96KHZ at 44.1khz

[SkipB]

I know this topic is somewhat confusing but I will explain what I mean. I'm finding it beneficial to record at 44.1khz with your A/D converter outputting 96khz. It took me some experimenting to get this setup with all my outboard gear,and the good part is I'm not creating any cumbersome files loading up my hard drive. I've done some extensive listening tests by first going 44.1 to 44.1 then 96khz to 44.1 and the difference is very noticeable. Ok first off I'm using a Lynx L22 running out Analog to an Analog Parametric EQ and Multiband Compressor then into a Mackie Big Knob,input. Then output from the Big Knob to the input of a 9624 Lucid A/D. Then digital out back to the Lynx Card. The A/D is clocked with a Lucid Genx6 96 and the lynx card uses its own internal clock. Ok from the Lynx digital out I'm inputting a Lucid DAC 9624 then analog out into another input of the Big Knob then Monitor output to Monitor speakers. I'm using WDM Drivers to actually be able to Play the song back through all this at 44.1 then once it reaches the A/D it upsamples to 96khz but my recorder is set to record at 44.1khz 32bit Float. I have to say that I was a bit surprised at how good the song sounded using this method. Is anyone here doing anything similar? I also have the Lucid SRC 9624 but it looks as though its gonna sit in the corner for now. NE way cheers to all and Happy Mastering. Regards Skip B.

[HarBal]

Hello SkipB,

Before taking this any further I'd suggest that you should re-do your experiment using 88.2kHz rather than 96kHz. Why? Because to do sampling rate conversion between 96kHz and 44.1kHz without introducing either frequency response anomalies or aliasing is very difficult. Many of the commonly available rate converters have poor measured performance doing that conversion. The reason why it's difficult is because the smallest whole number ratio that the conversion can be factored to is very large, which incidentally, was a deliberate ploy by the makers of the DAT format (48k to 44.1k conversion) to reduce piracy. 44.1k to 88.2k is simple because it's just a factor of two.

If after this experiment the results are still the same then I guess you may have more investigating to do to figure out why this might be. On the other hand, if the difference is not as dramatic than with the 44.1k / 96k test then I'd argue that most of what you are hearing is probably due to the limitations of the sampling rate converter, whether that is frequency response or aliasing or both.

Personally, I'd say the biggest problem with doing tests like this is knowing which source is which. You need a blind test where you don't know which source came from which process. Maybe you should get someone to rename the files for you and randomise the draw so you don't know what your playing. If you do know what you're playing then you'll likely be biased towards 96kHz because that must be better (higher rate must be better)! The same issues plague Hi Fi magazines because naturally the gear that cost 10 times as much must sound better.

I'd be reticent to draw the conclusion that processing at higher rates is superior because from a mathematical standpoint, such a proposition has no basis. Mathematically it is clearly inferior, but the real problem is what is added by the 96kHz processing may be audibly pleasing even though it may be due to distortion. So if you like the sound of what it adds, I have no issue with that standpoint, but if someone starts claiming that it is purer and more accurate then I'd be reticent to agree. People like adding distortions to sound. At the risk of offending many Hi-Fi purists, the essential reason tube amps continue to be made is to add pleasing distortion. That's fine by me, but I just cringe when people start claiming that it is because tubes are somehow technically superior. Measurement cannot demonstrate that assertion so it invariably ends up as an article of faith. I'm not a believer in that faith.

Different is different, not necessarily better.

Cheers,


Paavo.

[SkipB]

Hi Paavo, I understand your point,and normally when I did SRC it would be 88.2 instead of 96k. But is the math the same when you go through a device upsample but actually record at 44.1. The file is being played at 44.1 but passing through the A/D converter it takes a step up to 96khz and I'm capturing the sound realtime at 44.1. I have 2 Software Sample Rate Converters that I have used in the past but it was an actual 88.2 file that I was converting back to 44.1, or vice versa. I did notice that if I converted a file from 96khz back to 44.1 it would take a lot longer than say 88.2 to 44.1 due to a much more demanding calculation. Alot of Audiophiles have been doing this a long time by using CD transports with DAC's that upsample the music to 96 khz, then back to 44.1 for listening.I do put more weight into good cables and connectors,Calibration of your system,and good listening environment though. Regards SkipB

[HarBal]

The problems are the same irrespective of whether the source is recorded at 44.1kHz or not. The fact that the re-sampling is occuring in Hardware rather than software doesn't make the problem any different. In average hardware the same sort of short cuts in conversion processing is likely to occur in your average converter as compared to software implementations and such processing can only add noise, frequency response anomalies and distortion to the original source.

Mathematically, the perfect sample rate converter converting from 44.1kHz to 96kHz to 44.1kHz will reproduce the original source exactly! I repeat, bit for bit exact replica of the original source. The fact that it isn't an exact replica and that it sounds different is an indication that it isn't a perfect sample rate conversion! In fact, a perfect converter is not possible as it requires brick wall filtering which is not realisable. To make it realisable requires a finite transition band in the reconstruction filter implementation. That in turn means that sample rate conversion without aliasing will result in some loss of the very top end of the signal bandwidth and if it doesn't it results in aliasing!

It gets back to what I was saying before. I'm not doubting that you are hearing a difference in that processing but that difference isn't due to superior performance of that configuration. It's just a case of the process colouring the signal in a way that you like. There are a lot of pro sound and hi-fi consumers that buy into a lot of pseudo science mumbo-jumbo but if they happened to educate themselves on the science behind the technology they'd realise the folly of it all. Worse still, manufacturers in alliance with their marketing departments push those pseudo science explanations onto the public because it helps them sell their products but that doesn't make it true. People spend big money on Homeopathy too but that has a very shaky foundation when looked at from a rational scientific perspective.

Now don't get me wrong. If you like the sound you get from that processing then use it by all means and be happy with that. Just don't start making claims that it is generally technically superior to processing in 44.1kHz. It may actually be technically superior in a mix context if you are using plugins that internally up-sample to 96kHz, but that isn't because up-sampling is superior but because your upsampling is of better quality than the up-sampler in the plugin.

Cheers,


Paavo.

[Mister]

This post is more about something Paavo said rather than the present topic - it deals with converting a 96kHz file to the CD (44.1kHz) rate. I was fortunate a while back to attend an AES meeting in Toronto with Bob Ludwig as the guest speaker (It was, no doubt, one of the best attended meetings in memory). Bob passed on this tip regarding sample-rate conversion. He was actually passing on information from a colleague who he regarded as an expert in things digital. It was simply this: regardless of the equipment or software used to go from a 96kHz rate to 44.1kHz rate, always first convert to 48kHz, and then convert that 48kHz file to 44.1kHz. Basically the computational inaccuracies are less severe than going directly from 96kHz to 44.1kHz was the reason given. It may take a longer time (especially with WaveLab 6 excellent Crystal Resampler) but the results based on feedback from clients are worth it.

Karl