So...
As a musician and basically novice/sophomore engineer, I have continually had one gripe about recording to digital -- that cheap, distorted high frequency sound that I constantly have to listen for with every new piece of software that I add to the signal chain.

I love the clarity of digital. I love the definition of the transients. I love the transparency. And with very little recording experience (other than 4 track tape) prior to the digital revolution, my tastes are not fashioned by gear that I've become attached to so I feel that I listen with a fairly objective ear.

BUT.

I have been painfully aware of this cheap digititis since I began recording digitally back in the early 2000s. I have tried all kinds of warming/saturation/smoothing/oversampling plugins to avoid, aid or fix the problem but alas, most of all that I've tried has amounted to nothing. Tuning my room and upgrading my gear has allowed me to hear these gremlins better rather than fix the issue.

So... What I'd like to know once and for all is:
1. How do I avoid this cheap digital sound?
2. What gear/software should I look for?
3. What specs will tell me if this is going to be an issue?
4. What gear/software is there that is available that will help fix this problem?
5. What gear/software is there that is available that will help mask this problem?

Thanks.
Mattyy

--------------------
Just a fan of music...

Yes I've hated that sound and after many a conversation here I believe the problems have been ironed out. There are still some cheap sound cards out there but i believe it all comes down the the DACs.

I accidentally tried a higher sample rate, on my Scope Pulsar soundcard, the other day and the soft synth's tail had much less noise and more clarity. It was unreal. I thought that only applied to really high end stuff.

Quote Mattyy:

---

1. How do I avoid this cheap digital sound?

---

Avoid cheap, poorly designed converters, avoid situations that produce signals over 0dBFS, and avoid situations that involve very high levels of very high frequencies.

Quote:

---

2. What gear/software should I look for?

---

Most modern digital gear is fine within the provisos above.

Quote:

---

3. What specs will tell me if this is going to be an issue?

---

None, really. Products with 'half-band converters' can suffer aliasing issues if pushed, but most chips are of this type and the manufacturers don't generally mention them in the specs.

Quote:

---

4. What gear/software is there that is available that will help fix this problem?

---

None. Aliasing distortions can't be removed. Working at double sample rates will avoid the problem if the rest of the recording and mixing equipment can cope.

Quote:

---

5. What gear/software is there that is available that will help mask this problem?

---

None. Aliasing is a very unnatural sound which we pick up very easily.

In my experience most modern digital gear works fine if used sensibly. Aliasing is usually the result of misuse one way or another. Improper clocking is the most common reason. Excessively high levels of hf sources is another.

If your are having problems and everything is clocked properly, the use double sample rates and keep peak levels below -10dBFS.
H

--------------------
Technical Editor, Sound On Sound

Or solve the half band aliasing issue the other way, by applying a LPF at about 18K to all the inputs from the ADCs once in the digital domain.
The aliasing from the half band decimators will all lie in a fairly narrow band just below Fs/2 (The halfband filters tend to be sharp, they just go over a little too high), chop the top couple of K off and the aliasing goes with it (And odds are nobody past early teenage will hear the difference).

Or sample at 48 and downconvert to 44.1 using a resampler that has a sane impulse response (Which is to say NOT a halfband filter).

Or just use truly decent kit in the first place of course.....

A really good test BTW is to record a set of keys being shaken a inch or so from the mic, tends to shop up all sorts of issues.

Regards, Dan.

--------------------
Audiophiles use phono leads because they are unbalanced people!

+100 for keeping tracking levels down.

A couple of years ago I was helping out a 'back to tape' die-hard who was complaining about the digital artifacts he could hear since he'd gone to using a DAW.

He was hitting -3dBFS for his vocals using a 'less than ideal' aging audio interface and it did sound terrible.

We had to drop the peaks to -12dBFS before things improved noticeably. Once there his recordings sounded perfectly acceptable and his 'digititis' was cured.

--------------------
An Eagle for an Emperor, A Kestrel for a Knave.

My gear is of decent if not "Lavry" quality. My engineering skills on the other hand... LOL!!!
I'm learning
So - It's most likely a method problem - not an electronics one

That being said - this is what I've learned here:
1. Keep peak levels below -12db.
2. Record at 88 or 96K when possible.
3. If problems still arise - apply a LPF @ 18K - 24 db/8ve slope? Linear Phase?
4. Keep my high frequencies in check.

Thanks for the tip about the keys - I'm definitely trying that one

One last important question:
I most often notice this digititis (for lack of a better word) when using soft synths. A few are excellent in that they never sound this way but most sound REALLY harsh - especially when you open the filter up. Is this the same problem or something different? I know that the Virus has popularized (?) harder sounding synths but even it doesn't sound "cheap". Then again most presets are clipping the output so... What gives here??

Thanks again for your generous advice. I still read the magazine cover to cover but it's this board where I seem to find the most answers these days

Mattyy

--------------------
Just a fan of music...

Quote The Elf:

---

We had to drop the peaks to -12dBFS before things improved noticeably. Once there his recordings sounded perfectly acceptable and his 'digititis' was cured.

---

I'm not surprised... and the silly thing is that the 'digititis' was almost certainly analogitis, in fact.

I know Elf knows this already... but for the benefit of others, most diital converters are set up to produce +20 or +24dBu when fed with peak level 0dBFS signals. And that is entirely reasonable given that most analogue devices are also designed to clip at around +24dBu -- so the standard converter clipping level aligns with standard analoguie clipping levels.

However, most budget analogue equipment struggles with peak levels over about +12dBu, and will start to sound harsh, strained, edgy, aggressive and so on when the elvels go over that. The reason is that the analogue electronics are starting to become non-linear as they run out of headroom and head towards clipping, so produce increasing amounts of distortion.

High end analogue equipment tends to be able to cope rather better because of superior circuit designs, and so a lot of people think that's the only way to go.... But actually the problem lies int he fact that so many users of digital equipment work under the misguided 'knowledge'* that they should peak close to 0dBFS all the time.

The real solution is to work with traditional analogue nominal levels and headroom margins. Set your tracking and mixing to work with average levels around -20dBFS, with peaks no higher than -10dBFS and the whole thing suddenly sounds nice and analogue again. Funny that!

* Some of the misguided 'knowledge' for peaking high comes from the very earliest days of digital when converters struggled to reach 16 bits and didn't use dither properly (or at all!), but is mostly because people see commercial music peaking against the endstops all the time, and don't realise that the headroom-stripping madness is a function of historical mastering lunacy which, again, dates back to the earliest days of rubbish converters. The technology has long since rendered this policy completely unnecessary, but the industry inertia prolongs it...

H

--------------------
Technical Editor, Sound On Sound

In answer to your last point - synths are supposed to sound harsh when you open up the filter. Using the filter properly is one of the most important aspects of synth programming.

James.

--------------------
JRP Music - Audio Mastering and Restoration.
http://www.jrpmusic.net

Quote Mattyy:

---

1. Keep peak levels below -12db.

---

-12dBFS -- the lettters are important to the meaning -- but yes. That alone will make a big difference.

Quote:

---

2. Record at 88 or 96K when possible.

---

It is a useful option, but not one many of us feel the need to use all the time. Working at double sample rates obviously doubles the amount of data that has to be moved around, processed and stored, and that isn't always practical. I generally use the double rates only when working with material that involves high levels of high frequencies -- so close-miking strings and brass, or lots of loud bashed metal!

Quote:

---

3. If problems still arise - apply a LPF @ 18K - 24 db/8ve slope? Linear Phase?

---

Yes

Quote:

---

4. Keep my high frequencies in check.

---

Just be aware that high levels of HF near the filter cut-off will lead to aliasing if the filtering is less than perfect... and they all are!

Quote:

---

Thanks for the tip about the keys - I'm definitely trying that one

---

It can be quite a scary test. Jangling keys generates a lot of ultrasonic energy which will reveal non-linearities throughout the audio chain -- from the microphone, through preamp, converters, and power amps! The converters are likely to be the culprit if you hear nasty artefacts, but just be aware that other elements could contribute too!

Quote:

---

I most often notice this digititis (for lack of a better word) when using soft synths. A few are excellent in that they never sound this way but most sound REALLY harsh - especially when you open the filter up. Is this the same problem or something different? I know that the Virus has popularized (?) harder sounding synths but even it doesn't sound "cheap". Then again most presets are clipping the output so... What gives here??

---

If the output is clipping that's the problem! A lot of soft instruments and sample-based systems generate stupidly high levels. I recommend inserting a pad or gain-adjustment plug-in in the DAW channel for such instruments and sources to bring the level down to peak around -12dBFS before doing anything else.

H

--------------------
Technical Editor, Sound On Sound

Quote Hugh Robjohns:

---

Quote:

---

2. Record at 88 or 96K when possible.

---

It is a useful option, but not one many of us feel the need to use all the time. Working at double sample rates obviously doubles the amount of data that has to be moved around, processed and stored, and that isn't always practical. I generally use the double rates only when working with material that involves high levels of high frequencies -- so close-miking strings and brass, or lots of loud bashed metal!

---

Also, when it comes to digital audio synthesis, some very common processes add lots of harmonics: broard band generators or oscillators (square wave for example), distortion, overdrive, soft clipping, wave folders, frequency modulation, etc

There are 2 options:
- when programming the software, you make sure that the synthesized sound has no energy above 20000 Hz; elegant, but not always possible
- oversampling > proper DSP processing > anti-aliasing filter > downsampling

But with the oversampling solution, 96K is often low. Many times you want something like 16x oversampling. Oversampling the whole synthesis software would require an extreme computer in realtime though, so the idea is to work at 44100 or 48000 Hz, and use oversampling only where it is required (some parts require 2x oversampling, some up to 16-32x, etc).
Now I have no idea what plugin makers do in practice. I would say many just do nothing, so it may sound slightly better at 96K, but not a lot more, especially when comparing to the 44100 Hz + local hardcore oversampling solution.

Cheers

Brilliant stuff!!

So what new stuff have we learned today? To summarize for us punters out there:
1. Opening up the filters of a softsynth can "bumrush" the top end of the sonic spectrum causing issue.
2. Double sample rates are useful especially for material with a lot of high frequency content.
3. Simply dropping the output level on a softsynth to -12 can improve it's tonality.
4. Oversampling at 44k or 48k can often be more effective than working at 88 or 96 with software synths.

I apologize to H for not including the units here but as a musician, quite frankly, these confuse me - LOL!!! I guess learning these fundamentals a little better might be a more beneficial than simply demoing another distortion plugin no? LOL!!

So the million dollar question has to be:
WHY on earth do a few of the newer softsynths sound so sweet? *#coughDiva#@*??? Is it intelligent gain structuring inside the plugin? The presets are still often too loud and even with oversampling off there are at least a half dozen pieces of software that just seem to sound better...

Thanks again.
PS. Yesterday after I read this, I immediately finessed my template inside of Logic. Can't wait to start driving my newly tuned DAW - LOL!!

--------------------
Just a fan of music...

Oh.
And clocking was mentioned earlier. For the benefit of us newbies:

What is the best way to set up the clocking in a digital system

I'm sure that there's a few articles on this so a link would probably suffice. I was at a loss when I first set up my Logic/Motu system. The literature that came with both products just seemed to gloss over how exactly to do this seemingly VERY important duty.. I just set my interface as the master and left it at that but still I get the occasional audio spike when recording. My mic is of pro quality and my gain is peaking below -10 so...

Aside: I personally would like to hear an A/B of the SOS piano test - one sample recorded peaking at -3 vs. another recorded peaking at -12, both normalized (24/44 of course). More audio examples of this might convince people to clean up their acts. LOL!!

--------------------
Just a fan of music...

Quote Mattyy:

---

What is the best way to set up the clocking in a digital system

---

http://www.soundonsound.com/sos/apr03/articles/digitalclocking.asp

There can only ever be one clock master in a system. Everything else has to be slaved to that master.

Clocking is only critical where the audio signal is translated between the analogue and digital domains, with the A-D being the most critical (because you only get one chance to digitise things, but can use different D-As if necessary).

So as a rule of thumb, use whatever device hosts your A-Ds as the clock master -- which is probably your interface in a simple system.

Most devices perform worse when externally clocked.

H

--------------------
Technical Editor, Sound On Sound

Soo... does my DAW need to be synced to my interface?
My interface is an 8 in MOTU unit that uses SMPTE if I'm not mistaken.
My DAW is Logic Pro 9 (MTC).
Both manuals talk about being the master or slave but neither of them clearly mention a single interface/DAW configuration so I just left them both in internal (master) mode. I just read the SOS article and it doesn't reference this either.
Strange business this clocking. There should be a single standard/universal reference no? Maybe just the DAW/Software? Or is that technologically not feasible?
Questions....

--------------------
Just a fan of music...

Quote Mattyy:

---

Soo... does my DAW need to be synced to my interface?

---

No.

Quote Mattyy:

---

My interface is an 8 in MOTU unit that uses SMPTE if I'm not mistaken.
My DAW is Logic Pro 9 (MTC).

---

It seems you're getting confused between digital clocking (AES, SPDIF, ADAT, Wordclock) and sychronisation (SMPTE, MTC, MIDI clock) systems.

The former are very high speed clocks (e.g. 44.1 kHz) designed to keep digital systems in step as they pass bits of information to each other.

The latter are relatively low-speed (e.g. 25 frames per second) signals designed to keep things like video and audio in sync from the days of film. MTC was used to lock MIDI sequencers to multi-track tape recorders, for instance.

Hopefully now that you appreciate that they are completely different the manual will become clearer.

--------------------
An Eagle for an Emperor, A Kestrel for a Knave.

Whew!!! Thanks Elf. You're right - I was confused.
I often tend to get ahead of myself in these matters and then panic when I realize that I might have missed something. Now off for a reread of both manuals...

Mattyy

--------------------
Just a fan of music...