Continuing the 16bit Analog Input Debate....
Posted: Sat Jul 21, 2012 1:20 am
I was going to title this post as: "Save Yourself 8 Pages of Misinformation, Charts, Graphs and Acrimony" but it wouldn't fit.
I'm posting this as a separate thread to avoid 8 pages of confusion, misinformation, flawed testing and (eventually) uncalled-for acrimony.
I had a lengthy conversation this morning with the engineer who programmed the V3.1 fixes and the entire V5.x software revision. Not some low level guy. The top guy. The guy who would know. I have that kind of address book.
The subject of our conversation was the appearance of a 16bit signal being passed from the Mic/Line Pre to the digital outputs (Direct or 2Mix).
Let me start by how he prefaced his answer: by pointing out the theoretical dynamic range of 24 bits: it's 144dB.
What's the preamp's dynamic range spec of the D8B?
• Preamp dynamic range: 114 dB RMS max.
That said.....he continued:
The Cirrus A/D converter in the D8B is in fact, a 24bit converter. The original programming of V3.0 analog signal chain was brutally flawed. It passed a 16bit signal with an additional 8bits of a sort of "repeat" to form a 24bit word. This was corrected in V3.1 and further refined (to perfection, I am told) in V5.x.
So why the appearance of what looks to be a 16bit signal being passed from the Mic/Line>Direct/2Mix chain?
Plain and simple: the analog preamp's dynamic range. The D8B's preamp in front of the A/D doesn't have a large dynamic range. Keep in mind, this is a preamp using 1996 components producing a usable range of 60 dB feeding a converter developed somewhere around 1996.
What this means is the D8B can only take advantage of, on average, about 16-18bits when being fed a signal near the upper limit of the preamplifier. Really pushed? Maybe 20bits. He said that typically, the material filling up the last few bits is noise.
To compare, a signal being generated from a high end VSTi synth such as U-He's Diva can produce a full 24 bit signal in a DAW, pass it into the D8B, via Lightpipe, send it out at the Direct Out or 2 Mix Out and will (depending on the material) be measured as a full 24bit signal.
Plug a high end "boutique" A/D preamp into any of the D8B's digital inputs and it will produce a signal that is actually less than 24bits because the dynamic range still falls short of 144dB. Have a look at high end A/D preamps specs. Have a look at mid-line gear (RME, MOTU and the like) It's eye-opening. They are not producing full 24 bit words on average either. Just sayin'.
So, to sum up:
V3.0 software produces a 24bit signal with the first 16bit actually being the signal and the last 8 bits a sort of "repeat of the first 16 filler". It is essentially, a 16bit signal. It was "the dirty little secret". It is crap.
V3.1 software was debugged to produce a theoretically "real" 24 bit signal but doing so becomes limited by the dynamic range of the analog preamp. The process was improved but not perfectly.
V5.x software has additional tweaks over V3.1 to make this process as efficient as was theoretically possible within the constraints of the hardware elements (the preamp and A/D converter).
It's the analog preamp that's the bottleneck, not the converter. It really is a 24bit A/D converter in there.
P.S.: Version 5 is a lot more than a graphic overhaul. Plenty of useful new features and tons of bug fixes. It is well worth the upgrade fee.
I'm posting this as a separate thread to avoid 8 pages of confusion, misinformation, flawed testing and (eventually) uncalled-for acrimony.
I had a lengthy conversation this morning with the engineer who programmed the V3.1 fixes and the entire V5.x software revision. Not some low level guy. The top guy. The guy who would know. I have that kind of address book.
The subject of our conversation was the appearance of a 16bit signal being passed from the Mic/Line Pre to the digital outputs (Direct or 2Mix).
Let me start by how he prefaced his answer: by pointing out the theoretical dynamic range of 24 bits: it's 144dB.
What's the preamp's dynamic range spec of the D8B?
• Preamp dynamic range: 114 dB RMS max.
That said.....he continued:
The Cirrus A/D converter in the D8B is in fact, a 24bit converter. The original programming of V3.0 analog signal chain was brutally flawed. It passed a 16bit signal with an additional 8bits of a sort of "repeat" to form a 24bit word. This was corrected in V3.1 and further refined (to perfection, I am told) in V5.x.
So why the appearance of what looks to be a 16bit signal being passed from the Mic/Line>Direct/2Mix chain?
Plain and simple: the analog preamp's dynamic range. The D8B's preamp in front of the A/D doesn't have a large dynamic range. Keep in mind, this is a preamp using 1996 components producing a usable range of 60 dB feeding a converter developed somewhere around 1996.
What this means is the D8B can only take advantage of, on average, about 16-18bits when being fed a signal near the upper limit of the preamplifier. Really pushed? Maybe 20bits. He said that typically, the material filling up the last few bits is noise.
To compare, a signal being generated from a high end VSTi synth such as U-He's Diva can produce a full 24 bit signal in a DAW, pass it into the D8B, via Lightpipe, send it out at the Direct Out or 2 Mix Out and will (depending on the material) be measured as a full 24bit signal.
Plug a high end "boutique" A/D preamp into any of the D8B's digital inputs and it will produce a signal that is actually less than 24bits because the dynamic range still falls short of 144dB. Have a look at high end A/D preamps specs. Have a look at mid-line gear (RME, MOTU and the like) It's eye-opening. They are not producing full 24 bit words on average either. Just sayin'.
So, to sum up:
V3.0 software produces a 24bit signal with the first 16bit actually being the signal and the last 8 bits a sort of "repeat of the first 16 filler". It is essentially, a 16bit signal. It was "the dirty little secret". It is crap.
V3.1 software was debugged to produce a theoretically "real" 24 bit signal but doing so becomes limited by the dynamic range of the analog preamp. The process was improved but not perfectly.
V5.x software has additional tweaks over V3.1 to make this process as efficient as was theoretically possible within the constraints of the hardware elements (the preamp and A/D converter).
It's the analog preamp that's the bottleneck, not the converter. It really is a 24bit A/D converter in there.
P.S.: Version 5 is a lot more than a graphic overhaul. Plenty of useful new features and tons of bug fixes. It is well worth the upgrade fee.