Oh, that is a can of worms
For me a computer is a commodity device, that can evolve and get trashed after X years when it is not convenient anymore.
With a computer you go on the web, you install softwares, you install updates and viruses (which are often similar is how they impact the user), etc.
This sort of things is prone to instability. Of course you can avoid all this by only installing proven software, using linux, not go on the web, etc. But then you loose part of the convenience...
On the other hand an audio device should be reliable, stable, serve only a defined purpose, and not change over time.
You can do that with a computer, but that means having a specialized computer for filtering (for example with AES/EBU inputs and multiple analog outputs), and another for playing media through it, surfing the web, etc.
Of course this is just my opinion...
Thanks. I do agree with arguments presented here. A PC environment is indeed pretty dynamic unlike a dedicated audio device. But the performance and flexibility to be had with an HTPC followed by a multichannel DAC is quite compelling.
My original preference was to use Najda for simple FIR crossovers (just like passive XO, just infinitely more flexible) and the HTPC handling heavier FIR filters for corrections on the overall response. But that now appears to be harder to implement than it sounds.
IMO, the 8ch blackbird DAC i linked to is a very well executed project. It might offer performance advantage over DAC integrated in Najda. But HTPC+this DAC is several orders of magnitude more expensive than HTPC+Najda.
My original preference was to use Najda for simple FIR crossovers (just like passive XO, just infinitely more flexible) and the HTPC handling heavier FIR filters for corrections on the overall response. But that now appears to be harder to implement than it sounds.
IMO, the 8ch blackbird DAC i linked to is a very well executed project. It might offer performance advantage over DAC integrated in Najda. But HTPC+this DAC is several orders of magnitude more expensive than HTPC+Najda.
Since you have this conversation going about a dedicated dsp solution vs a PC computer my question is what is the minimum PC specification you would need for a dedicated music PC? I have some old desktop computer that is about to be put out to pasture and am wondering if it would handle music functions and take everything else off of it. It is an older Pentium 4 with a 3.2Ghz processor and 4G of ram. I see you talking about the I7 processor but what is really necessary just to work as a music server and do the FIR filtering and run your rePhase software? The computer currently has Win 7 on it but I can always install Suse Linux or something similar on it. I've been following this thread for a long time and you have got me curious now.
Thanks, Steven
Thanks, Steven
I wouldn't think so much about whether an old PC is strong enough for use as a dedicated audio PC. It probably is, so why not just try it out.
A bit of optimisation is needed to secure playback stability. This is a guide with a lot of good points: PC Optimization Guide for Windows 7 | SweetCare. Some of the more advanced options can be omitted, but it is important to do all steps related to powersaving. And make sure to turn off all system sounds, because they can be very disturbing and LOUD when amplified through a HIFI system.
A bit of optimisation is needed to secure playback stability. This is a guide with a lot of good points: PC Optimization Guide for Windows 7 | SweetCare. Some of the more advanced options can be omitted, but it is important to do all steps related to powersaving. And make sure to turn off all system sounds, because they can be very disturbing and LOUD when amplified through a HIFI system.
We are on the same page on this.
I can see a perfectly tuned PC being perfectly stable (ie never updated
On the other hand using a PC as a media player (local and web) and filter in the same time is too much a risk IMO.
I've been using my main PC for all my audio processing and regular PC stuff. It stays on all the time. I use this PC to surf the web, watch movies, play the occasional game, CAD/CAM, audio simulation sw, audio measurements, spreadsheets, etc. All this while doing the processing for my active speaker system. I've been doing this for 3-4 years now on a Windows platform and I haven't had any major issues. A few settings might get changed with updates on rare occasion but 99% of the time it just works. My (nearly) computer illiterate roomate can use the big active rig to watch Netflix when I'm not home without any problems. Not once has any of my filtering been broken, nor have I had any unwanted loud noises (I keep all Windows sounds OFF). I've yet to blow anything up and don't see that happening other than by my own occasional stupidity (crosses fingers
). Hell, I've had this thing outside for measurements many times.I keep my audio settings backed up to the cloud in case my HDD goes up in smoke, and I have an older W7 machine I do for speaker testing in the basement that can step in for this one if there's any hw failure. I recently upgraded to W10 and all my audio stuff works just as it did before.
Kindhornman - that PC will be more than adequate to get you started in PC audio.
+1 above, I run nearly all the same programs as Nate does on my i5 -2500 3.3 Ghz PC. JRiver hosts 3-way digital XO using 64 bit FIR filters with 65536 taps per filter. Using JRiver's ASIO line input, I can loopback REW through JRiver's Convolution engine to verify proper filter operation. I use JRiver's 64 bit digital volume control (with volume protection) with 6 DAC channels running directly into 6 amps.
Been doing this for 4 years daily without blowing anything up and very little in the way of hiccups. First on Windows 7 and then on Windows 10 - everything works fine with very little CPU overhead. With music playing and being processed, very rarely does it ever reach 3% CPU utilization. Even when I run digital multitrack using Mixcraft, with 24 channels of audio being processed with VST plugins and the FIR processing, rarely does the computer reach 10% CPU utilization. There are no audio streaming glitches of any kind. If you have a choice in your digital audio driver for your AD DA converter, use ASIO for lowest latency and best performance.
Like Nate says Kindhornman, that PC with be adequate to get you started in PC digital audio. Cheers!
Been doing this for 4 years daily without blowing anything up and very little in the way of hiccups. First on Windows 7 and then on Windows 10 - everything works fine with very little CPU overhead. With music playing and being processed, very rarely does it ever reach 3% CPU utilization. Even when I run digital multitrack using Mixcraft, with 24 channels of audio being processed with VST plugins and the FIR processing, rarely does the computer reach 10% CPU utilization. There are no audio streaming glitches of any kind. If you have a choice in your digital audio driver for your AD DA converter, use ASIO for lowest latency and best performance.
Like Nate says Kindhornman, that PC with be adequate to get you started in PC digital audio. Cheers!
Thanks guys, now I can start my PC music quest to learn all that you have all seemed to master. I'm real interested in learning how to do all this FIR for simple two way systems at first. I have Clio analysis hardware and software so I can see what the results are as I go. I'm starting from nothing so it will be a real learning curve for sure.
natehansen66 , mitchba, to each his own I guess
I know many people are running this kind of setup, but this is just not for me.
Today $300 can give you 2*6144 taps worth of 48kHz time-domain convolution with close to zero processing delay and a mere 2W power consumption.
This is enough for my current use (I am not a big fan of high Q LF corrections, and I am not interested in >20kHz reproduction either...), and things will only get better in the coming years.
I don't see myself going back to computer based FFT convolution anytime soon
I know many people are running this kind of setup, but this is just not for me.
Today $300 can give you 2*6144 taps worth of 48kHz time-domain convolution with close to zero processing delay and a mere 2W power consumption.
This is enough for my current use (I am not a big fan of high Q LF corrections, and I am not interested in >20kHz reproduction either...), and things will only get better in the coming years.
I don't see myself going back to computer based FFT convolution anytime soon
A few month ago I had to made a decision wether to use a complete hardware like MiniSharc or to build a PC for convolution. My requirement was a "black box" which acts as a FIR crossover with 8 output channels. I ended up with a small PC on Linux basis. It was cheaper and much more powerful than the MiniSharc. And it can directly stream my music. No jitterful S/PDIF or AD/Da conversion is involved.
As sound card I use a Xonar Essence STX II 7.1. I wanted an internal solution and no USB involved. So I can turn-on and shut-down the "black box" with one push the power button. And my little daughter can not pull out any USB cable.
And btw. the Xonar has much better D/A converter than the MiniSharc. And filter length is not restricted which is great in the low end.
I use a Debian Linux with command line interface only on a small SSD. MPD accepts the music stream and BruteFIR does the convolution. There is also a small program that does the volume control and is triggered by a HID remote. The whole signal chain is in 64 Bit floating point. At the end there is high quality dithering to 32 Bit.
This Linux box is very stable. I had no crash and no malfunction yet. And I don't need any display or keyboard to control it. The HID remote is enough and the streaming sender can be a smartphone or another PC in the network. It works very well and you can not mal-operate the box.
On the downside I had to spend much time to get it running that good.
PS: forgot to mention that the PC was cheaper than a MiniSharc with case.
Attachments
Last edited:
88.2 and 96 kHz. CD material is upsampled to 88.2 kHz, while 48 kHz runs at 96 kHz. Of course 88.2 and 96 kHz material run at the native sample rate. I always try to up- or downsample in integers. Sample rate is easy to set up in JRiver.
For anybody wanting to experiment with multichannel DACs I can recommend ESI Gigaport HD+ at about 150 USD. The sound quality is far better than much more expensive interfaces like RME Fireface. However my present MOTU 1248 is a giant step up from the ESI.
I also find the sound quality from ESI's 2 channel DACs to represent an extremely good price/performance ratio. Many studio DACs give very much value for money; AVID is another example. When you hit the price level of about 2000 USD it can be discussed how much extra funtion is gained from more money.
Thank you.
The PC is ready in about 15 s and shuts down in 2 s. I would call this pretty fast.
It has now such a stable state that I never have to connect to the command line via SSH to correct things. I just stream with another PC and foobar via an UPnP renderer. This works without drops or failures. Audio format is set automatically.
I even wrote a small script to route the analog input to the outputs using the complete convolving chain. One push on the remote control switches to "measuring mode" and I can take measurements.
Changing the filter files is not automated yet, but this is not a big thing. A simple share to the filter folder should to it.
Last edited:
folks,
I had a question regarding measurements and it applies to PC and dedicated DSP platforms.
Tools like ARTA use a full duplex interface for loudspeaker measurements, for sending the stimulus and receiving the measurements from microphone.
For example I use USB Focurite 2i2.
How does this work when using PC or say Najda for DSP XO.
ARTA's stimulus should flow through the DSP, how is that done? DSP/DAC output is multichannel to the amps driving individual drivers.
Now where does the microphone captures get back in from?
Can measurement software receive microphone's input stream from one USB port through focusrite, while it sends the stimulus through Jriver convolution engine and out through another USB port to DAC/amp and drivers.
There are two related parts, 1. measuring individual driver response in cabinet without any XO, 2. Full loudspeaker response measurement with XO and drivers in place.
thanks for sharing how you go about with this.
I had a question regarding measurements and it applies to PC and dedicated DSP platforms.
Tools like ARTA use a full duplex interface for loudspeaker measurements, for sending the stimulus and receiving the measurements from microphone.
For example I use USB Focurite 2i2.
How does this work when using PC or say Najda for DSP XO.
ARTA's stimulus should flow through the DSP, how is that done? DSP/DAC output is multichannel to the amps driving individual drivers.
Now where does the microphone captures get back in from?
Can measurement software receive microphone's input stream from one USB port through focusrite, while it sends the stimulus through Jriver convolution engine and out through another USB port to DAC/amp and drivers.
There are two related parts, 1. measuring individual driver response in cabinet without any XO, 2. Full loudspeaker response measurement with XO and drivers in place.
thanks for sharing how you go about with this.
Last edited:
XO usually belongs with the speaker. However in the HTPC approach they are in a software running on the OS.
Once the filters are implemented using rephase and setup using Jriver, can they be in the path of all music sources played in the PC?
1. music streamed from pandora, youtube
2. CD music from the bluray player on the HTPC
3. some high res music and CD-rips on the hard drive
Once the filters are implemented using rephase and setup using Jriver, can they be in the path of all music sources played in the PC?
1. music streamed from pandora, youtube
2. CD music from the bluray player on the HTPC
3. some high res music and CD-rips on the hard drive
jojip,
Can Arta be set as REW because following works for me measuring with REW and using JRiver as DSP for all sound streams.
Use JRivers virtual soundcard to listen into measurement chain, see JRivers help or website if virtual soundcard is not installed by JRiver setup. This soundcard if set as default in windows settings or manual is set for every program that create sound stream, will make sure all sounds be routet through JRiver DSP engine including web browser sound streams, but be sure JRiver program always is open or running in background, picture 1 is virtual sound card device. I don't view video on this computer its only used for audio but seen some articles annouced at JRiver website how to have video in sync with the delay that routing all sounds via JRiver DSP engine create.
Picture 2 is REW settings where JRiver soundcard is set instead of real hardware output device, and inside open JRiver program the real hardware device is set as output device.
Picture 3 is JRiver DSP window that always is open when measuring, example make a sweep tick off a filter or on to a new filter and compare plots, same goes for sum of multiway speaker verse single LF/MF/HF measurement just tick off/on channels and sweep again to compare.
Can Arta be set as REW because following works for me measuring with REW and using JRiver as DSP for all sound streams.
Use JRivers virtual soundcard to listen into measurement chain, see JRivers help or website if virtual soundcard is not installed by JRiver setup. This soundcard if set as default in windows settings or manual is set for every program that create sound stream, will make sure all sounds be routet through JRiver DSP engine including web browser sound streams, but be sure JRiver program always is open or running in background, picture 1 is virtual sound card device. I don't view video on this computer its only used for audio but seen some articles annouced at JRiver website how to have video in sync with the delay that routing all sounds via JRiver DSP engine create.
Picture 2 is REW settings where JRiver soundcard is set instead of real hardware output device, and inside open JRiver program the real hardware device is set as output device.
Picture 3 is JRiver DSP window that always is open when measuring, example make a sweep tick off a filter or on to a new filter and compare plots, same goes for sum of multiway speaker verse single LF/MF/HF measurement just tick off/on channels and sweep again to compare.
Attachments
Last edited: