TOPIC: Feature request ideas for next rePhase update

Hi Pos,
The latest v1.2.0 version of rePhase is absolutely, completely brilliant - but I had a few ideas for some new features, please:

1. MOUSE CURSOR VALUES
Okay, I'm working in "Paragraphic Gain EQ" tab trying to tweak my EQ settings...
As I move the mouse cursor around in the graph area, those little grey animated boxes slide back and forth along the graph's edge, following my realtime cursor position, giving me very useful live numerical readout of my position.
I can move my mouse cursor to a certain peak and see "13846 Hz" freq in animated grey box - that's good - then I go down to click inside the little EQ window and manually type "13846" to set my EQ band correctly. I have to repeat this task over and over and over for dozens and dozens and dozens of EQ freq points - hand jumping from mouse to QWERTY keyboard to type numbers, press enter, then back to mouse, etc. Takes me a long time.

What I'd really love is to be able to position mouse cursor on the graph, then right-click, and see pop-up menu "COPY FREQ" and that current numerical value "13846" gets copied to the clipboard, and then I go down to the little EQ window and right-click and get menu option "PASTE" and it simply pastes "13846" for me - without having to manually type keyboard and no risk of human typo errors.
I could then whiz through the EQ process far quicker with my right hand staying on the mouse nearly all the time, dragging the dB faders with the mouse too. I'd need to trim the Q values by keyboard -with left hand fingers toggling the up / down arrows for that, while watching the graph, and right hand stays on mouse and I can concentrate on what I'm doing.

2. IMPORT & EXPORT - from HOLM Impulse v1.4.2.0
I seem to have great success exporting a text file from HOLM using "custom format", "FFT-frequencies" which seems to make a .txt file approx 2.5Mb in size with maybe 65536 points which rePhase recognises for measurement import. It's a revelation to be able to do that so quickly and easily. Thank you.
But once I've made my EQ curve and/or crossover slopes in rePhase, I'd like to be able to export that freq response and phase response curve as txt file in the same format, so I can open it in HOLM and try the product and sum manipulations with various curves, etc
I can't seem to get rePhase to do this, although I know it can export a txt file for the generated impulse values, but the contents aren't the same thing, so HOLM can't import it, and rePhase can't import it again as another measurement. I'd love to see in the Measurement Tab area an "EXPORT" button, to allow "export rePhase's current freq & phase curve as txt measurement" facility.

3. MIDI CONTROL
This is a bit far-fetched idea, but it would be lovely to be able to use an external USB - MIDI hardware controller with nice rotary knobs or physical sliders to give hands-on adjustment of rePhase parameters, particularly for repetitive mouse and keyboard jobs, where you're tweaking EQ values around by trial and error.

Keep up the fantastic work, and many many thanks for making rePhase so wonderful!
Very best regards,
Richard

Hi Richard,

Thank you for the kind words
Richard wrote: Great idea, I will see how this could be integrated.
Richard wrote: That something that has been requested several times already, and is definitly on the todo list.
In the meantime you can import the impulse into HOLM and get the same result, the only addtional step compared to importing a frequency response is that you have to manually set the delay.

Here is how to do it:

1. choose txt as the output format
2. generate the impulse
3. take note of the impulse delay in samples
4. choose the "impulse" tab in HOLM's import window
5. set the correct sampling frequency and choose txt format
6. load you impulse (for some reason the brwosing box only shows wav files by default, so you need to set it to *.txt)
7. go to the option window and enter the correct impulse delay to get a valid phase trace

Richard wrote: That would be interesting, but not likely to occur I am afraid

Thomas

Hi Pos,

Thanks for your reply. You were right about importing text impulse back into HOLM, and it did work. Thank you, it helped a lot.
Here's a few more ideas for new improvements for rePhase...

1. Please include 6144 taps among the taps factory preset values.



When you click inside this box, you can type your taps value manually, or use up/down arrows to scroll through the factory preset values.
There's already a huge range from 1 to 1048576, doubling in size with each step. It's very interesting to watch how accuracy improves to lower bass frequencies as you step up through these when generating filters. You can see how many taps you'd really need for a good job!

Working with OpenDRC, of course, I'm always needing 6144 taps, so I would like "6144" to be inserted into that list of factory preset values, please, for convenience so it pops up inbetween 4096 > 6144 > 8192 and then it's never more than a few click away, and you can't ever forget the magic number...!
I'm sure many other readers on this forum are also OpenDRC users, who would also appreciate 6144 taps option always being there.
If it were up to me I'd have 6144 as rePhase's default setting. But I'm sure users of other machines will request their favourite taps be default presets as well.


2. Please include these Nyquist frequencies in the kHz high frequency range factory preset values.



I've highlighted my suggestions in red.
These are the Nyquist limit upper frequencies for the possible digital sample rates in popular use today. (44.1kHz up to 768kHz)
Working with OpenDRC system being fixed 48kHz sampling rate, the Nyquist limit will always be 24kHz for those FIR filters, so I'm always setting 24kHz as the graph upper range (rather than typical 20kHz audible) because I want to see everything that's going on in my 48kHz digital FIR filters. If it was a factory preset value it would be very handy.

NB. The Nyquist frequency 22.05kHz (for 44.1KHz sampling) must be typed as "22050 Hz" because rePhase doesn't seem to like "22.05 kHz" for some reason...? I think it doesn't like having two decimal places in kHz range..?


Thank you, that's all for now!
I've got many many more suggestions, but I'll have to post them another day.

Richard

Hi Richard,
Richard wrote: I am reluctant to add anything platform-specific to the software right now.
I hope miniDSP will be able to blow away that 6144 taps limit soon anyway
Richard wrote: Very good idea. I will probably not add all of those values, but 22.05kHz and 24kHz do make a lot of sense.
And I will look at that "22.05" bug
Richard wrote: Keep them coming!
But please keep in mind I cannot promise anything about when (and if) these features will be included.

Thank you for your suggestions.

pos wrote:
Oh! Have you heard something about a upcoming product or firmware release then?
If only they could implement downsampling, and make the 48kHz machine run internally at 24kHz sample rate or even 12kHz sample rate it would be amazing for woofers, giving effectively 12288 or even 24576 taps !!!

Speaking of long taps of convolution, I wonder if you've ever seen this amazing product? Logidy EPSi

It is a 32 bit floating point DSP guitar effects pedal designed for stereo convolution impulse reverbs, but will accept any user impulse file for convolution as any mono or stereo, 16 bit or 24 bit, 44.1 kHz WAV file, via any standard FAT32 formatted SD Card, as long as the wav file is inside the right named folder on card. It even has the SD card reader built-in.

According to the Instruction Manual Page 10 - it can run 65536 sample stereo impulses !!!! Which is 1486 milliseconds!
Compared to OpenDRC 6144 sample impulse = 128 milliseconds!

This thing only costs $199. but a shame it's only a guitar type FX stompbox with unbalanced jacks only(!) and cheap 9V PSU rather than balanced XLRs and AES/EBU for serious use. But it should process the woofers perfectly if you ran a big 65536 tap FIR on it.
Why can't MiniDSP make a box like this, or three or four linked together for a crossover, inside a 1U rack case with professional I/O.
I'd buy that!
pos wrote:
I totally understand. I'm just suggesting whatever ideas seem valid to me, and hopefully a few of them will be good ideas for everybody..
Here's some more suggestions...

How many iterations? How long to wait until all the iterations are done?

When you click the "Generate" button, you're often in for a very long wait while rePhase renders the filter.
Several minutes... sometimes half an hour... sometimes several hours... if like me, you have MAX settings and longest possible FFT length, it can take ages...!

There is no taskbar filling up, or count-down timer telling you "37 minutes remaining." like with most slow software rendering, eg. video.

It would be great if you could add a taskbar, but I realise it's unknown how long the software will keep optimizing because it depends on the shape of the user's filter which is an unknown arbitrary quantity.

The only guide is the number of iterations used.
I know you can interrupt them - "Stop And Save" - but I usually let it do them all, so I go away from computer and do something else then come back much later when it's finished the whole process.
The screen then shows this important info...



But it DOES NOT show how many iterations it actually reached when it completed. In my tests, my woofers can take about 465 iterations, and tweeters over 2000 iterations, depending on settings, but I have to sit and watch carefully and note the number. Very boring!

If I decide to re-do a slightly different EQ filter again with similar settings, it would be nice to know how long it's likely to take.
The only guide is the number of iterations that shows you numerically how far through the overall process you are, which is likely to be similar if you're using the same FFT length and windowing settings again.
You could sit watching like a hawk until the very last one - then make a note of it - then you'll know how many to expect next time with those settings, and you can plan your time easier, but the iterations number instantly disappears as soon process is complete! Many times I've missed it, or gone away for half an hour when it might have only took 5 minutes.
I'd like the info screen to show the number of iterations, like this...



NB. Due to limited space in that green text area I had to remove the word "response"

This final iteration number is very useful info to me. It's the only numerical measure of the progress elapsed, and if I can know (from previous experience) to expect 475 iterations and I can see I'm only up to 67, I know if it's close to the finish or not, and can estimate roughly how much longer I've got to wait. Not knowing is frustrating, and when the number instantly disappears, you don't really get the full benefit of each experience you could have learned something from.


I'll write some more ideas soon...

Richard wrote: I don't have any special info, just hope
Like you I would love a see a downsampled version of the openDRC.
32kHz would be a reasonable target as it would already double FIR "expressiveness" and still be compatible with most DACs on the market.
www.minidsp.com/forum/suggestion-box/10749-32khz-opendrc-plugin
This one is impressive indeed, but of course the increase in taps is not worth the hassle of having to deal with unbalanced analog in/outs of unknown quality...
I wonder what DSP this device is using...
Sadly this is impossible to predict.
That said your settings seems a little bit to extreme
FFT length for example is mainly useful for visually judging the repsonse down low with more precision.
In most cases default settings should do just fine as far as iterative optimization is concerned.
In some rare corner cases a lower FFT might even give better results

Also please consider the fact that iterative optimization only focus on the magnitude response, and as such the magnitude/phase relation might diverge from the target one.
This might cause preringing, especially with minimum-phase corrections.
I personally don't use it anymore for that reason.

I just hope miniDSP will soon release a new device with enough power to carry on time domain convolution (no FFT please!) at 96kHz sampling rates with enough taps so we don't have to worry about petty things anymore

Re: MiniDSP allowing us to increase the number of OpenDRC taps beyond the current 6144 limit....

1. Only necessary for woofers / subwoofers. I think most tweeters and midrange corrections work okay with 6144 taps.

2. Ideally MiniDSP should launch a new higher power OpenDRC-PRO machine with 16,384 taps or something...

3. Or... use Downsampling x2 method.
Allow an internal 24kHz sampling rate mode, accepts existing 48kHz A/D or digital input, throws away every other sample. Simple.
New audio bandwidth becomes 12kHz max (fine for woofers)
New filter length doubles to 256 milliseconds, equivalent to having 12288 taps to play with, (although really is just same 6144 at lower rate)
Convert output back to 48kHz rate with simple sample and hold repeat of sample values - don't even need to interpolate - no problem if user remembers to include brickwall 12kHz HPF in their filter for anti-aliasing purposes!!!
Why can't MiniDSP do this downsampling in a firmware upgrade?
At least x2 would be a start. Adding further downsampling x4 option would be the next step after this, just keep 1 sample in every 4.

4. There's another way....
It occurs to me that for LINEAR PHASE ONLY type FIRs, ie. not mixed phase filters, but strictly Linear Phase, the FIR impulses are always symmetrical and will peak in the centre of the impulse file at sample number #3072 out of 6144 total.
Because they are symmetrical about the centre, the sample data to the left of centre peak must be identical to sample data to the right of peak, only running in mirror image opposite direction.
So instead, let's make sample number #6144 the effective central peak of an effective 12288 length linear-phase impulse file, and read the data forwards up from #0001 t o #6144 for the first half, then immediately again in reverse direction same file data backwards from #6144 back to #0001 for the second half of the file. We then are reading 12288 samples of data, symmetrical about the centre. Convolving this whole 6144 sample file read twice over in both directions would be the same as one single symmetrical 12288 tap file read forwards only.

Using just 6144 samples of data means file's hardware memory requirements are halved though, so hopefully they can still use the same existing OpenDRC hardware, and still run at native 48kHz. Just needs a firmware update to allow this method to work.
Would this method work???? I think it has a chance.
Can MiniDSP do this????
Cue Jeremy Clarkson saying, "I mean, how hard can it be...?!"

Re: Iterations
I do realise it's impossible to guess how many iterations.
My suggestion was merely to display how many iterations were used once process completes, in the little green text info area. That would be useful info I think.

Another suggestion, would be option for linear Hz freq axis graph scale vs current standard logarithmic Hz axis graph scale.
This would be able to show how the FFT bins are evenly spaced in the Hz freq scale axis (on a linear plot) but appear sparse in the bass and bunched up tight together in treble on a logarithmic Hz freq scale axis.
Just an idea. HOLM already has linear and logarithmic freq axis options.

Some great request here...

Actually, after many months talking of it , we do have an update working for very large tap count.. a lot more than anybody could have expected (even us...ahaha)
The problem? We don't have enough flash memory to store such large banks on the miniSHARC as we never envisioned to load such large FIR banks as we'd need a BGA/Parallel flash. We use a test tool now that loads the configuration to the DSP RAM (i.e. reboot will clear it). But it's pure FIR, pure engineers tool, no plugin, no RMS, no volume.. For that reason, we think it's not exactly fit for our market yet..

Very basic but powerful.. potentially very beta too.. hahaha.. Hoping in the future, we can change this with an update somehow.. (e.g. load to SD card instead)...
@ Pos, Richard, if you want to become guineapeg feel free to let us know. (support.minidsp.com). Might not be ready yet but we can help to get you started.

DevTeam

I have since given up on IIR and switched fully to FIR. IMO, there is nothing you can't do with FIR that you can do with IIR. IIR only saves you processing power. You can even use FIR to emulate IIR...

I would love a plugin that focused entirely on FIR and gave us delay and level settings. I no longer use PEQ, compression or crossovers. I handle everything with FIR. FIR handles amplitude, phase and crossovers. All I need is a good delay to adjust between speakers and the ability to level match.

devteam wrote:
Hi DevTeam,
Yes please, count me in! I'd be delighted to be a guinea pig and beta test any new product like that, and find flaws and issues, etc.

How many taps do we actually need?





I'd be quite happy to settle for 16,384 taps. That matches the shape close enough to satisfy any audiophile's ear.
Going further beyond 24,576 it's already near perfect match, giving negligible real improvement, but just adding more latency delay to the audio processing.
Either 2x or 4x Downsampling would be able to work a treat on my woofers. (This plot is from a Faital Pro 15XL1400 fifteen inch woofer)
devteam wrote:
Yes, I would love the SD card option! It makes life very easy if we could load impulses straight onto device WITHOUT needing to attach to a computer and software plug-in. The Logidy EPSi is doing this SD Card method and it seems so simple and quick.
Adding a front-panel LCD display with alpha-numeric file names display would be very, very, very useful upgrade for this new OpenDRC model, please!

Are you trying to make such enormous big FIR that it's impossible to work with your existing flash banks? Needs completely new hardware?
Perhaps we'd rather get a more modest firmware upgrade sooner, eg. this summer, that works for our existing hardware, than wait many years for the major upgrade to enormous big FIR sizes that we don't necessarily need.
I'd sooner have 12288 taps working now...

Q1. Downsampling - can you make this work, first of all ? Is it possible with a firmware revision? Does it require hardware modification?

Q2. Reading 6144 forward and backward for single symmetrical 12288 tap impulse - is this possible?

Q3. Reading both Left and Right channels sequentially as one long mono 12288 tap impulse - is this possible?

Q4. Reading BOTH Left and Right channels forward AND backwards as one super-long symmetrical 24576 tap impulse - how about that?

The OpenDRC is seen as a stereo Room Correction device, but it's application for many of us is loudspeaker driver correction & active crossover duty, using rePhase filters. That is far more important task. User can also amalgamate any room correction if they want to.
Please consider making a multi-channel version of OpenDRC in one 1U rack-box with tweeter = 6144 taps, midrange = 6144 taps, woofer = 24576 taps downsampled x4 and that would be amazing!!!

devteam wrote:
Is it using FFT convolution ?
How much processing delay does it add?
Taking advantage of the sdcard slot and loading the FIR on startup is very appealing.
It could even let the user recall its preset from the sdcard.

I volunteer for guinea pig duty!
No need for any particular user interface or pluging: I would be perfectly happy if I could simply drop two FIR on the sdcard from my PC with specific names (alongside the corresponding rephase presets just for convenience, that the openDRC would just ignore...), plug it into the openDRC, turn it on, and.... enjoy !
Brutally simple and just what I'd need!

FIR names would let the openDRC know what channel mode to use:

• L.bin and R.bin would mean stereo
• L1.bin and L2.bin would mean L
• R1.bin and R2.bin would mean R

Anything else with a bin extension would just cause the openDRC to stay muted on startup.

No need for individual delay or volume as this can be handled within the FIR, just the plain ol' FIR POWER!

Hello!

I would like to see the group delay vs frequency of the filter as well. This gives you detailed view in the time domain as with the phase.
At the moment I import the impulse to REW and check the group delay.

This method can be seen here: www.minidsp.com/forum/suggestion-box/125...48k-plugin-for-2x4hd

One thing I think would be cool is to add a view instead of degrees of positive phase or negative phase it shows radians or ms.

Maybe when you snap to the crosshairs on a specific spot it shows where it’s at in ms , like a GD equivalence, relative to 0° And it’s frequency. Or , a way to view excess GD and use phaseEq to move it in a way that works with the display.


Or something that shows us a time vs phase.

Why I say that is I always get backwards. I know positive phase is the same as adding delay (Q&Pe) and negative phase moves the area on the peak phaseEq forward “in time”

When I have all my measurements overlay in REW and looking at data , trying to pick witch one to adjust , it gets overwhelming and I seem to always get it backwards when I have a lot going on.

Some kind of equivalents graph would be sweet.

Also, it would be cool if there was a way in rePhase to show the effects on a imported measurement if the user was to add time delay to the measurement. For example if I was building a fir and had a measurement imported and after I’m done and get the fir I want to use than send it to my dsp , all is good at that point but that speaker I intend to add let’s say 1.0ms delay to that speaker (let’s say it in a car and it’s the left side) and the right side has a very very similar fir ) so it would be cool to see the effect on the fir if it had maybe a way to compare to the prediction as far as adding delay. Basically want to see the com filters appear as if it was comparing to itself with delay added.


That would be super cool. Maybe not a possibility tho.

Thanks in advance,
Andrew

Hi Andrew,
denver8me4dinner wrote: Currently the values are relative to the cursor, not the curves. This is fine with freq/mag/phase, but not with GD, as it is proportional to the slope of the phase, so cannot be get from the position of the cursor alone.
a GD curve would be useful indeed, and you are not the only user requesting it, but graphic work is the most time consuming part in rephase development, as everything is custom made and the GUI library I am using is quite outdated...
Well, it is on the todo list
You can add delay to an imported measurement using the "time offset" setting in the measurement tab.
What you are looking for here seems to be the effect of adding two responses with a time delay.
This is a different thing and probably better executed using a measurement software.
For example in HOLM you can load two measurements in slot A and B (possibly the same measurement in both), then adjust delay in one of them (time zero offset setting), and then use the C=A+B function in the manipulation menu entry). I am sure REW would let you do the same, and more.