#1209968 - 04/11/13 07:08 PM
Vsr880/890 gurus: vsr mode 24bit why uses less space?
|
mutetourettes
Planeteer
Registered: 02/10/13
Posts: 38
|
Hello - so i've searched and googled and faq'd my little heart out... I *think* I've established that my vsr-880, used as a simple 8in/8out rack-recorder attached to a mixing desk that covers all the in/out/eq/mixing etc is best used in 'VSR' mode to keep the full 24 bit depth of the recordings in uncompressed format, and has to be at 44khz to allow the full 8 tracks..
My question is this - that mode seems to offer the same recording time as 'MT1', and much more than the 16bit uncompressed MAS and CDR formats... what's going on there then?
is 'VSR' mode data-compressed? if so is it lossless or lossy? If I rip the files out using the software (I've yet to sort out how to do that really) do i get nice 24bit wavs?
Related questions since you were kind enough to read this far:
a) if i use the vsr-880 to record a 24bit 44.1khz digital input, (perhaps from a very posh but now somewhat passe and therefore surprisingly cheap outboard analogue/digital converter) how do I establish concretely that the input mixer is recording bit for bit what i put in, and not messing around with it (attenuating/boosting etc)? ie how do I tell it: 'hands off my signal, just record it'
b) if i record a 16bit digital signal into a 'VSR' mode (24bit) song how is it handled by the vsr-880?
I assume that all this applies to the 890 too... anyone got any definitive answers?
Cheers for your patience...
Mark
|
|
Top
|
|
|
|
|
#1210011 - 04/11/13 11:20 PM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: mutetourettes]
|
bear
Planeteer
Registered: 10/25/99
Posts: 6626
Loc: abq,nm,usa
|
I am unfamiliar with the "VSR mode". All other modes I know of though (MTP,MT1,MT2,LIV etc) use roland proprietary RDAC compression which IS LOSSY - i.e, you can never exactly get out what you put in...
If you say the recording time is what it is for VSR mode, then logically it must also be using compression, and it almost surely is a form of RDAC.
MAS and CDR are the exception - they do true PCM encoding, with no compression, but as you see they have they own set of limitations.
In general though the consensus is that RDAC works pretty well, and sounds pretty good.
In general, feeding a 24 bit recording format with 16 bit digital signal will result in the lower 8 bits being set to zero
|
|
Top
|
|
|
|
|
#1210013 - 04/11/13 11:33 PM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: mutetourettes]
|
uptildawn
Planeteer
Registered: 12/15/01
Posts: 9106
Loc: on land
|
I believe I've read that Roland uses some form of their own special data compression in every recording mode..... except MAS and CDR, I guess.... Forgive me if I'm wrong.
Thing is, that no matter, the MTP (1680/1880) and VSR modes are 24 bit and sound quite pleasing to the ear. In my opinion and based on tests I'd performed with the 1680, MTP is smoother sounding than MT1 and MAS modes... I don't have CDR mode, but have heard it and didn't notice anything special.
Maybe the VSR mode has more recording time than the MAS and CDR modes because there IS some data compression being performed (as I thought), even though it's 24 bit. MAS and CDR modes are obviously made strictly for building a final audio CD-R.... I've never much cared to use it myself.
From my own experience, observations and bear's own comments, I think I'm safe in saying that you do, indeed get very nice 24 bit converted wav files from the ripper/VSWE combination. Not only that, but putting those wav files into a PC daw brings them to a new level of life, at least in all the projects I've used them in. Maybe that's because the Roland amplifier sections kind of choke some of that life out of the sound, when monitored through the VS beast.... just a guess. I've always been left lacking when listening to my "finished" mixes out the back of the 1680.
Not sure how to answer your (a) question.... I'm of the opinion that it doesn't really make a bit of difference in terms of what sounds good, or not. If you're pleased with the sound, what more counts? Of course, that's hardly likely to satisfy a technically minded mind..... sorry if that isn't helpful.... it's strictly my biased and crudgey opinion.
Your (b) question is very interesting to me and I've asked roughly the same in another thread going on in the 1680 forum right now..... I'm hoping someone like bear might shed some intelligent light on the subject.
I have put 16 bit files into 24 bit projects on the computer daw and felt that they sound slightly nicer than they do played through a media player over the same system and at the same volume (at least as near as I can match it using my ear and the output meters on my monitoring mixer).
Sorry, but I'm afraid these aren't definitive answers..... I could make an excuse that I have a cold, which I do..... 
And of course, I'm rattling at the fingertips so much tonight that bear had time to respond before me....
Edited by uptildawn (04/11/13 11:35 PM)
_________________________
uptildawn
|
|
Top
|
|
|
|
|
#1210121 - 04/12/13 01:57 PM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: uptildawn]
|
Hook
Admin / Slacker
Planeteer
Registered: 01/03/01
Posts: 6526
Loc: Venice, FL
|
Check out my FAQs, specifically Section I:
http://www.ps139.com/vs880ex/pages/FAQs.html#I1
I had researched this stuff quite a bit when I worked on the VS Recording Power book - here is some info from that:
Recording Modes
All VS workstations, other than the VS-2000CD, provide the ability to record data in either linear or compressed modes. Table 2.3 shows a list of the various recording modes. Linear recording modes can record 16-bit audio or 24-bit audio, depending on the VS workstation. The compressed recording modes use Roland's proprietary R-DAC compression scheme. The original version of R-DAC found on the VS-880 was a 16-bit routine. When Roland introduced the VS-1680, it included a new version of R-DAC written as a 24-bit routine.
TABLE 2.3
Recording Modes and Descriptions
M24 (Mastering 24-bit): 24-bit linear
MTP (Multi Track Pro): 24-bit RDAC with an average 3:1 compression ratio (also called VSR mode on older workstations)
CDR (CD-R Writing): 16-bit linear non-separable stereo
M16 (Mastering 16-bit: 16-bit linear (also called MAS mode on older workstations)
MT1 (Multi Track 1): 16-bit R-DAC with an average 2:1 compression ratio
MT2 (Multi Track 2): 16-bit R-DAC with an average 3:1 compression ratio
LIV (Live Recording 1): 16-bit R-DAC with an extreme compression ratio
LV2 (Live Recording 2): 16-bit R-DAC with a very extreme compression ratio
R-DAC
One of the things that made the original VS-880 revolutionary in 1996 was R-DAC (Roland Digital Audio Coding). According to Roland, R-DAC was a completely new compression technique, optimized for professional music recording, which was developed to reduce the amount of digital data needed to record an audio waveform. Unlike compression schemes that function in the "frequency domain", such as those used by MiniDisc and MPEG, R-DAC processes audio in the "time domain" where no frequencies are singled out or thrown away. The end result of using R-DAC is a doubling or tripling of recording time available. Remember, back in 1996, hard drives were much more expensive than they are today, and a 2GB hard drive was considered big.
_________________________
Hook
|
|
Top
|
|
|
|
|
#1210124 - 04/12/13 02:23 PM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: Hook]
|
bear
Planeteer
Registered: 10/25/99
Posts: 6626
Loc: abq,nm,usa
|
Ah - I had forgotten M24 mode, and I was unaware that MTP was called VSR on some machines or that MAS was called M16 on some machines....
Other than that, I think we are all in violent agreement here...
Physics is physics - if the recording mode takes up less space per time, then compression is being used, and in the case of Roland the compression is RDAC. The smaller the space, the more extreme the compression
|
|
Top
|
|
|
|
|
#1210149 - 04/12/13 04:50 PM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: bear]
|
mutetourettes
Planeteer
Registered: 02/10/13
Posts: 38
|
Thanks all - especially for the book excerpts Hook - the faq I had read, but i didn't know that VSR was 3:1 avg RDAC compression "claimed" lossless but Bear you seem convinced it isn't, and indeed why offer MT1 and MT2 if it was.. I wonder if whoever untangled the ripper/VSWE decoding software could shed some light on the lossy/lossless values of RDAC
I guess if the data's gonna be compressed anyway then my question 'a' is moot, there's not going to be any bit-for-bit recording...
and thanks Bear for confirming what i thought about 'b'.
Cheers
|
|
Top
|
|
|
|
|
#1210370 - 04/13/13 05:02 PM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: mutetourettes]
|
bear
Planeteer
Registered: 10/25/99
Posts: 6626
Loc: abq,nm,usa
|
RDAC is most definitively NOT lossless - if you encode a file and then decode it you DO NOT get 100% the same file...
On the other hand... it sounds pretty good. I have even heard people compare it to "tape saturation sounding"
|
|
Top
|
|
|
|
|
#1211923 - 04/20/13 08:28 PM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: bear]
|
Slabraton
Planeteer
Registered: 12/16/12
Posts: 16566
Loc: California
|
It's amazing how hard facts are to come by in the digital age. It's hard to see the steak for all the sizzle that's being sold.
"most Planeteers agree that 24-bit RDAC compression found in VSR and MTP modes sound better than the 16-bit uncompressed modes." I am dismayed to find that these are my choices on the VS 890. Somehow, I managed to miss overlook that tidbit.
Also, this comment startled me: "In general, feeding a 24 bit recording format with 16 bit digital signal will result in the lower 8 bits being set to zero".
I want my Studer back.
|
|
Top
|
|
|
|
|
#1244490 - 09/19/13 12:14 PM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: Slabraton]
|
mutetourettes
Planeteer
Registered: 02/10/13
Posts: 38
|
so here's a question... a 24-bit VSR-mode song can be converted to MTP mode for 'export' to a 1680.. and that MTP-mode data can be read nicely by Danielo/Randy's software, which has been established as a bit-perfect emulation of Roland's conversion process... The file size is the same VSR/MTP but the data/encoding is definitely different...
Question: Is the transfer from VSR>>MTP losing any further data/accuracy in that process?
Anyone know?
|
|
Top
|
|
|
|
|
#1244726 - 09/20/13 07:19 AM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: mutetourettes]
|
ricket
Planeteer
Registered: 04/19/99
Posts: 245
|
Question: Is the transfer from VSR>>MTP losing any further data/accuracy in that process?
Yes.
|
|
Top
|
|
|
|
|
#1244728 - 09/20/13 07:27 AM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: Slabraton]
|
ricket
Planeteer
Registered: 04/19/99
Posts: 245
|
Your Studer loses *fidelity* at higher gain levels from 2 sources:
- preamp distortion
- soft tape distortion/compression
|
|
Top
|
|
|
|
|
#1244756 - 09/20/13 01:35 PM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: ricket]
|
uptildawn
Planeteer
Registered: 12/15/01
Posts: 9106
Loc: on land
|
Question: Is the transfer from VSR>>MTP losing any further data/accuracy in that process?
Yes.
Is this something you know for sure?
How is it losing anything in the conversion, if both formats are the same sample and bit rates?
_________________________
uptildawn
|
|
Top
|
|
|
|
|
#1244769 - 09/20/13 02:07 PM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: uptildawn]
|
mutetourettes
Planeteer
Registered: 02/10/13
Posts: 38
|
indeed I'd love to hear ricket's 'Yes' explained in detail - If the VSR is a later, further developed version of RDAC than MTP then is it 'better' at packing 16x24bit samples into 16x16bit chunks?
if so then presumably transcoding VSR>>MTP might involve some loss.
If the VSR just makes wiser choices about equivalent packing patterns then there wouldn't be a loss.
(Of course 'better' might just mean it runs quicker in code or better on the processor or something, to justify the extra development, I suppose the later version might not be any 'better' in terms of waveform fidelity)
The process of conversion takes a good while... so it might be going decode/reencode via PCM.... that, I would expect, would be even lossier...
Another thought: Did the 'VSR' mode coincide with introduction of machines with true 24-bit A/D and D/A converters? (the 1680 was 20bit wasn't it?)
|
|
Top
|
|
|
|
|
#1244851 - 09/20/13 07:22 PM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: mutetourettes]
|
uptildawn
Planeteer
Registered: 12/15/01
Posts: 9106
Loc: on land
|
Yes, the 1680 has 20-bit converters... The 1880, which came out next in line (IIRC), has the first 24 bit a/d/a converters... don't quote me on that, as I can't check my facts at the moment.
However, I don't believe the a/d/a converters have anything to do with the conversion of VSR and MTP modes... one to the other.
In the a/d/a sense, the effect is on the source signals being recorded and/or played back. In the other sense, we're talking about converting digital data to digital data, with no conversion to analog in between... I actually doubt that the a/d/a converters have anything at all to do with THAT sort of conversion (bear might be better informed than I am on this, so again, don't take my word to the bank).
If both record modes are 24 bit, but just renamed according to Roland's own scheme, then there should be no difference at all, I would think. MTP mode was the only 24 bit record mode for a long while in Roland-speak. I don't know VSR mode, but if mutetourettes says it is, then I assume it is the same as MTP mode in other VS recorders.
The 1680 can record in MTP mode, which is 24 bit... yet it has 20 bit a/d/a converters... I assume then that I am correct in my assumption that those 20 bit converters don't come into play when speaking about the different bit rates of the record modes... MTP, MT2, etc.
________________________________
Edit..
guess the vsr880 was out before the 1880, so the 1880 might not have been the first with 24 bit a/d/a converters...
Although the 1880 manual (pdf) doesn't give a specifications page... and in either case, doesn't spell out the bit rate of any of the record modes... at least one can compare the estimated track record time on a 1gb partition for both MTP and VSR modes and draw conclusions about the bit rates from that figure... which is 808 track minutes for both modes.
Edited by uptildawn (09/20/13 08:31 PM)
_________________________
uptildawn
|
|
Top
|
|
|
|
|
#1245383 - 09/23/13 11:49 AM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: uptildawn]
|
mutetourettes
Planeteer
Registered: 02/10/13
Posts: 38
|
Hi - yes I was just suggesting that 24-bit VSR-mode might have been somehow optimised or adjusted for encoding the extra very-low-level data from the shiny new actual 24-bit converters, as opposed to the MTP mode which, while using 24-bit words was originally only receiving data from 20-bit converters and perhaps wasn't going to waste precious compression space on the bottom four bits... it's just a hypothesis...
the RDAC files for VSR and 'MTP-from-vsr' are similar but definitely different... I have some files if anyone fancies playing about with them... And they're both nominally '24-bit' modes - ie when decoded they dump out 24bit PCM data. I think it's been established elsewhere on this board that MT1-mode is just MTP mode chopped down to 16 bits (and so only using a subset of the possible encoding 'patterns').
Ok so it's news to me that the 1880 can do both MTP and VSR modes... that might prove helpful in trying to untangle the encoding/decoding.
|
|
Top
|
|
|
|
|
#1245384 - 09/23/13 12:14 PM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: mutetourettes]
|
mutetourettes
Planeteer
Registered: 02/10/13
Posts: 38
|
Hi - yes I was just suggesting that 24-bit VSR-mode might have been somehow optimised or adjusted for encoding the extra very-low-level data from the shiny new actual 24-bit converters, as opposed to the MTP mode which, while using 24-bit words was originally only receiving data from 20-bit converters and perhaps wasn't going to waste precious compression space on the bottom four bits... it's just a hypothesis...
but then the 2480 sticks with MTP anyway (not VSR), so maybe they decided that mode can handle proper 24-bit data juuuust fine...
|
|
Top
|
|
|
|
|
#1245391 - 09/23/13 01:23 PM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: mutetourettes]
|
uptildawn
Planeteer
Registered: 12/15/01
Posts: 9106
Loc: on land
|
...Ok so it's news to me that the 1880 can do both MTP and VSR modes... that might prove helpful in trying to untangle the encoding/decoding.
I never suggested that the 1880 has both VSR and MTP modes..... I only suggested that, the two modes both being 24 bit modes and Roland having referenced their total track time in minutes in their respective manuals, that a comparison and possible conclusion could be drawn from that info about whether or not they are the same. As far as I am aware, neither machine has both modes.
As to the rest of your post, I was never aware of a discussion that MT1 has been established to be MTP mode, truncated to 16 bits. MT1 mode was around before MTP mode, so maybe MT1 mode is an uncompressed 16 bit mode, making it similar to an uncompressed 24 bit mode..... in that sense. I wouldn't know, myself.
If RDAC files from VSR and MTP-from-VSR modes sound different, and different in a negative way, then that at least gives someone pause when wanting to make that conversion....
Don't have any idea about why VSR mode was offered...... Roland makes many decisions about their product lines that often leave one wondering...... as do a multitude of other companies.
_________________________
uptildawn
|
|
Top
|
|
|
|
|
#1245423 - 09/23/13 03:45 PM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: uptildawn]
|
mutetourettes
Planeteer
Registered: 02/10/13
Posts: 38
|
ok - I misunderstood re VSR/MTP on the 1880. The file saizes are indeed identical...
the MT1/MTP thing was revealed when Randy/Danielo were deciphering the code back in 2006..
I haven't worked out how to link a specific post but it's number 12513 on this page....
http://www.vsplanet.com/ubbthreads/ubbthreads.php?ubb=showflat&Number=12513
|
|
Top
|
|
|
|
|
#1246107 - 09/26/13 10:58 AM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: mutetourettes]
|
mutetourettes
Planeteer
Registered: 02/10/13
Posts: 38
|
Just got a message from the very helpful Roland UK, who asked roland Japan:
"Japan have responded and advise information on VSR Mode to MTP Mode is on page 30 of the owners manual.
There is loss during conversion unfortunately."
|
|
Top
|
|
|
|
|
#1246118 - 09/26/13 01:29 PM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: mutetourettes]
|
uptildawn
Planeteer
Registered: 12/15/01
Posts: 9106
Loc: on land
|
Well, that's kind of a partial answer. At least they were helpful and got back to you about it. Could they not have revealed a little bit of detail about why there is a loss?
According to pg 30 of the vsr880 manual:
"The sample rate and recording mode of the newly created song will be the same as the original song."
"If the recording mode of the VSR-880 song is 'VSR,' it is imported by converting it into the VS-1680 song with 'MTP.'"
Those two statements definitely lead one to believe and have confidence that there will be NO loss of quality, especially given that VSR and MTP modes are supposedly both 24 bit modes.
What "loss" is the Roland tech rep. refering to?
_________________________
uptildawn
|
|
Top
|
|
|
|
|
#1246293 - 09/27/13 08:19 AM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: uptildawn]
|
mutetourettes
Planeteer
Registered: 02/10/13
Posts: 38
|
Well both VSR and MTP are 24bit modes with 16 samples of 24 bits 'packed' (and approximated - hence lossy, and with some 'inaccuracies' added to make quantisation distortion more palatable) using pattern codes into blocks of 16*16bits (hence 2:1 compression).
Because VSR is using slightly different decisions about which patterns to use, and possibly a slightly different way of allocating/defining/restoring the data in the patterns *I suspect* there is no 1:1 match between VSR and MTP data blocks, so it cannot just transcode with no loss of information.
It might be doing a sort of 'fudged' transcode (where some loss occurs in the fudging process) or...
*Possibly* it's decoding the VSR back to full 24bit PCM samples, and then encoding into MTP again, and that second encode from PCM into MTP is of course also a lossy process, ie it's approximating/dithering again.. losing *even more* data.
Buuuuuuut all of this is conjecture..
I guess I need to establish with a wav file converted properly to MTP on a machine that's MTP-native just how different that is to the same file encoded to VSR and then converted to MTP..
here's what Roland said (in the same email) on the subject of divulging:
"We cannot disclose details information about RDAC However this technology is patented so you might be able to find the principles of this on the patent office websites.
RDAC is still used in some products…"
The patent is a fun and illuminating read but has no specific detail about the real-world RDAC formats... http://worldwide.espacenet.com/publicati...B=&locale=en_EP
side note: Why does this interest me? Because I'm trying to suss out the specific VSR encoding, and have been using Randy/Daniel's great work as a model. What this tells me is that I probably shouldn't aim to match the VSR and VSR>>MTP outputs, and had better stick to just recording the digital output of my VSR880 to compare with.
Edited by mutetourettes (09/27/13 03:34 PM)
Edit Reason: better version of the patent online
|
|
Top
|
|
|
|
|
#1246368 - 09/27/13 05:04 PM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: mutetourettes]
|
bear
Planeteer
Registered: 10/25/99
Posts: 6626
Loc: abq,nm,usa
|
... The file size is the same VSR/MTP but the data/encoding is definitely different...
I have been watching your conversation out of the corner of my eye, without really paying attention, but I finally got around to thinking about it...
I just am not sure you are correct that VSR and MTP are different.
I pulled up Randygo's source code - here is what he says in his VSR conversion routine - he uses his normal MTP conversion then adds a couple of lines
// VSR mode is like MTP, but seems to need this bit of post-processing.
for (j=0;j<16; j++) {
if (out[j] >= 65536) out[j] -= 65536;
else if (out[j] <= -65536) out[j] += 65536;
else out[j] = 0;
I have no experience with VSR mode files - but it appears Randygo did and he felt they were pretty much MTP with just the small boundary check he added
|
|
Top
|
|
|
|
|
#1246919 - 09/29/13 09:36 PM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: bear]
|
ricket
Planeteer
Registered: 04/19/99
Posts: 245
|
Question: Is the transfer from VSR>>MTP losing any further data/accuracy in that process?
Yes. Is this something you know for sure? How is it losing anything in the conversion, if both formats are the same sample and bit rates?
Let me ask you this ... Exactly how do you perform a VSR -> MTP conversion ?
[quote=mutetourettes]... I pulled up Randygo's source code - here is what he says in his VSR conversion routine - he uses his normal MTP conversion then adds a couple of lines
// VSR mode is like MTP, but seems to need this bit of post-processing.
for (j=0;j<16; j++) {
if (out[j] >= 65536) out[j] -= 65536;
else if (out[j] <= -65536) out[j] += 65536;
else out[j] = 0;
I have no experience with VSR mode files - but it appears Randygo did and he felt they were pretty much MTP with just the small boundary check he added
Is Randygo still around ?
The post processing looks like some kind signal level reduction.
Does the VSR-880 add any gain to a signal it's recording because the lack of preamps, and the hot output level pops the WAV specification ?
What are we talking here ? 4db, 6db, 12db cut ?
Did anyone generate a WAV from a VSR-880 and check what it looks like or just sounds like ? Like recording on a channel a fade-in tone, hold a few seconds, and fade out ... then convert to a WAV ?
Edited by ricket (09/29/13 11:11 PM)
|
|
Top
|
|
|
|
|
#1246936 - 09/29/13 11:10 PM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: mutetourettes]
|
ricket
Planeteer
Registered: 04/19/99
Posts: 245
|
Hello - so i've searched and googled and faq'd my little heart out... I *think* I've established that my vsr-880, used as a simple 8in/8out rack-recorder attached to a mixing desk that covers all the in/out/eq/mixing etc is best used in 'VSR' mode to keep the full 24 bit depth of the recordings in uncompressed format, and has to be at 44khz to allow the full 8 tracks..
My experience (Disclaimer: I'm not from Roland.) confirms.
My question is this - that mode seems to offer the same recording time as 'MT1', and much more than the 16bit uncompressed MAS and CDR formats... what's going on there then?
For performance and ressource reasons (lack of a general purpose cpu, lack of RAM) the VS has to minimize operations, doesn't have the time or the amount of memory needed to write data where it would like to to keep things tidy and avoid wasting space. The VS sees the disk like watching from behind a window a train of wagons rolling outside just a few yards away. The VS business is to throw the data it holds in it's hands whenever it spots an empty wagon. No time to figure if a partially full wagon will hold the data in his hands, and it can't risk losing any data, so it'll wait until the next empty wagon passes by.
Now depending on the size of the data, it will either entirely fill a wagon, or just partially. But for the same amount of time you spend throwing big data or small data, you're still filling in the same number of wagons which represent an amount of bulk space that you can't use anymore.
That's why, 16bit or 24bit data, the waste varies, but the used up space remains the same.
is 'VSR' mode data-compressed? if so is it lossless or lossy?
Until now I would have said uncompressed, but Randy's code suggests otherwise.
Sampling of any measure is *** always *** lossy.
It's like running sandpaper on a wooden carving.
Every time you sample you take some wood off.
If I rip the files out using the software (I've yet to sort out how to do that really) do i get nice 24bit wavs?
You might get a nice collection of 24 bit samples in WAV format.
Then you can run some math to draw a nice wave that will connect all your samples and might even look like the real wave you sampled. But there will always be tiny variations between the original wave and the synthesized wave. For example, that wooden carving above, if it was painted, the paint on the exposed places would be the first to go, and no maths designed to create shapes can create color.
a) if i use the vsr-880 to record a 24bit 44.1khz digital input, ... how do I establish concretely that the input mixer is recording bit for bit what i put in,
For digital input, you route straight to the converters, bypassing the preamps of your mixer.
For analog input, the VSR doesn't have input preamps, so you plug your source straight in one of the VSR analog inputs.
WARNING: watch your signal level, if your input is too hot, the converter clips. And no preamp to control the signal's level, so you'll have to control the level at the source.
b) if i record a 16bit digital signal into a 'VSR' mode (24bit) song how is it handled by the vsr-880?
The VSR will recognize the input is 16 bit and do what's right.
I assume that all this applies to the 890 too... anyone got any definitive answers?
Not really. For analog, the 890 has preamps you cannot bypass. You can shut off their effect but can't bypass them
|
|
Top
|
|
|
|
|
#1247047 - 09/30/13 10:02 AM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: ricket]
|
mutetourettes
Planeteer
Registered: 02/10/13
Posts: 38
|
So I used Randygo's little algorithm and added that into the VSWE code - it works very nicely! to my ear the output sounds pretty clean! So three cheers to Randygo, and to Bear for sending me stuff and saving me from reinventing the wheel...
Let me ask you this ... Exactly how do you perform a VSR -> MTP conversion ?
Thanks for all your responses ricket! In the vsr880 it's in the section on exporting to other machines - if you chose 1680 then it converts the VSR-mode data to MTP-mode.
The post processing looks like some kind signal level reduction.
Does the VSR-880 add any gain to a signal it's recording because the lack of preamps, and the hot output level pops the WAV specification ?
What are we talking here ? 4db, 6db, 12db cut ?
It's interesting isn't it? it's effectively chopping off 1/256th of the range of the signal, perhaps because shoving the bits further up the word makes the compression more easy/effective. I don't think it's about the converters particularly as it can accept/send spdif in/out too. I think the overall effect is it chops off the sampled values smaller than +/-65536 so the very quietest 256th of the signal (near zero) disappear.. I interpret this as effectively giving it a pretty big aliasing situation just in the 'very near zero' case and probably not audible... I've never quite understood translating linear PCM integer values to db so I'll leave that to someone with better maths.
It does seem (theoretically) to prevent it from outputting a full scale signal though - I'm going to have to do some recording from the spdif OUTPUT of vsr-encoded full scale data and see what's going on there..
Did anyone generate a WAV from a VSR-880 and check what it looks like or just sounds like ? Like recording on a channel a fade-in tone, hold a few seconds, and fade out ... then convert to a WAV ?
I've been futzing around with signals of 0dBfs (full scale) 440hz tones and pink noise. I'll have to do some running of the encoded stuff back out via spdif and compare actual values to see what's really gone on... my ears are impressed but I might be biased.
is 'VSR' mode data-compressed? if so is it lossless or lossy?
Until now I would have said uncompressed, but Randy's code suggests otherwise.
Since asking I've found out It's definitely compressed - in the same way as MTP is - various of the pcm samples are replaced with estimated values, then differences between the estimated and actual values are encoded, and then all are turned into mantissa/exponent values i.e. quantised to various degrees to take up fewer numbers of bits. then the differences are again compared and all the values are refined again. It's very clever stuff but some detail is still lost in this process.. The patent I linked to some posts above describes it all fully.. some of the choices around what amounts of estimation/quantisation they can get away with have been refined through listening (ie psychoacoustic) experience, but the compression is simple maths on the sample values, no FFT shenanigans..
You might get a nice collection of 24 bit samples in WAV format.
Then you can run some math to draw a nice wave that will connect all your samples and might even look like the real wave you sampled. But there will always be tiny variations between the original wave and the synthesized wave. For example, that wooden carving above, if it was painted, the paint on the exposed places would be the first to go, and no maths designed to create shapes can create color
Ah yes, that's a nice anology... but as I understand it if the filters in the analogue domain are properly removing the illegal frequencies above nyquist frequency (ie above 22050hz in this case) before the A/D conversion, then the sample values in the PCM code can be used as pointers to reconstruct the entire analogue signal, without any degradation whatsoever. That's what the nyquist frequency and filters are all about.. Still the 'tiny variations' above the nyquist frequency that are filtered out of the signal before A/D mean that you're correct, it's not quite the full original signal as bats/audiophiles would like, and the bit resolution will define a certain quantisation distortion that is overlayed onto the thus-filtered signal.. a full discussion of 'resolution' myths and truths is here - a very long read with some VERY knowledgable gear-designers weighing in... book yourself a week's holiday and dig in if you're interested: http://www.gearslutz.com/board/so-much-g...s-restored.html
For digital input, you route straight to the converters, bypassing the preamps of your mixer.
Thanks - yes I was wondering about the 'virtual' settings inside the roland, which are there even if I use spdif in - ie attenuation and input mix level etc - but I think I'm happy that 0db and 100 are decent 'straight-through' values..
I didn't know the 890 couldn't bypass the preamps, thanks for pointing that out. (the vsr880 does have two preamps for the front inputs, but they're definitely optional, and... er.. not too impressive).
So anyway - the good news is that it looks A LOT like Randygo's algorithm is right - maybe other vs890/vsr880 users knew this already and have been using reaper.dll happily with vsr projects for years... It remains to be seen whether I can package up a working version of VSWE for the rest of us VSR-folks - my test version works fine...
(clarification - I don't have a roland-friendly scsi device, so can't do backups etc, so Randygo's reaper.dll and rdac2wav.exe seemingly can't help me. I converted my VSR880 to using a SD-card as its internal drive, and VSWE (with a little persuasion) CAN read that, and grab the rdacs off there happily, and now it can convert them too)
Edited by mutetourettes (09/30/13 10:04 AM)
|
|
Top
|
|
|
|
|
#1247361 - 10/01/13 04:15 PM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: mutetourettes]
|
ricket
Planeteer
Registered: 04/19/99
Posts: 245
|
... It remains to be seen whether I can package up a working version of VSWE for the rest of us VSR-folks-
Don't forget the VS-890 users who could also benefit from your little hack.
|
|
Top
|
|
|
|
|
#1247519 - 10/02/13 08:13 AM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: ricket]
|
mutetourettes
Planeteer
Registered: 02/10/13
Posts: 38
|
Indeed I won't - hopefully Danielo will be ok with me releasing a botched about version! currently hampered because visual basic self destructed on my laptop the other night grrrrrrrrrrrrrrrrrr.. this is exactly why I want hardware studio gear!!!!
|
|
Top
|
|
|
|
#1247948 - 10/03/13 08:53 PM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: mutetourettes]
|
mutetourettes
Planeteer
Registered: 02/10/13
Posts: 38
|
ok - I've compiled it as an exe, calling it vswe 1.25 vsrversion...
so who (VSR880/VS890-owning or otherwise) wants a go? PM your email and I'll send..
then we can see what I forgot to sort out...
Would be helpful if non-vsr folk could check that it still works ok for them too...
Cheers
M
|
|
Top
|
|
|
|
|
#1248120 - 10/04/13 12:25 PM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: mutetourettes]
|
mutetourettes
Planeteer
Registered: 02/10/13
Posts: 38
|
oh - here's the thread where Randy figured out that VSR is just MTP with a 128k 'hole' in the middle..
post number 13078 in this thread...
http://www.vsplanet.com/ubbthreads/ubbthreads.php?ubb=showflat&Number=13078
his method does indeed make the files sound fine. Soon I'll have a good delve into the actual bits to see if they are exactly the same..
If anyone could suggest what would be Roland's reasons for /implications of this I'm all ears! especially how they'd hope to keep a full-scale signal accurately represented...
|
|
Top
|
|
|
|
|
#1248155 - 10/04/13 03:09 PM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: mutetourettes]
|
bear
Planeteer
Registered: 10/25/99
Posts: 6626
Loc: abq,nm,usa
|
WOW that was a trip down memory lane... I had totally forgotten I had that conversation with Randygo about all that.. so I do have some sample VSR mode files somewhere on my machine...I must - I sent him some back then
|
|
Top
|
|
|
|
|
#1248156 - 10/04/13 03:20 PM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: bear]
|
bear
Planeteer
Registered: 10/25/99
Posts: 6626
Loc: abq,nm,usa
|
You might as well send me your new version of VSWE so I can archive it. I really have no need for it now, but somewhere someone will. I will stash it with the the rest of my archive of the whole evolving project.
It would be nice to see where you changed the source also...
|
|
Top
|
|
|
|
|
#1248240 - 10/04/13 09:38 PM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: bear]
|
mutetourettes
Planeteer
Registered: 02/10/13
Posts: 38
|
On its way to you, Mister.
Thanks for your help. I'm happy to send the .exe to people too, lets have some testing..
|
|
Top
|
|
|
|
|
#1264665 - 12/12/13 02:20 AM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: mutetourettes]
|
ricket
Planeteer
Registered: 04/19/99
Posts: 245
|
I'm happy to send the .exe to people too, lets have some testing..
I'm gonna have a few days between xmas and new year to spend tweaking gear. May I kindly ask for a copy of this new improved VSWE to gve it a spin ?
Thanky-thanks!
|
|
Top
|
|
|
|
|
#1265949 - 12/17/13 05:42 PM
Re: Vsr880/890 gurus: vsr mode 24bit why uses less space?
[Re: ricket]
|
Frans
Planeteer
Registered: 08/31/02
Posts: 190
Loc: netherlands
|
Just pm'ed Mutetourettes..... Seems like a nice exe.
Frans
|
|
Top
|
|
|
|
|
|
