I'll be very interested in how you guys end up assessing what 2048 taps can do.
I've come to believe in a rule of thumb that says 500 tap works down to 500Hz.
And double the tap requirement for each octave decrease; halve it for each octave increase.
That's for 48kHz sampling. Of course, double all tap requirements if 96kHz.
Anyway, i'm thinking the 2x4HD's FIR is good to around 125Hz (assuming it runs at 48k).
Looking forward to findings, and how they compare to me thumb rule🙂
I hope you don't mind me posting this here but it's hard to pm images.
In your newest measurement taken at 90 degrees there is a massive spike at around 6.3 to 6.4ms it is very consistent between channels, it is only 5dB below peak. This is weird because it wasn't in the other ones. For now I will continue with the first measurement until you can work out what is going on. Works out to be a distance of about 2.1m.
In your newest measurement taken at 90 degrees there is a massive spike at around 6.3 to 6.4ms it is very consistent between channels, it is only 5dB below peak. This is weird because it wasn't in the other ones. For now I will continue with the first measurement until you can work out what is going on. Works out to be a distance of about 2.1m.
Attachments
In terms of distance, the 6,3ms = about 210cm = 2x distance baffle/front wall.
Not sure why it show now vs earlier measurements, the front have not changed - only mic orientation and cal file.
But cal file should not affect impulse right?
Not sure why it show now vs earlier measurements, the front have not changed - only mic orientation and cal file.
But cal file should not affect impulse right?
Last edited:
Our maths agrees 🙂
Cal file only affects spl graphs so that shouldn't be causing it. Could be a glitch as it is so loud. Perhaps try again and see, due to the change in orientation I would look at the ceiling but the arrays normally take care of that.
The miniDSP 2x4HD can handle a total of 4096taps, arbitrarily split over the 4 outputs. Minimum taps per channel is 6, so I can use 2042 taps each for two channels. (@96khz?)
Fluid have generated some test filters, which load to the DSP but apparently severely attenuate the output signal. Reason unknown.
The graphical filter display do not show this attenuation.
Work in progress 🙂
Last edited:
The filter won't run properly at 96K it needs to be at 48K. With the limited number of taps available the sampling frequency will need to be left at 48K.
The reason is because the filter is not normalised. Jriver's convolution normalises on the fly, MiniDSP must not be able to. The solution is to normalise the filters volume.
If you have Audacity you can do it yourself. Change the extension on the .bin file to .pcm and import the audio to Audacity, double click the waveform view to select all (you won't see much because the volume is low) then normalise. Export the audio as a raw file 32 bit float, (header-raw headerless / encoding-32bit float)
Change the extension back to .bin and make sure you turn the volume down because it will probably be quite a lot louder 45dB+.
If that sounds too hard I'll do it for you when I get home 😉
Most of the last paragraph above is probably wrong as the volume of the filter was quite high when I checked it, now I'm thinking sample rate.
I thought that the 2x4HD could be set to 48 or 96KHz like most of the older units could. I think I may have been wrong and it can only run at 96KHz because of the plugin it is using.
If that's the case all the files I sent were set for 48k. These won't work at 96K and could be the cause of the low volume. Seems the preEQ biquads are OK as REW knew they needed to be at 96K 🙄
I'll make some new files at 96k, but it will involve a bit of sample rate converting and it will effectively halve the tap count.
Seems ridiculous to me that MiniDSP would not give you option to set a lower sample rate with such a tap limited device 😡
I thought that the 2x4HD could be set to 48 or 96KHz like most of the older units could. I think I may have been wrong and it can only run at 96KHz because of the plugin it is using.
If that's the case all the files I sent were set for 48k. These won't work at 96K and could be the cause of the low volume. Seems the preEQ biquads are OK as REW knew they needed to be at 96K 🙄
I'll make some new files at 96k, but it will involve a bit of sample rate converting and it will effectively halve the tap count.
Seems ridiculous to me that MiniDSP would not give you option to set a lower sample rate with such a tap limited device 😡
Last edited:
Apparently 2x4H is 96khz only - MiniDSP : 2x4 HD (and DDRC24) sampling rate confusion (1/1)
I have not upgraded to DDRC24 yet.
I've seen reference of a double convolution workflows, but unsure if that applies here.
I have not upgraded to DDRC24 yet.
I've seen reference of a double convolution workflows, but unsure if that applies here.
Last edited:
That thread is quite amusing, my take would be that the quality is likely to improve by leaving the sample rate at 48K to give you more processing power. 96k filters should be on the way soon so hopefully that will give you something to test.
No fly 🙁
Managed to skip out from work in order to take a delivery at my house (270x TC9's...) and briefly tried to import the filters you have provided.
Managed to skip out from work in order to take a delivery at my house (270x TC9's...) and briefly tried to import the filters you have provided.
Attachments
Last edited:
270 TC9's? Does this mean I can order line arrays from you? Or are you planning on doing floor to stratosphere?
Haha 😀 6x25 per side perhaps? No, I have arranged a PowerBuy on behalf of a small group of line source builders 🙂
Unobtainium... PartsExpress out of stock, MOQ=2000units.
Same at Madisound and DigiKey. Also sent Peerless/Thympany a high count request (1000+) but no reply.
Same at Madisound and DigiKey. Also sent Peerless/Thympany a high count request (1000+) but no reply.
It would only take a couple of guys crazy enough to do 6x25 driver floor to ceiling arrays 😀.
Anyways, were there some drivers in that order destined for your own project? I know you've had a couple of damaged drivers, but I thought the new arrays were fully filled already.
Do you have alternative plans? 🙂
Anyways, were there some drivers in that order destined for your own project? I know you've had a couple of damaged drivers, but I thought the new arrays were fully filled already.
Do you have alternative plans? 🙂
I did get new ones for the 2x20 array, later decided to start 2x24 and was short again 🙄 So I have 56 new ones now, and the rest are spoken for 😀
I wouldnt mind making an mk3, but construction/estetique would be different.
However I would need to sell mk2 in order to fund an mk3.
Anyways, much too soon as mk2 isnt finished yet 😎
I wouldnt mind making an mk3, but construction/estetique would be different.
However I would need to sell mk2 in order to fund an mk3.
Anyways, much too soon as mk2 isnt finished yet 😎
Now you're just making me curious 🙂, what would a MK3 look like? 😀
Only hypothetically speaking 🙂.
Only hypothetically speaking 🙂.
Hmmm, you can see from the comparison graph that the 1020 tap filter was correct at generation as it was less accurate than the 2040 one but something has caused it to grow 4 times instead of two on re-sampling. Leave it with me....
I like the look of Mk3, maybe I'll make myself a Mk2 when/if I learn to drive the CNC 🙂