FAQ: Record L&R and Use Aux 6 for the PA

[RoadRanger]

Yah, as the DL1608 doesn't have a recording level adjustment and records WAY low it makes sense to record at 24 bits so you can boost it in post. And my target is web videos so either bitrate is probably OK - but yes, video tends to be at 48.

[WK154]

Hmm... I wonder if there would have been any real advantage to recording at 24/48 instead of 16/44?

Not without fixing bigger problems first. I converted the 136 AAC (not that it mattered) into a wav file. The peak values where around 10% of what 16 bits can handle (64K). That's about a 13 bit recording. Philips cassette's are around 12 bits as a reference point. It would however have given a little more detail and would have helped when normalizing (yuk) and down-converting to 16 bits for CD quality. Oh and 44.1k would have been a better choice (no time conversion involved). This is assuming that your target was 16/44.1k CD.

I thought about this a bit and decided to run some simple test to confirm my suspicions. Turns out there is no difference between 16 and 24 bit recording using Voice record. Sounds strange but in reality it's quite simple. The DL converts an analog signal via the 24bit A/D converter and deals with it internally with at least 24 bits. What bit depth you store it in only matters if the number is large, in this case greater than 16 bits. This is not the case with the DL recording. My settings were 0 meter and fader setting on input (using a stable 1kHz sine wave) and you end up with -24 dBFS rms on the recording. That's under 10% of a 16 bit depth, in other words the number 6400 peak out of a 64K range. The 24 bit depth would contain the same number (as it does based on the recording) in a number range of 0-16.7 Million. No more detail and other than using more storage no real value. The recording settings for Voice Record were 160 kbps (AAC-LP) at 48K sample rate and a bit depth of both 16 and 24, stereo and max encode quality. I adjusted both the input fader and the output compressor gain and could get a +10dB (input fader) or +20dB (compressor) gain to get a reasonable level for recording. That's a -14dBFS rms on the fader and -4.5dBFS rms on the compressor adjustments. None of these caused clipping although the meters would indicate otherwise. Back on the old Mackie forum I mapped the meters and found them to be off as much as 10 dB ( fairly useless for accuracy, good for ball-parking).
Sam and many of you may want to take a look at the explanation on headroom in the digital world here. Keep in mind that any mixer can still be measured for dBu in and out. They're just all unique.
http://www.sengpielaudio.com/calculator-db-volt.htm
Conclusion: Due to Mackie's false indication of input levels the recorded output is too low. The best  S/N ratio of course can be gained at the preamp settings not at the tail end compressor. Yes that means a "hot" input signal based on their meters even if it's not so. This will also give you max dynamic range. Remember that the meters are PPM meters not RMS or VU meters and show peak values. A while back there were a lot of complaints that recordings from CD's were to "hot" when played back on the iPad, they're right on the money but the meters not so much. Meters on the DL are problematic. If you have a good rms meter you can verify this.

[WK154]

RR if your target was YouTube then the current recommended specs are 16:9 720p at 44.1K and 16 bits. Currently the audio ends up at 128kbps AAC. The upload converters are giving this kind of results at present but it has changed over time as they improve things.

[RoadRanger]

I thought about this a bit and decided to run some simple test to confirm my suspicions. Turns out there is no difference between 16 and 24 bit recording using Voice record. Sounds strange but in reality it's quite simple. The DL converts an analog signal via the 24bit A/D converter and deals with it internally with at least 24 bits. What bit depth you store it in only matters if the number is large, in this case greater than 16 bits. This is not the case with the DL recording. My settings were 0 meter and fader setting on input (using a stable 1kHz sine wave) and you end up with -24 dBFS rms on the recording. That's under 10% of a 16 bit depth, in other words the number 6400 peak out of a 64K range.

-24dBFS in volts is ~1/16 so I'd expect a 20 bit signal in that 24 bit recording? Yes, it would only be 12 bits in a 16 bit recording. Internally everything is floating point AFAIK. On the recording a 16 bit recording would be the 24 bit recording with the bottom 8 bits thrown away - not the other way around where the 14 bit recording would be the 16 bit one with 8 zero bits added on top as that would make the 14 bit recording -72dBFS vs the -24dBFS of the 16 bit recording .

[WK154]

I thought about this a bit and decided to run some simple test to confirm my suspicions. Turns out there is no difference between 16 and 24 bit recording using Voice record. Sounds strange but in reality it's quite simple. The DL converts an analog signal via the 24bit A/D converter and deals with it internally with at least 24 bits. What bit depth you store it in only matters if the number is large, in this case greater than 16 bits. This is not the case with the DL recording. My settings were 0 meter and fader setting on input (using a stable 1kHz sine wave) and you end up with -24 dBFS rms on the recording. That's under 10% of a 16 bit depth, in other words the number 6400 peak out of a 64K range.

-24dBFS in volts is ~1/16 so I'd expect a 20 bit signal in that 24 bit recording? Yes, it would only be 12 bits in a 16 bit recording. Internally everything is floating point AFAIK. On the recording a 16 bit recording would be the 24 bit recording with the bottom 8 bits thrown away - not the other way around where the 14 bit recording would be the 16 bit one with 8 zero bits added on top as that would make the 14 bit recording -72dBFS vs the -24dBFS of the 16 bit recording .

OK the math is like this (the reason I gave the Sengpielaudio link above) and we've all fallen for the analog thinking about digital at one time. The DL has a 0dBFS at 21dBu as per spec. The -24dBFS works out in theory to -3 dBu (21-24=-3). This amounts to 0.54V rms or 1.5V PP. The above 6400 example (10% of 64K) had nothing to do with actual values just a explanation (sorry for the confusion). The point I was trying to make is that if you input a voltage that converts to 6400 (we don't really know what Mackie is using internally fixed or floating point both are available) and you loose nothing in internal processing in precision one would expect that same number out, converted to the same voltage. If however you exceed that max number internally you would clip. Now for reality. Mackie looses about 3dB from input to output (all fader at 0 and no processing input signal at 0) if you use the visual scale and on top of all that it's nonlinear (as much as +/-10dB) if you plot it against dBu or dBFS. The point I made on the old Mackie forum. Their centerpiece on the MF and they can't get it right. Your correct if you down convert a full 24bit value to 16 bits you loose precision the least significant 8 bits. I converted the 24 bit recording to 16 bit (in Audacity) and no precision was lost due to the low values and a smart conversion.

[RoadRanger]

I'm 99% sure they use floating point so no loss of precision internally.

[WK154]

I'm 99% sure they use floating point so no loss of precision internally.

I'm not that sure and both have the same precision so no loss either way and fixed still beats float for speed.
Correct that the fixed point is 32 bits (higher precision than 32bit floating point equal to extended 40 bit Float) but we can't create something out of nothing. 24 bits in is the max (A/D) internal math is not going to change that.

[RoadRanger]

It's actually 24 bits per input, so 2 inputs gets you 25 bits, 4 inputs gets you 26 as you sum - but I dunno where the noise floor is? In any case it's a LOT easier to do all the EQ and such in floating point so you don't have to worry about over or under flow and you retain precision no matter how far down you are. You can attenuate something by 200 dB and somewhere else amplify it by 200 dB and still have your 24 (or whatever) significant bits. No challenge at all LOL.

[WK154]

It's actually 24 bits per input, so 2 inputs gets you 25 bits, 4 inputs gets you 26 as you sum - but I dunno where the noise floor is? In any case it's a LOT easier to do all the EQ and such in floating point so you don't have to worry about over or under flow and you retain precision no matter how far down you are. You can attenuate something by 200 dB and somewhere else amplify it by 200 dB and still have your 24 (or whatever) significant bits. No challenge at all LOL.

I understand but if you look at the Sharc data flow structure It has a 64 bit data transfer bus and 80 bit multiplication registers. They made sure that most things won't run into a limit. So 8 bits or 2 to the power of 8 = 256 channels. You think 16 is a problem? If they were clever they would use both for the right job.

[WK154]

For those of you reading this don't get the impression that 24 bit recording is no better than 16 bit recording because it is. The key in this case was the under-utilization of the hardware.  The manufacturer sets the range of the A/D converter for the 24 bits. If the range is set too low then you are converting a lot of your noise floor. This does nothing for your dynamic range which is the max level the A/D or preamp can handle minus the noise floor. The trick here is to utilize as much of your equipments range as possible without hitting it's limits (High and low). The goal is to get 50-60 dB above noise floor for great dynamics but that's not as easy as it sounds since some music only has a 20 dB dynamic range. The advantage to 24 bit recording is that it is more forgiving in trying to reach that compromise in settings. If however your dynamic range is only 30 dB there is nothing either 16 or 24 bit recording can do to get it back or improve it, it's gone. Hope this helps if the stuff above is too techie.

[Wynnd]

So if I'm reading this right, we should be setting the trims as high as possible without clipping to get the most out of the dynamic range that exists.  (I've been getting crap for using db numbers on another forum.  I'm trying to break that habit, but sometimes it just seems the only way to actually talk about things with any accuracy.)   I've noticed the tolerance of the DL's pre-amps.  There sure seems to be a large range there.  Anyway with most recordings rarely ending up with much in the way of dynamic range, it might not be as important in the end.  (One very well done exception is "The Letters" by King Crimson from the Islands album.  That one probably has a 30 db range from the beginning at a whisper to the solos in the middle that are screaming.  Don't listen in a car with the engine running.  Best with headphones set barely audible at the beginning.  And they did that with analog equipment.  Amazing.)    It's hopeful that Mackie could massage more recording possibilities out of the DL.  I'm hoping they don't give up on software development, but unless they charge for significant upgrades, there isn't much incentive for development of this mixer.  (And we have all gotten used to free upgrades.)  I would be willing to pay for an upgrade that included multichannel recording.  (I'm happy with the improvements that have occurred already.)  If most of the products shown at NAMM come onto the market, Mackie's not going to have much incentive to continue working on the DL and instead start their next mixers.  On the other hand, what they've learned about programming this mixer should be easy to port to new products.  (How long have most of these companies had software development departments?)   Thanks for the discussion, even if parts of it were solidly above my head.  Very interesting.

[WK154]

Wynnd this was meant for your recordings if your looking for quality work. There is nothing that can be done for other peoples work it is what it is. This of course leads into the "Loudness war" discussion something for another time and place.

[Wynnd]

The loud band is gone from my life.  Currently only working on a low volume project.  Might actually keep my hearing for some years to come.  (Hope so.  My Mother who never was exposed to loud music had two hearing aids at the end of her life.  She couldn't hear the phone ring without them.) 

[robbocurry]

Going to try this over the weekend, my own band on Friday and a very loud tribute act on Saturday.
Saturday will be easier (recording wise) as only the vocals will be actually coming out of the rig, if past experience is anything to go by, but everything will be mic'd up.