I'm using a Sound Blaster Live for recording in a church environment. This isn't critical listening stuff, it's mainly voice but sometimes music. Some of the results get broadcast, and some are put on CDs. Budgets are constrained, so it's preferable to make do rather than buy new stuff. I'm running this card in a P3-733 with Linux (LFS 6.1) and ALSA.

Here's my problem. When I try to set the input levels, I keep getting what appears to be analog clipping at some point less than full scale. If I attempt to raise the input level high enough to get a full-scale on the meters, all I get is clipping. I never can achieve full scale, and the recorded waveform is badly distorted, with major clipping, but the waveform isn't cleanly cut off in a horizontal line like digital clipping does, it's ragged, starting high and going downhill. If I drive it hard enough, it will spike at the top of each waveform, but I still cannot get a single sample at 100%. What this means is I have to run my levels low, because if I'm getting up into the -6 VU range or so, I start worrying about the potential of distortion, because my clipping indicator never lights to let me know if I'm too high.

Is this something which can be fixed? Is there a mixer setting I'm missing? I've heard rumors of there being some kind of input limiter on these cards, but if there is, can it be bypassed? I've never used this card under Windows, so I don't know if the mixer settings are labeled similar or not, but in alsamixer there are two controls which affect recording levels, labeled AC97 and Capture. AC97 seems to be a digital control, and I normally leave it at 100%. The Capture control I keep close to 50%. If I adjust the Capture control high enough, or raise the output of my mixer board high enough, I get this apparent analog clipping. If I then adjust the AC97 control, the clipping is not affected at all, the clipped waveform just goes higher or lower on the scale. Lowering the AC97 control to just a couple percent will give me a very tiny, but still clipped waveform.

If anybody has any information on this, I appreciate your assistance!

Creative included this nifty feature that, while anathema to any reasonable recording, allows complete boobs to avoid much clipping. There is very heavy limiting or compression, some such nonsense, built into the cards. This is very active in the upper 3dB. It is intrinsic to the cards so the only way to avoid it is to keep levels below where it functions. I've seen recommendations that vary from -3dB to -6dB. I suspect that the -6dB figure is more realistic for the best linear operation. It probably only becomes really evident in the last 3dB but it almost has to start somewhere lower.

The Windows mixer should be set at maximum and the level controlled in the analogue domain.

This is a well known issue with the Ensoniq 137x chipset, which is what's used on a variety of Creative cards, including the SB Live. It's some kind of saturation or overload in the analogue input circuitry.

The bad news is that there is no way around it as far as I've discovered. In other words, you'll never get recordings peaking up close to 0dB.

The good news is that the saturation/overload seems to happen in a part of the circuitry *after* the mixer level setting. The exact level where the clipping starts varies from one soundcard to the next, but I've never heard of one that clips beneath about -2dB. This means that if you back off your recording levels to peak no higher than about -3dB, you'll get clean recordings (which can be normalised in software later if you wish). Of course, losing the top 3dB means you're throwing away about half a bit of resolution, but that's a great deal better than the awful distortion you get by trying to push the levels higher.

EDIT: Andy beat me to it. Although I would reiterate that in my experience the problem is limited to about the top 2dB, nowhere near as widely as Andy speculates.

I'd have to disagree with you on the chipset. I'm familiar with the Ensoniq 1371, I have one of those cards, called a SB16 PCI... I used a SB16 ISA card until I upgraded the computer and no longer had a free ISA slot for it, so had to upgrade. Or thought I was upgrading! Back then I naïvely assumed the newer card would be better... No matter where I set it, it just sounded "dirty". There's some major distortion going on somewhere in the signal chain, and I never could figure out a way around it. But the SBLive is using the EMU10k1 chipset, which seems to work pretty well. I have no problems with the way the digital part of the card works, except for the resampling part which doesn't affect me much as I only use analog I/O. It's the analog input circuitry that messes me up. Leave it to Creative to take a halfway decent chipset and put a cruddy input on it, with no way to bypass it! I wish they'd just give me a wire to the ADC!



Even if it's only the top 2 db, I still have to stay well clear of it to be sure I don't get distortion, so practically speaking, I'm losing close to a full bit. Which is only 1/16 of the total, but audio perfectionist that I am at times, I hate to lose that! The real issue though is the difficulty in metering to be sure I don't overload. It's mainly an inconvenience, but it's a really irritating one because it prevents my overload indicator from working, and I feel like I'm back in the analog tape days, watching the meters to see where they end up, and hoping it's not peaking somewhere in between meter updates. I guess if there's no way around it, I'll have to see if I can do an upgrade.

I thank you both for your help! Anybody else know if there's a way to disable this? And if not, does anybody have some good recommendations on a not too expensive (like to keep it below $200 US, but you know how that goes...) audio interface with good sounding analog inputs, preferably balanced (TRS or XLR) +4 db inputs? I like the idea of getting the analog electonics outside the computer, but don't know if you can do it in my price range.

All I can say is that the early SB Lives did use the 137x chipset, and suffered from the saturation/overload problem you've got. If later SB Lives use the EMU chipset, and they *still* have this problem, then it would appear that Creative are doing this deliberately for reasons that elude me.


It sounds like a lot when you say you're losing "1/16 of the total", but if you think about it as a 6dB loss of dynamic range, then it's not catastrophic. It all depends on what standards you're aiming for.


Does your recording software have any facility to configure what it regards as an overload? Perhaps you could do some experiments to find out exactly where the onset of distortion is, then configure the s/w to fire the overload indicator at that level.

I'm trying to find a way of saying this kindly. I fear I'm going to fail, so don't take this the wrong way Corax...

You're going about this the wrong way! You shouldn't be aiming to get anywhere near digital full scale when making a live recording from microphones into a digital recording device.

Real professionals pick a level, typically -12dB, -18dB or -20dB relative to digital full scale to hit on average. This leaves plenty of headroom to ensure that the signal never clips. This is much much more important that losing 1-bit of resolution. Of course professionals are using 24-bit recording, but unless your soundcard is very noisy (and your input signal is very clean - unlikely I'd suggest), then even you can happily throw a bit or two away and no one will ever notice.

CliveB will probably give better advice than me, but if the signal is peaking at -6dB, that's fine. That's where I aim for when recording from old records. In a live environment, I would be even more careful, and keep levels lower into the sound card. I wouldn't worry about noise from the sound card - I would concentrate more on the noise from cheap microphones, pre-amps/mixers etc. I would concentrate even more on microphone placement, and maybe a little compression. The 1/16th which you are worrying about is irrelevant.

Are you using a mixing desk? With a little care, you should be able to "calibrate" (in the loosest sense of the word) this to the PC input, and then just look at the meters on the desk to monitor levels.

You might choose to dynamically compress the results a little in software. You will almost certainly normalise the results in software.

Hope this helps.

Cheers,
David.

P.S. As for recording levels: I let someone else record a wedding onto MD once. I transferred this digitally into my PC. There was a section where the person on the mixing desk had reduced the levels sent to the MD by accident. It peaked at -48dB. Strictly speaking, that's an 8-bit recording! However, I was able to amplify it, de-noise it, and I was totally amazed by the result. The final result (from a hand-held wired mic) were still better than an unprocessed recording from a decent radio mic. The mixing desk had a good quality output, the MD had a very quiet input, and I was lucky!

As far as replacing the SB, there are many choices that are better. The M-Audio Audiophile 2496 was reduced from $250 to $99 when newer hardware came on line, and I've seen some sales at $69. It is much better than SB although its input is -10dBV RCA. The Echo Mia has balanced TRS at +4dBu for $200 or so, but the difference in recordings from the Audiophile will not be easy to notice. The Mia is more sensitive to motherboard chipsets than most cards, so check Echo's recommendations against your computer before buying.

I find the internal computer noise will my PCI Audiophile tolerable at -98dB on noise floor recordings. I suspect you will have to pay significantly more for an external interface that does better.

Thanks guys for your help! 2Bdecided, you did very well... I probably am being a bit too picky, but I guess it's more of a control thing than anything else. I feel like the card just isn't working right, and I'm accustomed to being able to control the whole signal chain from -96 to clipping. I don't like having part of my system (clip indicators) broken by my hardware, and I can't set my software (Audacity) to compensate. It just doesn't quite "fit" the way I like. But I may have no choice for now.

Anyway, just to give you a better picture, I'm using Shure SM58 and Audio Technica Unipoint microphones most of the time, sometimes a pair of Behringer B2 Pro mics for special occasions, a Mackie SR24-4 mixer board, and everything runs through an Alesis NanoCompressor before going to the computer. I have the compressor set to keep everything fairly stable on the levels, so I can put it in the -6 db range and it stays quite close to that most of the time. The program material I record doesn't have a lot of dynamics to it anyway, unless the preacher gets too close to the mic when saying "Let's P(op)ray!"

AndyH-ha, thanks for the recommendations. One of the reasons I'd like to get a better card is because consumer-grade connections are almost unworkable in a professional environment, which is close to our situation. Things get used too much, wires get moved around and wiggled, signal levels are too hot (literally!) and those little 1/8 inch soundcard connectors get scratchy, and so do RCA jacks. I hate them with a passion sometimes! Nothing bothers me more than to look up and see my meters running low, check all my level settings, then have to start wiggling sound card connectors to fix it... especially if I'm recording something important! So RCAs are out, and so are mini-jacks. 1/4 inch jacks are okay, and XLRs are tops. I guess Echo finally released driver info for their cards, so us Linux users can use their cards now. The bad publicity must have gotten to them... I should check them out. I'm sure there's something we can afford eventually! At least it's not an emergency, though I'm keeping my fingers crossed!

Not true for the line input level when you have got an SB Live. With mine, the digital full scale of the ADC is mapped around -1 dB near the middle position of the slider, and it is clipped if the slider is set higher.

Pio how did you do this kind of measurement?