Harrison Mixbus - Virtual Harrison Mixer plus full-featured DAW for OS X

[hugol]

Come on Seablade please do stop exaggerating!

There is very little you can do differently with this stuff. It is just basic maths and almost all DAWs use either 32-bit or 64-bit floating point for all operations internally. There really isn't anything in it. Sure PT-HD is different, let's not go there.

Precision isn't the issue or variable you're making out. Summing and panning and gain staging are very very simple, nothing complicated and very little that could be done differently. Yes we could dither liberally, fine no bad thing.

Mixbus may be a perfectly fine product, that's cool, but there is a lot of marketing going on here. It's like trying to get straight answers from a politician around here!!

[Tui]

Originally Posted by Seablade

You said yourself, the moment you added in panning you didn't null anymore.

I said, with panning engaged, I could get the nulling down to about -60dB.

If you know anything about mixing, as a rule of thumb, anything below -40dB is inaudible in a complete mix. If you don't believe me, try it out.

Back to my test. What little remained of the signal, which was heavily distorted and barely recognisable, I put down to minute differences in volume between Logic and MixBus when panning. Interestingly enough, on this thread

http://www.gearslutz.com/board/new-p...uno-6-a-2.html

I was asked to do another null test, this time with Logic's ESX24 sampler and Kontakt. Initially, identical signals played back through both samplers did not null completely. I had to lower the signal in Kontakt by -0.21dB, to achieve complete cancellation. If someone were to perform a similar test, but without this small adjustment, they might conclude the signals were different.

I'm happy to accept that panning in MixBus and Logic differ ever so slightly, with respect to volume levels, however this has no bearing on the resulting sound quality.

Again, we're talking about a residual, not completely nulled signal around -60dB. If Harrison were to base their advertising on this microscopic difference, they would make themselves look rather foolish.

[Tui]

Originally Posted by HugoL

Mixbus may be a perfectly fine product, that's cool, but there is a lot of marketing going on here. It's like trying to get straight answers from a politician around here!!

Let me help you out.

"MixBus is change you can believe in!"

[Tui]

Originally Posted by PascalC

i'm sorry but as a non professional user, 80 dollars are NOTHING for such a nice soft + plugin. NOTHING.

You wont hear me complain about the price for what you quite rightly label a plug-in. Albeit, a plug-in you can't plug into anything.

[Seablade]

Originally Posted by buncker

hmm... I know something about programming =)
There is no any problem to implement summing with, for example, 256 bit wide mantissa to store result. It will no be a hard load for modern cpu (SSE is good thing =) )
any student will code it without a problem...

Correct, as I said repeatedly, not an example for programmers It is easy to work around that particular example. However when dealing even with floating point math you will often have instances where you are forced to lose some precision, heck even that basic example had one.

Now consider how that loss of precision can add up rapidly when talking about thousands of calculations(Understatement of the year) from 40 to hundreds of tracks with complex automation curves.

So I don't think that there is some "magic" in tracks summing.
stereo is just two mono tracks =)

That is what people seem to be missing. It isn't that there is a single outright 'magic' in Harrison's approach, but rather that the particular way they are implementing the summing engine is particular to them to gives a pleasing sound in the end.

And if you know, when it works better comparing to "usual hosts" - why just not tell us?

Why we need magic?
Thanks!

For the sake of clarity, Fixed small typo above.

I personally am not of the opinion you 'need' magic. I would think it would have been obvious by now, you NEED whatever works for you. As I said, I find Mixbus works for me because I can get the sound in a mix I want with speed and ease.

I do believe this in part has to do with the signal processing built in, the compressors, EQ, and Saturation. But I also believe that it has to do with how they have implemented their particular summing engine, which not only encompasses the '1+1=2' but also encompasses how the algorithms are set up to handle things like panning, automation, etc.

Seablade

[Seablade]

Originally Posted by HugoL

Come on Seablade please do stop exaggerating!

There is very little you can do differently with this stuff. It is just basic maths and almost all DAWs use either 32-bit or 64-bit floating point for all operations internally. There really isn't anything in it. Sure PT-HD is different, let's not go there.

Precision isn't the issue or variable you're making out. Summing and panning and gain staging are very very simple, nothing complicated and very little that could be done differently. Yes we could dither liberally, fine no bad thing.

Mixbus may be a perfectly fine product, that's cool, but there is a lot of marketing going on here. It's like trying to get straight answers from a politician around here!!

Actually you missed my entire point of my post. Math in computers is not perfect, and where, when, and how to work around that is what is going to be part of what define the summing engine.

Seablade

EDIT: ...is going to be PART of what defines...

[Seablade]

Originally Posted by Tui

I said, with panning engaged, I could get the nulling down to about -60dB.

If you know anything about mixing, as a rule of thumb, anything below -40dB is inaudible in a complete mix. If you don't believe me, try it out.

Back to my test. What little remained of the signal, which was heavily distorted and barely recognisable, I put down to minute differences in volume between Logic and MixBus when panning. Interestingly enough, on this thread

http://www.gearslutz.com/board/new-p...uno-6-a-2.html

I was asked to do another null test, this time with Logic's ESX24 sampler and Kontakt. Initially, identical signals played back through both samplers did not null completely. I had to lower the signal in Kontakt by -0.21dB, to achieve complete cancellation. If someone were to perform a similar test, but without this small adjustment, they might conclude the signals were different.

I'm happy to accept that panning in MixBus and Logic differ ever so slightly, with respect to volume levels, however this has no bearing on the resulting sound quality.

Again, we're talking about a residual, not completely nulled signal around -60dB. If Harrison were to base their advertising on this microscopic difference, they would make themselves look rather foolish.

You tested a very limited number of tracks with no automation whatsoever. The more those minute differences add up, the larger differences they will make.

Seablade

[PascalC]

Originally Posted by Tui

You wont hear me complain about the price for what you quite rightly label a plug-in. Albeit, a plug-in you can't plug into anything.

ok it's a pack, a complete pack. But imo you don't know what OPEN SOURCE philosophy IS. Ardour needs Harrison and Harrison needs Ardour. For 80 dollars you have both. Where is the problem ? / Use jack + other softs and you can connect everything to everything IRT.

bye.

[Tui]

Originally Posted by Seablade

You tested a very limited number of tracks with no automation whatsoever. The more those minute differences add up, the larger differences they will make.

Seablade

So now it's not the summing anymore, or the panning, suddenly it's automation and increased track numbers that will make MixBus sound different from other DAWs?

Jeez.

I feel like I've run into a cult.

[Seablade]

By the way, since it seems I need to repeat things i said over and over again throughout the thread...

I do NOT work for Harrison, I am just another user of Mixbus in the end.

Seablade

[Seablade]

Originally Posted by Tui

So now it's not the summing anymore, or the panning, suddenly it's automation and increased track numbers that will make MixBus sound different from other DAWs?

Jeez.

I feel like I've run into a cult.

Lets see... how many times have I said this, I know it has been several times in the past few days(Quite a bit by me). And quite a few before this in the thread...

The SUMMING ENGINE is more than simply adding two numbers together. I would consider it the handling of anything that will modify the signal mathematically. This includes how to handle automation curves, panning, etc. The number of tracks has to do with the fact that you noticed already differences while panning, albiet at low level(How audible they are depends on many factors, but for the moment we will stick with you noticed to low level to really notice). But now take those and have them increased as the number of tracks increase. Small differences add up and can become audible is my point with that.

Seablade

[hugol]

Originally Posted by Seablade

Actually you missed my entire point of my post. Math in computers is not perfect, and where, when, and how to work around that is what is going to define the summing engine.

Seablade

No I think I got it and that's fine. I just don't believe these things to be the issues you are making them out to be, certainly not for simple maths on the mix bus. Providing you keep your numerical representations consistent I don't see the issue - it's converting between different types of numerical representation that usually causes the problem.

By the way your 2/3 vs 2.0F / 3.0F example was hardly surprising. Of course I can't represent a fraction with Integer maths. It's also misleading as if I was representing a sample value in Integer my scale is going to be completely and utterly different. i.e. in FP 1.0 may represent 0dBFS - but in 32-bit Integer maths it might be 2^32.

Also fine some complex transforms may benefit from greater precision maths, it's true, but individual plug-ins doing these operations are free to use greater precision maths internally.

[PascalC]

When I open the same sound with Sonar or Samplitude without using ctrls or vst or plugs etc, i have the same sound at the end. Are they bad DAWS ? No. Ardour is not superior to Samplitude but i have another "color" when i use it and i like it. (i like sam too) : thx to the "plugs" or features created by Harrison imo.

[Seablade]

At this point I am fairly certain that we are going to start going in circles in this conversation. I have said what I had to say, others have said what they have had to say. In the end it comes back to using what works for you, as I have also said repeatedly.

Seablade

[Tui]

Originally Posted by Seablade

Small differences add up and can become audible is my point with that.

Seablade

Yes, however when summing, there is no difference. Nada, zilch, nothing. You can record in MixBus at unity, or -3dB, -19dB, or any other value, you'll end up with the same sonic result as with other DAWs.

If you still want to include MixBus' EQs, compressors and saturation effect in your mixes, you can use MixBus as if it was a plug-in. However, it's quite fiddly to do so, and Jack uses up more CPU cycles than I'd be happy with.

[Seablade]

Originally Posted by HugoL

No I think I got it and that's fine. I just don't believe these things to be the issues you are making them out to be, certainly not for simple maths on the mix bus. Providing you keep your numerical representations consistent I don't see the issue - it's converting between different types of numerical representation that usually causes the problem.

By the way your 2/3 vs 2.0F / 3.0F example was hardly surprising. Of course I can't represent a fraction with Integer maths. It's also misleading as if I was representing a sample value in Integer my scale is going to be completely and utterly different. i.e. in FP 1.0 may represent 0dBFS - but in 32-bit Integer maths it might be 2^32.

Keep in mind, I specifically did not go into a very detailed example as I specifically was trying to keep it simple.

Also fine some complex transforms may benefit from greater precision maths, it's true, but individual plug-ins doing these operations are free to use greater precision maths internally.

Correct, but then they also don't have control over what the DAW does with those results and thus can't really replace a summing engine completely.

Seablade

[Seablade]

Originally Posted by Tui

Yes, however when summing, there is no difference. Nada, zilch, nothing. You can record in MixBus at unity, or -3dB, -19dB, or any other value, you'll end up with the same sonic result as with other DAWs.

Which as Ben mentioned, means that neither are broken. This is a 'good thing'

But you are still referring to summing as merely the addition of multiple signals. This is a semantic difference that can mean a great deal. I consider the summing engine to mean much more than that, and thus why your pan results in particular interested me.

Seablade

[buncker]

Let me say the final word.

please read this -

Originally Posted by Harrison Press

Traditional DAW mixers, designed by companies with no pedigree in music mixing, all suffer from well-known flaws. Harrison eliminated these flaws by completely redesigning the DAW's internal mixing engine and applying proprietary algorithms.

it sounds ugly.

All I want to know - what is these "well-known flaws." ?

If anyone can deliver that rare knowledge, I will be happy.
Please, without words "may be", "somehow".

About summing math, sorry - but it is not a rocket science
Precision is the key. multiply and add first (if you have a headroom, and we have it today), divide last (so that error will not be accumulated)
And today we can have any precision we need.
intel core2duo is not a Z80 or 8086. But even on z80, offline mixdown with ANY precision you can dream of could be done...

I have to add, that DAW could use its own math function, not to be constrained by processor hardware multiplier / divider.

So - math in computers is as perfect is it needs to be.

PS old consoles rarely had automation =) may be manual fader-riding? =) So if we are talking in terms of "old analog sound" - it is "static" mixing in general.

[bluesky]

Originally Posted by buncker

Let me say the final word.

please read this -



it sounds ugly.

All I want to know - what is these "well-known flaws." ?

If anyone can deliver that rare knowledge, I will be happy.
Please, without words "may be", "somehow".

About summing math, sorry - but it is not a rocket science
Precision is the key. multiply and add first (if you have a headroom, and we have it today), divide last (so that error will not be accumulated)
And today we can have any precision we need.
intel core2duo is not a Z80 or 8086. But even on z80, offline mixdown with ANY precision you can dream of could be done...

This was posted two pages ago by Ben Loftis in response to Tui:

From our FAQ:
Quote:
What are the "well known flaws" of other workstations?
It is our opinion that the "gross" defects in many workstations include internal clipping, incorrect dither, multiple format conversions, out-of-control gain stages causing plugins to work outside their intended range, routing choices that cause latency/timing errors, Inability to see meters such as compressor gain reduction without opening the plugin dialog, bad ramping of plugin coefficients, and poor user-interface integration.

Of course different workstations will exhibit these problems to different extents. Our goal was to design a mixer using the "best practices" that we have developed over the course of 30 years. Multiple subtle design decisions, accumulated over a long period of time, are required to make a truly world-class mixing engine.
Mixbus is the result of a LOT of effort to alleviate some issues that bothered us about other DAWs (obviously, to different extents depending on the DAW). We haven't made any claims that the product doesn't deliver on.

Best,
-Ben

[hugol]

Originally Posted by Seablade

Correct, but then they also don't have control over what the DAW does with those results and thus can't really replace a summing engine completely.

True, but I don't think there is a need to replace the summing engine in any (?) DAW - as that bit isn't broken unless you're after analogue mix-bus emulations. I was of course talking about e.g. high quality EQ plug-ins that are doing a lot of transforms where a loss of precision along the way could be an issue.

Anyway Mixbus seems to be a fine and popular product, which as I understand it has been fine tuned in a number of areas. That's totally cool.

It's just the marketing hype and vagueness that is annoying... and it's probably unfair to direct this solely at this product, but I can see why frustration exists - the pro-audio market seems worse than the audiophile market at times! More snake oil and marketing than science sometimes.

[Seablade]

Originally Posted by HugoL

True, but I don't think there is a need to replace the summing engine in any (?) DAW - as that bit isn't broken unless you're after analogue mix-bus emulations.

'Broken' I suppose can be different things to different people. If it works for you then it probably isn't what you would consider broken. But if a different implementation works better, how would you describe it?

I am not certain I would consider any summing engine 'broken' myself that is true, but Harrison obviously felt the need to implement their own rather than rely on what was already there, even in Ardour. What exactly goes on in there I can only guess, as I don't have source code for it. But I can see different ways that implementation can make a difference, even very slight, that may be beneficial, which was of course my only point... that the implementation can be different, possibly better, from what might be initially obvious.

Seablade

[bluesky]

Originally Posted by Tui

So now it's not the summing anymore, or the panning, suddenly it's automation and increased track numbers that will make MixBus sound different from other DAWs?

Jeez.

I feel like I've run into a cult.

Automation and track numbers ARE a part of the summing process. This is also a big issue with Ableton live and all the allegations concerning the quality of Live's summing engine. Setting all parameters to unity also resulted in complete nulls between hosts. Yet, many claimed that Live sounded inferior (and indeed many still do). At some point in the discussion at the Live form, Ableton made known that the way a sequencer handles automation does indeed effect the summing engine and the sound quality. I'm not sure if they addressed this issue, but they did make some changes to their automation implementation.

I am an Ableton Live and Mixbus user. A hobbyist, I am not affiliated with either company in anyway. When I bought Mixbus I also ran some tests, and at unity settings could not tell a difference between Live and Mixbus. However, when I began to mix, I noticed differences in Mixbus favor.

In my opinion, there is nothing that indicates that Harrison is being deceptive and fraudulent. In fact, I find their responses to be very professional and there product to "deliver" at least as far as I am concerned. For $80, I find Mixbus to be a bargain, much more so than other plugins costing 3-4x as much that make similar claims.

As others have mentioned in this thread, actually mix in Mixbus. Until then, our null tests really doesn't mean anything, imho.

[Seablade]

Originally Posted by HugoL

It's just the marketing hype and vagueness that is annoying... and it's probably unfair to direct this solely at this product, but I can see why frustration exists - the pro-audio market seems worse than the audiophile market at times! More snake oil and marketing than science sometimes.

I can understand this absolutely. Then again, to flip this around, sometimes there is also a concentration on the technical side of things in audio, rather than just getting the work done. One of the live mixers I am working with right now has this problem, he wants to understand things and it tends to get in the way of him actually getting the work done when he needs to during a show. Many people have said his work has improved since I started working with him and getting him to try something rather than think through if it will work or not, which took longer than him actually just trying it.

I think the same thing can apply here as well. There is a desire to understand exactly what is happening, that I can sympathize with. But as others have posted, sometimes it is best to dive in head first and try the thing for work, and see what the result is. For instance, if half the time spent debating in the past several days over the details of the summing engine was instead spent on mixing down a song or two and seeing how it sounds, how might this have affected things? In the end it comes down to, will it do the job you want it or not, possibly better than what you already use? For me, and for others apparently, that answer is yes. Does that mean it will be for everyone? Absolutely not, and I can understand that. But the difference is actually using it to see how it sounds, rather than just measurement.

Seablade

[hugol]

Originally Posted by Seablade

I am not certain I would consider any summing engine 'broken' myself that is true, but Harrison obviously felt the need to implement their own rather than rely on what was already there, even in Ardour. What exactly goes on in there I can only guess, as I don't have source code for it. But I can see different ways that implementation can make a difference, even very slight, that may be beneficial, which was of course my only point... that the implementation can be different, possibly better, from what might be initially obvious.

Ok cool, thanks for the honesty. So you're not sure what they've done either.

I definitely agree that are always variables and things that could be slightly different. Although I'm struggling to think of that many. Still we're not going to get actual answers from Harrison.

And to be fair if people love using the product and are getting great results, for whatever reason that's the key thing.

Edit: Haha - I see we kind of got to the same conclusion in the end.

[Seablade]

Originally Posted by HugoL

We're not going to get actual answers from Harrison though.

Just because you said it, not trying to pick on you and know you probably realize this...

What do people want exactly? An exact description of the math that Harrison is using? That wouldn't be beneficial to Harrison at all, because then it just gets reimplemented by someone else and they have lost what amounts to years of research on their end. That is likely why they have been vague I think, as the more detail you give the more you help your competitors as well.

All that being said, I can say I have had several conversations with Ben that have been enlightening for me, and I do think they try to help people understand when they can.

Seablade

[dawhead]

Originally Posted by Tui

No I didn't, not once. Reread my posts.

Is this your typical rhetorical style? Completely ignore the substance of the post? Let me repeat: "Nobody has suggested that MixBus sounds any different with everything at "unity", a signal that isn't close to clipping and all parameter static. You've set everything at unity, ignored the pan, ignored the saturation, ignored the parameter ramping, . What were you expecting? distortion? phase shift? bass boost? what?"

What is it that you expect to determine with you test?

There are a lot of people in this thread who have gotten excessively tied up with term "summing". Its used with varying levels of precision to mean different things by a lot of people. If you imagine that your test is going to establish whether two DAWs both add 1+1 and get 2, then sure, its a reasonable test, and it establishes that both Logic and MixBus can do math, and do it in the same way.

But go talk to someone who actually builds consoles or a DAW, and the term "summing" is frequently used to mean a lot more complex and "distributed" than adding up signals. I had a long discussion with a fairly well known plugin company several years ago about a new "summing engine" (their terminology) that would still have added 1+1 to get 2, but the entire architecture was designed to generate an entirely different sound given certain kinds of signals and mixing techniques. However, had you done your test with that (hypothetical) engine and Logic, you would also have gotten full cancellation. The phrase "summing bus" has a hand-waving quality to it that doesn't adequately capture what is going on in a contemporary console - something Ben Loftis has tried to explain repeatedly. The nuances of a real console are not demonstrated by static "unity" mixes - you will frequently hear engineers comment on how it "feels" as parameters are modified, or signal levels change in certain ways. The details of a compressor curve. Many DAW users, because of their experience with a plugin architecture, have come to imagine that you can build a console-like experience merely by stacking plugins one after the other in a track, and feeding the tracks to a bus, and that's all there is to it. The resistance by some people (not necessarily you) in this thread to the claims of people who work for one of the oldest, most respected console makers that this is not true is ... well, its quite remarkable.

So perhaps you could explain again what it is that you think that your cancellation test establishes? Can you imagine how BAD it would be if ANY DAW failed the test that you performed? (Some used to, by the way - they've been fixed subsequently)

[Kenton]

Originally Posted by Tui

Most of my tests were null-tests.

#1 I streamed 7 very busy stereo tracks from Logic into MixBus. The tracks were complete songs, consequently the resulting summed signal was total cacophony. However, I reckoned if MixBus really does something special to the signal when summing, it should be immediately apparent with complex material.

Result: The 7 stereo tracks streaming from Logic nulled completely with the same tracks going live through MixBus.

#2 I bounced the 7 tracks realtime in Logic, resulting in one bounce track. I then recorded the same 7 tracks simultaneously in realtime into MixBus. I imported the Logic bounce track into MixBus, and lined it up with the 7 stereo tracks I recorded earlier. I routed the 7 tracks to the MixBus Master bus, to see if summing inside MixBus is different from that in Logic. I routed the Logic bounce track directly to my RME HDSPe Mixer (= TotalMix), bypassing the Harrison "summing engine".

Result: The 7 stereo tracks summed inside MixBus nulled completely with Logic's bounce track.

I have just re-read this and I have to say that I think you have made an error somewhere in your test.

Are you saying that you summed 7 tracks through the Harrison Master Bus (Tape Saturation and all) and managed to get a null with the same 7 tracks running in Logic?

I don't think so.

You should be getting a residue that is (at the least) the distortion components produced by the Tape Sat on the Master Bus (which you can't bypass even when fully turned down).




Sorry, I call shenanigans.

[Tui]

Originally Posted by HugoL

It's just the marketing hype and vagueness that is annoying...

Well, yes. The thing is, if MixBus was the product of some no-name outfit, I wouldn't have bothered. Spend 80 dollars on something I can't demo, yet it promises to revolutionise the DAW market?

Please.

However, if a prestigious company like Harrison tells you they've done something, you'd expect it to be true. Right?

Wrong, as it turns out.

[Tui]

Originally Posted by Kenton

I have just re-read this and I have to say that I think you have made an error somewhere in your test.

Are you saying that you summed 7 tracks through the Harrison Master Bus (Tape Saturation and all) and managed to get a null with the same 7 tracks running in Logic?

I don't think so.

You should be getting a residue that is (at the least) the distortion components produced by the Tape Sat on the Master Bus (which you can't bypass even when fully turned down).




Sorry, I call shenanigans.

I didn't say any of that. Where did I mention I engaged tape saturation? I didn't, so please stop making stuff up (seems to be a popular pastime on this thread).

But since you mentioned it, I noticed an inconsistency during my test. On the 4 dedicated mix busses, tape saturation can always be turned off completely. However, on the master bus, I found when I started with my testing, on the first day, tape saturation was always applied to a small extend. On the following day, this did not occur.

However, I do remember reading that MixBus' behaviour had been changed during the course of later revisions, and that V 1.2 (which is the only version I ever had) does indeed allow for complete disengagement of the saturation effect on the master bus.

At any rate, signals passed through audio channels, busses and mix busses do cancel completely.

Try it out. It won't hurt, I promise.

[Tui]

Originally Posted by dawhead

IYou've set everything at unity, ignored the pan, ignored the saturation, ignored the parameter ramping,

No I didn't. Do you have a comprehension issue?