Topic: Sony PCM-M10 audio recorder - NEW

[jamesmcn]

One more thing:

The M10 names your recording with the current date and a recording number. If you delete recordings, the M10 keeps incrementing the recording number. If you format the unit, as you would after moving data to your PC, the number is reset back to zero.

[Idle Wind]

also - this has a digital limiter, the r09hr says it has an analog limiter.  not sure what the difference is..

http://digitalmedia.oreilly.com/pub/a/oreilly/digitalmedia/2008/01/10/sony-pcm-d50-recorder-review.html?page=2
The O'Reilly D50 review explains how the sony digital limiter works. In a nutshell, Sony records at both the normal level and several dB lower. When it detects clipping, it smoothly and seamlessly splices in the quieter wave. I have not verified this on my M10, though.

that is SO cool....

[Idle Wind]

not sure if the limiter circuit is the same as the D50.

from the m-10 instruction manual

http://www.fullcompass.com/common/files/11777-PCMM10%20Operation.pdf

=============================
z On the limiter circuit
• The limiter circuit functions to keep the signal level below the
maximum input level. When a loud sound is input suddenly, the
excess part of the sound is automatically set within the range of
the maximum input level in order to prevent distortion.
• The limiter circuit of the PCM recorder does not compensate
for clipping when audio over 12 dB is input. In this case, sound
may be distorted.
===================================

[illconditioned]

not sure if the limiter circuit is the same as the D50.

from the m-10 instruction manual

http://www.fullcompass.com/common/files/11777-PCMM10%20Operation.pdf

=============================
z On the limiter circuit
• The limiter circuit functions to keep the signal level below the
maximum input level. When a loud sound is input suddenly, the
excess part of the sound is automatically set within the range of
the maximum input level in order to prevent distortion.
• The limiter circuit of the PCM recorder does not compensate
for clipping when audio over 12 dB is input. In this case, sound
may be distorted.
===================================

This sounds the same.  I'm still not sure how this works in practice.  I've got both the PCM-D50 and M10, so I will try it eventually.  When I get time.

What I'm wondering is if the signal suddenly drops by 12dB or if it is gradual.  And if there is any marker in the file.  Ideally a marker in the file would allow us to reconstruct that lost dynamic range, effictively converting the 16bit recorded file to a 18bit file, where the 2 extra bits are the headroom afforded by the limiter.

  Richard

[Idle Wind]

apparently, it patches the -20db piece, then normalizes it to 0db.

pretty cool.

http://digitalmedia.oreilly.com/pub/a/oreilly/digitalmedia/2008/01/10/sony-pcm-d50-recorder-review.html?page=2

=====================================
One of the most talked about features of the expensive Sony PCM-D1 is its unique limiter. It works like this: the recorder creates two audio files during recording. One is written to memory, another — recorded 20dB down — is held in a buffer. If peaks exceed zero (i.e., maximum digital level), the recorder grabs a portion of the safety track and writes it to memory. I am happy to say that the D50 shares this feature.
PCM-D50 Limiter Waveform
Fig. 1: Instead of distorting when you record too loud, the D50 deftly swaps in a parallel safety track captured at a lower level and then normalized to zero (top arrow).

How does sound? To be honest, it works so well I didn't know it was engaged. I'm so used to hearing poorly implemented dynamics processors drastically ducking the audio and then sucking the levels back up that at first I thought my review unit wasn't functioning properly!

To test it, I purposely slammed the levels, causing the red LEDs to flash like Rudolph's nose (hey, it was just before Christmas). Listening back, I didn't hear any clipping, but I didn't hear any pumping, either. So I imported the files into my DAW and inspected the waveforms under the microscope. I expected to see a 20dB drop whenever the limiter wrote the safety track to memory; when I didn't, I contacted Sony. It turns out everything was just fine; when the safety track kicks in, it's boosted until its peaks are just below zero. (See Figure 1.)

Sure, slam the level too hard and you'll clip the waveform. And with no user-selectable threshold and only three release times (153ms, 1 second, and 1 minute) you have limited control. But the PCM-D50 limiter isn't a conventional limiter (dynamics signal processor). As a result, it doesn't require extensive parameter controls (threshold, release, hold time, etc.) to operate transparently, and that makes it easy to use. This is far and away the most useful limiter I have ever encountered on a portable device.

Fig. 1: Instead of distorting when you record too loud, the D50 deftly swaps in a parallel safety track captured at a lower level and then normalized to zero (top arrow).

[illconditioned]

apparently, it patches the -20db piece, then normalizes it to 0db.

pretty cool.

http://digitalmedia.oreilly.com/pub/a/oreilly/digitalmedia/2008/01/10/sony-pcm-d50-recorder-review.html?page=2

I believe it is -12dB.

Anyway, what do you mean by "patch in" and "normalize"?  I'm looking for a precise description (ie., algorithm).

From what I can tell by the circuitry (I saw the schematic from the service manual), it has two ADC streams, a regular one and a -12dB attenuated one.  If the levels clip, it just "switches in" the lower sensitivity version, presumeably for a fixed time (1sec, 10sec, 1min, selected from the menu).  I was hoping to be able to be told where the switch is done.  Maybe you can guess from the abrupt "step" in the waveform.  I tried this, but could not find it.

  Richard

[Idle Wind]

all that info came from the oreilly review (link above).  I don't really know how it works.

see highlighted revisions.  apparently he went looking for the level drop in the wave form and couldn't find it, because the alternate -20db patch somehow gets normalized to 0db on the fly....(see waveforms in Fig 1 above)

 

I'm interested in finding out if this circuit is also in the m10

[Idle Wind]

the limiter is described somewhat differently in the d-50 and m-10 manuals. 

also, promotional materials describe the d-50 as having a "dual path limiter" - not so in the m-10 materials,

so I suspect the m-10 does not have the same limiter circuit,  which is a bummer...

If Richard or someone could get the m-10 service manual, I suspect there'd be only one A/D stream.

[darktrain]

anyone know how many volts in the plug-in power?

also - this has a digital limiter, the r09hr says it has an analog limiter.  not sure what the difference is, and if you'd even use one in a stealth situation.  I never used AGC, but I did use the clip setting on my rockboxed IRiver, which actually works.

I was all ready to pull the trigger on an r09hr, until this came out, now I'm on the fence.  My sense is either one would be great, but I'd use this primarily for stealth - so ease of use, low key lighting, easily visible controls, etc.  you've got to love the IRiver remote, which shows you the levels right on it.

The M-10 looks like it has forward facing mics with clip indicators right on them, so if you're clipping, you'd be blinking.  I'd generally use separate mics (AT853) but I'm surprised at the results people have been getting from the built ins...

You can disable all the led's

[Ozpeter]

If the limiter in any recording device is operating in the digital domain in almost any way, there's not much point in using it unless you do not intend to post-process the recording (in other words, if you are simply going to listen to it as is, then there's a point, but as soon as you pass it through a DAW, then there's no point).

For any 'dual path' system to be worthwhile to someone intending to post-process the recording, then it would have to operate in the analog domain, which would mean that you'd have to have two preamps per channel, one working with less gain than the other.  Then if the preamp with the higher gain clipped, you could use the one with the lower gain.  I believe this is done with some systems but only for mono recording - you have the right channel operating at (say) 12dB less gain than the left channel, so you have something to fall back on if the left channel clips.

In the R-44 (for comparison) the limiter works by the lowering the preamp gain by 12dB (whenever the limiter is switched on) and then in the digital domain, digital limiting is applied followed by 12dB of digital gain.  There is absolutely no logical difference between this and simply setting the analog gain 12dB down manually yourself, and then applying a VST limiter in your DAW afterwards with 12dB of makeup gain.

Bear in mind that once the signal has been digitised in the recorder it's simple for the processing to look ahead at the data and limit the level entirely seamlessly.  Limiting in the analog domain is that much harder to achieve (perfectly) because the analog circuitry typically doesn't "know" that a peak has been encountered until after the fact.

I don't want to restart the endless arguments about the merits of 16 bits vs 24, but it's generally agreed that 'under-recording' using 24 bits, then normalising later, means that any need to limit or otherwise compress normal musical signals toi avoid analog clipping goes out the window, due to the very large dynamic range in the digital domain that 24 bits provides.

[xpander]

Both D1 and D50 have two A/Ds (AK5358s in D50). Right after the usual analog input stages, the signal is split into two paths, both running their own A/D stages, one set for normal recording levels, one running 20dB lower. System controller selects the appropriate line to be committed into memory according to limiter settings.

Like Richard already pondered, I don't think there's any normalizing happening there, but rather a switch between the stages, depending on the (clipping) levels. Also, manual states that levels over 12dB (?) will clip even while the limiter is on....which makes sense in that configuration.

This extra set of A/D circuitry was the first thing I thought they must've taken out of the M10 in order to get it physically smaller and even with lower power consumption than D50...those features don't come for free, after all. Service manual would confirm this, but I quess nobody has one yet?

[Ozpeter]

Both D1 and D50 have two A/Ds (AK5358s in D50). Right after the usual analog input stages, the signal is split into two paths, both running their own A/D stages, one set for normal recording levels, one running 20dB lower. System controller selects the appropriate line to be committed into memory according to limiter settings.

It's hard to see the utility of that system - the problem that remains when recording using 24 bits is not digital clipping but analog clipping.  I'm not saying it doesn't work (in typical scenarios) but I don't see how it can! 

[dogmusic]

One more thing:

The M10 names your recording with the current date and a recording number. If you delete recordings, the M10 keeps incrementing the recording number. If you format the unit, as you would after moving data to your PC, the number is reset back to zero.

That incremental recording number attached to the date will automatically start from "_01" again on the next day, even if you don't format.

[xpander]

Both D1 and D50 have two A/Ds (AK5358s in D50). Right after the usual analog input stages, the signal is split into two paths, both running their own A/D stages, one set for normal recording levels, one running 20dB lower. System controller selects the appropriate line to be committed into memory according to limiter settings.

It's hard to see the utility of that system - the problem that remains when recording using 24 bits is not digital clipping but analog clipping.  I'm not saying it doesn't work (in typical scenarios) but I don't see how it can! 

Hmmm. I don't have all the details exactly, as I don't have the service manuals myself...just two little pieces of schemas. Anyways, as far as I can tell..and what I was trying to tell is that the input limiting is "built in" at analog stages in D50, right in front of the ADs.

Nothing but external limiter in front of the recorder would prevent clipping the input stages if you'd have strong enough source. Next option would have something built into pre-amp stages like here, in a way. And the last, manipulating the already digitalized signal, which is kinda moot. Right?

[jamesmcn]

I cooked up a simple sine wave in Logic this morning, and ran the signal out from my macbook to the M10. I adjusted the output volume on my mac so that the M10 "over" lights lit up during the loudest part of the test tone.

You can examine the test tone and the M10 recording.

I don't see any deformation in the shape of the sine wave in the M10 recording. The is some "bounce" to the amplitude of the sine wave at the beginning and end of the loud section.

I'm not sure what (if anything) this says about the M10s limiter.