X1800 Sample Rate Discussion

Greetings,

I’m a new X1800 user and use it along with my MBP and Virtual DJ. I have all my channels set to USB and mix with a combination of mp3 and ALAC tracks that are all ripped with a Sample Rate of 44.1. I also use VDJ to record the mix internally to a 44.1 WAV file before using Audacity for mastering (and fixing mistakes). Looking in the X1800 utility, I see the default Sample Rate for the X1800 is 48. My question is, should I leave it at 48 or change it to 44.1, in order to match the songs that are being inputted to the mixer? Personally, I’ve always tried to keep audio processing to a minimum when working with digital audio and I’m not sure what the point of upconverting 44.1 to 48 then back to 44.1 when recording internally using VDJ. I understand that there is a benefit to upconverting temporarily in a DAW if you’re gonna be editing or processing but it seems that that would be different from what the Sample Rate setting suggests in the X1800 utility setting. Can somebody set me straight?

1 Like

I believe Atomix recommends you set your gear to 44.1 if you can. That’s what VDJ was originally designed for and operated at, anyway, with downsampling of higher rate files eventually added. Version 8 now supports the ability to resample to higher rate output settings, but I would assume it’s most stable if you just put it at 44.1 still. I think their only exception to this recommendation is if you’re specifically mixing stuff at higher sample rates, like music videos or studio masters where realtime downsampling of them is going to lead to more CPU usage and degraded sound.

The X1800 does seem to do a better job forcing the CoreAudio/ASIO driver end to the right settings to match the mixer’s own settings. I notice the X1700 sometimes gets a mismatch and you have to go in there and fix it manually. So I assume one of the reasons you’re asking this is right now you’re probably not having any overt issues.

Yeah no problems here. When I plugged in the X1800, I set it to 44.1 out of habit because that’s what I had to do in order to get my old Denon DN-X600 to work in the same setup. I just wondered if using 44.1 is the best practice since it’s set at 48 by default.

Yes, for VDJ it usually is.

The SPDIF output, by the way, is a separate setting. You’re probably not even using that, though.

Any kind of conversion will cause degradation of the contents of an audio file. Even for post processing in a DAW, my personal belief (as a certified sound engineer) is that it is better to stay within the native sampling frequency/bit rate of any track. With 44.1 being the CD-standard, it is safe to say that this quality is high enough for the most critical listener on a high end hifi system. Us DJ’s at not always great PA’s, at high volume levels and with drinking customers who are not there to “listen” to the music, it is a total non-issue. Again, in my humble opinion.

Sidestep: Only if you record in 48K (and if I know I am gonna end up with CD-quality anyway, I’ll actually opt for a 44.1K recording setting) is it worth processing in that rate. Frankly, if you record, doing so at 88.2K makes way more sense. Down-scaling would be a precise 2:1 ratio, relatively easy to calculate. As opposed to recording at 96K, which brings the ratio to approx. 2.18:1. Clear how that requires a lot more calculating and “guessing”.

If you go from sampling 44.1K times per second and upscale to 48K (only a relevant rate because the DVD boys decided they wanted something else for DVD audio), the computer will try to insert the missing bits. At best it’s an educated guess based on the previous and next values and the steepness of the waveform. So, if you do post-production in a DAW, you are manipulating stuff that wasn’t really there to begin with. When you then down-scale to 44.1K, the computer needs to take out bits here and there to make it “fit” again. You can see how that wouldn’t enhance the quality, it would sooner be a degradation.

End of the day, for any and all DJ purposes (including publishing your work on soundcloud etx, streaming/broadcasting it, distributing it on CD - should you still be inclined to do that)), CD-quality 44.1K is fine. Just stick with it throughout your recording to mastering process and you’ll be absolutely fine.

1 Like

Great information, this section was really helpful. I must confess that I don’t get booked anymore but I still regularly grind out mixtapes from the safety of the bedroom. I’m a budget audiophile and a DJ which is, admittedly, a bit of a conflict. I mostly consume mixes via my mid-fi portable headphone setup so I guess I make mixes for the same type of listener these days. That’s why I take care to produce the best-sounding mixes that I can without being too anal about it, lol. For instance, long ago, I culled low bitrate mp3’s out of my collection because garbage in = garbage out.

Sample rate is not important unless you want to use the content for film production or DVD, in which case you should work on 48 kHz. For mainstream audio content susceptible to be distributed online or on CD it is best to work with 44.1 kHz sample rate. The important aspect here is the bit depth of the material to be recorded. If you want to work on your mix afterwards, “improving” it’s sound signature, normalizing (including taking care of true peaks in analog domain) and setting the cue sheet in a DAW, then, I suggest recording at a bit depth of 24 bit. After you are done with the finishing touches you can convert from 24 bit to 16 bit, applying dither.

I’ve heard some great sample rate conversions using advanced interpolation over the years.

Some of the original engineering work for Red Book strongly suggested it should have been 20bit/56khz, but they could barely even mass produce 14bit/40khz DACs at the time and they didn’t want to hold out for chip improvements.

Sample rate does make a difference still, though nowadays less so than decades previously with the modern use of massive oversampling and also delta-sigma architecture as methods of mitigating 44.1’s limitations from calling it so close and having just barely enough sample rate to get by. The first CD players were truly awful and I don’t think people nowadays realize just how bad. A lot of this stems from the issue of real-world implementation of filters, which is also a problem in sample rate conversion. ALL REAL FILTERS HAVE AN EFFECT BEYOND JUST REMOVING HIGHER FREQUENCIES.

While technically, for instance, going from 88.2 to 44.1 is easier than 96 to 44.1, you still have to low-pass first. Even with digital filters, there are tradeoffs between sound quality, latency, and slope steepness – think of it like a line extending out in both directions from a pivot point that is the corner frequency. Shannon-Nyquist Theorem is only perfect on paper working under zero constraints, like not having to worry about having either infinite time or infinite processing power to convolve it according to the actual math proof. It requires that AT LEAST the top half of the sampling range be lopped off and be outside the audible spectrum for it to even work. This is necessary both on the AD and the DA ends. The optimal method is just to use a higher sample rate to begin with and combine it with a better-sounding real-time filter with a gentle slope that’s higher up in corner frequency. Gentle slopes affect earlier but do less later (think that line on a pivot). With 44.1, you’re calling it very close… unless of course you want roll-off beginning at 10 or 12khz on your CD playback.

Now, if you don’t need real-time filtering, it’s not such a big deal. With powerful enough computer chips you can use some windowing functions to feedback forward audio for some FIR math wizardry and get both a steep slope & good sound quality with just some accompanying laggy delay (because you have to wait for the data loop to do its thing). Fine if you don’t care how long it takes for the music to come out of the sound system. Not so good if a band is playing into the sound board or you’ve got a DJ doing their thing.

1 Like

That depends on what audio material is used by the DJ. Usually we work mostly with 44.1 kHz material coming either from CDs or from other internet sources. It doesn’t make sense to work with high sample rate and bit depth material unless this material it’s your own creation (mixing and mastering). I don’t think many of us mix live from studio quality source’s. So, if we usually mix 44.1 kHz material, there is not much sense in recording the mix at higher sample rates. But we can use the highest bit depth possible during recording to enable us to retain as much detail as possible to work with afterwards if we desire. Of course there no problem if we want to record everything at the highest sample rate possible so that afterwards to be able down sample as desires using good non integer Sample Rate Converters.

Even if you are playing back 44.1 material, there is a benefit to higher formats if the digital gear is processing already at higher formats. It will usually sound best if it’s recorded in the native format the gear processes at. If you’re all analog gear and you’re going to 44.1 to post it on the net or release it commercially, then there’s little benefit to sampling at a higher rate unless you’ve got poor AD converters or something.

With higher formats, variation in the quality of both AD conversion and DA conversion is also less important. While it’s true you can fool most people into thinking a 44.1 recording played back on 44.1 equipment sounds as good as a native 192khz recording on 192khz equipment, to do so is an engineering feat that relies on hiding the flaws of the format from human perception and is not a given. It’s much easier to engineer the higher format gear to sound better than the lower format gear. It’s one of the reasons video like DVD and Bluray tends to use 48khz and above: more consistent results from hardware to hardware. Which is also why it’s preferred in studios, too. The defects of many 44.1 conversion hardware is right in the presence region and even 48khz bumps it up a little above that.

Working at studio quality standards always benefit from high sample rate and bit depth values. But, let’s not forget that many mobile DJs want to just record their performances for promotional purposes and without too many complications. Very few people work with sound and mastering engineers because of the costs involved.

You don’t have to be working with a mastering engineer to hear the benefit. If you want to listen to the recording of a mix you’ve done with the original sound quality, recording the 24/96khz SPDIF output from even the old DJM800 is going to sound better than recording it at 44.1, partly because of inherent characteristics of the format and partly because with 44.1 you’re either relying on DA conversion or sample rate conversion. If you don’t care about preserving the original sound quality to hear it later at your convenience and all you care about is a functional internet release or something, then yeah, no point in bothering. But there is a sound quality difference.

High sample rate and bit depth recording and processing always ensures best sound quality. But ultimately the distribution format will end up almost always in 44.1 kHz, to be able to play in most consumer devices.

Even if that’s all you care about, I’d rather record the original 96khz work signal and then use better non-real-time post-processing like iZotope’s stuff than relying on inferior real-time conversion. Even a lot of non-real-time post-processing stock/default sample rate and bit depth conversion in many DAWs and editors are not great… and the real-time variety is usually even worse, both in hardware gear and software packages.

If it feels better to record at 96 kHz or higher sample rate it’s ok. Your ears won’t hear it anyway. What your ears will hear though is the improved signal to noise ratio (dynamic range) due to higher bit depth.

Not really. You’re more sensitive to temporal variations with a consistent, repeatable pattern, especially in the presence region where you get upper midrange vocal harmonics and sibilants than to the difference between -80dB noise floor and -120dB noise floor. It’s also why vinyl with its high noise floor of sometime only around -50dB has such a surprisingly easy time sounding relatively pleasing, natural, and lacking etch, assuming you’ve set things up right and are using good quality gear & relatively unworn pressings.

Then there’s the myths surrounding dither (random noise), which does little beneficial unless you have an incredibly minimalist recording (hardly anything going on in it except like a triangle tinging in the distance) with a very low noise floor already for quantization distortion to need masking. The pre-existing noise floor of the audio combined with the music itself usually sufficiently masks the distortion from a bit depth conversion.

The only other time dither is useful is masking other forms of periodicity distortion already present besides the quantization distortion issue from bit rate conversions, but it isn’t quite as effective at that. If you’ve got such issues, though, running a DSP at something like 64bit double float and then doing aggressive dither has some effect. Even though dither is usually not helping, because slight changes in noise floor or larger changes with already very good S/N ratios is not particularly noticeable, it usually doesn’t really hurt to do it, either.

This has been an extremely informative discussion and I appreciate all of your technical knowledge. I studied broadcasting in college and learned a little about digital recording and editing back in 1998-2002 (the dark ages) so I learned a little about this stuff back then but I love brushing up.

I guess this leaves me with a question, why did Denon choose 48 as the default sampling rate?

This conversation has me thinking that, for my needs, I should set the X1800 at 44.1 but consider using the second USB-out on the X1800 to record to my PC with Audacity, since Audacity works at 32-bit float before downsampling to 16-bit when exporting the final mixdown after I’m finished editing. Until now I’ve usually just recorded a wav file within Virtual DJ (my preferred DJ software) where the only bitrate options are 16 or 24.

I don’t know anything about the USB ASIO on the Denon being 32bit or not. For sure the SPDIF isn’t 32bit. It’s not going to hurt anything to record the samples at a higher bit depth, though I’d recommend making sure you’re using a bit perfect method like ASIO, WASAPI, or CoreAudio if on a mac. I think you would be fine just recording the 24/44.1 in virtual DJ with an X1800 channel back.

Denon probably set the default to 48khz because some effects on the X1800 don’t work in 96khz mode. The processing or resampling quality at 96khz also does not sound as amazing as the 96khz mode on the X1700 to my ears, so there doesn’t seem to be a whole lot of benefit to bumping up to that unless your source gear is running at that and everything else later is, too.

On an unrelated topic, the on-board resampling is still probably better than what a lot of streaming apps do internally (like Wirecast or OBS), so it’s probably best to also put the mixer at the rate any streaming app might want to use, which is dictated most by what the site is you’re streaming to.

The SPDIF output sample rate does not appear to be a separate setting on the x1800 as it was on the x1700. Not a problem, just correcting my prior statement.


Another correction: FIR filters feed forward, not backward. IIR filters also feedback.


Also, here are some resources on the topic:

https://ccrma.stanford.edu/~jos/resample/Free_Resampling_Software.html

And an amazing bunch of tests and comparisons that will let you see the amount of aliasing getting through on a sweep, distortion/noise, when the roll off begins and how much ultrasonics get through, phase/temporal changes per frequency, and the impulse response of varying implimented techniques. You’ll be able to easily see the tradeoffs, like impulse with no pre-ringing/echo comes with all sorts of issues like more post-ringing/echo or more noise/distortion, or more roll-off & phase problems, etc. Also don’t miss their FAQ and HELP links. Particularly fascinating are the sweep tests where you can see the aliasing geometrically reflected back into the audible range if the low-pass hasn’t sufficiently done its job prior to decimation, and then quantization distortion is also apparent when the black background gains a tint. Make sure your screen contrast is adjusted properly to show subtle amounts of the latter.

http://src.infinitewave.ca/

1 Like