ShinOBIWAN said:
The little I know about NOS,are you sure the fittings aren't leaking. That would explaine the laughing feeling.
😀
Very nice work with the speakers,design and all are very stylish honestly!!
pikkujöpö said:
The little I know about NOS,are you sure the fittings aren't leaking. That would explaine the laughing feeling.
😀
Very nice work with the speakers,design and all are very stylish honestly!!
There's a pressure release valve on each tank which gets rid of excess NOS should their be too much in the lines running to the injectors at the front.
After I had it fitted and the guy was going through the instructions, he said "if you hear a hissing coming from in the back, then put the windows down otherwise you'll be laughing ya t*ts off." 😀
One down, three more to go...
Neeeed the 8" Flex units like yesterday. Not sure what Per is doing, its been a month and half now. I think its a test.
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.
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.
Neeeed the 8" Flex units like yesterday. Not sure what Per is doing, its been a month and half now. I think its a test.
Cheers dudes.
There's about 12 hours of work there so they'd better look good. The finishing is a killer. To top it off the portable radio, that I listen to during this, only picks up Radio1 and if I hear about the soddin' Artic Monkeys once more I swear I'm going to hurt something.
There's about 12 hours of work there so they'd better look good. The finishing is a killer. To top it off the portable radio, that I listen to during this, only picks up Radio1 and if I hear about the soddin' Artic Monkeys once more I swear I'm going to hurt something.
Member
Joined 2002
Grumpy_Git said:
"Sir, Please step away from the loudspeaker cabinets!"
Nick
After all that hard work I think people would realise its rude to start wiping their hands on them. So I'd hope good taste and manners would prevail. But if I do catch someone touching then I'll beat them with a copy of the Arctic Monkeys album.
ShinOBIWAN said:
After all that hard work I think people would realise its rude to start wiping their hands on them. So I'd hope good taste and manners would prevail. But if I do catch someone touching then I'll beat them with a copy of the Arctic Monkeys album.
Better make sure you've got that album on 200 grams vinyl then.
😀
Cabinets look absolutely fabulous.
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.
An externally hosted image should be here but it was not working when we last tested it.
This may have been discussed, but what is that foam exactly? Is it aesthetic or does it really serve a purpose?
Its F3 grade felt rather than foam and not really going to impact performance to any significant degree because its too thin for that considering how far the mid reaches up. So consider it aesthetic.
The same felt will be applied to the tweeter and that will have an effect however.
The same felt will be applied to the tweeter and that will have an effect however.
Care to elaborate on your spraying equipments and steps?
Absolutely stunning work 😎
I'll check back for updates 🙂
Absolutely stunning work 😎
I'll check back for updates 🙂
sqlkev said:
Here's the equipment and paints on this page:
http://www.diyaudio.com/forums/showthread.php?postid=888407#post888407
This is link to some paint related ramblings from me to Vik, which he kindly reposted in another thread. They're spead out over a couple of pages I believe:
http://www.diyaudio.com/forums/showthread.php?postid=908705#post908705
Some better daytime shots:
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.
An externally hosted image should be here but it was not working when we last tested it.
Shin,
First off, stunning work! Can't stress that enough. The things you make are simply drool-worthy. One day I'll document my own speaker project(s) I'm sure. Might have to wait until they're worthy of it!
Now to my point:
Have you decided what you want to do for your crossovers, yet?
I've been following you since your early Perceive days, when you inspired me with your posts on PC-based crossovers. I'd had the same idea already, but didn't realize how far it could be taken - I was using my Soundblaster Audigy2 with the kX drivers at the time. In all fairness, that was a hell of a steal - the whole thing cost like £35.
Your setup using Console was a revelation. Offloading the calculations to a powerful CPU and using a quality filter system (Waves) was something that I didn't know was possible. So I tried it, and I liked it!
However, as with most of the folks on here, I couldn't leave well enough alone and began to experiment. First off, after hearing about the adventures of people using BruteFIR on Linux, I ditched the LinEQ plugin and decided to try Convolver (www.convolver.sf.net) - this is an excellent open-source plugin based on the same FFTW as BruteFIR. It also implements the famous "partitioned convolution" - if you have a powerful enough PC, you can reduce your latency down to only a few ms.
Using Convolver also allowed me to experiment with designing my own filter kernels in MATLAB, which was a lot of fun. LinEQ had no flexibility at all in this regard.
I considered this a major step forward in flexibility. It also used less CPU power and had much lower latency. All-round good.
Then.
I started my industrial placement year with Meridian Audio last summer (am still there now) and was lectured all about the evil that is "Floating Point Numbers In Audio". I read the AES papers on precision and dither and realized that (not wholly surprisingly) they were absolutely right - floating point numbers are quite simply wrong for audio. They are used because they run quicker on an x86 processor - and that is the only reason. Dedicated recording systems such as ProTools could use whatever number format they like - and they use fixed-point. It turns floats are often good enough for certain applications in practice, but fixed-point maths always does a better job.
To that end I began to write VST plugins to replace Convolver (which is 32-bit floating point) and the various Waves plugins (which frankly could be any old number format) to use inside Console. It quickly became apparent that a single-precision floating point number could encapsulate a 24-bit integer value, and so you can losslessly use 24-bit fixed-point maths inside your individual plugins dithering back down whenever an operation produces a 48-bit result. At the very end you put your nice 24-bit dithered result back into a 32-bit float and send it out of the plugin.
I ended up with VST plugins for multichannel gain, biquad filters and FIR filters. I measured them extensively on an Audio Precision System Two Cascade instrument and they worked exactly as I wanted. Which was excellent.
Nearly there now.
The latest development was the discovery of an cross-platform open-source audio backend called PortAudio (www.portaudio.com). It's quite simply fantastic. Using this library you can create a ground-up audio processing application with about 50 lines of code. I found this about three weeks ago and leapt on it. One of the secondary benefits of PortAudio is that you can get 32-bit integers directly from the ASIO driver of your sound card, and give back 32-bit integers at the end - no messing around with floats.
Just now I'm fine-tuning the code, but I have a complete 3-way speaker crossover system, comprised of:
- FIR filtering between tweeter and mid (128 taps),
- Two cascaded biquads (making a fourth-order butterworth) filters between the mid and the woofers.
- Variable gain at the end of the chain
- TPDF dithering down to 24-bits for final output.
The individual modules all have a dithered 32-bit noise floor, which would be some -220dBFS if I could measure it (which I can't). The final dithering to 24-bits produces a pristine flat noise floor of around -175dBFS, with perfect linearity and no noise modulation.
(I used a 16k-point FFT and an input frequency exactly on an FFT bin - thus no need for a window.
I measured the whole thing on the AP this evening after work and it's all working perfectly!
For reference, at 44.1kHz the program runs using about 75% CPU of a PIII-700, introduces ~130 samples of latency (2.9ms) and consumes around 2.5MB of RAM while running. The executable file is 17kB.
Now, I realize that I've rambled on at quite some length! My original point was that if you're interested and you still have your Apogee/Lynx/RME kit (can't remember what you ended up keeping!) I'd be happy to let you - or anyone else interested - have the code for this. It's not too tweak-friendly at the moment - changing coefficients requires a recompile! - but I find it to be the best-sounding system I've heard so far (I'm using the Seas millennium/W18EX001 combination, plus a pair of side-firing 8" Seas woofers per speaker). It's also quite basic in use - I've never programmed a Windows GUI before and so it's just a console app at the moment. Not ideal, but definitely functional.
Still, if anyone's interested, let me know 🙂
Thanks,
First off, stunning work! Can't stress that enough. The things you make are simply drool-worthy. One day I'll document my own speaker project(s) I'm sure. Might have to wait until they're worthy of it!
Now to my point:
Have you decided what you want to do for your crossovers, yet?
I've been following you since your early Perceive days, when you inspired me with your posts on PC-based crossovers. I'd had the same idea already, but didn't realize how far it could be taken - I was using my Soundblaster Audigy2 with the kX drivers at the time. In all fairness, that was a hell of a steal - the whole thing cost like £35.
Your setup using Console was a revelation. Offloading the calculations to a powerful CPU and using a quality filter system (Waves) was something that I didn't know was possible. So I tried it, and I liked it!
However, as with most of the folks on here, I couldn't leave well enough alone and began to experiment. First off, after hearing about the adventures of people using BruteFIR on Linux, I ditched the LinEQ plugin and decided to try Convolver (www.convolver.sf.net) - this is an excellent open-source plugin based on the same FFTW as BruteFIR. It also implements the famous "partitioned convolution" - if you have a powerful enough PC, you can reduce your latency down to only a few ms.
Using Convolver also allowed me to experiment with designing my own filter kernels in MATLAB, which was a lot of fun. LinEQ had no flexibility at all in this regard.
I considered this a major step forward in flexibility. It also used less CPU power and had much lower latency. All-round good.
Then.
I started my industrial placement year with Meridian Audio last summer (am still there now) and was lectured all about the evil that is "Floating Point Numbers In Audio". I read the AES papers on precision and dither and realized that (not wholly surprisingly) they were absolutely right - floating point numbers are quite simply wrong for audio. They are used because they run quicker on an x86 processor - and that is the only reason. Dedicated recording systems such as ProTools could use whatever number format they like - and they use fixed-point. It turns floats are often good enough for certain applications in practice, but fixed-point maths always does a better job.
To that end I began to write VST plugins to replace Convolver (which is 32-bit floating point) and the various Waves plugins (which frankly could be any old number format) to use inside Console. It quickly became apparent that a single-precision floating point number could encapsulate a 24-bit integer value, and so you can losslessly use 24-bit fixed-point maths inside your individual plugins dithering back down whenever an operation produces a 48-bit result. At the very end you put your nice 24-bit dithered result back into a 32-bit float and send it out of the plugin.
I ended up with VST plugins for multichannel gain, biquad filters and FIR filters. I measured them extensively on an Audio Precision System Two Cascade instrument and they worked exactly as I wanted. Which was excellent.
Nearly there now.
The latest development was the discovery of an cross-platform open-source audio backend called PortAudio (www.portaudio.com). It's quite simply fantastic. Using this library you can create a ground-up audio processing application with about 50 lines of code. I found this about three weeks ago and leapt on it. One of the secondary benefits of PortAudio is that you can get 32-bit integers directly from the ASIO driver of your sound card, and give back 32-bit integers at the end - no messing around with floats.
Just now I'm fine-tuning the code, but I have a complete 3-way speaker crossover system, comprised of:
- FIR filtering between tweeter and mid (128 taps),
- Two cascaded biquads (making a fourth-order butterworth) filters between the mid and the woofers.
- Variable gain at the end of the chain
- TPDF dithering down to 24-bits for final output.
The individual modules all have a dithered 32-bit noise floor, which would be some -220dBFS if I could measure it (which I can't). The final dithering to 24-bits produces a pristine flat noise floor of around -175dBFS, with perfect linearity and no noise modulation.
(I used a 16k-point FFT and an input frequency exactly on an FFT bin - thus no need for a window.
I measured the whole thing on the AP this evening after work and it's all working perfectly!
For reference, at 44.1kHz the program runs using about 75% CPU of a PIII-700, introduces ~130 samples of latency (2.9ms) and consumes around 2.5MB of RAM while running. The executable file is 17kB.
Now, I realize that I've rambled on at quite some length! My original point was that if you're interested and you still have your Apogee/Lynx/RME kit (can't remember what you ended up keeping!) I'd be happy to let you - or anyone else interested - have the code for this. It's not too tweak-friendly at the moment - changing coefficients requires a recompile! - but I find it to be the best-sounding system I've heard so far (I'm using the Seas millennium/W18EX001 combination, plus a pair of side-firing 8" Seas woofers per speaker). It's also quite basic in use - I've never programmed a Windows GUI before and so it's just a console app at the moment. Not ideal, but definitely functional.
Still, if anyone's interested, let me know 🙂
Thanks,
Interested, I think it would be worth starting a seperate thread for this. Would you want to create a small tutorial showing some detail on the steps you used to create your crossover? Thanks very much.
- Status
- Not open for further replies.