Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[greg charles]

I have not used them in the same room at the same time. But I did use them on the same project. When I used the L-2 hardware for the first time I initially thought I couldn’t recognize any sonic differences.  I was really expecting the L-2 hardware to be "better" in some way but I couldn’t honestly tell myself that.

Yes using both in the same room at the same time would help.

The only possible difference I thought was maybe you could push the L-2 hardware a bit further before the nasty....

Anyone done any tests to find differences and similarities?

[lagerfeldt]

Basically the same in terms of sound to my ears.

[Bob Boyd]

Before I sold my hardware L2 years ago, I confirmed with Waves that the software version was the same algorithm.

I didn't notice a difference when I switched.

[TotalSonic]

The algorithm in both is exactly the same.  The hardware and TDM versions use 48bit fixed point math - the native one uses 32bit floating point.

The hardware one has the ability to be set for dual mono instead of linked stereo though - and many people report for some material this gives less artifacts.  It's extremely easy to set this same thing up in a DAW though by using two instances of the L2 (one processing left channel only - the other processing right channel only).

Best regards,
Steve Berson

[Ronny]

Where did you hear the native ones use 32 float, Steve? Even the software L1 uses 48 bit double precision. The L1 and L2 on the Waves y56k and y96k cards also use 48 bit.

[Jerry Tubb]

TotalSonic wrote on Wed, 25 January 2006 12:41

The hardware one has the ability to be set for dual mono instead of linked stereo though - and many people report for some material this gives less artifacts.

We run our hardware L2s Un-Linked (dual mono), thought it sounded mo' better.

Also run our Manley Vari Mus Un-Linked... more open sounding.

We also have the software versions L1, 2 & 3... but I like knobs better.

peace

[TotalSonic]

Ronny wrote on Wed, 25 January 2006 18:46

Where did you hear the native ones use 32 float, Steve? Even the software L1 uses 48 bit double precision. The L1 and L2 on the Waves y56k and y96k cards also use 48 bit.

Just looked at the manual and looks like you're right Ronny.
Some hosts will change the 24bit integer returned figure to 32bit floating point math though before passing it on to the next process in the chain.  

Best regards,
Steve Berson

[bobkatz]

Ronny wrote on Wed, 25 January 2006 13:46

Where did you hear the native ones use 32 float, Steve? Even the software L1 uses 48 bit double precision.

This is incorrect, Ronny. You can't compute fixed point in a floating point processor. All native plugins use floating point at this point in time.

Quote:

The L1 and L2 on the Waves y56k and y96k cards also use 48 bit.

Now this may be true, and the reason is those cards use the Motorola chips.

But every other software plugin in the Waves line that is Native (and not TDM) uses floating point internally. Waves' literature sometimes gets this mixed up as you will find a manual sometimes that refers to double precision, but this needs to be clarified as it mixes up the "facts" with the "fiction."

BK

[Ronny]

bobkatz wrote on Wed, 25 January 2006 22:26

Ronny wrote on Wed, 25 January 2006 13:46 Where did you hear the native ones use 32 float, Steve? Even the software L1 uses 48 bit double precision. This is incorrect, Ronny. You can't compute fixed point in a floating point processor. All native plugins use floating point at this point in time. Quote: The L1 and L2 on the Waves y56k and y96k cards also use 48 bit. Now this may be true, and the reason is those cards use the Motorola chips. But every other software plugin in the Waves line that is Native (and not TDM) uses floating point internally. Waves' literature sometimes gets this mixed up as you will find a manual sometimes that refers to double precision, but this needs to be clarified as it mixes up the "facts" with the "fiction." BK

Understood Bob, I have the L2 hardware manual and it says 48 double precision. I also have the L1 software plug, so I checked that pdf manual and it said 48 bit double precision. I was under the assumption that even processing in 32 bit, the plug in processes at 48 and than returns to 32. That's what the y56k and y96k do, they receive 32 bit from the consoles internal buss, process at 48 and return to 32 bit. The L1 on the y56k does this, it has no IDR and is designed for single multi track processing. The L1+ OTOH, has IDR and is recommended to use on the output buss. The L2 has IDR and is also meant to be used as a last process on the output buss. The digital outputs are all 24 bit, the processor works at 48 bit and than the IDR dithers the quantization to 24 for the AES, Tdif, ADAT or SPdif 24 bit outputs. I haven't seen a Waves manual yet that says the software plugs work at 32 bit, so naturally I was confused as to why I hear people saying that the plugs (non-TDM) process at 32 as the manuals all say 48 double precision.  

[Bob Boyd]

I think it simply boils down to the fact that the Motorola DSP 56k series processors (TDM cards, Waves hardware & cards) are fixed point processors and the Mac & PC processors are floating point.

Therefore for the software to run on the 56k series DSP, it has to operate at 48 bit fixed and for the "native" internal processing, it runs 32 bit float.

I think 48 bit "double precision" only applies to the fixed point processing.  

Originally, L1 (and Q10 and the like) was a 24 bit plug.  That's what it was running back when I had a digidesign AudioMedia II card which was 16 bit in, 16 bit output, 24 bit internal.  Waves updated all of their software later.

[Bob Olhsson]

The double precision native plugs have always been 64+ bits float. L1 also has the "analog" setting that prevents intersample peaking. It's nice seeing that some others have finally picked up on that issue, what is it now, ten years later?

[cerberus]

we know that a 16 bit system can be emulated by a 32 bit system. perhaps the same algorithm can run in native by splitting the 48 bit word into 2 24 bit words... this is a strategy i've heard of. i think that some systems exist that work like this.

when i requested to waves that their limiters should output to floating point so I could use the limiters more flexibly in mixing as opposed to only for mastering, they said that the design did not make that option practical; that these particular limiters differ from all the other waves plug-ins:  the maximum bit depth of their output in native is 24 fixed, I suppose that they don't just drop our exponent in float to get to 24 fixed?  Otherwise they could easily not do this and make an output of 32 float which a lot of us could prefer on drum busses in mixing etc...but apparently this is not practical if the internal math was converted to fixed point, which I assume is why the output must be fixed point as well.  Certainly 56 bit processing is happening on an HD Accel system but L2 is the same algorithm on Mix+ which is a 24 bit system and technically cannot do higher than 24 bit math.

also the summing buss registers in my clean sounding DAW mixer are more than 32 bit float.. this is certainly mentioned in many companies' literature who make mixers that run on 32 float native CPU, so these kinds of calculations must be able to take place.. probably by splitting the word up for processing, i'd presume.. such as: 1200*2=((1.0*10^3)*2)+(2*(1*(10^2)))       well.. I am very poor at math, but it's obvious to me how it could be "possible" in this world to perform high precision calculations on lower order DSP and still be able to deliver the results in "real time".

jeff dinces

[Release]

What do you guys think of the UAD-1 Precision limiter in comparison with the L2?

I once did a simple comparison between the software L2 and the Precision Limiter. I thought the PL sounded much better (ie more open, more depth in the sound).

I always thought the hardware L2 was a completely different story, thought it was much better than any software limiter. Did anyone compare the PL with the hardware L2?

[bobkatz]

cerberus wrote on Thu, 26 January 2006 01:14

when i requested to waves that their limiters should output to floating point so I could use the limiters more flexibly in mixing as opposed to only for mastering, they said that the design did not make that option practical; that these particular limiters differ from all the other waves plug-ins:  the maximum bit depth of their output in native is 24 fixed

That has to do with legacy code and perhaps the fact that they are using C++ to code both the fixed and floating point versions. But it is perfectly possible for them to remove the dithering from the processor and pass on floating point to the next piece in line.

In fact, when I capture the output of my Waves IR-1 to a file on my SADiE I always make a 32 bit floating point file. I have to examine that file on my Wavelab's bitscope to see if it's got the extra bits, at least in the low level decay. Unfortunately, SADiE can read other party's floating point formats but it can only write to its own proprietary floating point format, so I may not be able to analyze this file.

Perhaps someone with Wavelab and the IR-1 can look at its output on the Wavelab bitscope and let us know.  But frankly, as long as the output is within the 24 bit (144 dB) dynamic range, it is academic whether or not it's in 32 bit float format. We may not see anything on the output of the Plugin that indicates it's got the exponent operating. It seems to me the so-called advantages of floating point only become meaningful if the Exponent bit is active, and that won't happen unless you "overload" or "underload". And you won't hear reverb decay that's below about -60 (ok we'll even say "90" if you want to stretch it) dBFS in most practical situations, so it has a long way to go before that exponent bit has any need.

BK

[Bob Olhsson]

cerberus wrote on Thu, 26 January 2006 00:14

when i requested to waves that their limiters should output to floating point so I could use the limiters more flexibly in mixing as opposed to only for mastering, they said that the design did not make that option practical; that these particular limiters differ from all the other waves plug-ins:  the maximum bit depth of their output in native is 24 fixed,...

I think my test showed the plain vanilla no dither control "native" L1 does have a 32 bit output. My test was to put in a 16 bit signal, change the gain, change it back and then count the number of bits in the result.

Bob, as far as I understand from my old buddies at Stanford, Motorola 56k DSP routines can't be written in high level languages. Certainly that situation may have changed.