Forum Replies Created
-
8 November 2024 at 8:54 pm #940In Second view/group delay of your filter example I see group delay is same in whole band for each band(HF, MF, LF), but on different levels. For HF it is less then 1ms, for MF it is around 3ms and for LF it is around 20-30ms. Can bands be adjusted to same group delay ?
Yes. The different group delays relate to the different “delay” parameters in each of the FIR filters. To align these, additional delay needs to be added to the MF and HF so that their group delay matches the LF. “FIR Designer M” automatically tells you what delay to add in the “FIR Alignment Delay” on the System “Export” tab, and this delay can be included in the delay blocks in the processor (and not in the FIR filters). This alignment delay is already included in the “Wavelet” plot when viewing the wavelet of the combined system response.
Another problem that I can’t solve is very steep crossover between LF and MF area(about 300Hz). Even Brickwall is chosen, in export system response graph it seems slope is not brick wall like at 2kHz(MF/HF point). At 300Hz it looks more like LR4 slope shape. I tried to get brick wall shape at 300Hz, but I couldn’t. How I can achieve brick wall shape at 300Hz crossover ?
On both the channel target and channel mag adjust tabs, either set the Brick-wall filter to be longer – by increasing the time length in “ms” – or use one or more higher order filter prototypes. Also, on the channel export tab you will need to set the filter to be longer, to ensure the filter can realise the steeper crossover shape. Remember, steeper crossover filter shapes, implemented in FIR filters, require longer FIR filters. It’s also worth considering the capabilities of the processor you will be using. For example, if the processor can only run 4096 length filters, then you need to adjust your crossover steepness to be realisable within 4096 taps.
4 November 2024 at 9:05 pm #938>> Group delay in this case is maximum around 2-3ms for HF and around 4ms for MF. It
>> seems too high compared to maximum around 1ms for LR4. According to some studies
>> more then 1ms in area of 500-5000 Hz could be audible. My idea is to make very
>> steep crossover around 72db/octave for both crossover points in 3way and Group
>> delay not higher then 1ms in 50-20kHz band. Could you suggest which type of
>> crossover I should choose. Maybe instead brickwall I should use LR12The potential audibility of group delay relates to the audibility of the variation in group delay across frequency. Linear-phase filtering inherently has no variation in group delay across frequency; i.e. the group delay of a linear-phase filter is pure delay only.
The Brick Wall set to 20ms should get close to what you need. Also, we provide LR up to 8th order, which is 48 dB/oct. In the Target tab and the Mag Adjust tab, you can combine an 8th order LR with a 6th order LR to give ~72db/octave – both set to linear phase. Or even try the Elliptic filter prototype – set to linear phase – and experiment with the filter order and stop band attenuation to achieve the slope you want.
Remember, the steeper crossover filter shapes, implemented in FIR filters, require longer filters.
As to your requirements, if you set up the processing to essentially be linear phase from 50 Hz to 20 kHz, then any variation in group delay across frequency is essentially zero, but the filtering will have a bulk delay. Again, take a look at the wavelength view to see this.
If you want to use minimum-phase filtering to achieve your goals, I can’t recommend any particular approach. You will need to experiment.
EA Support
25 October 2024 at 6:54 am #934Hi Mark,
Version 4.3.9 is now available. Restart FIR Designer and it will alert you and provide a link.
Regards,
EA Support
25 October 2024 at 5:42 am #933Hi Mark,
Our apologies. We know what the problem is and we’ll have a fix out later today.
Regards,
EA Support
11 October 2024 at 8:57 pm #930Version 4.3.8 is now ready. Start FIR Designer and watch for the update notification.
Regards,
EA Support
11 October 2024 at 7:11 am #927Thanks. I’ve found the problem. (It only started in 4.3.6.) We’ll have a fix out shortly.
Much appreciated.
EA Support
11 October 2024 at 5:27 am #925I just tested it here again and it appears to be working. The “Range” controls the calculation that updates the “Magnitude: max…” and “Phase: max.. ” numbers, directly below the “Range”. It doesn’t update the plot display. Since the error mostly occurs at low frequencies, if you increase the range low freq. number, the error numbers should reduce to close to 0 dB and 0 deg. Can you confirm this?
Regards,
EA Support
10 October 2024 at 6:21 am #923What version of FIR Designer do you have? Are you on macOS or Windows? Are you running the demo version or a licensed version?
Regards,
EA Support
6 October 2024 at 7:32 pm #919In my experience pre-ringing is fairly subjective, and it’s audibility can depending on frequency range and listening ability. Some people claim to hear pre-ringing more clearly over headphones than through loudspeakers.
The linear-phase brick-wall crossovers in some commercial speaker processors, and available in our software, have visible pre-ringing but are generally considered inaudible.
Regarding your design, if you’re concerned, try setting up a blind listening test both with your loudspeakers and maybe over good quality headphones.
Regards,
EA Support-
This reply was modified 1 year, 8 months ago by
EA Support.
26 September 2024 at 7:43 am #915In the globcon software, go to Help -> Documentation and search for CSV. It has a section “CSV File Format” which states the CSV file must be text with 2 columns – the first column frequency and second column magnitude – and with the the first line “FREQUENCY,MAGNITUDE”. “NOTE: CSV files to be imported in Custom EQ currently cannot contain any phase information; only magnitude curves are valid. Import of complex custom filters with phase information might be implemented in the future.” You can make these files yourself for testing.
Our products can import responses as CSV files or TXT files in freq|mag, freq|mag|phase or freq|mag|phase|coherence and use them as the basis for the filter design. Our products can also export FIR filters as CSV and TXT, but these are a single column of coefficient values. On eclipseaudio.com, see the “Import|Export” menu for more details.
We’d love to collaborate with DirectOut and we spoken with them on a few occasions, including at InfoComm. If you have contacts and can nudge them along, we’d be very appreciative.
Regards,
EA Support
19 September 2024 at 7:28 am #913With ‘FIR Designer M’, the desired crossover responses can be included in the per-channel target responses, in which case the Auto IIR will push towards the crossover target. However IIR filters tend to minimum-phase, so, depending on what type of crossover target you are using, the result of the Auto IIR might have the desired crossover magnitude but not the desired phase.
Best,
EA Support
18 September 2024 at 6:45 pm #911Hi Mark,
We don’t prefer any particular method over the others. The multiple methods are provided in case one method struggles with a particular spectrum.
Each method various in whether they use the power spectrum or log (dB) spectrum, and how they place each filter’s centre frequency.
We do recommend smoothing to 3rd or 6th octave (in the Auto IIR window) before it places filters, otherwise the calculation will use way too many filters.
Regards,
EA Support
29 August 2024 at 7:31 pm #907Glad you found the latest version. (The issue was in the QT toolkit and was fixed quite some time back.)
Regards,
EA Support
24 August 2024 at 1:02 pm #901Hi,
Sorry to hear this. This is first crash report we’ve had in a while.
Re 1.
– What ASIO buffer size are you using? I suggest something larger than 512 samples at 48 kHz.
– Can you try a different audio interface? We’ve done most testing with RME interfaces and their drivers are very stable and well maintained. We’ve also used MOTU and a few others.Re 2. Can you record a video clearly showing the UI prior to and as it crashes? There may be clues in exactly what stage the UI is at before the crash.
Re 3. I can’t explain this but it’s likely related to the audio interface issues.
Regards,
EA Support
23 August 2024 at 7:21 am #897Sorry to hear this. We’ll need some more information. If you have a particular system file you’re using at the time the crashes happen, please email it to us.
How many channels are you using?
What are the sample rates of the channels?
What IIR mode are you using?
Do you have any input position IIR filters enabled?
Is 2nd view open?
What tab or window are you working on when it crashes?
Are you using the IIR reference editor? (IIR mode converter)Regards,
EA Support-
This reply was modified 1 year, 10 months ago by
EA Support.
-
This reply was modified 1 year, 8 months ago by
-
|