A better way to needledrop? (Using Voxengo Curve EQ)

KISS method always worked for me.

 

Sure - it depends what you result you want - I think for needledroppers who spend a lot of time editing their needledrops to get a high quality/archive result, the Voxengo method is simpler as well as giving a higher quality result. For those who are happy with needledrops that sound just as if they're playing the record itself, it's KISS* all the way!

*Keep It Simple, Stupid. (I Googled it.)

 

Thanks for posting those samples, darkmass. I listened to them on my PC via Adobe Audition and headphones and couldn’t hear any obvious difference, which of course is the desired result. I think they’re very good indeed: one thing that struck me right away is that your samples are slightly brighter than my samples and that’s clearly because I was using Steve's Rhino LP as the source for the EQ, played via my cartridge that rolls off a bit, while you were using the DCC CD. There are probably slight differences in the mastering, but in any event the CD is bound to me more accurate than the vinyl/cartridge so I think your result is objectively “better”. I do quite like a warm sound though so I think I can live quite happily with mine!

I took the liberty of running your two samples against each other via Voxengo CurveEQ and was interested to see that, although I doubt very much whether anyone could detect it by ear, there’s a slight discrepancy in the bass area (which of course is something you talked about before):




I only mention this because I’ve just made a discovery. Let’s say you use Voxengo to do an EQ match. Afterwards, you run Voxengo again to check whether your target file now matches your source file. Now, let’s say you get a result like the one above - it’s very good, but it’s not perfect. My discovery is that (assuming you’ve still got your target file open) you simply hit the “apply” button AGAIN!!! This will apply the new “difference” curve to your target file. Basically, you’re just fine-tuning it. And bingo, look what happens when I do this with yours:




One step closer to perfection! In fact, you can run Voxengo again and again and each time it gets closer to perfection until it reaches a point where no further gains can be made.

It’s fairly obvious and you’ve probably already discovered it but I thought it was worth mentioning.

 

Edit: if applying Voxengo CurveEQ again and again as just described, I wouldn’t recommend saving and closing the target file each time or there's a risk of degrading the sound quality. The thing to do is to keep the target file open in Adobe Audition all the time so it remains in 32-bit floating point during all stages of the processing.

The way to do this is, each time you apply an EQ curve, you save a WAV copy of the target file somewhere convenient like your desktop. You can then use the WAV copy to do a new EQ check against the source file. Then apply the curve to the target file. Repeat until you’re happy. Simple.

Another tip: to any Voxengo users who might not know this, you don’t have to play the whole file to get an EQ curve. Provided the file is a WAV file, Voxengo will read the EQ of the entire file instantaneously without any need to play the whole track. It only does this in WAV, not in FLAC or any other format.

I normally store my files in FLAC so this does mean flipping to and fro between WAV and FLAC. There are lots of converters on the market but I use the dBpoweramp Music Converter which comes bundled with the excellent dBpoweramp CD ripper. The dBpoweramp Music Converter is incredibly quick and simple to use.

One other tip: changing the EQ of a track will change its overall volume level – sometimes, a lot. When I’ve finished EQing, I use Adobe Audition to volume-match every finished track with the same track on the LP I used as a source. That way, I’m copying all the mastering engineer’s original intentions – not only his EQ but also his judgments on the best volume balance between the tracks. Though sometimes, I can’t resist tweaking a little!

 

Oh no, I myself am not like that in the slightest!

First off, I liked your needle drop and EQ match quite a bit. Something that struck me about both of them was what I think of as a feathery light brushing of the strings in some locations where the guitar itself is just strummed (and distinct from the dulcimer). There is nothing quite like that in either of my examples. I found that "brushing" very enticing. Overall it was difficult to distinguish between your examples just by listening...though in the bass string section near the end, a couple of the bass notes on what I took as the HDCD sourced version seemed to have a bit stronger leading edge to them...likely because that particular version had some of the right channel mixed into the left. But that's in no way a bad sound, it's just there.

Here's how I believe your two renditions break down: "01" was sourced from the HDCD, "02" from Steve's vinyl. My versions break down as "03" from the DCC CD, "04" from HDTracks. Looking at the dynamic envelopes of "01" and "04", it's clear to me they were both derived from a common source. Naturally, Steve's vinyl and CD versions used a common source as well, and it was certainly not the HDCD/HDTracks source.

I appreciate the screencaps you provided and your thoughts. I fold in all information encountered so that I can move in improved directions. As you once replied to one of my rambles, it's about the education.

While I provided a DCC CD segment for you and a downrez of my work, that downrez was a convenience but not my actual destination. My true destination remains 192k/24 files that take on the important characteristics of Steve's DCC of Blue. The following link should enable you to hear that end segment of "All I Want" at 192k/24 from my completed result:

https://drive.google.com/file/d/0ByoJu9I97JBmMUJPY2VMaFkwTGM/edit?usp=sharing

Okay, I have a few things to show you...

Below is the CurveEQ comparison of my completed 192k/24 HDTracks of "All I Want" and my working uprezzed DCC target--precisely the HDTracks that I downrezzed and took the end from for the flac you listened to (and the end segment of the true 192k/24 version is the flac link above) . That's a 0.9 dB lift at 20Hz, with each point of the flat line portion reading as a 0.1 dB depression. Oh, I used 60 data points throughout my 192k/24 work and also used a "Block Size" of 4096 throughout.



It seems the downrezzing to 44.1k/16 had an adverse effect on the EQ comparison with the similar DCC CD segment, but there it is. On my first pass through Blue a few months back, once I downrezzed I re-EQed to the CD--and worked out significant differences at that rez. I credit that to the really significant source differences between Steve's work and the HDTracks. On the other hand, after downrezzing my DAD derived Blues for Thought work to 44.1k/16, I also re-EQed to the respective CD, but virtually no re-EQing was necessary. It's particularly clear at this point that all the Blues for Thought files truly originated from a well-behaved source.

Next picture... This is the left and right channel spectrum analysis (from Sound Forge Pro) of your provided segment of Steve's vinyl and Steve's DCC, RMS normalized to a common value. Green trace is the vinyl, orange trace is the DCC.



You've said your cart has a high frequency roll-off, but that seems to not quite be the case in comparison to the DCC CD. So why does the DCC CD sound "brighter"? I think it's a matter of relative response in the range above about 600 kHz. If the DCC CD was "slid up" in volume, overall, to more closely match the vinyl in the region from 600 Hz to 3000 Hz, the DCC would be stronger in the region from 3000 Hz to roughly 8000 Hz. The region from 10000 to 20000 Hz would also more closely correspond, however the response in that region is also falling pretty rapidly and probably does not make a significant difference in overall tonality. Oh, the small peak shown in the vinyl trace at the far right side of the graph is centered just above 20 kHz--I would think it's nothing meaningful.

Final picture, traces of all four of our 44.1k/16 files...you can probably figure out which trace is which.



Generally, all four match exceptionally well between 100 Hz and about 8000 Hz...with your pair consistent above 8000 Hz and my pair also having their own consistency above 8000 Hz. From 100 Hz down to about 35 Hz the four traces are not exact matches, but they also don't deviate tremendously from each other. Below 35 Hz, the HDCD file seems to take a rather significant turn, but you also have mentioned needing to work on bass regions of your EQ work. CurveEQ doesn't even show what happens below 20 Hz, so you may have had no idea this was going on. The "Block Size" parameter in CurveEQ controls the number of Fast Fourier Transform samples used in the CurveEQ spectrum analysis, if you haven't already explored the effects of various settings of that parameter, you might do some experimenting with it. For my 192k/24 work, I found a value of 4096 to give me the best bass results. However, there might be a more reasonable value than that for 44.1k/16 EQ.

And one further interesting observation... While zooming way into one of your files next to mine while in Vegas Pro (done so I could cut off an end segment that corresponded to yours), I saw that our two files were exactly 180 degrees out of phase with each other. Looking further, I saw your two files were phase consistent with each other, and my two files were phase consistent with each other. I see no right or wrong here--it's sort of the thing I might expect if one of us was working in the Southern Hemisphere and the other of us in the Northern Hemisphere (I'm actually in California as it turns out).

Oh, I wanted to mention this. As part of my process, and all at 192k/24, I separated left and right channels from each other, for both HDTracks and DCC, and EQed right channel to right channel and left to left. I did this because of the weird HDTracks left channel. Following that, I brought things back to stereo, then did a stereo EQ. The results of that stereo EQ are what I've shown in my CurveEQ screencap above. I've tried the "Dual Mono" setting in CurveEQ, but whatever it does is not convincingly to my liking. By physically separating right from left, I felt I had the control I needed. For EQing well behaved files, I would never, ever do such a separation.

 

Darkmass, I can't reply to the whole of your post right this minute but here are some initial thoughts.

First, I just wanted to say, the tips I was offering were aimed at Voxengo users generally and not at you because there's nothing I can tell you about Voxengo - as I've said before, you're way ahead of me on all this stuff!

I also think you have a better ear than I, or certainly more patience, because you pick up tiny details in sound that I would not have noticed, such as the feathery brushing of the strings you mentioned.

I've downloaded your latest sample and I think it's awesome - it has such clarity, such purity and such delicacy that it's hard for me to imagine how it could ever be improved upon.

I still think mine is slightly duller than yours, either because Steve's CD and LP masterings were slightly different or because my cartridge rolls off. Here's what I'm seeing, comparing my sample 01 with your sample 05 (I used block size 2048 for mine and 8192 for yours to reflect the different sample rates):



In this chart my 01 sample is green and your 05 sample is white. If anyone else is watching, that big green hump at the left isn't music, it's rumble from the LP because my 01 sample is in fact Steve's Rhino LP. The 05 sample has a bit of extra treble from 3kHz to 7kHz and my 01 has a big dose of sparkle added at the top end of the frequency range. I reckon the combination of less treble and extra high-end sparkle could account for the feathery brushing of the strings.

Here's an oddity, though. To repeat, as is clear from the hump at the left, the 01 sample is in fact the LP and the 02 is the HDCD. You thought 02 was the LP because 01 looked as if it came from the same source as 04, which was the HDtracks. It turns out that it's Steve's LP that looks similar to the HDtracks! But it must be a coincidence because obviously they are not from the same source in reality.

I'll just very quickly cover the 180 degree phase difference. I would use the term "inverted polarity". I don't know if you find this but I find nearly every time I compare the same album in LP and digital form, they have opposite polarity. I haven't yet found a reliable way of determining which is correct so I tend to go with the one that looks most convincing. In this case Steve's LP and the HDCD had opposite polarity and I decided Steve's LP looked the most convincing, so I inverted every track on the HDCD to match the LP.

I will have to stop now but will cover other points later. Goodnight!

 

OK, darkmass, I'm now moving on to the second part of your post. First, I need to make a point I should have made before.

As you know, of the two samples I posted, one was from Steve's Rhino LP and the other was from the HDCD. They are virtually identical because they have been EQ matched using Voxengo, but they are not completely identical for the following reason.

With vinyl, I find you almost always get a big hump in the sub-bass like the one shown in the chart above. This is vinyl rumble, not music - usually, it's particularly bad towards the outer edge of the LP and lessens as you get closer to the centre.

Naturally I don't want this hump included in the EQ match because it's non-musical and shouldn't be there. So when doing the Voxengo EQ match, I manually iron out any obvious rumble hump in the EQ curve taken from an LP, to prevent it being transferred to the target file.

So, if I have done this properly, you should find that the two samples I posted are identical except that 01 (which is an untouched recording from the LP) has the big rumble hump and 02 (which was sourced from the HDCD, then EQ matched to 01) does not.

I think this information puts the charts you produced in Sound Forge Pro in a very different perspective. For example, in your second chart, the blue line has an enormous hump in the low bass which must have puzzled you, since you thought it was coming from the HDCD. But as you now know, the explanation is very simple - you're just looking at the rumble in the LP!

Another thing that your charts show very clearly is the extreme top-end sparkle that's present on the LP mastering. I don't know whether this was added by Steve or subtracted in all the other masterings but in any event I have allowed it to transfer to the HDCD and as you see, both my samples follow each other faithfully in this respect (but differ quite noticeably from your samples).

Finally, we're left with the difference in brightness in the 3kHz to 7kHz range. My Voxengo chart immediately above shows this very clearly (and it was the first thing I noticed with my ears, before I'd looked at any charts). This difference is not so evident in your charts. However, in your charts, it's difficult to see with any precision what's going on in the 3kHz to 7kHz area and I wonder whether, if the charts were given more space to breathe, we might not see the same pattern as we see in Voxengo?

Just a footnote: my rule of thumb with block sizes is, the default 2048 for standard 44.1k sampling rate, 4096 for 96k sampling rate and 8192 for 192k sampling rate. That gives the best matches, I find. Best of all though, as you say, is to get the files to the same resolution before doing any serious EQ matching.

 

I'm here for just a brief second, back2vinyl, I'm way, way behind on getting to bed, and I still have miles to go before I sleep.

I can set parameters defining the spectrum analysis window pretty much however I want, so I'll do something with that. I do think it can show pretty much the same thing as Voxengo...and to a degree the "slid up" part of my comment was really a confirmation of the difference we both heard. Maybe just a different way of looking at the same thing.

On the rumble, the big hump did look to me like vinyl rumble, but I also had this preconception that if Steve had his hand in my DCC and the Rhino LP, they would have the same general characteristics towards the end of the selection. Unfortunately, that severely skewed my thinking. Now I know better.

In fact, I've just had my mind turned around in a LOT of ways. I needledropped side 1 (at 96k/24) of my Reprise vinyl of Blue. This thing is as original as it gets. Let's just say it was purchased new, and probably no more than a couple of weeks after the album was released. (Really, it may have been purchased no more than a couple of days after first release.) And what did I see in the end segment of "All I Want"? It had exactly the same bass string characteristics as the HDTracks and the Rhino LP. That rather makes me wonder where the source for my DCC and your HDCD came from--maybe it was an earlier version of the tape used for the initial LP. I find it easier to believe that things were added to the left track at some point rather than being subtracted at some point. I suppose the "subtraction" tech actually exists, however.

Oh, "inverted polarity" is the right term to use. Along those lines, my needledrop showed the same polarity as the HDTracks, my DCC, and the original polarity of your HDCD.

 

My main conclusion from all of this is that we're both getting extremely satisfactory results using the Voxengo CurveEQ needledropping technique. Our results aren't the same because each of us is using a different source but you're getting a superb recreation of Steve's DCC CD in a much higher resolution than the original whereas I'm getting a superb recreation of Steve's Rhino LP with all the vinyl imperfections removed. BTW mine ends up in the rather unusual resolution of 24/44.1 since I used a decoded HDCD as the target file!

Going back to the bass, it's very odd, what you've discovered. Let me see if I've got this right. On some versions, the bass (actually the bottom string of an acoustic guitar, we think) is spread across both channels, while in other versions, it's mainly in the right. The both-channels versions are: the original Reprise LP, the HDtracks and my sample 01 which is Steve's Rhino LP. The single-channel versions are: Steve's DCC and the HDCD.

If that's correct, it's interesting to note that both LPs are in the both-channels category. I believe it's normal practice for mastering engineers to centre the bass before cutting vinyl, to avoid the problems that can result from having the bass in one channel only. So it would make sense that both vinyl editions had centred bass frequencies. It would also make sense that the DCC and the HDCD, being digital formats, did not have centred bass, and this would tend to confirm that they were in fact taken from the original master tape, as one would hope. What's unexplained is why the HDtracks version has centred bass - the only logical explanation is that the HDtracks version was in fact taken from an EQ dub prepared for an LP release. This wouldn't be surprising - after all, only a minority of HDtracks albums are fresh masterings from the original master tape, and most are simply digitised EQ dubs taken from the record label archives - so this seems to me the most likely explanation.

 

Oh, you have put everything together. I was already impressed by you but, man, this is something else! It all makes perfect sense.

Thank you for the comments on my "05", I appreciate them, but I also have to say there is a lot I like about what you have done with your "02". And the vinyl qualities you've picked up (call it the characteristics of your cart, I don't care, it works) I wish I had in my result. To that end I've been working on CurveEQing the drop of my Reprise "All I Want" with the respective HDTracks file. Given the end segment characteristics, it's even a reasonably natural fit. It's been my first attempt at dealing with the rumble aspects, and it's like to have made me crazy. I'm a long way from developing your skill, though it could come in time...maybe. But after attempting to deal with the rumble with CurveEQ, I cheated and that worked for me. My Diamond Cut DC8 (software developed following engineering efforts to recover and restore Edison audio test pressings) has an RIAA filter package, and though I did nothing with the RIAA characteristics, that particular filter package contains a rumble filter that worked well for my Reprise. While my table and cart (not to mention the Reprise itself at this point in its history) have little to recommend them, I was able to pick up a trace of what I found appealing in your work. I even tried a hybrid version (50% DCC contribution and 50% Reprise contribution), furthering my "anti-KISS" tendencies, and I rather like it.

Okay, I had a try at conjuring up a Sound Forge spectrum analysis that would be somewhat similar to the Voxengo display showing "01" and "05" respective characteristics. It's NOT as clear as the CurveEQ showing, but it's something at least.



I've once again enjoyed our conversation--more than enjoyed it, I've also learned a lot. Should we pick up our conversation in the future, that would be a good thing. I myself am going to defer further Blue work till after I wrestle with a few of the projects that have been pacing impatiently in my queue. But at least I have a good clear roadmap for Blue that will finally lead to results I am completely happy with.

Till later.

 

This has been very informative for me and good fun, too. That's a nice chart, by the way, and I'm happy to see it accords nicely with the Voxengo result.

Best of luck with Blue and I look forward to comparing notes on other projects!