When I mix in Cubase, my mixed-down songs do not sound as good in playback as they did before I saved them. I listen before and after saving; in the same room with the same monitors; with the same quality conversion, and without any digital clipping on any busses. I usually mix 32 channels down to eight stems, and then into a stereo output buss.
Does analogue summing provide a more accurate representation of the multi-channel mix, an enhancement to the sound of the mix, or both? Can I get to the first option by simply capturing the pre-save monitor output with an accurate stereo A-D converter? And can most of the benefit of analogue summing be derived from creating stems in the DAW before saving, then routing just the stems out for analogue summing? Alternatively, can those benefits be gained by making all adjustments to channel levels in the software, and then running the software-adjusted signals through a summing device that makes no further gain adjustment? If so, are the gain controls on most analogue summing devices necessary?
Technical Editor Hugh Robjohns replies: Wow, that's a lot of questions to get through! Bear in mind that I'm not a Cubase expert, so I can only talk to you from the analogue summing perspective.
The first query, about an audible difference between a 'live' session and a replay from the saved file, is an interesting one. The audio components of a session in progress will typically be working within a 32-bit floating-point domain, whereas the output D-A will only accept a 16- or 24-bit fixed-point signal. So a conversion process is required which will involve some form of dithering. It is well known that some dithering algorithms have a more audible effect on some material than others — particularly where some kind of noise-shaping is involved. So it is possible that the differences you hear are the result of the conversion from the floating-point format to a truncated and dithered fixed-point one.
However, it is probably more likely that the differences you hear are due to the way you are monitoring the 24-bit mixed file. If you are using a different program to replay these files, that alone may explain the change — perhaps there is some kind of internal processing going on that has an audible effect. If you are monitoring through Cubase, it might be because the file is being transcoded back into a 32-bit floating point format within Cubase, and then converted yet again to 24-bit for the monitoring output. That kind of multiple conversion process is likely to result in a sonic degradation, no matter how subtle.
Moving on to analogue summing, I would say it generally sounds 'different' rather than 'better.' Analogue summing is inherently less technically accurate than digital summing (when performed correctly), but often those technical imperfections are musically and sonically attractive — in much the same way that the high levels of harmonic distortion created by some valve preamps are pleasing to the ear. So the answer to your question is: yes, in some circumstances analogue summing enhances the sound of a mix.
One thing that analogue summing provides for free is mix headroom. A huge number of people 'mixing in the box' do so with very high peak levels on each source channel; typically well up above -6dBFS. This results in the mix-buss accumulator having to accommodate signals with sample values above 0dBFS and the master output fader has to be pulled down to compensate. Although this absence of mix headroom shouldn't be a problem in theory, it appears that some systems (and plug-ins) do have a problem with it, and a great many users have found that leaving more headroom on the source channels avoids the problem completely, producing far sweeter-sounding mixes as a result. The old analogue concept of having 20dB or so of headroom was arrived at for a very good reason and has stood the test of time. The concept translates perfectly to the digital world, with the same practical and sonic advantages.
You asked if you could capture the pre-save monitor output with an accurate stereo ADC. Well, yes, you could, but it won't improve anything on a technical level. The output from Cubase is transcoded internally from 32-bit floating point to 24-bit fixed-point samples. These are then passed to the digital output and in turn to the D-A converter that feeds your monitoring. Routing that analogue signal back through yet another A-D converter to get it back into the digital domain gains nothing at a technical level and is likely to degrade rather than enhance the signal. Of course, it is possible that you may like the sonic character imposed by that degradation, in the same way that people like valve preamps.
Next you asked if most of the benefit of analogue summing can be derived from first creating stems in the DAW before saving. The answer is a qualified yes. The advantages of mix headroom and the sonic colouration will be the same whether you mix individual channels or stems in an outboard analogue summing box. However, if the mixing of stems created within Cubase produced samples greater than 0dBFS, the quality loss resulting from a lack of mix headroom in the digital domain would remain.
Finally, you asked if most of the benefit of analogue summing could be derived by making all adjustments to channel levels in the software, and then running the software-adjusted signals through a summing device that makes no further gain adjustment? The answer here would be yes again, because by adjusting the level of source channels before mixing, the advantages of mix-headroom are preserved, and the mix may well sound better as a result.
The gain controls on those analogue summing devices that include them are there to allow the user to optimise the gain structure for whatever source equipment is being used. Some D-A converters operate with peak levels of +26dBu, while some only reach +12dBu, so an input gain trim to optimise the gain structure is both necessary and important if you can't determine the source input levels.