Things are a bit untidy, so I’m going to build a system that will fit to the left of the TV, on the white shelf. There is a space about 26cm deep and 36cm wide, with no real height restriction.
The system I have planned is as follows:
Intel NUC as player
Asus Xonar u7 sound card
3 x Chi-fi TA2020 amp boards
The Heybrook Sextets, suitably butchered
The NUC should be capable of FIR filters, running a convolver in Reaper or JRiver. The sound card is capable of 24/192 (although the NUC probably won’t be able to process at that rate, we’ll see) and sounds pretty good. The amps are low power (about 6W really), but I generally like the sound of TA2020s, and this system will never go loud. It’s my Sunday morning system.
Progress will be slow, not least because the amps will take a month to arrive.
There’s an issue with one of them not being quite balanced. The drive units all make noise, but one is quieter at some frequencies. I could measure and try to work it out, but instead I’ll just bypass the XO and see if that fixes it. It’s probably an old cap. I hope!
OK I have all of the parts except for the one I need to get started! I decided to buy the plywood cut to size, and found this place: http://www.woodsheets.com/ Seems ideal - it cost £23 for seven small bits of ply cut to size. Their cutting will be way more accurate and clean than anything I could do. Unfortunately they still haven’t got round to doing it yet.
So instead, ten minutes on rePhase and building a linear phase crossover. I’ll talk much more later about getting an accurate crossover. To start, I just want something that works.
I need to create three files, which I’ll call B441.wav, M441.wav and T441.wav. These are the bass, mid and treble files at 44.1kHz. I’m also doing 96kHz for hi-res stuff. A linear phase crossover uses a convolver which effectively multiplies the music signal by another file that contains the crossover shape, producing a band-limited version of the music.
The options chosen in rePhase are very simple:
I’ve chosen a 48dB/octave (8th order) Linkwitz-Riley crossover. Shown above is the bass (low-pass). I’m crossing over at 250Hz and 4kHz, which is about what the design points were AFAIK.
I’m using 32,768 taps; more taps means a more accurate crossover. This number is overkill for the crossover I’m using here, but I want to see what the computer can do. I will need the taps when I start doing more processing (especially in the bass) later.
The centering and windowing are the defaults and are the best settings for what I need. The output format is what is needed to work with JRiver.
You can see in rePhase how accurate the output will be - the target is a blue line, and the output is the red line that almost completely obscures it. The errors are at -130dB and below, so way below noise level, let alone audibility.
Once I have the rePhase files I need to create the config files to tell JRiver how it works. These look like this:
There is also a 96kHz version, which is exactly the same but 441 is replace with 96 (including in the name)
The important parts here are as follows:
Name - JRiver will automatically change between 44.1 and 96kHz sampling rate if the files are named correctly.
Top row - sample rate, number of channels in (2) and out (8). I use 8 as that is what the sound card has, obviously two of them will just be empty in a three-way (six channels needed).
Bass, left channel - row 3 is the file to convolve, row 4 is delay (I’ll do that in JRiver not here), row 5 is the input channel (left) and row 6 is the output channel (main L).
It then repeats for the six channels I’m doing anything with.
And that’s it. Next I need to set up the computer (which needs everything installed) and configure JRiver.
The wood will be arriving tomorrow, although being school holidays I doubt I’ll be able to make much progress on anything.
Decided to get the computer up and running. Installed JRiver and it can run the convolver at 10x real time, so that’s all good, and that was while importing the music library and various update/security things running. Sweet.
That’s the box overall, there are two shelves to go inside it. The bottom shelf has the mains inlet (which will switch only the amps) and a fan inlet; this section is the only one that is closed at the back because mains voltage. It will house the power supplies.
The next shelf has the three amplifiers, next up the DAC and an external HD, and the PC will sit on top (mainly so I can access its power switch).
Looking on the software side, I had an interesting and annoying issue - playing music remotely. I have no idea why this is a problem nowadays, but it is. I want to be able to play my own music, and Qobuz (or any other streaming service I fancy). Qobuz does not have a connect facility like Spotify Connect, which simply allows you to play on a remote device while browsing, selecting and generally controlling from your phone or tablet. Neither does Tidal IIRC.
I considered using a Chromecast Audio, and routing the digital output from this through JRiver as crossover. It would work but be rather clunky. Actually it would be fucking barking to use such a system in 2019 (or 2024, when I might actually have finished this thing). All of the streaming services seem intent on adding devices, not improving the system in their Windows programs.
As is often the case, the best solution is achieved via Logitech Media Server, the old Squeezebox system. I have installed Squeezelite-X from the Microsoft Store (it’s free), and this is a player that works with ASIO, so can play bit-perfect to JRiver. There is full Qobuz integration (and Spotify, Deezer, Tidal…), so it Just Works, and can be controlled from any phone or tablet. I’ve set it up on my main PC to try it out, and it’s working flawlessly, with the sample rate changing as expected so I know it’s bit perfect. Sweet.
I had Reaper running a while back but moved to JRiver when they brought out their WDM system. With newer versions it also changes sample rate very well.
Do you have any views on digital processing at higher sample rates? Is it worth doing?
It was before the asynchronous usb chips were more common, like the xmos etc.
I run an analogue source (tt) and with other setups the input clock and output clock would lose sync and cause grief. Not an issue these days really.
But really, it works, was fairly high end so made well and I cant be bothered to change.
Specs are adequate too:
Sample rates: 44.1 - 48k, 88.2 - 96k (+/-10%); 176.4 - 192k
• Frequency response: 10 –20k (+/- 0.2 dB) at 44.1k
• Analog max levels: +6dBV, +20dBu, +24dBu max (set by internal jumper)
• Dynamic range: 114 dB A weighted (AD + DA) • THD+N: -105 dB (AD), -103 dB (DA)
