Topic: Edirol R-44 - 4 Channel Recorder (Part Tres)

[Ozpeter]

We know how the limiter works from the email I had from Edirol and which I posted earlier - from memory, the limiter switch cuts the analog gain by 12dB and the limiter itself then works in the digital domain.  Then, 12dB of makeup is digitally applied before the bits go to the card.

As for the smooth knob - I have had an idea for a test which would determine how it works and I may be able to post further in a few hours' time.  I accept that my view of it is based on zero level = zero bits, and I also accept that it would be possible for them to arrange digital mute at the bottom of the knob's travel.  But did they?

[Ozpeter]

The reason to use as little front end gain as possible is to both reduce the odds of overload and reduce the amount of hiss.

As you increase the gain, the noise floor does not increase as much as the gain does (see my previously posted tests) - except at one point in the knob travel.  Therefore, at higher clicky knob settings, you are getting a better signal to noise ratio - so long as you don't overdo it and clip, of course.

I think I'll repost those noise floor readings later too.

[Ozpeter]

OK chaps, brains in gear... I've been doing some measurements on my 'stock' R-44 (24 bit files).  All figures are average RMS below full scale in dB.

Firstly I investigated whether, when you go from click to click on the clicky level knob (the preamp knob), what difference there is from step to step.

To do this I simply recorded a steady tone at each step, going up from the lowest gain, and the results are as follows.

Average RMS
Clicky 24 bit tones
-78.07
-71.82 = 6.25
-65.68 = 6.14
-59.42 = 6.26
-47.09 = 12.33
-40.88 = 6.21
-34.8 = 6.08
-28.66 = 6.14
-22.38 = 6.28
-16.74 = 5.64
-10.71 = 6.03

The figure after "=" is the difference from the figure before.  So from lowest setting to next one up, gain goes up by 6.25dB.  Then by 6.14, etc etc.

Note that from the postion "-14" on the R-44 display of gain to the position "-20" the gain increase is actually 12.33dB not 6dB.  The rest are ballpark 6dB jumps. 

Now let's see what we can tell about the smooth control, which I suspect is just digital gain.  At its lowest setting, when recording a steady tone at a fixed clicky gain setting, you get digital silence (all bits zero).  As you increase the setting to some arbitrary points you get these pairs of signal and noise results (noise measured with a 150 ohm resistor plugged in) -

Smooth 24 bit tones & noise
-39.86 signal
-106.58 noise

-23.85 = 16.01
-90.48 = 16.1

-17.03 = 6.82
-83.67 = 6.81

-13.77 = 3.26
-80.39 = 3.28

-11.07 = 2.7
-77.78 = 2.61

-8.95 = 2.12
-75.63 = 2.15

Again, the figures after the "=" sign are the change from the previous values, and you can see that signal and noise pretty much increase by the same amount as you raise the knob from its lowest setting to its highest setting (as represented by the last pair of figures).  This to me is still the kind of behaviour you would expect from digital amplification - signal and noise increase by the same amount as you go up.  Put it this way - it shows that there is no particular point in amplifying the signal this way vs doing it in the DAW as you don't get an improvement in signal to noise ratio either way.  Analog gain change vs noise change tends to not be linear (see next lot of figures).

Lastly, below are some figures just for the noise floor at each position of the clicky gain knob measured across 150 ohms.

Noise 150 ohm
-107.42
-106.24 = 1.18
-104.42 = 1.82
-101.68 = 2.74
-106.65 = -4.97
-104.26 = 2.39
-100.42 = 3.84
-95.35 = 5.07
-89.48 = 5.87
-83.85 = 5.68
-77.84 = 6.01

 This shows the noise floor being at -107.42dB with the knob at its lowest setting.  When you increase the gain by one click from the bottom (gain increase of 6.25dB, see first table of figures) you get a noise increase of only 1.18dB (to -106.24dB).  So that's a better signal to noise ratio if you don't get clipping.  And you can read off the figures for other clicks of the knob. 

Note that when you go from the position labelled "-14" to that labelled "-20" on the display, you are getting a gain increase of 12.33dB but a fall of noise floor of 4.97dB, so there is a very significant improvement of signal to noise ratio at that point.  I would interpret that as being the setting from which which Edirol assume you will be using mics rather than line inputs.  At each point above that, you get more gain increase vs noise increase with each click of the knob, though the differences become small for the last four settings.  One could argue that recording with the clicky knob higher than the setting labelled "-38" on the display is not really worthwhile in terms of improved signal to noise ratio - you might as well raise the level in your DAW if need be, as that too will raise the noise and the signal equally, but with the knowledge that you won't get unforseen clipping.

Bottom line - have the clicky knob at "-38" or below (because you might as well use your DAW to amplify above that figure) and avoid going as low as "-14" when using mics as the loss of gain when you go down one more click is 12dB, while the noise goes up 4.97dB at that point.  And you might as well keep the smooth control at noon as if it isn't a digital gain control, it certainly seems to act like one.

Whew!  And goodnight.

[Gutbucket]

Nice work, that sheds a lot of light on what's going on.  Thanks for doing the measurements!

[Roving Sign]

This statement in the Oade quote is interesting...

The key point is the analog signal does not pass thru the rotary controls, they are digitally controlled but of an analog signal path.

Seems he is implying the gain is analog - but the control is digital...which doesnt sound like the same thing as saying it's "digital gain" (ala normalize).

[Ozpeter]

Seems he is implying the gain is analog - but the control is digital

That's not so unusual these days (digital control of analog gain) - eg Motu Traveler works that way for one.

[Gutbucket]

That's what I was getting at as well. 

Just to clarify what I stated earlier about the practical aspects of the smooth knob changing the metering so that the onset of clipping may not be as obvious until the channel indicator blinks (ignoring noise issues)-

The danger is in setting the smooth level knob below the 12:00 position.  In that case the input stage will clip, indicated by the blinking channel marker, before the meter display reaches 0db. So there is no fore-warning, just a indication once input stage clipping is happening.

From an avoid-any-clipping metering standpoint there isn't a risk if setting the smooth knob above 12:00 which causes the level to the ADC to clip before the analog input stage.  At least in that case an overly hot level is obvious on the display.  Running it that way is sort of like dialing in forced headroom in the input stage, or setting your clock ahead to trick yourself into being on time. I like accurate clocks.

With Ozpeter's measurements I'll shoot to keep the 'clicky' knob in the center 5 positions and the smooth knob at 12:00.  Conveniently, it makes intuitive sense that the best settings are near the center of both the knobs' travel.  It's nice when intuition and practice agree.

To sum up, the smooth knob is best left near 12:00, can be used above that position without risk even though the benefits of doing so are questionable, but should not be used below 12:00 to avoid clipping the input stage without warning.

[kirk97132]

Thanx for the detailed test results Ozpeter.  Gutbucket in theory and on paper I agree with what you summarize but in reality I have never been able to run my deck with the outer knob more that 3 clicks up which would be -14, even then it was because of the line level signal needing more gain.  Granted I have never run my mics straight into the R-44 so I cannot speak as to the practicality of those setting for straight mic in.  I do run a sound Devices MP-2 in front of the unit and with the MP-2 set as low as it can go I have to run the outer(clicky) knob as far down as possible or risk a clipping indication on the channel display.  FWIW that display, in my mind, is like a peak indicator on a mixer and should be avioded at all times.  Like Louie said Blinky ANYTHING is bad.  It would be nice if Edirol was to update things so there was a peak hold and a clip indicator that held too instead of returning to normal after a short time.  I like the idea of a lower noise floor but can't see being able to reach those knob setups while recording amplified rock and roll.  Those are my experiences in the last 11 months. 

[Roving Sign]

How does using this gain system compare to a traditional mixer/channel strip approach? (this might have been referenced in the Oade quote)

Seems like they were trying to replicate the channel strip (trim + fader) approach using a "digitally controlled analog signal path" - but perhaps the controls dont work as intuitively as they might have liked? - and you dont get a trimmer peak light - instead the downstream clip light (assuming this is the A/D input) not as useful...once you've lost control of the gain.  A -3db peak light on the input would be more telling...

(hope Im not lagging off topic here too much - but Im still on "Team Lust" - so, been following this thread a while...)

[Gutbucket]

Kirk,

Of course with a hot board feed or preamp output you need to accommodate the signal level however you can.  My comments are based on going mic-in.  The only preamp I've run in front of the R-44 so far is a CA-UGLY and that was for powering PIP mics.  I do plan to run the V3 in fornt of it but I'll 'prolly connect it digitally.

Roving,

Using the channel strip analogy, input trim =  the clicky knob and the trimmer peak light is the blinking channel indicator (indicating clipping of the input stage).  The smooth knob = the fader, the level meter is then post fader.

[kirk97132]

Well right of course levels need to be set appropriately , and like the V-3 I do also use a digi in from my UA-5 a lot too. I do like the digi input option which then negates the whole gain conversation.  in fact I use my UA-5 on most board feeds I get.  I have  calibrated  an SUV-1 meter that makes setting levels on the UA-5 a snap.  The other thing is the MP-2 signal runs on the really hot side even turned all the way down.  We even used this once for a second set of mics with a 744T and had to back down the gain a lot compared to the stock pre's in the 744.  I've never run my BM2p+ analog out into this deck,  one because of the A/D >D/A conversion and 2 because I  like the digi in and like to think that the UA-5 has a slightly better A/D section than the R-44 or at least I've heard that...somewhere can't recall where...although I'm not sure I could tell the difference personally.  One day I'll have to try and run mics directly in and see what I think of the stock pre's.  I hear that they sound good but have never run mics that way in all this time I've owned the unit.  For me most recording done nowadays is a mic/sbd mix with mics almost always being on stage lip.  In fact I've only done two audience recordings with it. 
    Now that I know the -20 setting has a lower noise floor I will try to get to that setting if possible.  I'd like to think that with that big a difference it would be something you could hear.   It does seem that conversations always seem to fall into agreement that the metering is the one area where we all wish it was a little better than it is.   Of course I'm not going to get rid of me deck because of it and even knowing that still think that pound for pound, feature for feature and most importantly dollar for dollar  there is not anything better out there right now.  Every time I use it I am glad I bought it, Kirk

[Gutbucket]

Now that I know the -20 setting has a lower noise floor I will try to get to that setting if possible.  I'd like to think that with that big a difference it would be something you could hear.

As you are no doubt well aware, it really depends on what you are recording.  For most of what people are recording here at TS, the ambient noise of the venue will be way above the equipment's noise floor.  If you are recording something like delicate chamber music that needs loads of gain then these efforts to optimize the noise contribution of the preamp might have more merit.  Even then, the self noise and sensitivity of your mics is likely to be a bigger contributing factor.

[kirk97132]

Gutbucket, good point...couldn't see the forest because the trees were in the way

[BayTaynt3d]

The danger is in setting the smooth level knob below the 12:00 position.  In that case the input stage will clip, indicated by the blinking channel marker, before the meter display reaches 0db. So there is no fore-warning, just a indication once input stage clipping is happening.

I haven't been around in a while, but in reading this whole thread, GutBucket nailed it in my book. After using this thing in the field weekly for a year or more now, and after my own tests, I agree with this assessment, and this is how I rationalize drawing the conclusion that the inner smooth knob is purely digital (not just digital control of analog gain). I pretty much always run at 12 'o clock+ these days on the smooth knob. I actually find that reversing moves on the outer knob in post to be pretty easy, especially knowing the steps are roughly 6db. Makes it super easy to reverse a move. BUT, that 12db gain move is really interesting, and good to know. Thanks for finding that!

[Ozpeter]

BUT, that 12db gain move is really interesting, and good to know. Thanks for finding that!

I double checked (did the whole test twice) and the result ties in with the odd change in noise floor, but if anyone had a moment to do any sort of test on their unit to confirm that it's the same with theirs, it would be good to have some assurance that it's not something odd about mine - unlikely though that might be.