A how to for a PC XO.

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
diyAudio Member
Joined 2004
Originally posted by ewildgoose Hi, I have spent *many* years building a succession of quiet PCs and I agree it's VERY hard!!

The only machine which I owned which was really terrific was an Antec Sonata. I put in a fanless PSU (which runs pretty hot and I tried several because they all failed fairly quickly...). The other fan was a 120mm big boy on the back which spins really slow and was absolutely silent.

The Sonata was near inaudible in my room from very close and totally quiet from a distance. Also the cooling was incredible with temps at 30c or so with only the single 120mm fan at the back (compared with my Coolermaster boxes which have 7+ fans in them and the machine is much hotter!). Clearly design plays as much of a factor as number of fans!

The worst case I owned was the Claritas Strata fancy HTPC case. I have one for sale if UK buyer wants it. Lovely looking box, but zero silencing.

Also although most of the desktop style cases look best for HTPC the cooling is much, much worse than with a tower case. Also the tower case usually ends up on the floor, but the HTPC case ends up nearer ear height (and high freq whine is more audible in straight lines). So if looks aren't important then tower is the way to go

Antec seems to be a really solid choice if you just want something which works! I was very, very pleased with my Sonata case. I sold it and bought the Claritas case....

I'm using an Antec Sonata 1 case right now and whilst its not bad on the noise level its still way too loud. I am however using the standard truepower PSU that came with the case and the fans are just simple antec 3 speed ones. Hardly using it to its full potential really.

I also heard from the HTPC quarter that the desktop style cases were a no go for silence and decent cooling. Much better is a tower case. I almost went for a Silverstone case but swapped it for the Antec at the last minute! Glad I did now.

To be honest I think the "no fans" thing is a risk if you care about drives failing though. I added a small low speed fan just to get a small amount of air moving inside the case. Even a tiny fan will drop the temps 20-30C so it's worthwhile. Anyway, it's currently the best choice for me

I've got a couple of 120mm fans running at 5v just to keep the air circulating in the PSU/Harddrive chamber and the CPU section.

OK, so now tell me: How have you found the Lynx compared with the RME 9652? Seems to me that they are going to be identical with the RME being 1/4 the price? Do you hear an difference bearing in mind you are spitting out digital?

I haven't actually got the PC up and running so I can't comment yet.

The Lynx is certainly a better card than the RME though, it supports AES and wordclock256X for a start. The analogue side of the Lynx is also obviously better but I'm not using that (yet). Again though it provides me with more expansion options for sure.

You've got to bear in mind that I was running two RME cards - one HDSP9632 and the other a 9652. I've sold both of these and could virtually buy the Lynx with what I made from the sales of those. So really its cost me nothing over what I was using before.

How are you doing volume control with so many DACs? What would be perfect is if the Apogee had a master volume control (which didn't work in the digital domain)... I had to build a multiway attenuator using a DACT stepped switch - I have a remote control on it which allows me to change volume control from the couch. In practice I set it roughly to max volume that I need and control the volume using digital volume control in the player app - sure it looses some precision, but the RME 9632 DACs are around >110db so I hardly care if I loose a few bits at low volumes!

The Lynx supports volume linking in the current release, so you can simultaneously control multiple outputs. Very similar to the RME solution.

The RME total mix is also 40bit precision so attenuation general results in far less quantization than most solutions. The Lynx is even better in this regard since it doesn't simply truncate the data like the RME but rather it dithers and uses noise shaping techniques to move the artifacts caused attenuation even further into the 'inaudible' frequency range.

Its not ideal because the DAC's aren't getting a 0dB or reference level input but I plan to later add a series of analogue pre-amps after the Apogee DA16X to do the attenuation.

What kind of filters are you running now? I am using Brutefir at 44Khz with 65535 tap filters and it's tiny amounts of CPU compared with the amount needed for resampling. Brutefir does partitioned convolution and running at 2048 sample latency is ok, but dropping to 1024 sample latency would really raise the CPU. What kind of latency are you getting out of your convolver?

Still using LineEQ for FIR, Voxengo Pristine Space for convolution( with Sbragion's DRC) and Voxengo Audio Delay. Here's a shot of my routing scheme when I was using the RME cards:

An externally hosted image should be here but it was not working when we last tested it.


An externally hosted image should be here but it was not working when we last tested it.


Really love the setup and I just couldn't use anything else now.
 
Interesting.

Just a note, but the HDSP 9652 should have AES output? (did you have the other one?) Also some other flash outputs like TDif etc? It has a wordclock input as well, but perhaps the Lynx is technically better there. I doubt it makes all the much difference though?

How much CPU are you chewing in that config? Does Voxengo have a variable delay trade off against latency?

I'm curious that you are using (steep) linear phase crossovers, especially right up at 4Khz. Does this not "ring" and pre-echo like crazy?

I have not heard too many results from people trying linear phase crossovers so this is interesting.

Commenting on the GUI for a sec. That all looks horrendously complicated... Is there a market for an all in one solution if it's much, much simpler to use...?

Looks great anyway

All the best

Ed W

Edit: My mistake - you are correct that the 9652 doesn't have AES, it has ADAT. There is a new card though which has AES: http://www.rme-audio.com/english/madi/hdspaes32.htm - don't know the price mind...
 
Not having used Shinobiwan's recommended software, I can't comment on the complexity of his solution, but, I can say that I do believe a market for a simpler, all-in-one solution could be created. At the moment, the only way to do what we're trying to do requires the use of several pieces of software, all of which are doing things they were not really written to do - there is no single, end-to-end solution available on the PC. The requirements for mastering (which is what most relevant pro audio software is meant for) are much different, and require various settings and controls that are irrelevant for home audio room correction, XO and EQ. Trying to tweak software or hardware for purposes for which they weren't really intended is, in my experience, a sure recipe for several strong headaches.

Having said this, I believe that Doug at ETF Acoustic (www.etfacoustic.com) is on his way to creating the first purpose-built solution for the PC with R+D. IMO, it's only a matter of time before PC-based solutions besting both DEQX and Audyssey are available. It seems as though the capability to do everything we're trying to do on the PC exists already, it's just going to take some small company to put it all together in a single, purpose-built package. Then comes wave field synthesis and PC's with 32- or 64-channel soundcards...
 
Do bear in mind the economics of this though. Such a bit of software would be a niche market and for a full time developer don't expect to see it cost less than a few hundred quid (euros/dollars, etc).

Remember that Tact sell just a few hundred of these RCS boxes a year, so this sets the size of the market. Divide your expected annual earnings by a few hundred (or be ambitious and go for a few thousand), either way you end up with a fairly large number

The way out is to get it into the mass market, but now you have a bit of a struggle unless you are capable of doing the high level maths and making it *your* product

Anyway, the punchline is it's a difficult product to market


You mentioned ETF R&D though. I only gave it a cursory glance and although it's interesting it doesn't seem any more sophisticated than what you can do with Excel and a Behringer DEQ 2496? Seems like a very simple form of room correction to me? Certainly well below what is possible with DEQX/Tact and *much* less sophisticated than DRC.

There are a couple of commercial options about to hit the market very soon. Keep your eyes peeled...

Ed W
 
diyAudio Member
Joined 2004
ewildgoose said:
Interesting.

Just a note, but the HDSP 9652 should have AES output? (did you have the other one?) Also some other flash outputs like TDif etc? It has a wordclock input as well, but perhaps the Lynx is technically better there. I doubt it makes all the much difference though?

Sorry but like I said before, it isn't up and running yet. The new PC is around 3/4 finished. Then I've got the software to install and config not to mention experimenting with the fans for optimum cooling/silence. So still plenty to do before I even think about the performance.

The HDSP9652 only support 2 channels of AES and the TDIF doesn't work with the DA16X so I was forced to use 3 x ADAT pipes with dualwire for 96Khz/24bit capability over 12 channels. The DAC's in the Apogee are 16 x 192Khz/24bit capable so there's wasted potential. With the AES I've got no problems using upto 192/24bit.

Also, never underestimate clocking. I think the reason why its so underated is because people simply take it for granted and there's a small degree of ignorance relating to the fact that its digital so its bit-perfect right? Wrong! Clocking and the DAC sections are where it all goes wrong so its these area's that provide the biggest and most noticable gains.
When I switched from the rather poor RME clock to the Antelope Isochrone OCX I was pretty astonished at the extra detail, realism and soundstaging. Then I decided to do the DAC's and found an integrated masterclock and DAC solution in the Apogee DA16X. Again the DAC's in the RME aren't poor but Apogee highlighted the fact there was even more to be heard.

How much CPU are you chewing in that config? Does Voxengo have a variable delay trade off against latency?

With all that running for complete loudspeaker management and DRC over 6 channels, CPU usage is around 40% on a P4 3.2Ghz. And around 25% for 2 channels with all the FIR filters for 3-way.

I'm curious that you are using (steep) linear phase crossovers, especially right up at 4Khz. Does this not "ring" and pre-echo like crazy?

Not at all.

Waves LineEQ allows you to tailor the FIR window function to either maximise the steepness of the filter slope or increase(lower) the passband ripple rejection. I personally use the 'low ripple' method which offers -100dB of passband rejection before the usual FIR ripple begins.

Here's a snip from Waves LineEQ manual:

"Lets now look at the task of creating a Hi-Cut (Low-Pass) filter. When we design a Hi-Cut filter, the Method specified will determine the accuracy of the slope vs. the gain in which the slope stops its accurate descent and a further descending ripple begins. This point is also known as the stop-band. Lets create a Hi-Cut at 4kHz. The Q control will specify the desired slope with Q-6.50 being the steepest slope possible. Now as we toggle between Methods you will see that the Accurate Method gives a near brickwall drop at the cutoff frequency but the accurate descent will stop at about –60dB and from there upwards in the frequency domain, a slowly descending ripple will occur. The Normal Method will yield a more moderate slope or a lower dB per Octave value. The stop-band will occur in a higher frequency but at a lower gain of about –80dB. This same difference will be even more extreme using the Low-Ripple Method. The slope will be even more moderate and the stop band will happen at a higher frequency but at a lower gain of under -100dB."

A great deal of analogue and digital crossover don't offer over 100dB of rejection and certainly not with 150dB+/octave slopes.

Commenting on the GUI for a sec. That all looks horrendously complicated... Is there a market for an all in one solution if it's much, much simpler to use...?

Its very easy to use once you get around the fact that you've got a visual representation of the signal flow between various plugin components. In all I'd say its greatly intuative.

What you saw in my post above is a pretty extreme example that shows the routing scheme of a complete multichannel solution featuring DRC for each channel, a 3-way stereo XO with FIR filters, digital delays for each individual channel and finally bass management (highpass filter) for the surround and center channels.
A two channel configuration would look far less intimidating. Nontheless its a set and forget program that, once configured, you never have to touch again until necessary.

Edit: My mistake - you are correct that the 9652 doesn't have AES, it has ADAT. There is a new card though which has AES: http://www.rme-audio.com/english/madi/hdspaes32.htm - don't know the price mind... [/B]

Yes the AES32 was considered but again the Lynx just offers more. The price BTW is around £500 I was told.
 
ewildgoose said:

I am working with an Italian firm to distribute a commercial filtering standalone box. It has a nice gui and comes with varying amounts of outputs. If anyone is looking for an off the shelf solution (and this is DIY ground) then please do get in touch.

Count me interested, but your e-mail is blocked. Were I to pick today (and I will need something soon), I'd go with the DEQX. Shin's solution looks very capable and he obviously enjoys the software application side. That's great and it's fun to follow the progress, but that approach is not for me at this point. Something between the two might be fine.

Sheldon
 
Two questions and an alternative approach

1) What is the best way to pipe SACD into such a PC-based XO? I have been looking at the dvdupgrades.ch S/P-DIF add-on board for multi-disc players, in conjunction with an add-on board for the LynxTwo. However, if there are more cost-effective solutions, I'd like to hear about them.

2) What is the best way to enjoy DVD video with 5.1 soundtracks while also processing through a PC-based XO? I wouldn't mind using a pair of LynxTwo cards in a single PC but according to LynxTwo technical support, I must choose between 5.1 sound on the one hand or XO applications on the other -- I can't have both.

3) Although I already own a LynxTwo-B, I have also been looking at aftermarket modifications of Behringer's digital crossover that give it digital output (S/P-DIF x 3). Some people have been using thus modified units together with Panasonic SA-XR series receivers with good results. BUT the people who have done this have not been using all the available channels of the SA-XR series receivers -- rather, they have been using just two channels because there is just a single coaxial (or optical) digital input rather than three. (Digital access to all channels requires going through HDMI in the case of the SA-XR70.) So for a three-way system they must use three of the SA-XR series receivers! This strikes me as a huge waste and I wish there was some sort of a mod that could be done to the Pannies so all channels could be directly accessed via additional S/P-DIF ports. It is really a shame since the Pannies will do a good job of amplifying whatever they are fed, all the way up to 24/192. I have been browsing through the Texas Instruments documentation for the PurePath chips in these amps and I can't see any reason why all channels should not be able to function like the pair of channels that are currently available through S/P-DIF. Now, it is also possible that even cheaper solutions might be available as it seems that basically similar PurePath-based receivers are available very cheaply on eBay (e.g., Koss KS5192/KS5190/KS4192, Insignia, and JIS brand "1000W" home theater systems, etc.). So one way or another, one could end up with six fully digital, fully usable amplification channels (100W each) for under $300. Still, the thought that in each receiver some channels are going to waste grates on me. BTW that last bunch of el cheapo receivers seem to be characterized by substandard QC and there are occasional duds (at less than $100 for a complete home theater system -- sans TV -- it is probably cheaper to just replace them rather than repair), and no doubt there is a lot of room for tweaking and modding to improve the sound. Anyway, one alternative to the PC-based XO would be a multi-disc standalone player and modified Behringers feeding into PurePath-based amplifiers. A sort of poor man's TaCT, and maybe not quite as good as the PC-based approach but with a very high cost/performance ratio.
 
Hi Shin!

During these Xmas days I am also looking for an "all in 1 PC" for HC, Game et maybe Xover ;-)...
I must say that we arrived to the closely the same conclusions for the hardware!

Anyway, my choice for the PSU would rather be a Seasonic, see the test to see why! The problem of the Phantom is that the fan once it has started never stops and in this case, wich is really likely to happens if your're a gamer, there are better solutions...

http://www.matbe.com/divers/zoom.php?img=000000023951.png
http://www.matbe.com/divers/zoom.php?img=000000023949.png
http://www.matbe.com/divers/zoom.php?img=000000023947.png

@+
Maiky
 
Regarding the Panny SA-XR's and HDMI.
Sony just came out with a new vario HTPC that has a HDMI output, I do not know if the HDMI on this machine outputs both video and sound as is part of the spec. My guess is that we will see video cards with HDMI real soon.
such a card plus a sa-xr70 would be the ultra cheap ticket to a digital pc x-over and 6 channels of pure digital power !
 
Thanks for the hot tip on the VAIO XL1 Digital Living System. Very interesting! It has been quite some time since I have seen a PC comprised by two distinct sections with a combined weight of 40kg. And the support for DSD audio is also very interesting. All in all it seems like a pretty reasonable price; I think one could easily spend more money for less performance and poorer looks than that.
 
Shinobiwan

I can't help feel it is impossible to get a truly quiet PC that is able to do the kind of processing you are trying to get it to do without reverting to a £600 Zalman water cooled case

Why not dump the concept of having the pc in the same room as the hifi? I've looked into this and it looks promising:

- powerful noisy pc to do all drc / xo
- stream the audio using an app such as Wormhole - http://plasq.com/wormhole over a standard ethernet network, unless you are upsampling in which case you may need to use gigabit
- silent pc with a low spec processor such as a VIA plus the expensive soundcard in the lounge

Trying to silence a powerful PC to make it quiet enough not to be noticed in a listening room is just going to turn into an expensive uphill quest

Z
 
It is difficult, but not impossible.

a) If your requirements are "quiet" then you have lots of choice

b) If your requirements are "silent" then forget anything which is labelled silent and buy only things with ZERO fans and heatpipes. Also you will *NEED* to wrap the hard-disk in some kind of insulating box. Personally I then put back into the box a single low speed fan just to move some air around - it drops internal temps by 20C but raises the noise level to "very quiet".

Many of the Via boards are available in fanless, but they are quite low powered.

If you want fanless and more horsepower then you have to shell out for a Pentium-M and a heatpipe really. I have a Pentium-4 and a heatpipe, but it's a bit hot really. Basically NO fans and then it becomes silent, but plan to have the HD fail more frequently than if you put it in a server box with a hoover sucking air through
 
Hi All,

For music lovers on budget a (cheap) MDF built box lined with acoustic felt (or foam tiles) may be a solution for silent PC . PC should be located in the central cavity and sealed (front to back) so that the cool air intake (front) doesn't mix with hot air out (back). A dust filter may be placed on the front slot.

This arrangement prevents mixing of hot and cool air and may well improve PC cooling at a fraction of the cost of some hi-tech solutions floating around.

Regars,
 

Attachments

  • box.jpg
    box.jpg
    3.4 KB · Views: 671
I stress. If you want "silent" then you have to take out ALL the fans and put the HD in a silencing box. Anything else will be at best "quiet" and to be honest the annoyingness of it is related to the sound/pitch of the fans rather than the volume. For example I am sitting next to my server which makes a pleasant whoosh noise - it's quite loud, but easy to ignore. In contrast my older HTPC made a barely audible whine from the HD and it would drive me nuts...

I personally found the Sonata when modified with a fanless PSU to be the best case I ever owned. It was very, very quiet. I am now using an HFX fanless thing which is totally silent, but then I added back a small fan to move a little air (it's now very, very quiet)

Ed W
 
Ed,

If your post was in response to my post, I can only assume you're talking about the Reserator. There are no fans on the Reserator -- just a large heatsink used to cool the water. There are cooling blocks for the CPU and the GPU, so, no fans needed there, either. The only noise would come from the HDD or optical drive, assuming I haven't misunderstood the product and the Reserator itself is quiet. This is why I posted in the first place -- I'd like someone who has had experience with the Reserator to confirm my understanding that it is indeed silent.



Brad
 
diyAudio Member
Joined 2004
Zodiac said:
Shinobiwan

I can't help feel it is impossible to get a truly quiet PC that is able to do the kind of processing you are trying to get it to do without reverting to a £600 Zalman water cooled case

Trying to silence a powerful PC to make it quiet enough not to be noticed in a listening room is just going to turn into an expensive uphill quest

Z

Hi Z

I'd have agreed to a certain extent before I set out on this PC build.

However as it stands I really have to strain to hear the PC from the listening position. I've also added one of the large Artic Coolers to the VGA card and the whole thing has really suprised me.

I use an Optoma Themescene H79 DLP projector and its rumoured to be the quietest of it breed and its certainly pretty quiet, however I'm finding that I'm more distracted by the gentle hum of the fan when its on than anything comming from the new PC.

All in all its a remarkable sucess and I'm very happy. I think the combination of every component carefully selected for noise emissions and the judicious use of acoustic materials and the most quiet fans available running low RPM at 5v means that you'd only get it quieter with water cooling. The noise really is virtually zero from listening position.

Still have no idea on performance as I'm having some issues with installing windows using RAID controllers for the harddrives. Hopefully that's all sorted though now I've had a bit of advice.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.