 
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
|
|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
In Reply to: Eric, I am not finding any signs of peak limiting in Diana Krall Look of Love posted by Christine Tham on June 19, 2004 at 03:00:41:
Hi Christine,sorry, I was away for a few days, no access to the Internet. I'll check my recording more carefully, because I have kept the file. I didn't check into individual samples, only on the pattern for the whole 45mn sample. I'll let you know if I get the same pattern when zooming to the full window size for that song.
I don't think my card does any limiting, and I didn't use any filters when normalizing. I have both normalized and un-normalized versions so I will carefully check. Another option : maybe it could come from my Sony player (it was set to audio only, coax out off, etc, as per the usual inmates recommendations for that model.
I'll check it out tonight or tomorrow (tonight is music fest :) and will let you know
Cheers
Eric
Follow Ups:
you will find that the limited peaks will now be unlimitedaccording to my data, if i set the gain to what you've set it at, the recording would clip.
I am not sure I can set the gain to 6dB less precisely, because I cannot change the gain on my card, it's either -10dB (normal) or +4dB (balanced), so I have to avoid clipping and amplify or normalize at 32 bit depth. If you are looking to a specific level I can do that, by lowering or increasing the gain to match the total RMS you have. Just let me know what level (average RMS and total RMS?) you need and I'll try to compute it.Anyway, the picture you have is correct if I look at the original ripped file. I looked briefly last night and you are right, the clipping is on my file, not in the original file, which looks more like yours so the mistake appeared either when recording with the Sony player or when I did the processsing. I don't know what happened and maybe the mistake is on my part. I'll have to redo the pictures and the statistics for track 4, hopefully it will work out.
I haven't checked the entire file for track 4, but I have the same missing frequencies now that you have in your file, so that's probably an artifact that can be reproduced. Also in the changes, I don't have the "peaks" of ultrasonic noise anymore, but I have a lot more ultrasonic noise throughout the spectrum.
I'll do the screen captures and statistics tonight and post them tomorrow if I have time. I can also send you the Excel file if you want.
Cheers
Eric
I find it strange that you cannot set gain for the analog input to your sound card. Perhaps because it is intended for studio use, where it will typically be front ended by a mixer that controls gain.if you, i would strongly recommend that you buy a mixer and not run signals directly from your player to the sound card. Some consumer equipment (including sony players) run "hot" outputs - i.e. greater than "nominal" voltage, sometimes by up to 10%. these players will surely cause your sound card to clip unless you adjust gain in a mixer.
also, i would discourage you from normalizing by RMS - you can't predict whether the signal will clip or not. you should normalize by peak digital value (preferably to -3dB FS or less) - at least for these sort of recordings anyway.
i'm not sure what you mean by "missing frequencies". are you talking about the lack of content above 35kHz in the spectral view? or are you talking about the histogram? if you are talking about the histogram - that is not a histogram of frequency distribution - it is a histogram of sample amplitude distribution.
interestingly, my recording of track 4 does *not* suffer from the "missing samples" phenomenon. maybe this is an anomaly in my sound card? i now find there is no correlation between whether there are any missing samples or not and the source of the recording (LP, DVD-A, SACD, CD, ...)
Chris,I don't have a mixer (apart from my old DJ mixer :) , I need a new amp and I'm broke, so I will have to do with a direct input for now. I could be using my preamp, which has balanced outputs, but I'm not sure that would be a better solution.
The question of the 4dB input gain was discussed on the Lynx forum several times in the past, and the opinion was that as long as you're working in 24 or 32 bit depth, the original level is not essential because you can adjust gain later on. Most SACDs are very quiet but with very high peaks (the worse I have seen so far is Michael Jackson), so I have to stay at 4dB
Re: normalizing. I am usually careful not to normalize by RMS, to avoid clipping, as you describe. Unfortunately, if I need to normalize a large number of tracks or albums (6 or 7 at a time, usually), I have to do it through the batch processing so I need to do it by approximation.
As for the results, I still cannot post them, because I need to check a few things. The "missing samples" problem I now have, but even worse than on your chart (and it doesn't go away entirely with a bessel filter), but not only that, now I also have a major peak at 35kHz, so I need to check whether this is a problem with my card, with my player, or a problem with that disc in particular. I'll have to move my card around the various PCI slots to check whether something is playing tricks in my setup.
I'll post the pix as soon as I have checked all that, probably over the weekend.
Cheers
Eric
i await you results with anticipation.if you use the +4dB setting, does that mean that on average your levels will be -14dB below nominal? if so, that is a significant loss (~2.5 bits out of 24). it's a pity because your soundcard obviously is much better than mine with a much lower noise floor (dynamic range 117dB vs my paltry 102dB).
perhaps rather than buying a mixer, which i agree is expensive, you could build your own gain control. unfortunately i am not an electronics expert (i dropped out of elec eng at uni) but i'm imagining it will be relatively simple to build using just a few parts.
i think your missing samples problem is directly correlated with the spikes you are seeing.
the 35kHz peak is probably the analog tape bias.
suggestion: have you tried recording at 96/24?
Chris,You are correct, I plan to record at 24/96 also to be able to match your charts and measurements (and I can record the entire album that way).
The Lynx card is very good (better than any digital source I have, that's for sure), but I have this problem with unstable results which are probably related to my PC (sometimes missing samples, sometimes not, sometimes worse, etc) hopefully that should be eliminated this weekend. I had done many recordings but your research has forced me to look more carefully into the technical results, I think this is good.
Building a gain control is something i could not do myself (I dropped out of anything math very early :) but I could buy one in kit from an electronics company, I have seen those in the shops, I will look into it.
You'll get the whole thing after the weekend :)
Cheers
Eric
Sounds good. I'm currently experimenting with comparing my recordings of LPs with their CD, SACD and DVD-A equivalents, and will have some interesting results to post soon too!I would strongly encourage you to build or buy a gain control, if you can.
To give you some sort of perspective on the difference it will make, look at it purely from a resolution/dynamic range perspective.
From memory, your soundcard has a dynamic range of around 117dB. That corresponds to a resolution of around 19.5 bits - very good and pretty close to state of the art.
In contrast, my soundcard has a dynamic range of 102dB - corresponding to a resolution of 17 bits (hardly better than 16 bits!).
However, if you record at -14dB FS, you will lose the extra resolution and bring your card down to a dynamic range of 103dB - around 17 bits. In other words, by not using a gain control and not setting gain optimally, you have reduced the resolution of your sound card to the same resolution as my sound card.
So I don't agree that not using a gain control and adjusting/normalizing after the recording doesn't make a difference. (that's a triple negative!) It makes all the difference in the world.
Thanks for the explanations Christine, I didn't realize that the loss in gain is the equivalent of a loss in resolution, I should have thought of that.I'll try to catch one of those gain control thingies tomorrow.
But in case I can't, do you think using my preamp would be OK?
Cheers
Eric
if your preamp offers gain control, sure, use it - it should be okay.if you do build your own gain control, make sure you use the highest quality parts. noise is a problem with any analog circuit, and if you are not careful you could easily lower the dynamic range to 80dB or worse, in which case you are better off not using a gain control.
To be save it's better to use the +4dB setting.The cd/sacd players output is far higher than the standard -10dB input setting.
However the +4dB input is a balanced input so you need to tie the inverting input (pin 3 on the XLR) to the common. (pin 1)
I use a cinch to xlr cable with the inverting input connected to ground in the cinch connector. This way any interference entering the cable is still rejected by the balanced input.
Frank
This post is made possible by the generous support of people like you and our sponsors:
Top of Page ]
[ Contact Us ]
[ Support/Wish List ]
[ Copyright © 2024, Audio Asylum, all rights reserved ]
