A/D input levels with 24-bit - need to max out?


I use a USB Pre A/D converter with a Grado PH1 phono stage to record vinyl records in 24-bit 44.1Hz stereo. I'm using Bias Peak to record. I use Cool Edit and ClickFix to process. Back in Peak again, I raise the gain to the highest unclipped level before dithering to 16 bit to make CD's or FLAC archives.

I record a lot of records these days, so to avoid re-recording due to clipped levels on the USB Pre, I set the levels conservatively. As a result, the peak for some records might be 80% or so instead of 95% more. I do try to record loud records in a set, etc., and adjust the level accordingly. But the Pre's level knobs are tricky to set evenly by eye (wide-spaced markers) so I may adjust downward more than I would otherwise.

My rationale has been that raising the input level doesn't help much, since I'd also be raising the noise level of the phono stage and any noise generated by the analog portion of the converter and the signal to noise ratio would not change.

Now, I may be answering my own question here, but if I understand digital audio correctly, more bits are given to louder passages, so by keeping the level lower, I'm getting fewer bits for the quiet parts than I should be getting.

But, since I'm recording at 24-bit anyway, before editing and then dithering (to 16 bits) after raising the gain, does it really matter that much? The 16-bit result has the highest level sound, it's just the 24-bit initial recording that is lower.

Thanks in advance for reading this and any advice you might have.

Don't worry. There is NO difference on the digital side. Your 24 bits are way more than enough to capture all the dynamic range from your source. It does not matter if you digitize at lower levels - it changes nothing in the accuracy of what you record (digits in the bottom half of the bit stream or the top half are all just as accurate). What you lose if anything is a slight bit of "potential" dynamic range, however, I honestly doubt you can produce anything from your source that has the full dynamic range of 24 bits....so just forget about it!

Far more harm could come from inadvertent clipping by having the signal too high....so you are wise to stay away from the upper limit.

Remember Digital is NOT analog. In analog media everyone tried to stay close to the peak levels. This was because analog media has a very narrow or limited dynamic range with a very high noise floor....so it made sense to record peaks close to the maximum of the tape media in order to eek out the most dynamic range from the recording. In digital the rules change....in fact clipping in anlaog media was quite benign and sounded nice....in Digital clipping counds terrrible....so you really need to stay away from the limits in order to ensure no clipping and to preserve good sound.
You may not need to use all 24 bits, but too low an analog signal into an A/D converted can be a problem. You could be using only the lowest 10 bits of your 24 bit hardware, and that would certainly degrade a 16 bit source.

The performance of my Behringer DEQ2496 equalizers was definitely improved when I adjusted analog gains to as to fully drive the 24 bit hardware in them. (Gain before the DEQ2496 was increased, and gain after the DEQ2496 was reduced, so that the overall gain was the same, but the signal was stronger as it went through the DEQ2496).
You could be using only the lowest 10 bits of your 24 bit hardware, and that would certainly degrade a 16 bit source.

It would only degrade if the signal has more dynamic range than can be properly represented by 10 bits. Of if by playing the signal louder from the source it is being rasied further above the noise floor. To me this risk of clipping (flat topping) by trying to position the peaks in the music at too close to the maximum signal of the ADC is a much greater risk.

It is not the dynamic range you need to worry about. Even the best and the cleanest vinyl only has about 45 db of dynamic range, well within the 96 db red book capability. But if you record at too low a level, you will start losing resolution. You loss 1 bit of resolution for every 6 db down in level. So for optimum recording, you should adjust the input level such that the loudest passage reaches 0 db or just below it. If that is too difficult to achieve than make sure the loudest passage is within -6 db range so that you won't loss more than 1 bit of resolution.
Shadorne...As mentioned in another thread I have found that EVERY CD, at some peak signal point, comes within 2 or 3 dB of maximum, but never more (ie: no digital clipping). I'm told that this is no accident, and that when CDs are mastered the signal level is adjusted to achieve this end. So every CD uses all 16 bits, and represenring the signal using fewer bits, eg 19, results in loss of resolution. In other words, the least significant bit would represent more analog voltage at your speaker.
Obviously I meant 10 bits, not 19. 9 and 0 are side by side on the keyboard.
Thanks, a variety of opinions here, basically all over the map, I think.

On one hand, with 24-bit, I can afford to not worry about maxing out the gain, but on the other hand I could lose a a lot.

Just to underline, I'm recording at 24-bit. I know that at 16 I'd be crazy not to try to get the highest gain, but don't I have enough bits at 24 to get by? If I could record at 96 would that be even better?
I suggest you buy this Digital Audio Explained book.

It explains it all clearly. It will cost you very little and it will put your mind at rest. An ordinary 16 bit CD has 96 db of dynamic range. 24 bits give you 144 db dynamic range (this is really only needed by studios, as they mess with signal levels on each track and it gives them a bigger sand box to play in).

Since a turntable will give you at most 60 db dynamic range (well known fact) then you have a lot of room to play with even with 16 bits: 96 - 60 db = 36 db of extra dynamic range, which is about 6 Bits that really are never needed to faithfuly reproduce the entire output of a very high quality LP/TT rig.

Add to this the fact that recordings with such extreme dynamic range as 60 db are extremely rare to almost non existant. Try Shefield labs drum track and you may get close to 60 db. I have the CD and I can hear the print through of the kick drum from the analog tape very clearly. It sounds like tiny clicks which echo the real recording only a second or so later and it stems from the way the tape is stored on a reel. (analog tape has limted dynamic range of about 60 db too)

I really think you are worrying too much.
Shadorne...It's not just about dynamic range. The size of the digital steps (quantization) is the LSB (Least Significant Bit). Even if a program is loud all the time (little dynamic range) the accuracy with which the digital data represents the analog waveform depends on the quantization.
Ok, folks, while some of you feel I don't need to worry, given 24-bits, with maximizing the gain, the rest of you maintain that it still matters.

I'm sure it does from a purist's point of view. If I was recording one or two records, I'd do it as many times as necessary to get close to max without clipping, but I have a tone to do, so I can't afford to be too picky. ;-)

That was the point of this query. I want to verify that if I'm down a few db, 24 bit will make up for it a little.

I know that when, after cleaning clicks, etc., I raise the gain (using Bias Peaks anti-clipping tool), I'll just be padding with zeros.

What is your opinion about what happens with I dither down to 16?
What is your opinion about what happens with I dither down to 16?

In theory this will increase the noise floor but I expect your data will already have such a high noise floor due to the analog source (as I eplained above there us no way it exploits 24 bits of dynamic range or 144 DB). So in practice the difference between your 24 bit and 16 bit data should be totally inaudible. 16 bit 44.1 KHz was well chosen 30 years ago - it is perhaps one of the reasons SACD has struggled.