Forum Replies Created
-
19 June 2025 at 4:27 pm #1075What is the effect of running a 96kHz filter on the 93675 hZ DSP of Hypex? Will it work with a slight frequency offset in the magnitude response?
93.75 kHz is 97.66% of 96 kHz. So the net effect of running a 96 kHz designed filter at 93.75 kHz is that the effective frequency response will be 2.34% lower in frequency. (For example a response peak designed for 18 kHz will actually be at 17.59 kHz.)
Regards,
EA Support
17 June 2025 at 11:26 am #1073Take a look at sections 5 and 7 of
I always set Filter-Delay to 0 in FIR Designer ….. This is to slightly increase the resolution of the low frequency control within the limited number of filter taps.
The overall filter length is what mostly affects the low frequency resolution and control. The delay setting mostly relates to the type of filter you are making. With a delay of 0, you’re making a minimum-phase filter, which you could also do with IIR filters, or a mix of both IIR and FIR.
Regards,
EA Support
16 June 2025 at 6:51 pm #1071That sounds like a question for Rational Acoustics. It’s beyond the scope of our software support.
Best,
EA Support
13 June 2025 at 7:41 am #1068Hi, yes they support FIR with 1500 taps per channel, 3 channels in case of DSP3-213.
Great, that’s 16ms per channel. So to clarify, what you need is FIR Designer to be able to create FIR filters at 93.75 kHz?
We would consider enabling 93.75 kHz if we could also emulate the Hypex IIR filters, as part of complete preset creation. For that we need IIR filter information from Hypex – see the “FIR Designer Brand Integration” link above. If you have contacts at Hypex, please reach out to them. We have not had any success contacting them in the past.
Regards,
EA Support
8 June 2025 at 6:23 pm #1065However, if I import RTA data with banding set to “None” and then run ‘Octave smoothing’ the tilt does not change.
That is correct. FIR Designer works fundamentally with transfer functions, not power spectrums. So when a RTA spectrum is imported into FIR Designer, it’s shape is essentially treated as a TF spectrum (but without any phase), which is why the smoothing function does not act like “banding”.
We provide RTA importing into FIR Designer for convenience, since often it’s helpful to design EQ based on a RTA measurement. So when importing a RTA measurement, we recommend applying the “banding” in SMAART before importing.
Please add “stop generator after Capture’ beside of ‘Signal generator”
‘Signal Generator’ buttons are too far from ‘Reset Time averaging’. Please move to above ‘Signal generator”We will consider this but no promises. 🙂
Regards,
EA Support
6 June 2025 at 2:25 pm #1062I don’t know why. The perception that “FIR filters are always linear-phase” is more apparent in pro audio and hifi audio than in other fields, like telecommunications. It may be due to audio folks historically wanting low-latency minimum-phase filtering which is typically done with IIR filters OR linear-phase filtering which can generally only be done with FIR filters. And therefore the fact that FIR filtering can also do mixed-phase and maximum-phase filtering has not been considered until recently.
Best,
EA Support
5 June 2025 at 2:57 pm #1055Here are a few suggestions re your latest system file:
– the Gain Polarity Delay tab still has some common delay across all channels. Change the LF, MB and HF delay to 5, 0 and 10 samples. (This reduces some of the bulk delay that the Auto Phase tab is trying to remove in the HF.
– the Auto Phase tab in the HF is still being used to undo bulk HF delay. On the Import tab, shift the timebase of all channels by setting the “Delay: samples” to -468 samples.In addition to the above, I’ve made a few tweaks across all tabs and shortened the FIR filters. I just emailed you the file.
Also, if you haven’t tried this already, use the “Second View” window to watch later parts of the workflow while working on earlier parts of the workflow.
Regards,
EA Support-
This reply was modified 1 month ago by EA Support.
3 June 2025 at 12:29 pm #1053Hi richardsllc,
I’ve receive the files and looked them over. Here’s some initial thoughts.
I had assumed that you had obtained magnitude and phase measurements of the processing for an existing EV preset, and that you were trying to replicate this processing using IIR, FIR or IIR+FIR. (This is very straight forward with FIR Designer M and a safe path for getting a good result, benefiting from all the manufacturers expertise.) However the files appear to be individual driver measurements of a cabinet (or cabinets).
Before I talk about the FIR Designer M system files, I need to talk about making presets. Designing multi-way presets, particularly for pro audio cabinets, involves optimising many factors including response on axis, response in different directions within a desired coverage angle, consistency of directivity across frequency, and system robustness at high levels. For example a DSP design which gives a good on axis response at low levels could damage a driver at high levels or (if using limiters) cause some drivers to limit earlier than others (causing dynamic fluctuations in response and directivity). This may then require moving crossovers higher in frequency – to reduce the burden on the MF or HF – which then affects directivity, and so on, and so on. All this to say, I would be very nervous trying to make a preset from on axis measurements, without having the ability to measure the system at other angles (e.g. using a turntable or multi-axis robotu) and without being able to test at high levels (and be willing to burn drivers).
Anyway… regarding your FIR Designer M system file.
IMPORT tab
The measurements you provided all have a time offset of roughly 9.5ms but the offsets are not all the same. Were the measurements captured with a common timebase, or are they completely independent? The slight time delay differences seem to match the different physical depths of the drivers from a XVLS photo I found online, but it’s important to be sure. If they weren’t captured with a common timebase, then you will need to take more measurements.
In FIR Designer M, the ~9.5ms delay is present in the loaded measurements in each channel. If you remove this delay, much of the high phase slope goes away, making the workflow easier. For example, on the Import tab “Delay Adjustment” “Delay: samples” to -458 samples.
For each channel, turn on “Hide phase” and for LF set “>” to 1000 Hz, for MB set “>” to 5000 Hz and for HF set “<” to 200 Hz. Now the magnitude and phase plots become easier to view as you work through the workflow.
GAIN/POLARITY/DELAY tab
All channels have some non-zero delay. The G/P/D tab is about adjusting the alignment between the channels. Any delay common to all channels is just bulk delay added to the whole system, which isn’t necessary and can be removed.
IIR FILTERS and FIR tabs
There are IIR HPF and LPF filters enabled AND HPF and LPF filter prototypes enabled on the Mag Adjust tab. These are causing major problems with the preset. Choose one or the other, but not both. Generally HPF and LPF filters are more efficient and effective as IIR filters, leaving the FIR to deal with more delicate mag and phase adjustment. If you definitely want linear-phase crossover HPF and LPF’s, then turn off the HPF and LPF filters in the IIR filters tab.
The Auto-phase that is enabled on the HF is unnecessary. It’s just trying to undo bulk delay and wasting FIR taps.
I have just emailed you two new systems. One is IIR only and the other is FIR only. Please study them carefully in terms of how to use the workflow, but don’t rely on the actually settings… see the “Before..” paragraph above. 🙂 Our support is focused on how to use the software.
Regards,
EA Support
30 May 2025 at 1:33 pm #1050Please send your text file, the FIR Designer M system file that you have created, and your FIR Designer M license ID number to the email address ‘info (at) our website name’.
(Note, here you’ve replied to a different topic. Best to either reply to the previous topic that you started, or create a new topic.)
Regards,
EA Support
22 May 2025 at 4:06 pm #1045Version 4.4.2 is now available and fixes the issue.
Regards,
EA Support
22 May 2025 at 4:06 pm #1044Version 4.4.2 is now available and fixes the issue.
Regards,
EA Support
22 May 2025 at 9:51 am #1042Ok, we’ve found the problem and it has nothing to do with the sample rate. We need to chase this issue with Rational Acoustics as it’s something inherent in the TRF measurement at very high frequencies. Like I said above, below 20 kHz, everything is ok so you can continue with the filter design process.
Regards,
EA Support
22 May 2025 at 8:24 am #1040Files received, thanks. The magnitude, phase and coherence of the file is loading correctly, so you can continue through the FIR Designer M workflow as normal. The issue appears to be in the sample rate conversion from the file’s 44.1 kHz sample rate to the FIR Designer M channel sample rate – e.g. 48 kHz – and only affects frequencies near the half Nyquist frequency (22.05 kHz to 24 kHz). Note: if you set the channel sample rate to 44.1 kHz, the impulse response is clean.
We’re looking into this now.
Regards,
EA Support
21 May 2025 at 7:37 am #1038Hi hulkss,
Please email a few of the TRF files to info (at) our website name, along with a screenshot of what you see in SMAART, and we’ll take a look.
Regards,
EA Support
17 May 2025 at 9:04 pm #1035Hi Bido,
Firstly, the filter length of 400 samples at 48 kHz (8.3ms) is much too short to achieve the magnitude changes and phase correction that you’d like. Take a look at the “FIR Filter Frequency Response” and “Total Error” graphs in your FIR Creator project. To achieve the EQ your desire, particularly at 100 Hz, you need a FIR filter length closer to 2048 samples with a delay of ~400 samples. Take a look at Section 5 on https://eclipseaudio.com/fir-filter-guide/
Secondly, looking at the original measurement – your 1st Smaart screenshot…. The phase rotation at ~760 Hz is an acoustic cancellation – most likely from a reflection – as evidenced by the corresponding dip in the magnitude response and dip in the coherence. No amount of DSP (in magnitude and/or phase) can fix this. Just considering the magnitude response, if you try to boost this with EQ, very little will change when re-measured. In your case, adding the 2nd order phase rotation in DSP is just widening the acoustic cancellation region. (Try removing the 2nd order phase filter and the remeasure, and I think you will see a mag dip similar to the unprocessed measurement.) The other frequencies where the coherence dips – e.g. 90 Hz, 390 Hz and 440 Hz – also correspond to magnitude dips which look like cancellation effects, likely from reflections.
DSP can address phase effects inherent in the loudspeaker, like from crossover filters, but little can be done about cancellation from reflections.
Best regards,
EA Support -
This reply was modified 1 month ago by
-
|