Post Your Needledrops, Pt. 2

Ok, so I've had time to download your sample and check this out. It really does seem to be a drum transient, but this is the kind of stuff that a mastering engineer might tame with proper limiting (perhaps analog, but not severe compression/brickwalling) so that the overall level could be increased. I went in and selected it using the Shift+O method I mentioned above then reduced the peak by 4dB. The drum sounded fine. When I used the Fix Single Click Now, the result was a slightly duller drum transient. I find the manual reduction method to be much more effective when a transient is involved. Below is a quick screen grab. As you can see, there is stilla nice, dynamic peak, it's just tamer.

 

Some good comments, Antares. I just listened to these two clips as well, and there's definitely potential there, just a bit of tweaking to the method.

@onlysleeping:
Audigy soundcards are known for having chips that operate natively at 48kHz so if you do anything at rates other than that, you'll be forcing some kind of sample rate conversion either by hardware or by a standard software driver that may negatively affect your sound. I'd recommend recording at 32/48kHz in Audacity and follow Antares' advice about levels. Then use the highest quality SRC (sample rate conversion) software you can get your hands on to convert it to 44.1. Finally apply dither and convert to 16 bits. I think you'll find the results will be better. I'm not sure what kind of SRC comes with Audacity, but the free program called SoX does tremendous SRC. It's command-line based, so you might prefer to use something else such as the SoX SSRC "component" you can add to the freeware Foobar2000 program.

 

Cool Stefan, thanks for the attention! The more I listened to that part, you can really hear the snare ring. Can you tell me how you reduced the peak?

Even on my very modest stuff, I can hear what's going on (and with stuff like Nile, with a huge wall of guitar of distortion, it ain't easy), but it's hard! I had to undo/redo a dozen times. But going over this file with a fine-toothed comb has made me appreciate the record - I'm surprised the drums are left with such dynamics intact. I wonder if this record was made from the mix-down and mastered with little attention. I don't have the CD, but I'm sure all peaks are gone.

 

Here's one I did the other day. It's from The Blues a Real Summit Meeting on mfsl. http://cid-697784bca0c008aa.skydrive.live.com/browse.aspx/.res/697784BCA0C008AA!136

Benz Ruby2>well tempered record player>DIY phono stage>Maudio 610>Reaper.

 

Hey, thanks for listening to the samples! I thought the audigy did 96khz, but that is only output, not input.

I recently mounted a Stanton 681EEE on the table. I had a bit of difficulty getting the weight adjusted right because of the longhair brush. I figured I'd keep the brush in hopes that it would keep dust from getting picked up. The problem is that my Shure tracking force tool isn't high enough and the brushes hit the mat, so the stylus isn't really resting in the gauge.

I'm also trying to figure out the anti-skate. I don't have a test record so I've just been using a blank record to adjust with, but I've since read that's not optimal. From looking at it, it looks straight to me though.

In changing the cart, I also noticed that the wires from the headshell to the cart were very brittle. In one case, there was literally one thread connected to the connector! I solder them all up and have a much better connection (even with my lousy soldering job).

I'll try to post some clips with the new setup. Sound-wise, the highend isn't as loud with the new cart.

Regarding the audigy, I don't have the audigy package installed, only the drivers. For adjusting the levels, I was just clicking on the speaker icon in XP, then changing the level on the aux input.

 

I'm at work and my Audition is at home or I'd do a screen grab. I do have Photoshop here, so I grabbed your graphic and did a quick simulation and added comments as you can see below. Note that this is on one channel. You can just do both channels at once (by omiting step 2), but when a peak is too high on both channels, I tend to do them separately as extending the selection with Shift+O with both channels selected tends to extend the selection too far (there are fewer common zero crossing points between both channels for each side of the selection). As I said before, practice, and it becomes really easy. I find this method preserves the original sonics of transients much better than interpolation (using Fix Click Now or other similar approaches, such as Manual Declicking in izoTope's RX). Of course, when the peak is not musical and is due to a glitch, interpolation is a better solution as the above method would simply make the glitch less loud!

This is also the method I use for deleting clicks that just won't clean up well with interpolation. I select to zero crossings on one channel, then reselect both channels (Ctrl+B in Audition) then delete. The channel that doesn't necessarily have the selection exactly at zero crossings, should be ok due to Audition's smoothing feature for edits (provided you have it turned on in the program preferences).

 

For the Audigy, you might consider trying out the free KX drivers available at (http://kxproject.lugosoft.com/). These require a bit of a learning curve, but they will absolutely get the best possible sound out of your Audigy card. When I had a Soundblaster Live and then an Audigy 2, I used them a lot. They tend to provide ways around some of the limitations of these soundcards and also expose such limitations so you'll be aware of them. Even if your particular model only does 16-bit/48kHz natively, you can still get quite good needledrop sound. Just watch your levels, and as soon as you have the file recorded, convert it to 32-bit format before you do ANYTHING. Then work in the same format and sample rate until all editing is done. Finally, do the highest quality sample rate conversion you can to 44.1. and apply dither as the last step before converting to 16-bit resolution (if you're using Audition for the dither, it combines dithering and converting to 16 bits if you have the Dither checkbox checked when you press F11. Plus for basic work, the sample rate conversion in Audition is not terrible providing you set the quality slider high enough). Not that if you use some sort of dither plugin, you may have to convert the file to 16 bits even after applying dither. In this case, do not check the dither checkbox in Audition (F11) because you'll end up with two passes of dithering, which is too much.

 

Sounds lovely, Dale, very live.

Any anecdotes about the Ruby?

 

Yes, it sounds great, Dale. I saw your post when I was at work this morning and really wished we didn't have a filter blocking download sites as I was anxious to hear it. Nice, balances tone and really natural sounding vocals.

 

OK. Definitely a natural transient, as I first suspected. You probably ruined hundreds of transients after using ClickRepair. ClickRepair is a great program, but using it in automatic mode, with a Declick setting as low as ONE will kill transients. I demonstrated this fact in the previous "post your needledrop" thread, although I don't think anyone bothered to do the simple test, or maybe just didn't feel like commenting about it.
This one stray click that is concerning you probably slipped through undetected, that's why it stood out in your post CR waveform. But looking at the audio clip you just posted, with only about 10 seconds of music, there are many transients, some of them almost as high as the one concerning you.
Take a look at the one @ 0.560 seconds, the right channel peaks at -4.3db. The one concerning you (at 5.514 sec, right channel) peaks at -2.9db. So there's only 1.4db difference in these peaks, and this is only a 10 second sample of music.

Anyway, it's all moot as you don't have a problem. I see no reason for you to be concerned about this transient, there's no point in lowering it, you won't gain anything by lowering it. I don't know why Stefan is bothering with different ways to reduce the peak either. This is exactly what the loudness war is all about, and it's silly. You cannot change the signal to noise floor ratio of the original analog mastering. If you want it to sound louder, turn up your volume knob.

 

Perhaps some of us didn't agree with you but didn't want to bog the thread down arguing.

Sorry but I completely disagree with you. I use Clickrepair almost every day at settings ranging from 10-40 and in only a couple of cases have I ever noticed a degradation of transients, and in only one case was this musically less acceptable than the click and pops I was trying to get rid of. I also know a lot of others who do needledrops and who use ClickRepair all the time and have never complained.

As most competent professional recording or mastering engineers will tell you, the dynamic range of most live music is greater than what can be captured in a recording. Decades ago, limiting was invented to tame transient peaks. This has very little to do with the "loudness war" and everything to do with presenting an asthetically pleasing recording. If it's silly, then fifty or more years of recording is also silly. You'll notice by the way that I'm not the only one who was providing advice on reducing the peak.

 

I understand what you mean, believe me. The guy I bought it from recorded it at 16/441 and had no such sharp peaks. I guess 24/96 picked up a lot more of the transients? Which actually I think is cool, if that's the case; not bad for a cheap Audigy soundcard.

When I recorded, I tried to keep most of the peaks below -6db (just by sight), as recommended by Barry D and others here. I'd like to take advantage of 24/96 and boost the level a little for the end result, which will be Redbook. I don't have a hi-fi system really; this is for car listening more than anything else. I guess I can try both ways and see which one I like more.

CR clearly affected that transient when set at 25; at 20 it seemed to leave it alone. Anyway, for a record like this, with a huge wall of guitars, probably cut pretty hot (it gets pretty distorted in several parts), I don't know how much it matters, but food for thought.

 

I guess you didn't do the simple test yourself, because there's nothing to disagree or argue about. It's a fact, why not download the very small flac files I posted, run them through ClickRepair (with a Declick setting as low as ONE), and watch the transients get reduced.
I don't see why you should feel the need to get defensive or feel offended about this. It's a demonstrable fact. If you can demonstrate a flaw in my simple test, I'd be glad to see it.

I'm not complaining. As I said, I think it's a great program, and I too use it almost everyday. It's the only program I'll use for declicking a needledrop (but I use it in manual mode).The point is, still, that it will reduce transients with a Declick setting as low as one. If one simple test doesn't prove it to you for some reason, I can come up with more examples (kick drum, snare drum transients, etc) from different styles of music that still demonstrate this.

I'm a professional recording and mastering engineer, some people think I'm competent, but it's beside the point and I'm not one to toot my own horn. Anyhow, many competent mastering engineers are ruining music today by using compression and limiting, no need to go into details on this forum.

Decades ago, limiting was used get more signal to noise ratio on a vinyl record, which is a good thing. But this is moot, because as I said, you cannot change the signal to noise floor of this needledrop by reducing transients and then raising the overall level. The ONLY thing you'll accomplish by this is slightly dulling the sound by reducing the transients. If that's aesthetically pleasing to you, you're welcome to it.

 

Great observation about the transient not even being captured at 16/44.1. If you zoom in on that transient peak, it's roughly about 25 samples in duration. The difference between 44,100 samples and 96,000 samples is obviously huge, so easy to understand why that transient might not get captured in the first place, if done at the lower resolution.
I agree with Barry'd advice to keep the peaks low when setting the levels, one good reason is that different programs display waveform peaks at slightly different levels. If I do a needledrop and the peaks are around -4db, I usually won't even normalize it.
I'd recommend the following CR settings for a file like the one you were working on: pitch protection & reverse ON; method WAVELET, DeClick 10. This is if you are going to do it fully automatic. If you still hear clicks afterwards, run it through again using the SIMPLE method, with same settings. I've found that it has a lower threshold for detecting clicks, and the repairs that it does are not as big in size. The settings X2 and X3 do much bigger (more extreme) corrections, respectively.

 

Thanks SuperFuzz, I'll try those settings, too. The guy I bought it from had cleaned it with record cleaning machine and was still noisy. I think it's gotten noisier just from my playing it several times. My AT cart picks every little particle in the grooves.
Even at 20, 25, several big pops did get by CR - they were buried in a sea of crushing guitars so I guess that makes sense. I've been playing around with "remove single click now" in Audition on those.

 

Where are files? Provide links and I'll run your test. If I seemed defensive it's because I and many others I know use CR regularly and quite frankly the notion of a setting of one ruining transients is to my mind ludicrous. I haven't seen CR dull transients on material at any setting below 25-28 or so. Thus, when I see someone repeatedly claim a setting of 1 will do it, I immediately believe that person is deliberately trolling and is not serious. So give me the file please and I'll check it.

 

I got the Ruby 2 (used of course) for less than half what I paid to get my Van Den Hul MC1 rebuilt last year. When I played the first record I was blown away by the improvement over the MC1 but perception is a powerful thing. After a week of tweaking I've found that the cartridge seems to be very sensitive to set-up. I can say that the Ruby seems to have a lot more resolution than the MC1 and has a way of making it easier to forget about the system and imagine the sound is not a recording.

The ruby measures neutral if frequency response as does the MC1 but the ruby is a db flatter with better channel balance and more high frequency extension. If you can afford such things new, which I can't, I think the Ruby may have actually been worth twice the cost of the MC1 when they were both being manufactured. The MC1 has been my reference cartridge for 15 years but the Ruby is on my best table for now anyway.

 

That recording is one of the best sounding live records I've ever heard. It has sounded awesome on every system I've had since I picked it up about 10 years ago.

How do you like that test record? I find the azimuth and pink noise tracks very useful. The channel balance on the 1 kHz reference track seems a little off on my copy though.

 

The link is in the previous "Post your needledrop" thread mentioned earlier. You responded to the test in that thread but apparently didn't have time to verify it for yourself. The file is still there. Here's the post:
http://www.stevehoffman.tv/forums/showpost.php?p=4621830&postcount=1573
I'm not really sure what trolling is but I was serious, I figured some people might find this information very useful. I could post examples with other sounds if you like.

 

I thought I'd give another example of Clickrepair 'fixing' false positives with a declick setting of one.
http://rapidshare.com/files/295116437/MalagaVirgen-SAMPLE2.flac
This is a 30 second clip of a Brand X tune from the album Livestock. The well recorded kick drum, with a nice snap to it, causes many false positives. With a declick setting of one (other settings: pitch protection & reverse on, wavelet method) slide the automatic slider to off, and the processing will stop on transients at 03.689, 09.376, 25.097, and 29.478. If you raise the declick setting to 15 (which some people consider low) it will stop on about a dozen other transients.

If people are learning how to use the program, I'd suggest setting the automatic slider to about 10, so you can visually inspect (most) all corrections before accepting. If you can't tell by looking if it's a true positive, keep the original file open in another program and listen to it to check. Very time consuming but does give optimum results.

 

"Perception is a powerful thing" - I like that.

Enjoy that gem (and share another clip or two when you get a chance will you)!

 

Yes, you're right, there are two sounds that ClickRepair removes. I hesitate to call these false positives really because musically they sound like no harmonics I've ever heard either as a musician who's played guitar and some bass for over 30 years or as a music fan. It almost sounds as if your cartridge was unable to track them properly and is perhaps adding some sort of distortion. It shows up in a spectral view as such, but the waveform view looks more like music. I noticed on the first one, when I used the Remove Single Click Now in Audition after highlighting the bar in spectral view, the resulting sound was much more natural. For the second one, this didn't work as well.

Of course, it may well be the recording, but it's not something I personally would find pleasing. I'd always be annoyed by it because it sounds like a click to me. To each his own.

As for the new sample you posted, yes, when I run CR on it with a setting of one, then null the resulting file to the source (I prefer running full auto and then checking with a null inversion. Manual mode is far too tedious for my tastes), there are indeed a handful of clicks removed but when I listen to the resulting file, I hear no overall dulling of the file. In fact, when I listen to the noise output in CR, the few ticks and pops I do hear are things I wouldn't want in the recording. Again, this is my personal preference. They may well be part of the original recording, but they are not, to my taste, musically desirable or relevant. There are still plenty of dynamics. Saying that might be sacrilege to some who frequent this forum, but we are all entitled to our opinions.

Could it be there you're overly focused on uncovering individual cases where CR is catching false positives? I know I've been guilty of this at times to the point where I couldn't enjoy the music because I was almost obsessed with achieving perfection. As Brian Davies wrote in the manual that comes with CR, in any declicking situation, there is always some sort of trade-off. You have to arrive at a level of trade-off where you're comfortable. When I wrote that I've only encountered one or two cases where CR overdid things to an unacceptable level, I was talking in a more drastic and global sense. One case I remember was pointed out by someone on this forum, the intor to Van Halen's Jump where the bright sawtooth waveform on the synth seems to wreak havoc with CR's detection algorithm. Of course if I went through my day-to-day needledropping, I might find the occasional transients that were mis-identified, but I have many more things to do in a day then put each click under the microscope. If the resulting declicked file sounds basically like the original but without all the clicks, pops, and other assorted junk that was distracting me from enjoying it, then I consider CR a success. That's a trade-off I'll gladly make.

 

I'm quite happy with it as I was able to dial in the azimuth much more precisely than in the past and I notice a significant improvement on imaging and focus, especially for high frequencies. I double-checked mine last night and th 1kHz reference track seems ok.

My only disappointment was with the frequency "sweeps" in that they are stepped and not true linear sweeps. Actually both would have been useful. The steps are fine for identifying specific frequencies, but I was looking for a complete sweep I could look at in a frequency analysis. I ended up using the pink noise track for this by matching it with pink noise generated in Audition and then developing a compensation EQ curve. Strangely enough, I had to back off on the bass compensation as the result was a bit too bass heavy.

 

If you prefer a duller sound, so be it. But no, it is not distortion, or a mistracking problem. I think my system reveals transients especially well (as evidenced in the samples provided). That's the way Percy Jones's fretless bass harmonics sound, and it's a beautiful thing.

What ticks and pops are you talking about?! There are none in this sample. As I stated in the previous needledrop thread, listening to the "noise" output while the file is being processed is useless... no human being can hear a sound less than 1ms in length and determine if it's "noise" or a normal part of the music. The noise output will always sound like little clicks. The second sample posted is from a well recorded and mastered record, with lots of transients and dynamics, beautifully needledropped - or so I like to think.

Yes we are, and the whole point to this isn't about personal preference or what people find desirable in sonics. It was about the fact that ClickRepair (again, my favorite declicking program that I highly recommend) will kill transients with a Declick setting as low as one.

As for your judgment of me being overly focused on perfection - guilty as charged. However you want to Declick your needledrops is your business, and I wouldn't pass any judgment on it. Of course my method is tedious and time consuming, as I stated before, and I don't fault anyone (especially inexperienced users) for using the program in automatic mode, which still does give very good results. Anyway, I think the point has been made and perhaps a couple of people reading this might find my findings useful. The anal retentive perfectionists among us.

 

and now for some music.. here's a sample of the Gene Vincent rockabilly tune "I Sure Miss You" done by Robert Gordon and Link Wray from their self titled record from 1977 (cut by Greg Calbi).
http://rapidshare.com/files/295565610/ISureMissYousample.flac (24/96 flac)