I know there has been some ferocious debate over at Gearslutz over the benefits of analog summing, and forum regular Mixerman is one of its most vocal proponents. I read his book "Zen And The Art Of Mixing" and he waxes lyrical about it in there as well. I'm not experienced enough to have any kind of credible opinion on the matter, but one point he makes in the book is that not all analog summing is the same: 
 

"There's one other way to gel your mix, and that's by slamming your 2-bus with the mix. This method allows the electronics on your 2-bus to act like a limiter. If you're mixing in the box, this isn't an option. You have to have an analog 2-bus to do this. Furthermore, not all 2-buses are created equal. The 2-bus on an SSL will crumble in short order, yet you can absolutely hammer the 2-bus on most of the 80 series Neve desks to the point that the meters are pinning at all times as you work."

 
 He also recommends Slate's VCC as an alternative to those without access to an analog 2-bus. How seriously you can take a plugin endorsement from someone like Mixerman is up to you I guess. The book's an interesting read anyway. 
 
The other side of the coin was expressed recently by Graham Cochrane of The Recording Revolution fame, who wrote an controversial article called Analog Summing And Why You Shouldn't Care. I guess his point is not that analog summing is useless, or bad, but that if you don't have access to it then you shouldn't worry in the slightest because you can still get the same results. That article had Mixerman taking a swipe at him in this Gearslutz thread....
 
Anyway it's definitely a controversial subject for sure, along the lines of CD versus vinyl in the audiophile world. 
 

That is rubbish about the SSL being driven hard on its two buss and crumbling in short order. The SSL can produce something like +24 dBu at its output before clipping. So that is 20 dB of headroom over +4 dBu. That is NOT crumbling in short order.
 
You might push the two buss compressor in an SSL in order to get dostortion but if you let that one off though the mixer itself has massive headroom.

Well I guess it depends on what Mixerman means by "crumbling." 

Jeff Evans
I agree Drew re the blind testing and have not taken this up to that level. I have only felt this was the case. I have checked in terms of any processing etc and it still happens when there is no additional processing in the digital mixer, just summing.  I agree that summing digitally should sound the same but to my ears it simply does not.

 
If it doesn't sound the same and you aren't imagining it (and I'm not insinuating you are), then I'd expect there must be something more than just summing going on. Because unless the Yamaha is really noisy, well, 2 + 2 = 4 and all that. 
 

  I would go so far as to say that a blind test might actually reveal it.  It might just take the listeners a little while longer to hear the differences though I would imagine.
 
The good people at Yamaha are not incompetent and I am not imagining this either. Dave O’Donnell has the exact same opinion (I am sure you respect him)  

 
Don't take this wrong, but I don't respect anyone - or at least try not to.   You can't claim to be objective if you let the people you like and/or respect (or agree with) slide. The point is that casual listening tests don't produce reliable data regardless of who is doing the listening. 
 
It is interesting though.

The Yamaha is noiseless trust me on that. There is no processing either of any form going on. The mixer that I use is also a different model to the one Dave O'Donnell used as well.
 
I do not know anything about Dave O'Donnell and have never heard about him until I read the article about James Taylor's new album.  So I am not biased in anyway towards him.  But I guess James Taylor would not have employed him if he was not good.  I just found it interesting he made the same observation as I in regards to summing stems digitally out of a DAW into a Yamaha digital mixer.  It just sounds better.  When I get time I might organise a blind test for sure.
 
What I might do is upload the two files and let you guys hear for yourselves.  It would also be good to try this with a different digital mixer as well to rule out what you are saying about the Yamaha.

I would say that I come from the school that not everything has to be proven (or thoroughly investigated) if we can accept that we aren't 100% sure of things and behave accordingly.
 
But, and it's a technical nuance kind of point, if the Yamaha isn't just adding the samples together and it isn't an attempt to emulate analog summing somehow and it isn't just noise/errors from calculating at too low a bit depth or bugs and you aren't imagining it or hearing something unrelated to the summing, then some other processing must be going on by definition. Because if one isn't just adding the samples together then one isn't doing "summing". The point being that there's no magical DSP cloud when it comes to summing - the only difference is that you can do various programming things that can affect the level of calculation error. But something is very wrong if audible errors are occurring from adding a few dozen 24 bit numbers together.

Jeff Evans
That is rubbish about the SSL being driven hard on its two buss and crumbling in short order. The SSL can produce something like +24 dBu at its output before clipping. So that is 20 dB of headroom over +4 dBu. That is NOT crumbling in short order.
 
You might push the two buss compressor in an SSL in order to get dostortion but if you let that one off though the mixer itself has massive headroom.

I've mixed on nearly every console you can name, and if you push the 2-bus on an SSL E, G, J, or K, the mix will crumble. Like you need to be careful about it because the 2-bus can't handle excessive level. You can quote specs all day. All I can tell you is what happens, and it's certainly no secret. You can spank a Neve8068. You can pin those meters all days long. You can't do that on most SSLs. 

As to analog summing, yes, it makes a big difference. IF YOU HAVE THE SUMMING IN YOUR ROOM AND YOU'RE MIXING THROUGH IT. If you just have someone run some tracks through a summing mixer, that's just a waste of time and money. The benefit is in how it helps you mix. It's a process benefit, which manifests into a results benefit. But it isn't something that you just do blindly and can expect will improve your results without mixing through it. If you get more depth, width, and punch happening while you're mixing, it makes mixing easier and faster. This keeps you focused on what's important, the music. 


It's also true that I recommend the Slate VCC if you can't afford a summing box. And if you're a hobbyist, you don't need any of that stuff (although the VCC isn't all that expensive).

I don't officially endorse Slate products, although I may as well given my relationship with Steven and everyone over at his shop. But I can tell you, I don't ever recommend products for money (believe me there's no money in that!). I recommend what I use and believe in. Even with my stance on summing, I'm super clear that I think everyone should TRY IT for themselves, and I lay out a specific methodology for it. My beef with Graham's article is that he has never even tried summing (which he admitted) yet he's telling us all how superfluous it is. Please. If you haven't tried something you have no business discussing it or telling others they don't need it. I suppose his message is that all that's important is the music. Do you really need someone to tell you that you can record great music without analog summing? My advice in regards to summing is for people who get paid to mix and want to make their life easier in the process. It's not about whether artists and musicians can make music that touches others regardless of how they sum.

New book coming out. #Mixerman and the Billionheir Apparent. Google is your friend.

Enjoy, #Mixerman
 

Yes that is interesting.  I have actually repaired these consoles and then after run tones through them and measured output levels and looked at the signal on a CRO for example.  In fact the SSL spec is something like output before clipping is  +26 dBu which is enormous.  It is one of the reasons why we pay so much for this level of console design.  They have massive headroom.  Have you tried this without the 2 buss compressor on though.  That does not have the same headroom as the mixer does by itself.  It could be the bottleneck.
 
But I must say I have never pushed a mix really hard in a console like that so I guess I take what you say is interesting.  Of course tones are one thing but a full mix is another.  The fact an SSL can’t go as high as the Neve before distorting may be due to other parts of the signal flow perhaps.  I have always been someone that gets a full mix producing around the nominal rms level at the output eg + 4 dBu. I am in the clean, transparent camp and not pushing the console.  It certainly can do that though if you want it to.
 
I have done lots of mixes both ways and find in the end I still end up with a very similar result.  It is nice to mix on a console because the workflow is quicker for sure.  But I don't agree it is necessarily a huge amount easier.  I find mixing ITB very easy too.
 
In reality it is all about how you move the listener emotionally at the end of the day and nothing else.  If someone is moved and they love the music they are not going to be wondering how you summed the mix believe me.  It is irrelevant.  But if analaog summing gets you there easier then I say go down that path for sure.  I don't mind analog summing because I have done tons of it too in the past and I never fail to get a great mix that way either.  But I find I can still great a great mix without it.  What is wrong, is to say that you can't great a great mix unless you do it.  Because you can.
 
What I am finding interesting though is the fact that summing stems digitally in a digital console actually sounds better to me than ITB and that one is hard to explain.  Got any ideas on that?  The Yamaha digital console I am using is doing no processing or anything yet it just sounds slightly different for some reason.  If I can I will try this on another digital mixer and see what happens.  As Drew points out there may be something going on in there I don't know about.

According to some engineers, e.g. Ken Scott, it’s not so much about analog summing being so good, it’s more about digital summing not sounding right. A digital recording or sample of a single instrument can sound perfectly fine, but keep adding together more and more tracks digitally and it becomes increasingly harsh and edgy.

Kev999
According to some engineers, e.g. Ken Scott, it’s not so much about analog summing being so good, it’s more about digital summing not sounding right. A digital recording or sample of a single instrument can sound perfectly fine, but keep adding together more and more tracks digitally and it becomes increasingly harsh and edgy.

No. This is nonsense. Digital summing with an appropriate level of precision does not add distortion* or change frequency or phase response or anything else like that. And though I've argued that with 32 bit summing any calculation errors will never be audible, with Sonar we don't even need to have that discussion because if one turns on the 64 bit double precision engine it's simply impossible for any errors to make it to a 24 bit output.
 
Note that this does not mean analog summing can't sound better, but if it does it's because it's adding pleasing analog coloration - not because of any fault with digital.
 
 
*EDIT: with fixed point summing it is of course possible to clip, but I don't know if any currently available ITB systems still use fixed point. With floating point you can't clip when summing, but if you don't attenuate to <= 0dBFS you will clip when converting to 16 or 24 bit. But I would call clipping of this sort "user error" rather than a problem with digital summing.
 
And knowing Jeff, let's just say I find it kind of unlikely he's clipping his Yamaha.   So if there's anything going on there with the Yamaha I might guess some kind of background processing or something, but I don't really think it makes sense to speculate - if we really want to know enough to be worth the bother, we should just run some test signals through and see what (if anything) is up.