A Confused streaming system - Mutec / SOtM Ultra

[Confused]

I'm not even sure if it is beta 14 that rocks, maybe. But beta 14 is the first time that AIR was working and the first time I have tried Roon. And the combination of beta 14 with Roon/AIR sounds better here than Roon with the Mutec kit. So it can be Roon/AIR that brings the difference.

Yes, this makes sense.  When I first tried Roon integrated AIR with the Pro, I was very surprised just how good it was.  OK, we can discuss is source A better than source B until the end of time, but I think that in absolute terms, Roon Integrated AIR is very good, plain and simple.

Also, looking at the Munich 2018 thread, it seams we now have confirmation from Devialet that the Pro's are utilising the superior clocks that are used in the CI board.  So this should take Roon integrated AIR up yet another level.  Of course, the better clocks may help improve sources via AES/EBU or USB as well, but whatever you are trying to do, it is always up against decent opposition when comparing with Roon AIR.

[Confused]

Time for an update.  Events have been progressing very slowly, in recent weeks I have only been able to get any serious listening time at the weekends, and some weekends (last weekend being a good example), the listening time has been fun, with company, but not that serious!  Meanwhile, things have been progressing, albeit slowly, as I have been trying various “HAF” convolution filters as provided by Thierry’s Home Audio Fidelity service.  

Let’s jump back a few weeks and summarise key events.  This is a long thread and many of my trials and tribulations have been dead ends or distractions, the following is a summary of the important bits.

I added the SOtM bundle, first impressions were disappointing, the SOtM kit sounded a little thin, treble and presence range seemed too prominent.  After much analysis, I could tell that the SOtM kit sounded good, but I did not like the overall sound.  I had quite simply stopped enjoying my system, with everything sounding a little thin and too bright with the SOtM kit in place. I was convinced that the SOtM kit would be measurably brighter when compared to AIR or my old microRendu.  I measured using REW and a calibrated UMIK-1, and found that AIR, the mR and the SOtM kit measured more or less identically in terms of the frequency balance.  Let’s be clear here, subjectively the SOtM kit sounded maybe +3dB brighter, but the REL SPL curve indicted this was not the case.  This was puzzling.  I spent much time switching between the mR and SOtM kit, listening very carefully to different aspects of the presentation.  The conclusion was that the SOtM kit sounded better in all areas, more realistic bass, more accurate treble, more detail in the mids, I could go on and on, but overall, I would still rather listen to the mR, somehow nicer to listen to, more full bodied, and without the (un-measurable) prominence in the highs.  This was both puzzling and frustrating.  Many of us are interested in making those small improvements, someone declares component A performs better than component B, others try this, everyone seems to agree that component A is the one to have.  So you try component A in your own system, you can see (hear) why it is better, but overall, your nice new (and expensive) component A has made your system sound worse.  Where do you go from here?  Logically, if you then get something that performs even better than component A, is this not just going to exacerbate things further?  It was about this time that it dawned on me that maybe this was not just an SOtM issue, that I had been experiencing this already, albeit to a lesser degree.  This was with the upgrade from my old D800 Expert to the 1000Pro.  For me, the 1000Pro is clearly the better machine, far more realistic, more detail, higher resolution, better bass power, the whole lot.  Yet thinking back, with some material I was finding the 1000Pro a little more difficult to live with than the more relaxed D800.  The key point here is that these niggling negatives for the 1000Pro were exhibiting themselves in exactly the same areas that the SOtM bundle was spoiling my listening enjoyment.  So maybe the SOtM kit was not having as big an effect as I first thought, this was cumulative, with the SOtM kit adding to something that the switch to the 1000Pro was already invoking.  So the impact of the 1000Pro I could live with, the SOtM kit was simply pushing things over the edge.

I concluded from the above that if every improvement I make to the electronics is making things sound worse, at least in part, then something else must be wrong.  It was at this point that I decided to try Thierry’s Home Audio Fidelity service.  The logic being that the issue was most likely to be related to my room, and if Thierry’s convolution filters could correct this, then maybe the improvements in the electronics could shine through.  It would at least provide some data and experience to try to understand what remains a puzzling issue.

Two things happened almost at the same time.  For Thierry to make the filters, you need to make many REW measurements for the left and right channel independently, both at the listening position and spaced around the listening position, so this is maybe 14 or more measurements.  I made the measurements and sent them off to Thierry.  The next day, I had my first convolution filters.  I loaded these into Roon, and had a listen.  I could hear many things I liked about the filters, but there was an issue with the bass.  Wondering around the room, the bass sounded fine.  As an example, standing a meter or so behind the listening position the bass sounded great, but at the listening position the bass was very lean, quite good bass but nowhere near enough of it for me.  For the record, the listening position is at a point in the room that normally has a bit of bass reinforcement, so a little more bass at the listening position apparent than elsewhere in the room.  Anyway, some correspondence with Thierry, he explained why the bass sound as it did, the next day he provided some new convolutions with changes to phase & low end correction.  There was something interesting here though, intuitively, you would think that some convolution filters that reduce bass would be the last thing you want if your system was a little bright, but even with these bass light convolutions, the treble issue seemed to have pretty much vanished.  My assumption was that maybe the convolutions had reduced the HF in some key areas, hence providing this result.  I started this paragraph stating that two things happened at the same time.  The other thing I was referring two happened after I had listened to the first convolution filter, but before Thierry had sent me the update.  It was about Sunday lunchtime, I had listened to the filter, and emailed some feedback to Thierry.  Now it was a case of simply waiting for Thierry’s response.  Without going into all the reasons why, I was sat at my computer and was listening to some semi-background music, this was from a playlist in iTunes feeding the SOtM kit via Shairport.  OK, I know this is not the best way to use the SOtM kit with respect to SQ, it was just convenient for what I wanted to do at that time.  Anyway, some music was playing, and yes, it did sound a bit harsh and unpleasant in the presence range.

This was a little annoying and it got me thinking.  I opened up REW and loaded in the very many measurements that I had made to send to Thierry.  So this is 14 measurements, taken in 7 different positions for both left and right channel.  Using REW I produced an average curve from all these measurements.  Looking at any individual measurement curve, you might see a bit of a peak at 4750kHz or something, other curves taken in different positions might show a peak at a different frequency, but the averaged curve from all the readings actually looked very good in the mid to high frequencies, no obvious peaks and troughs, and a very fractional downward slope.  What was more obvious was a bit of a trough between 420 and 720 Hz and more prominent bass null at 125 Hz, followed by bass reinforcement above about 70Hz.  I was still listening in iTunes, so just as an idle experiment, I opened up the equaliser, nudged up the 500 Hz region (about +5dB) and nudged up the 125 Hz region.   To be honest, this was just a bit of idle fun, I was not expecting much.  Well I was wrong, this tweak on the iTunes equaliser killed the apparent presence / treble issue dead.  I then sat down for a serious listen, in doing this one thing was clear, going from Roon / HQPlayer to iTunes Shareport, this with iTunes doing some of it’s own processing with the equaliser invoked, is a clear step backwards.  So not a solution, but with iTunes and equalisation it was listenable.

Taking these two results, the iTunes equaliser solution is perhaps easy to understand, adding a touch more bass and then a little body at 500 Hz takes the presence / treble range into balance.  I can only assume something similar is happening with the first convolution filter from Thierry, but in this case it is a little harder to understand intuitively, as this filter offered a clear drop in bass.  I guess similar forces are in play though.  As my next experiment the same Sunday afternoon, I created a parametric equaliser curve in Roon, using the Q values to try to match the 420 to 720 Hz dip as accurately as possible, plus a degree of correction for the bass null.  This actually sounded pretty decent, and a mile better than the iTunes tweak.

After the above, three weeks have gone by.  In this time I have received revised HAF convolutions from Thierry, tried them, given feedback, and then Thierry supplies a tweaked version.  This process has taken far longer than it should.  This is not Thierry’s fault, he has been incredibly helpful and responsive.  It has just been bad timing, as it has just happened that whenever Thierry sends a new convolution, it has been at a time when I have been away from my system for a few days so it has taken a while before I can try the update.  So whereas it is theoretically possible to try 3 or 4 tweaks in as many days, it has taken about a week per iteration.
So to last weekend, this did not offer any time for any quiet listening on my own, but there was much music played with others in the house.  This did include one very extensive listening session.  In the interests of full disclosure, this session included two things, company and a modest (?) amount of alcohol, so more of a time to enjoy music, not analyse hifi kit.  This session was running Roon HQPlayer and Thierry’s latest HAF filter.  All I can say here is that I was quite simply enjoying the music, I was pretty much forgetting about my recent woes and enjoying the music enormously.  There were a few wow moments with the music too, safe to say things were sounding very good indeed.  I did drift back into audiophile mode at one point.  Armed with the iPad it is possible to turn the Roon’s DSP engine on and off, or even change settings.  With a mate, I played about 2 minutes of one track, and switched between DSP off, my own Roon Parametric EQ effort, and Thierry’s latest HAF filter.  I referred to these as one, two and three, so my mate was blind.  After a very short while, my mate said something like “it sounds best with three, leave it at that and stop (word omitted) about with your system.  So not exactly a scientifically sound level controlled double blind listening experiment, more of a 10 minute bit of fun with a slightly inebriated person, but a victory for Thierry’s latest HAF filter nevertheless.  Interestingly, at this point the SOtM kit did not sound thin at all, and the HF is actually becoming a highlight of the systems performance, incisive crashing cymbals and similar actually providing some of those aforementioned wow moments.

Then to today.  I have had the house to myself, I have consumed nothing stronger than coffee, and I have had time to listen to both the latest HAF and HAF crosstalk reduction filters. This was some careful listening with some familiar material, and the system sounds superb.  The best way I can think to describe the difference with the HAF filters is to simply state that listening with DSP off is like listening to some obviously very high quality very resolving electronics with some very revealing speakers, DSP on and the HAF convolutions in play is like listening to superbly reproduced music.  So thoughts of the system drift away, enjoyment of the music takes it’s place.  Everything sounds far more real, closer to the original and further away from reproduced music, sound staging is vastly improved also, allowing you to drift further into the music.  This is a seriously good result!  I now feel that I just want to sit down and listen to stuff. 
 
So, is this the end of the tail?  It is certainly the end of a very difficult chapter, but there is more to come. Thierry’s latest filter is now getting very close to where it should be in terms of frequency balance, but I do not think it is 100% there.  So I think I shall be going back for yet another iteration, I am sure at some point the next iteration will be worse than the one before, then I shall stop.  So far, I provide feedback to Thierry, he provides a new convolution that is just a little bit better.  It has to be said, the service and expertise that Thierry has provided has been absolutely first class, truly exceptional.  If you have occasional thoughts about trying his service, I would say go for it, I cannot recommend enough.
I also have a whole list of other things that that I plan to do, but this post is already too long, I shall save these for another day.  Safe to say for now that some changes to my room and room treatments are on this list, along with some hardware stuff to try.

As for the rest of this weekend, it occurs to me that I am still running Beta 12.  There were rather mixed reports regarding Beta 14, but some claimed better bass and similar, so it occurs to me that if I am fine tuning with Thierry’s filters, now would be a very good time to switch to Beta 14, I may as well do the final tweaking with the very latest firmware.
Tomorrow I think I will take stock and try to get some idea if all this effort with the SOtM kit, the REF10 and MC3+USB was really worth it.  Today it seems so, tomorrow I will try a simple head to head between Roon AIR on latest Beta 14 and the SOtM / Mutec kit.

Should be interesting.

[ogs]

I have been alternating between Roon/AIR and mR/MC3/AES on Beta 14 lately. It will indeed be interesting to read your opinion on this!

[thumb5]

Fascinating reading, as always. I honestly don't know how you find the time to do all that listening, never mind writing it up for everyone else's edification and enjoyment. Thank you!

[Hifi_swlon]

... 
Thierry’s latest filter is now getting very close to where it should be in terms of frequency balance, but I do not think it is 100% there.  So I think I shall be going back for yet another iteration,

Wow Confused, you’re really getting your moneys worth from HAF! I felt guilty asking for a third round of tweaks (to fix Roons grouping problem with filters longer than 500ms). That said I've added a PEQ layer which I’ve tweaked over months and really makes everything much better, so would like to incorporate back into the convolution at some point.

[Confused]

... 
Thierry’s latest filter is now getting very close to where it should be in terms of frequency balance, but I do not think it is 100% there.  So I think I shall be going back for yet another iteration,

Wow Confused, you’re really getting your moneys worth from HAF! I felt guilty asking for a third round of tweaks (to fix Roons grouping problem with filters longer than 500ms). That said I've added a PEQ layer which I’ve tweaked over months and really makes everything much better, so would like to incorporate back into the convolution at some point.

I think I was the same, and started to feel a bit guilty at about the third iteration.  I actually mentioned this in an email to Thierry, where I stated that I would be happy to pay a fee for further updates, I received a very positive response from Thierry, where he stated he was happy to do the iterations, to quote "I don't like giving up and it's always interesting & knowledgeable to understand customer needs."  I am sure there is a limit though!    For the record, the HAF filters I mentioned in post #202 are my fifth iteration, I have emailed some feedback to Thierry on these, and I am now waiting to see if he is happy to do a sixth version.

I am sure I will be giving Thierry more business in the future though.  I mentioned earlier about making some changes to my room and room treatments, all of these would have me going back for some new convolutions.  As I said before, the HAF service is the price of a modest USB cable or similar, or half the price of a modest LPSU, but the effect of HAF is an order of magnitude more than you might get from the latest Lush cable, PSU upgrade, or whatever.  It is seriously good value for money.

It is interesting that you use PEQ on top of the convolution.  I can see how this could work very well in terms of providing a little user fine tuning on top of a reasonably satisfactory convolution.  I would worry that this is a little extra processing though, something else in the signal path.  I am a bit of a purist, so part of me does not like the idea of adding convolutions, let alone PEQ on top.  To be honest I do not know enough about the inner workings of Roon to know if there are any detrimental effects of adding PEQ once you are already using convolutions.  Maybe it makes no difference once the DSP engine is engaged?  I just don't know enough to be sure on this point.  (Does anyone know the answer to this?)  Irrespective of all this, I could see a situation where you have a base setting of running the convolutions with PEQ off, and then have a small selection of PEQ's with specific uses.  Say one for those thin and scratchy sounding 80's CD's you sometimes get, this kind of thing.  For me, my aim is to get a Convolution that is as close as possible to what I think is correct, then just forget about it, if you then stumble across a recording that is a bit thin or whatever, just accept it for what it is.  There is a risk here of getting sucked in and end up worrying about PEQ rather than just enjoying the music.  I know that I might be susceptible to this, which is the key reason I would like to get the convolutions 100% as I would like them, then when satisfied these are optimised, it can just be set and forget.

[Hifi_swlon]

....and I am now waiting to see if he is happy to do a sixth version.

Ok, well I guess I’ll go back for my PEQ adjustment guilt-free then. TBH I’ll probably remeasure and get a new set as I’ve tweaked positions slightly

It is interesting that you use PEQ on top of the convolution.  I can see how this could work very well in terms of providing a little user fine tuning on top of a reasonably satisfactory convolution.  I would worry that this is a little extra processing though, something else in the signal path.  I am a bit of a purist, so part of me does not like the idea of adding convolutions, let alone PEQ on top. 

I used to think that but now I just enjoy the flexibility. In my imperfect room, having the ability to play with the balance far outweighs any SQ processing worries. Also once you’re applying convolution you’re along way from purist signal path and may as well just embrace the DSP functionality. I’ve found I tweaked much more at the start, and have ended up with something I rarely touch. It’s tricky to do by ear obviously as adding bass on some material means bloated bass on others, etc etc, but after quite a few rounds of adjustments over several months you settle on something that works best for all music as a compromise. This is pretty impossible to do from graphs and is the only aspect of HAF I don’t like - ie you’re locked out of user adjustments. So PEQ is the only way to do it really. You also learn a lot about what frequencies do what, many of which I guessed wrongly. And can experience phenomenon like adding bass to remove bass and vice versa.

Just to add even after asking for a more steeply slowed curve (more bass) I still added bass, but also removed a range of bottom end frequencies. I then titled the treble down slightly more. Probably a 3db slope increase. Maybe 6 in a small bass region. I also found Thierry’s filter removed a little too much room, which in my case gave too dry a sound compared to Dirac etc. May have been correct but I preferred to keep a bit of the room.

Anyway I plan to send Thierry back my PEQ and see if he can incorporate into the main convolution. It just feels ‘better’ to have the bulk of the processing in one place. FWIW some people say the PEQ engine in Roon lowers SQ (even if enabled but no settings). Personally I haven’t found either to be true but my system and room are far from perfect which is why I’m doing so much DSP.

The killer for me is going to be if I ever get the CI and use other sources direct (which would be nice). I really don’t like the raw sound in my room, so having to drop the convolution and PEQ will be harsh. At that point I might be looking at other options anyway.

There is a risk here of getting sucked in and end up worrying about PEQ rather than just enjoying the music.  I know that I might be susceptible to this, which is the key reason I would like to get the convolutions 100% as I would like them, then when satisfied these are optimised, it can just be set and forget.

I think you might be slightly susceptible

Anyway you’ve got a lot more to worry about, and on that note will revert back to your topic in hand.

[Confused]

I have realised that what I stated in post #206 was not quite right, there was one time that I provided some feedback to Thierry and he sent me two sets of HAF filters to try, so it is actually one less iteration than stated above.

@Hifi_swlon - Yes, quite.  Thank you for your kind words Sir!

Anyway, this morning I loaded up firmware 12.2.12 Beta 14.  With a quick listen, I can't say that I noticed any significant differences to Beta 12.  Let's put it this way, if I had not known the firmware had changed, I certainly would not have picked this up from sound quality changes.  One thing I did notice, is that with Beta 14 I seem to need a lot more gain.  I have not been paying much attention to volume levels recently but did today when doing a AIR versus SOtM / Mutec trial.  Volume levels (gain) seemed much higher than I have needed to use in the past.  For good order, I will note this in the 12.2.12 thread.  I might even try reverting back to Beta 12 to compare properly, but I am out of time today.

So to a comparison of Beta 14 Roon AIR versus SOtM bundle, Mutec REF10 and MC3+USB.  This was a fairly quick A/B comparison and something I would definitely like to try again if I have an idle hour or so free.  This comparison was done with both scenarios using the latest HAF filter, so in terms of how things sound this offers a bit of a system reset in both cases.

One thing is clear to me, in absolute terms Beta 14 Roon AIR is superb, and is in my view very close to the SOtM / Mutec kit.  Taking the PC out of the equation (the PC is a constant), there is a £7K delta between Roon AIR and the SOtM Mutec set up.  There are differences between the two, but they are very subtle.  With the SOtM / Mutec some sections of music has a little more atmosphere and is a little more engaging, a touch more dynamics, but to be clear, these differences are very small.  I think you would need to have a very good system, very well set up and be very fussy, for the SOtM / Mutec kit to be worth £7K to you.  This one was far closer than I expected.  One other point.  Prior to trying this, I did wonder if with the SOtM kit out of the equation and running AIR if the sound might be a little more full-bodied, even if AIR was a touch behind in other areas.  What I actually found was that they did, in fact, sound very similar.

[Confused]

I was going to leave today's posting there, but hot off the press, I have just received another set of HAF filters from Thierry. So I have not broken him yet! Plus, this is essentially a 24-hour turnaround. Excellent service yet again.

With this, something occurs to me. Historically I have been very negative towards room correction, it somehow does not fit with my purist sensibilities. So it is interesting that adding the SOtM kit has been the catalyst for me even trying it for myself. From where I am now, if I had the choice of Roon AIR + Thierry's convolutions (Euro 159) or the SOtM Mutec kit without the convolutions (£7K), I'd take the former. Enough said for now I think.

[zdenes]

I haven't done much room correction except the DSP in my subwoofers but my room is acoustically treated.

On the other hand, I have done extensive experimenting and A/B and A/B/C, etc. testing of power supplies for my Mutec MC3+USB (converted to accept DC) and also on the SOtM gear, testing a number of vibration control devices on the Ref10 and clock cable testing. They all matter to a great deal and once you optimize those you may find that the subtle difference between AIR and the SOtM based source is not that subtle. I find it a little difficult to comprehend why you spent 7 grand on the equipment and not the last few hundred to get things to perfection and instead countless hours on DSP optimization. The HF glare could have been solved just by getting PH power supplies.

Also, £7K is a lot of money for the subtle difference but then again, the cost of the Devilet + speakers + cables + power filtering + rack + room treatment, etc. is around 60K in my case, so I still found it justified having spent a bit over 10% of that for a clearly audible improvement and the hassle-free operation that SOtM streaming offers.
In my case (maybe unique) my home 2.2 home theatre system also goes through the Mutecs and they bring tremendous improvement in the sound quality of the movies. But I understand that for most people here, it may not be important.