dB, theoretically in 24 bit about -144 dB but since thermal noise is around -120 dB that is the real bottom limit which computes to about 20 bits.

So in the case of the M10 or other recorder that offers 24 bit recording but generates self-noise above -96 dB, the extra bits simply takes up room to contain noise. This is a common situation with inexpensive portable recorders.

I invite you to consider those cases where ambient noise is way above -96 dB, which is almost certainly the case for all live taping situations.

Fran

[voltronic]

The one possibly adverse thing that I've found about the M10 is that its 24-bit recording mode doesn't seem to offer any wider dynamic range than its 16-bit mode, or at most just a dB or two. Somewhere in another thread I posted spectrum analyzer photos of the unit in both settings, using the line inputs, with the recording level knobs set to low-ish settings.

If anyone here has determined any differently, I would certainly like to know about it (different settings or approaches might yield different results for all I know), but otherwise, it seems to me that we could all spare ourselves 1/3 of the storage-space hassles and just use the recorder in its 16-bit mode.

--best regards

I remember reading this as well. What exactly do you think the 24 bit mode does, then?

Do you understand what 24 bit recording does? Folks talk about "resolution" and "headroom" but the real difference is the level of quantization noise - the noise that is generated as part of the sampling process. Quantization noise in 16 bit is at -96 dB, theoretically in 24 bit about -144 dB but since thermal noise is around -120 dB that is the real bottom limit which computes to about 20 bits.

So in the case of the M10 or other recorder that offers 24 bit recording but generates self-noise above -96 dB, the extra bits simply takes up room to contain noise. This is a common situation with inexpensive portable recorders.

I invite you to consider those cases where ambient noise is way above -96 dB, which is almost certainly the case for all live taping situations.

Fran

Good point about the ambient noise in live taping.  I think it's worth mentioning that the M10 does measure better than other recorders though.  Guysonic's measurements seem to show that the M10's self-noise is around -120 dB for MIC IN through most of its range: http://taperssection.com/index.php?topic=124639.msg1722123#msg1722123

DSatz, I'd be very interested to see the spectrum plots you did to compare...

[OMA]

I would suggest using the limiter. That's different from auto level control--it will only kick in if you're about to overload. It's like an airbag, there for emergencies.  Other people will say oh no no, but I just have to disagree. It will do no harm whatsoever and could save you.

Won't I get noticeable volume drop downs when an incidental loud noise kicks in? (maybe even some clapping from the audience).

Meanwhile, when recording, if you can see the recorder you should see green lights for optimum levels and red if you're overloading. You can even see them if  you have the recorder stashed in a pouch at your waist--glance down and adjust the volume accordingly. I like mine to be where the green light is blinking occasionally,  so the green-light range is at the peak of the music's volume.

Well, I like to forget about my recorder after I've started "rolling the tape", be it because I might be playing in a session myself or just because I want to enjoy a concert I'm attending without being concerned about levels and whatnot. That I could do with my old Sharp MD. I always put it in recording level L9 out of 30 (if you ever had one, you might remember it had L and H in addition to the number), and recordings always came up sounding quite good, with no distortions at all.

You're comparing apples and oranges to compare the built-ins with the DSP-70. The DSP-70 is a very, VERY low-sensitivity mic with a built-in low-cut--the lower end of its frequency response is 100 Hz. The built-ins also taper off the bass but not that much, and they are probably more sensitive.

I suppose you mean DS70P. Well, yes, I understand they're different mics. I just wanted to test the internal mics for my first recording, but was surprised that I could get clipping with something that was just moderately loud.

I know that the DS70P doesn't go all the way down to 20 Hz, but I've read that might be a good thing for concerts, as you don't overload the input with excessive bass, is that true?

You could always just plug the DSP-70 into Mic-In and see what you get.

Well, yes, that will be my next test, mostly because the DS70P is the only mic I have . And it also passed the test of time. I've been using it since the beginning of the year 2000. So 13 years and a half already! . Another great Sony product

Or...sorry about your wallet...upgrade your mics and get a battery box...

I've sometimes considered that option, but I don't like having to carry yet another device (battery box) and with so many mic options, I'm not sure what to choose.

So in the case of the M10 or other recorder that offers 24 bit recording but generates self-noise above -96 dB, the extra bits simply takes up room to contain noise. This is a common situation with inexpensive portable recorders.

So you mean there's not any benefit in recording in 24 bit with the M10 and we should record at 16 bit instead?

[Ozpeter]

Use 24 bits only to avoid digital clipping.  You won't hear any other benefit.  In practical use, this is the case for ALL 24 bit recorders, regardless of cost and type.

[OMA]

Use 24 bits only to avoid digital clipping.  You won't hear any other benefit.  In practical use, this is the case for ALL 24 bit recorders, regardless of cost and type.

Well, yes, avoiding clipping is the only reason I'd use 24 bits, since I know that for an ambient recording no one will notice any quality difference between 16 and 24 bits.

But, since DSatz said that 24 bits didn't help with clipping either... Check his quoted message below:

The one possibly adverse thing that I've found about the M10 is that its 24-bit recording mode doesn't seem to offer any wider dynamic range than its 16-bit mode, or at most just a dB or two. Somewhere in another thread I posted spectrum analyzer photos of the unit in both settings, using the line inputs, with the recording level knobs set to low-ish settings.

If anyone here has determined any differently, I would certainly like to know about it (different settings or approaches might yield different results for all I know), but otherwise, it seems to me that we could all spare ourselves 1/3 of the storage-space hassles and just use the recorder in its 16-bit mode.

--best regards

[it-goes-to-eleven]

Far too much speculation and armchair talk.  Far too little actual testing of what sounds best.

[Gutbucket]

I agree it could save your ass, but with 24bit recording, why not just run lower levels

Because if you set the signal level too low it can encroach on, or slip below the recorder’s noise floor in the quiet parts.  If the analog input portion of the recorder is only capable of 17 maybe 18 bits of actual bandwidth between the recorder’s noise floor and full scale digital at the analog to digital converter, then setting it to record 24 bit files will at best provide the capability to capture one or two bits more dynamic range than setting it to record 16bit files.  The remaining bits at the ‘bottom’ of the 24 bit file are simply random numbers corresponding to noise.

Each additional bit provides approximately 6dB of additional dynamic range, so one or two bits more is significant.  That’s worth the additional file size to me just for peace of mind. Probably the best we could hope for in a machine of this type with excellent specs would be a real range of 20bits anyway (4 additional bits over 16), you’ll never going to see all 8 bits additional bits of usable signal switching to 24 bit files from 16 bit files.  The bottom bits will always be noise.  17 or 18 bits is plenty for my purposes anyway.

Do you understand what 24 bit recording does? Folks talk about "resolution" and "headroom" but the real difference is the level of quantization noise - the noise that is generated as part of the sampling process. Quantization noise in 16 bit is at -96 dB, theoretically in 24 bit about -144 dB but since thermal noise is around -120 dB that is the real bottom limit which computes to about 20 bits.

So in the case of the M10 or other recorder that offers 24 bit recording but generates self-noise above -96 dB, the extra bits simply takes up room to contain noise. This is a common situation with inexpensive portable recorders.

Yes, but quantization noise is completely taken care of by the analog to digital converter chip in the recorder which dithers the last bit of information, allowing sounds to degrade gracefully below the noise floor instead of degrading into quantization noise.  Assuming both the room's noise floor and the recording systems analog noise floor are lower than the dither level (which is a big assumption!) if you really crank it up on playback you can hear the dither hiss, but can still detect signals dropping lower in level below the hiss.  So in reality it’s the dither noise, not quantization noise that becomes the ultimate bottom limit of the usable dynamic range.  However, both the recording chain's analog noise floor and the recording environment’s noise floor may be higher than that.

I invite you to consider those cases where ambient noise is way above -96 dB, which is almost certainly the case for all live taping situations.

This is misleading.  It’s adjustable.  -96dB is the lower limit of a16bit file relative to 0dBFS, but that’s the range of the file not of the sound levels in the room.  That 96dB of usable range is adjustable up or down depending on the sensitivity of your microphones and how you set your input gains.  You can set things so that the noise floor of the room is either above or below the -96dB level of the file.

For instance, say your recorder is capable of recording 16 bit files and its analog portion is capable of passing that full 96dB of dynamic range, and let’s say you are recording a rock concert that has peaks up to115dB SPL.  You decide to leave 10dB of headroom in case it peaks louder than that at some point, so you set your input gain so that 0dBFS on the recorder = a 125dB SPL peak in room.  96dB below that, the dither noise in the recorded file then corresponds with an analog input of 29dB.  That’s probably far lower than the noise floor of the room. 

But if you set things so that 0dBFS = 150dB SPL, then your file’s noise floor would be at 54dB which might be higher than that of room.

[fguidry]

...

I invite you to consider those cases where ambient noise is way above -96 dB, which is almost certainly the case for all live taping situations.

This is misleading.  It’s adjustable.  -96dB is the lower limit of a16bit file relative to 0dBFS, but that’s the range of the file not of the sound levels in the room.  That 96dB of usable range is adjustable up or down depending on the sensitivity of your microphones and how you set your input gains.  You can set things so that the noise floor of the room is either above or below the -96dB level of the file.

For instance, say your recorder is capable of recording 16 bit files and its analog portion is capable of passing that full 96dB of dynamic range, and let’s say you are recording a rock concert that has peaks up to115dB SPL.  You decide to leave 10dB of headroom in case it peaks louder than that at some point, so you set your input gain so that 0dBFS on the recorder = a 125dB SPL peak in room.  96dB below that, the dither noise in the recorded file then corresponds with an analog input of 29dB.  That’s probably far lower than the noise floor of the room. 

But if you set things so that 0dBFS = 150dB SPL, then your file’s noise floor would be at 54dB which might be higher than that of room.

To my mind you've inverted my position. Let's take your example - 115 dB peak, 10 dB headroom, 96 dB dynamic range. Already the theoretical and practical noise floor of the recorder is well below the ambient noise and therefore "wasted." Switch to 24 bit, get actual 20 bit 120 dB dynamic range - your recording now has extra bits containing the ambient noise with no improvement in the quality of the recording.

Fran

[Church-Audio]

The one possibly adverse thing that I've found about the M10 is that its 24-bit recording mode doesn't seem to offer any wider dynamic range than its 16-bit mode, or at most just a dB or two. Somewhere in another thread I posted spectrum analyzer photos of the unit in both settings, using the line inputs, with the recording level knobs set to low-ish settings.

If anyone here has determined any differently, I would certainly like to know about it (different settings or approaches might yield different results for all I know), but otherwise, it seems to me that we could all spare ourselves 1/3 of the storage-space hassles and just use the recorder in its 16-bit mode.

--best regards

I remember reading this as well. What exactly do you think the 24 bit mode does, then?

Do you understand what 24 bit recording does? Folks talk about "resolution" and "headroom" but the real difference is the level of quantization noise - the noise that is generated as part of the sampling process. Quantization noise in 16 bit is at -96 dB, theoretically in 24 bit about -144 dB but since thermal noise is around -120 dB that is the real bottom limit which computes to about 20 bits.

So in the case of the M10 or other recorder that offers 24 bit recording but generates self-noise above -96 dB, the extra bits simply takes up room to contain noise. This is a common situation with inexpensive portable recorders.

I invite you to consider those cases where ambient noise is way above -96 dB, which is almost certainly the case for all live taping situations.

Fran

Good point about the ambient noise in live taping.  I think it's worth mentioning that the M10 does measure better than other recorders though.  Guysonic's measurements seem to show that the M10's self-noise is around -120 dB for MIC IN through most of its range: http://taperssection.com/index.php?topic=124639.msg1722123#msg1722123

DSatz, I'd be very interested to see the spectrum plots you did to compare...

You can't look at spectral noise graphs without taking into consideration the average RMS level. Looking at the graph alone and saying the noise floor is -120 because the graph is at -120 is misleading.

Chris

[Gutbucket]

To my mind you've inverted my position. Let's take your example - 115 dB peak, 10 dB headroom, 96 dB dynamic range. Already the theoretical and practical noise floor of the recorder is well below the ambient noise and therefore "wasted." Switch to 24 bit, get actual 20 bit 120 dB dynamic range - your recording now has extra bits containing the ambient noise with no improvement in the quality of the recording.

Fran

Yes, if you switch bit rates and do not change your input signal gain, that is correct.

It's somewhat semantic, but the point I was trying to make is the -96dB point is variable depending on how you set things.  In the second example, those extra 4 extra bits would be useful, lowering the noise floor of the recorder to 30dB SPL which stands a good chance of being below the ambient noise floor.

I was mostly trying to clarify this for others reading the thread.  Appologies for any confusion!

[earmonger]

Won't I get noticeable volume drop downs when an incidental loud noise kicks in? (maybe even some clapping from the audience).

Good levels are obviously the best approach.  But I'd rather have a volume drop down than the distortion of clipping.

Well, I like to forget about my recorder after I've started "rolling the tape", be it because I might be playing in a session myself or just because I want to enjoy a concert I'm attending without being concerned about levels and whatnot. That I could do with my old Sharp MD. I always put it in recording level L9 out of 30 (if you ever had one, you might remember it had L and H in addition to the number), and recordings always came up sounding quite good, with no distortions at all.

You'll learn the set-it-and-forget-it levels on the PCM-M10, too. They'll just be different.

I suppose you mean DS70P. Well, yes, I understand they're different mics. I just wanted to test the internal mics for my first recording, but was surprised that I could get clipping with something that was just moderately loud.

I know that the DS70P doesn't go all the way down to 20 Hz, but I've read that might be a good thing for concerts, as you don't overload the input with excessive bass, is that true?

To me it's better to get an accurate recording--even with the bass-heavy mixes that are all too common at concerts now--than to have a one-size-fits-all filter applied to every recording. What saved your recordings with the DS70P (yes, sorry) was the combination of its low sensitivity, the Sharp's low sensitivity setting and the lack of bass, not just the bass cut-off. The internal mics are more sensitive than the DS70P because Sony probably intended them to be used for interviews and other relatively quiet sources.

I'd suggest you do a fairly low-volume test, with a home stereo, with the DS70P and then the internal mics recording the same source--matching, just by eyeballing, the incoming recording level. Then play them back and see how you like the sound quality.

Better inexpensive mics than the DS70P include the tiny Sound Professionals BMC-2 or, more expensively, the slightly larger Church Audio mics; either way, add a battery box.  (Church Audios often show up in the Yard Sale here.) If you try them, sorry to say, your old recordings may sound squashed or claustrophobic.

Battery boxes aren't big: 9V ones are barely larger than the 9V battery, and Sound Professionals and Microphone Madness both make 12V versions that are even smaller.  The MM Classic Mini, really small, has  a lifetime warranty, but the batteries are expensive and I've had to get mine repeatedly repaired.

 I've had better luck, and the battery is cheaper, with the SPSB-10 with short detachable cables.

http://www.soundprofessionals.com/cgi-bin/gold/item/SP-SPSB-10

I use a belt pouch, LowePro Ridge 30, that holds the PCM-M10, battery box and remote quite neatly.

[Ozpeter]

What is almost never mentioned when dynamic range of recorders is discussed is that even using a cheap 16 bit recorder with modest mics, it's not practically possible to reproduce the dynamic range available, in a normal domestic environment. 

I don't know whether this sample is still online, but at https://docs.google.com/file/d/0ByTVNpxnwA3DdHI0UGpkUEVCZW8/edit?usp=sharing there's a recording which to me makes the whole discussion one of theory rather than practice.

The first part is an excerpt from a binaural street recording I made recently (using Roland in-ears into a Sony SX750 - somewhat downmarket compared to the M10). It starts with me descending some car park steps, then you're in the street (with a street musician conveniently at the foot of the steps). At 1'09" traffic moves off as traffic lights turn green (including a Melbourne tram). At about 2' 15" I turn towards the musician and at about 2'26" you hear the chink of a coin I tossed in his hat. He was of course quite unaware he was being recorded but none the less he was entitled to his fee! In the original wave file the chink was actually the biggest peak.

Then at 2'47" I've spliced on some ambience recorded in a bedroom at home. The recording level was the same as for the street recording - in preparing the file I haven't messed with the levels at all, just converted to mp3. Right at the start of this quiet section you hear a bit of my tummy rumble (!!) and later you hear a phone ring elsewhere in the house. However, throughout this section you should be able to just hear a clock tick (every second) which, in this bedroom, is just about audible.

What's interesting to me about this sample is that if you set your playback level comfortably for the first part in the street, it's impossible to hear the clock tick in the second part in the bedroom, unless you turn up the playback volume somewhat. Or, having turned up the playback volume so you can just hear the clock tick, then play back the street section, you'll probably get complaints from the neighbours. At least that's how it seems to me.

So what practical benefit in terms of usable dynamic range you'd get from using a higher-end recorder and 24 bits escapes me.

Edit - at http://www.audiomastersforum.net/amforum/index.php/topic,8668.0.html there's a discussion along similar lines based on another such dynamic range sample I posted elsewhere.