24 bit padded to 32 bit decoding seems to do weird things to the sound

[Patrickhat2001]

Ever since I got my EMU 1212m a week or so ago I've been very pleased with the sound except something seemed wrong with the bass--the sound of the card was very detailed all throughout the frequency spectrum except for the bass which seem a little blurry in comparsion. I was a little dismayed by this since I originally thought it was the fault of the card but a few days ago I realized that it was the 24 bit padded to 32 bit decoding that was causing the problem. Upon switching the decoding back to 24 bit fixed-point (this is all in Foobar, by the way) the bass cleared up and is now as clear and detailed as the rest of the sound spectrum. Also switching back from regular 24 bit has gotten rid of some of the articficial brightness that I was noticing in the midrange, especially with string instruments.

I just thought I'd share this since I've read some other people complaining about the bass of the 1212m being blurry. Perhaps this has been the problem? I never noticed blurry bass using 24 bit padded to 32 bit with my other soundcards before but upon going back and doing some comparisions with between 24bit and 24bit padded on the Terratec EWX 2496 I noticed that, again, the clarity in the bass improved (although it was much more slight) when using just regular 24 bit. I've also found that I like the way 32 bit decoding sounds with the 1212m (something which never really seemed to sound good with my other soundcards--they just end up sounding overly bright). I know that technically the 1212 (alone with all other modern soundcards) is only rated for 24 bit decoding but for whatever reason 32 bit decoding does work with it (alone with all of the other soundcards as well). 32 bit decoding sounds much more detailed than 24 bit decoding although it can be too thin and really doesn't allow the sound to bloom as much as it should naturally. Still, if you want to know exactly what chords your favorite guitarists are playing I recommend giving 32 bit decoding a shot.

I guess the real point of this post is just don't follow the recommendations of others blindly; experiment and find what's right for you. For the record I've also given up on upsampling with Foobar since I found the sound more with upsampling disabled. To each their own I guess. Computer audio can be maddening sometimes since their are so many settings to play around with.

 

[Publius]

If true, this would most definitely be a major driver problem and should show up in loopback tests - it's all 24-bit once it gets to the DAC. Could you post a few loopback recordings to verify this?

 

[Patrickhat2001]

Yeah, I know, but 32 bit decoding works just fine with my 1212 and with all of the other soundcards I've tried (which are pretty much all the major PCI ones, check my signature) and definately gives a more detailed sound than strait 24bit or 24bit padded decoding. Strange thing, too, when I select my 1212m within the Kernal Streaming plug-in for Foobar it states that the optimal resolution is 32bit while, if I select the Terratec it says the optimal resolution is 24 bit. Any ideas anyone? All I know is 32bit decoding sounds pretty good with my 1212m (although still imperfect) as opposed to all the other cards I've tried it with.

I must admit I'm pretty ignorant as to how to perform feedback recordings.

 

[Asmo]

Are you using ASIO ?

 

[Scrith]

With ASIO I have to pad to 32-bit. I've tried 16-bit padded to 32-bit and 24-bit padded to 32-bit and haven't found a significant difference (yet). Since my recordings are lossless and should be fine in their original 16-bit format, I'm using 16-bit padded to 32-bit.

 

[Patrickhat2001]

Actually, as of late I've been using kernal streaming. ASIO will only permit variations on 32bit decoding (fixed or padded up to that amount) but with kernal streaming I can use anything I want.

 

[Imyourzero]

When I set up my AV-710, I used the guide in Mr. Radar's sig and set the output to 24 bits padded to 32. What I want to know is WHY that setting is recommended. When you choose it, Foobar tells you that it will not improve the sound quality, so why not just choose regular 24 bit with no padding?

In your case (with the 1212m), you found that 24 padded to 32 actually decreased the sound quality. I wonder if it's doing the same thing with my AV-710. It currently sounds fine, but a lot of things sound OK until you hear something better...then you realize what used to be "fine" is no longer fine and is inferior.

 

[Asmo]

Quote:

Originally Posted by Patrickhat2001 Actually, as of late I've been using kernal streaming. ASIO will only permit variations on 32bit decoding (fixed or padded up to that amount) but with kernal streaming I can use anything I want.

Did you compare Kernel Streaming to ASIO ?

 

[Patrickhat2001]

Funny you should mention that. Actually I have and they do sound a little different to my ears. Right now I'm prefering kernal streaming to ASIO (while holding the decoding rate constant, of course). Both KS and ASIO clean up the sound relative to using strait waveout but to my ears ASIO just seems to clean things up too much leaving things sounding overly sterale and too transparent.

Actually I've found that using strait waveout really isn't all that bad (at least on the 1212m anyway). Sacrilege you say? Let me explain first before you start throwing rocks. People usually use kernal streaming or ASIO to bypass kmixer in Windows because kmixer degrades the sound--by adding distortion is the typical explanation given. Well, what else adds distortion to sound? Tube amplifiers, and some people really like the sound signature that tubes give. Now, true, using strait waveout does make the sound overly dense and lacking a little in detail but it has a sense of groove that just can't be denied (at least by me anyway). Also in some ways certain instruments actually sound more detailed using waveout. How in the heck is that possible? Well the added densness of the sound makes them stand out a bit more--I find this to be the case with many instruments such as acoustic guitars, which, while sounding great and increditibly transparent while using ASIO much of the detail is hard to decern because the sound is just so transparent.

On the other side of the spectum from waveout is using ASIO to output the sound. Using ASIO makes the sound very detailed and transparent, but to my ears it can sound a bit too transparent and sterile. Kernal streaming strikes a compromise between these two providing a sound which is about in the midway between them in detail and transparency. Overall it's a compromise I like (well, really because of the compromise in the density of the sound, I always appreciate more detail). I half wonder if this is so because (at least with the EMU cards) ASIO bypasses kmixer completely where perhaps kernal streaming only bypasses some parts of it. I don't know.

You gotta love how flexible computer sound is, though. There are oh so many options to play around with.

 

[Asmo]

Quote:

Originally Posted by Patrickhat2001 Funny you should mention that. Actually I have and they do sound a little different to my ears. Right now I'm prefering kernal streaming to ASIO (while holding the decoding rate constant, of course). Both KS and ASIO clean up the sound relative to using strait waveout but to my ears ASIO just seems to clean things up too much leaving things sounding overly sterale and transparent. Actually I've found that using strait waveout really isn't all that bad (at least on the 1212m anyway). Sacrilege you say? Let me explain first before you start throwing rocks. People usually use kernal streaming or ASIO to bypass kmixer in Windows because kmixer degrades the sound--by adding distortion is the typical explanation given. Well, what else adds distortion to sound? Tube amplifiers, and some people really like the sound signature that tubes give. Now, true, using strait waveout does make the sound overly dense and lacking a little in detail but it has a sense of groove that just can't be denied (at least by me anyway). Also in some ways certain instruments actually sound more detailed using waveout. How in the heck is that possible? Well the added densness of the sound makes them stand out a bit more--I find this to be the case with many instruments such as acoustic guitars, which, while sounding great and increditibly transparent while using ASIO much of the detail is hard to decern because the sound is just so transparent. On the other side of the spectum from waveout is using ASIO to output the sound. Using ASIO makes the sound very detailed and transparent, but to my ears it can sound a bit too transparent and sterile. Kernal streaming strikes a compromise between these two providing a sound which is about in the midway between them in detail and transparency. Overall it's a compromise I like (well, really because of the compromise in the density of the sound, I always appreciate more detail). I half wonder if this is so because (at least with the EMU cards) ASIO bypasses kmixer completely where perhaps kernal streaming only bypasses some parts of it. I don't know. You gotta love how flexible computer sound is, though. There are oh so many options to play around with.

Interesting read, I did a short comparisson when I first got the card and prefered ASIO, maybe I'll play a bit more.

Did you compare different rates btw? 44.1 vs 192 or 96?

 

[Patrickhat2001]

A little but it's hard because my computer is barely fast enough to support 96 Khz with slow mode enabled and besides that it's a pain to change sample rates with the EMU cards. During all of my posts above I had resampler disabled in Foobar. I do plan to play around with it more, though. With my other soundcards I found that I didn't like what upsampling did to the sound (made things less harsh and a little larger sounding but also made the sound stuffy and lacking in attack at the same time) so I've had it disabled for the past few months. However with my other soundcards I found the max rate I could upsample at was 88.2 KHz (with slow mode enabled) but, for some reason, I can actually hit 96 Khz with the EMU with slow mode and all and have things still run. Go figure. I'll play around with it some more. I just wish that you could upsample at 88.2 Khz (and at 176.4 Khz, for that matter) with the EMU cards but, unfortunately the EMU patchmix software doesn't seem to allow it.

 

[Asmo]

Quote:

Originally Posted by Patrickhat2001 A little but it's hard because my computer is barely fast enough to support 96 Khz with slow mode enabled and besides that it's a pain to change sample rates with the EMU cards. During all of my posts above I had resampler disabled in Foobar. I do plan to play around with it more, though. With my other soundcards I found that I didn't like what upsampling did to the sound (made things less harsh and a little larger sounding but also made the sound stuffy and lacking in attack at the same time) so I've had it disabled for the past few months. However with my other soundcards I found the max rate I could upsample at was 88.2 KHz (with slow mode enabled) but, for some reason, I can actually hit 96 Khz with the EMU with slow mode and all and have things still run. Go figure. I'll play around with it some more. I just wish that you could upsample at 88.2 Khz (and at 176.4 Khz, for that matter) with the EMU cards but, unfortunately the EMU patchmix software doesn't seem to allow it.

I prefer 44.1khz on my 1212m also.

Also, you don't need to enable slow mode, that dosn't do anything, except hog cpu resources

 

[Patrickhat2001]

In my experience enabling slow mode always seems to make things a little clearer relative to upsampling without it enabled. But just disabling upsampling entirely results in the best clarity which is why I prefer it disabled. Too each their own.

 

[lan]

Yes I also prefer 44.1 as I find upsampling with these programs reduces the quality as it makes it blurrier for me. I use 32bit output also.

It's great that there's so much flexibillity with a computer.

 

[Scrith]

I just tried playing around a bit with Kernel Streaming (16bit and 16bit padded to 32) vs. ASIO output (16bit padded to 32bit) and couldn't really find a difference between the two.

One thing I noticed, however, is that Kernel Streaming causes the output to go to the Wave L/R strip in PatchMix. The problem is, my Windows sounds (like the beep you get when adjusting the volume in the sound control panel) are also coming out through the Wave L/R strip. Which means somewhere the two audio signals are getting mixed before they get to E-mu's Patchmix, probably somewhere in the Windows audio code. Right? And, given that E-mu's software is designed from the ground up for high quality audio, whereas Windows is not, I think it might make more sense to have a pure ASIO signal coming into the E-mu software to perform the mixing, rather than leaving it up to Windows.

Does anybody have any thoughts on this?