• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Obsessive compulsive manual DRC in REW/rePhase

Any crossover that causes a phase shift will be harder to design. I said "harder", not "impossible". For example, here is a simulation of a minimum-phase 4th order Butterworth XO at 80Hz:

1731519994092.png


Red = LPF, Green = HPF, Brown = summation. All IIR crossovers, and all passive crossovers with the same 4th order BW will have that 3dB hump at the XO region. And that's the electrical crossover only. Once the driver is involved, there will be additional phase rotation so who knows what the result is going to be. The linear phase version sums perfectly flat. And because you can manipulate phase independently of amplitude, you can also make the summation with the driver flat.

You should understand what you are getting into. In a nutshell:

- Passive crossovers: minimum phase. Wastes amplifier power as heat. If you wish to do more correction, you increase the complexity of the network and waste even more amplifier power and component prices go up. There is no possibility for room correction. And furthermore, every driver is different thanks to manufacturing tolerances. Passive XO's have a very limited ability to deal with that.
- digital IIR crossovers: the digital version of minimum-phase passive crossovers. (Minimum phase means that the amplitude is inextricably tied to phase). It is a huge step up. Steeper slopes and more complex corrections can be achieved without wasting amplifier power. You can do room correction, but your ability to do so is more limited than a lin phase FIR.
- digital linear phase FIR: these do not exist in the natural world. A constant delay is applied to the signal allowing independent manipulation of amplitude and phase.

MiniDSP's have their place. They are inexpensive, do a great job, convenient, robust, and they measure extremely well. It's a great product. But what you can do with them is somewhat limited, and if you like to tinker you will start running into its limitations. The versions with Dirac automate the procedure for you and make filters easier to design, but there are many reports on ASR and elsewhere that Dirac often screws up and you get strange corrections.

IMO, for the best possible quality, manual correction with linear phase FIR is the way to go. I am not disparaging MiniDSP, because if it is used well it is likely to be extremely close to a linear phase FIR and the difference may not be audible in some setups.
 
Thanks a lot Keith, really good information here. I will need to start playing with all these starting from a PC FIR convolver. Would you recommend equalizer APO and rephase or something else?
 
If you want linear phase FIR, these are your options from easiest to use to hardest to use:

- Dirac. The most automation but also the least flexible. Requires Dirac's own convolver since it outputs proprietary filters, not standard .WAV files. Very expensive. Lots of reports of screwy behaviour and odd corrections.
- Audiolense. Less automation than Dirac. Lets you control more aspects of correction. Lack's Acourate's flexibility.
- Acourate. This is what I use and recommend. This has a mix of manual correction and automation. Although some parts of Acourate are automated or macro based, you can do everything manually if you wish. IMO this is the most powerful DSP tool on the market, it is extremely flexible and you can do nearly anything. For e.g. Audiolense can only generate one type of crossover but with different orders. Acourate lets you generate any crossover, any order, in minimum phase or linear phase.
- REW+RePhase. I haven't explored this combo's more advanced features so I can't tell you what it can or can't do. From what I can see, it exceeds Acourate in its measurement features (e.g. it has waterfall and spectrogram, Acourate does not), but may not match it when it comes to corrections. Lots of people on ASR use this combo. This is harder to use than Acourate because it completely lacks automation, and it is fully manual so you really need to know what you are doing. It requires two different software programs and the integration is not seamless, e.g. you have to export files from one program to another and the user interfaces are very different.

There are others on the market, e.g. Focus Fidelity which is quite new and lacking some features. Until recently it did not have the ability to take measurements. It is still under development. Eclipse Audio's FIR Designer is another option, but I don't know anybody who uses it.

The four that I mentioned are the packages most commonly used on ASR. I would avoid Dirac because of its lack of flexibility, expense, and tendency to screw up. But you might prefer it because you can use it without having to learn the nitty gritty of DSP. I also do not recommend REW+RePhase for beginners for the opposite reason, you need to learn A LOT about DSP before you can use it. But it is free, and nothing beats free. Acourate and Audiolense would be my recommendations. One holds your hand more (Audiolense), the other less (Acourate).
 
If you want linear phase FIR, these are your options from easiest to use to hardest to use:

- Dirac. The most automation but also the least flexible. Requires Dirac's own convolver since it outputs proprietary filters, not standard .WAV files. Very expensive. Lots of reports of screwy behaviour and odd corrections.
- Audiolense. Less automation than Dirac. Lets you control more aspects of correction. Lack's Acourate's flexibility.
- Acourate. This is what I use and recommend. This has a mix of manual correction and automation. Although some parts of Acourate are automated or macro based, you can do everything manually if you wish. IMO this is the most powerful DSP tool on the market, it is extremely flexible and you can do nearly anything. For e.g. Audiolense can only generate one type of crossover but with different orders. Acourate lets you generate any crossover, any order, in minimum phase or linear phase.
- REW+RePhase. I haven't explored this combo's more advanced features so I can't tell you what it can or can't do. From what I can see, it exceeds Acourate in its measurement features (e.g. it has waterfall and spectrogram, Acourate does not), but may not match it when it comes to corrections. Lots of people on ASR use this combo. This is harder to use than Acourate because it completely lacks automation, and it is fully manual so you really need to know what you are doing. It requires two different software programs and the integration is not seamless, e.g. you have to export files from one program to another and the user interfaces are very different.

There are others on the market, e.g. Focus Fidelity which is quite new and lacking some features. Until recently it did not have the ability to take measurements. It is still under development. Eclipse Audio's FIR Designer is another option, but I don't know anybody who uses it.

The four that I mentioned are the packages most commonly used on ASR. I would avoid Dirac because of its lack of flexibility, expense, and tendency to screw up. But you might prefer it because you can use it without having to learn the nitty gritty of DSP. I also do not recommend REW+RePhase for beginners for the opposite reason, you need to learn A LOT about DSP before you can use it. But it is free, and nothing beats free. Acourate and Audiolense would be my recommendations. One holds your hand more (Audiolense), the other less (Acourate).
Thats brilliant, many thanks Keith! I would really like to know the theory behind advanced dsp and started reading dspguide.com and accurate sound reproduction using dsp by Mitch Barnett.
Recenlty finished reading Dr. Toole’s book.
 
Acourate and Audiolense would be my recommendations. One holds your hand more (Audiolense), the other less (Acourate).
In order to use these don't we need a PC in the audio chain? I run an Eversolo A8 into minidsp Flex to active speakers and use REW only at this point.

How can I implement your two suggestions for my system?

Eversolo's DRC is unproven at the moment and I already have the Flex with REW settings. I've never heard of rePhase. Perhaps I should move down this road.
 
You have a few options:

1. Continue using your MiniDSP Flex to do the DSP. However you will be limited to minimum phase IIR's + FIR's with 1024 taps per channel. Acourate can output to MiniDSP, you can get a cheaper version of Acourate (AcourateDRC) that does this. I am not sure about Audiolense, but I believe it is possible.

2. Insert a PC between the Eversolo A8 and the MiniDSP Flex. Your PC will need an optical input and a way to route signal from the input, through the convolver, and out to the DAC. If you use Windows, you will have to use WASAPI Shared, so I would avoid Windows. With Windows you "need" WASAPI Exclusive or ASIO to avoid the problems with WASAPI Shared. The best way to route signal on a Windows PC is to use an interface, that way you can keep it in ASIO.

The no. 1 reason you should keep using your Eversolo A8 streamer is: remote control. It is possible to implement remote control on a PC, but it is difficult, requires more hardware and software, the solution is not as elegant, and you are limited to three streaming services (Spotify, Tidal, and Qobuz).

How the DSP would work: you can either use your Flex as a "dumb DAC" and do all the DSP functions on your PC, or you could use your PC to do the FIR functions (you can get up to 256k taps) and not use the MiniDSP's FIR function (only 1k taps).

3. Replace the Eversolo A8 with the PC, and the PC goes into the MiniDSP Flex. Same advantages/disadvantages as above.

4. Replace both Eversolo A8 and MiniDSP Flex and buy an interface (like a Motu Ultralite Mk. 5, RME Babyface/Fireface, etc). This has the greatest learning curve of all since you have to replicate the functions of both the Eversolo A8 and MiniDSP with software. The no. 1 reason to use a PC: you can use linear phase FIR filters with a high tap count.
 
Hey Keith,

I followed your advice and got the acourate drc, my question is, how can I easily stream music wirelessly from my pc to my airplay endpoint denon avr (via a streaming service say tidal) that will use the convolution engine of the equalizerAPO?

Is this possible or do I need something like roon to do this?

Many thanks and Happy New Year
 
The first step is whether your PC recognises your Denon AVR as a sound output device. And if so, how many channels can you address, and whether it is in ASIO or WASAPI.

If it does, it's all sweet. Simply run the Tidal app, set the output of Tidal to point to a convolver of your choice, then set the convolver to output to your AVR.

It it does not (and I highly suspect that it won't work) then you are stuck. You will have to make hardware changes.
 
  • Like
Reactions: CK.
The first step is whether your PC recognises your Denon AVR as a sound output device. And if so, how many channels can you address, and whether it is in ASIO or WASAPI.

If it does, it's all sweet. Simply run the Tidal app, set the output of Tidal to point to a convolver of your choice, then set the convolver to output to your AVR.

It it does not (and I highly suspect that it won't work) then you are stuck. You will have to make hardware changes.
Basically my pc does not recognise my avr as an output but the streaming software does (eg spotify). My avr supports tidal connect and is roon tested.

Does that mean that if I use roon and its conv engine I can stream wirelessly to my avr taking advantage of my FIR filter with no hardware changes?
 
Does that mean that if I use roon and its conv engine I can stream wirelessly to my avr taking advantage of my FIR filter with no hardware changes?

I don't know. I don't use Roon, and I don't have a Denon AVR. Best if you download a trial version of Roon and see if it works.

BTW there may be a way to hack together a wireless USB solution. All this should work IN THEORY only and I make no claims or guarantees that it will work.

Step 1: connect your PC to your AVR with a USB cable. Check that it appears as a sound output in Windows. Also check Denon's website and see if there is an ASIO driver for your AVR. Finally, set up your convolution pipeline and see if music comes out your speakers.

Step 2: replace the USB cable with a wireless USB adapter like this one.
 
  • Like
Reactions: CK.
Thank you Keith appreciate it! Will give it a try and see how it goes!
 
Hello,

For anyone that might be interested, I was able to use a convolution filter in roon and stream tidal music over airplay to my denon avr.

@Keith_W I understand that you have experience with Acourate room correction, I have used it in different setups and rooms and in all cases the ICCC always comes back negative after correction (eg -6%). Is this worriesome?

Also, in a typical untreated living room with assymeties and reflections, is it beneficial to apply the correction full range or limit it? I.e is full range correction best suited in rooms with less acoustic issues?
I know acourate uses FDW but not sure if that is still enough

Many thanks
 
Last edited:
@Keith_W I understand that you have experience with Acourate room correction, I have used it in different setups and rooms and in all cases the ICCC always comes back negative after correction (eg -6%). Is this worriesome?

ICCC is between 0% (zero correlation between channels) and 100% (both channels are absolutely identical). A very good system will have ICCC > 90%. I have never heard of a negative ICCC. If Acourate is giving you a negative ICCC, I would ask Dr. Bruggemann in the Acourate forum.

Also, in a typical untreated living room with assymeties and reflections, is it beneficial to apply the correction full range or limit it? I.e is full range correction best suited in rooms with less acoustic issues?
I know acourate uses FDW but not sure if that is still enough

I will put it this way - speaker correction and room correction are different. The aim of speaker correction is to correct the minimum phase response of the speaker alone, which means you need to take a reflection-free measurement of the speaker. You also need to consider the off-axis response. The aim of room correction is to correct the response of the room's influence on the speaker, which means that you need a measurement which includes room reflections at the listening position. As wavelengths get shorter, the area over which the correction applies becomes smaller and smaller. At some point, it becomes smaller than the distance between your ears and the correction is completely unrealistic because it is correcting for a single, extremely specific microphone position.

So you could:

- Do something like FDW 15/1 - meaning 15 cycles of windowing for the bass, and 1 cycle of windowing for treble.
- Or you could choose to omit the upper frequencies from correction altogether.

A few days ago I published a free Acourate guide which tells you how to perform these procedures in Acourate.
 
  • Like
Reactions: CK.
Thanks for the guides just downloaded them!

For the ICCC I actually meant that the difference between the non corrected result and the corrected is negative, meaning the corrected response is worse (?) - sorry I was not clear on that.

I actually have anechoic measurements for my speakers (spinorama.org) and thus before Acourate I was applying PEQ above Fc based on that and PEQ below Fc based on a harman like target curve.

I was missing the time domain correction (step response) and this is why I wanted to try Acourate.

So I was wondering if there is any benefit to do a full range correction in Acourate or do a limited correction and just apply PEQ above fc based on spinorama data. Meaning, if there was any additional benefit that I would be missing.
 
Thanks for the guides just downloaded them!

For the ICCC I actually meant that the difference between the non corrected result and the corrected is negative, meaning the corrected response is worse (?) - sorry I was not clear on that.

Without looking at the result (the test convolution) it is hard for me to say what went wrong. Acourate is very manual and requires a lot of thought about what you are doing. I hope that my guide will help you.

If it is still no help, you can post your Pulse48L/R.dbl here (you need to zip it first) and tag me so that I see them. I will have a play and see what I come up with. Alternatively you could post on the Acourate forum and Uli himself will probably look at it.

I actually have anechoic measurements for my speakers (spinorama.org) and thus before Acourate I was applying PEQ above Fc based on that and PEQ below Fc based on a harman like target curve.

I was missing the time domain correction (step response) and this is why I wanted to try Acourate.

So I was wondering if there is any benefit to do a full range correction in Acourate or do a limited correction and just apply PEQ above fc based on spinorama data. Meaning, if there was any additional benefit that I would be missing.

With Acourate, all I can say is - do not be in a hurry to get a perfect result. Over time, your approach to DSP will change. I have owned Acourate for 10 years now and I am still trying to improve my setup. I read something new and ask myself if I can replicate that in Acourate. Quite often, it is a lot of effort and time wasted but at least I learnt something from it!

If you have a Spinorama of your speakers, and they look good to you, then I suggest to either leave the top end alone (i.e. use Archimago's flat line method with SplitNJoin - see the advanced section in the first guide) or do what Toole suggests, a "broad tone control-like treble tilt" which means an FDW of 1. You can experiment and see which result you like the most.

Also remember that if you have an improper measurement technique, it will mean that what you measure is NOT what you hear. And if you take a bad measurement, you will make a bad correction. Learning how to take measurements has some importance for most of us in this hobby, but if you are going to perform DSP on the measurement it is ESPECIALLY important!

In the full length version of the book (which I am still writing) there is a very extensive chapter on how to take measurements. But if you can't wait, I suggest you read this: Jeff Bagby on how to take measurements.
 
  • Like
Reactions: CK.
Without looking at the result (the test convolution) it is hard for me to say what went wrong. Acourate is very manual and requires a lot of thought about what you are doing. I hope that my guide will help you.

If it is still no help, you can post your Pulse48L/R.dbl here (you need to zip it first) and tag me so that I see them. I will have a play and see what I come up with. Alternatively you could post on the Acourate forum and Uli himself will probably look at it.



With Acourate, all I can say is - do not be in a hurry to get a perfect result. Over time, your approach to DSP will change. I have owned Acourate for 10 years now and I am still trying to improve my setup. I read something new and ask myself if I can replicate that in Acourate. Quite often, it is a lot of effort and time wasted but at least I learnt something from it!

If you have a Spinorama of your speakers, and they look good to you, then I suggest to either leave the top end alone (i.e. use Archimago's flat line method with SplitNJoin - see the advanced section in the first guide) or do what Toole suggests, a "broad tone control-like treble tilt" which means an FDW of 1. You can experiment and see which result you like the most.

Also remember that if you have an improper measurement technique, it will mean that what you measure is NOT what you hear. And if you take a bad measurement, you will make a bad correction. Learning how to take measurements has some importance for most of us in this hobby, but if you are going to perform DSP on the measurement it is ESPECIALLY important!

In the full length version of the book (which I am still writing) there is a very extensive chapter on how to take measurements. But if you can't wait, I suggest you read this: Jeff Bagby on how to take measurements.
Really appreciate your help and guidance on this as well as your contribution to the community!

I already started reading your guides, so far I found it very illuminating and precise/to-the-point. Will post the result if indeed I am not able to get to the bottom of this.

Regarding the measurement technique, I followed the one in Mitch's book but I was surprised that acourate accounts for only 1 measurement position.

I think I will try FDW of 15/5 ~ 15/1 and see how it measures/sounds.
 
Regarding the measurement technique, I followed the one in Mitch's book but I was surprised that acourate accounts for only 1 measurement position.

Acourate doesn't account for only one mic position. It is a toolbox, it can be used any way you want, including a dozen mic positions if you wanted to. Mitch described a single microphone position for simplicity. If you read the advanced sections of the guide, there is a description on how to use different measurement techniques for correction. Do not think of Acourate as similar to Audiolense or Dirac - other software has a prescribed process that you have to follow. Acourate lets you do anything you want, including all sorts of ridiculous hare-brained things I came up with that turned out to be very bad ideas.
 
Really appreciate your help and guidance on this as well as your contribution to the community!

I already started reading your guides, so far I found it very illuminating and precise/to-the-point. Will post the result if indeed I am not able to get to the bottom of this.

Regarding the measurement technique, I followed the one in Mitch's book but I was surprised that acourate accounts for only 1 measurement position.

I think I will try FDW of 15/5 ~ 15/1 and see how it measures/sounds.

There's a moving mic recorder now. You can also average separate sweep measurements "by hand", saving each measurement, then loading them up into Acourate to average them. The procedure is the similar to the beamforming example given in Mitch's ebook. But by doing spatial averages or using a moving mic you lose the ability to do phase correction, so you would only use the average to do the inversion step, which would then be used along with a single-point measurement to do the final correction.
 
  • Like
Reactions: CK.
There's a moving mic recorder now. You can also average separate sweep measurements "by hand", saving each measurement, then loading them up into Acourate to average them. The procedure is the similar to the beamforming example given in Mitch's ebook. But by doing spatial averages or using a moving mic you lose the ability to do phase correction, so you would only use the average to do the inversion step, which would then be used along with a single-point measurement to do the final correction.
Ok I see thanks I will look into it.

On another note, by limiting the correction in acourate to below Fc, do you also “sacrifice” the correction of the impulse/step responce and phase?

I wonder if there is a way to incorporate the unechoic data magnitude corrections (since i have them from spinorama.org) to the acourate correction process and the resulting convolution filter. Thinking out loud…
 
Back
Top Bottom