Advice on DSP + Amp Upgrade for 2-Way Build (RSS210HF-4 & SEAS DXT)

[TomSR]

I made a little spreadsheet of the options and cost for the ideas that are out now. Any DSP, amps or amp combinations I'm missing that I should really add? Any reason to go for the Mini DSP Flex over the 2x4HD?

 

[mcdn]

Very curious at this. Would the flex be worth it looking at it's DAC's etc?

If you’re looking to put it in a rack and forget about it there is no upside to the flex apart from balanced outputs. The 2x4HD already has 100dB SINAD and the exact same DSP capabilities as the Flex.

But if you want something that acts more like a preamp then you need the Flex.

Either way there aren’t many other DSP options than those two, so pick one and spend the rest of the budget on the amps.

 

[Keith_W]

Oh, one requirement I forgot to mention is that the latency can't be more than 20ms and preferably switchable to low latency mode with less phase correction so you'd be able to play instruments through your DAW.

You can calculate latency of a linear-phase FIR filter with this formula: latency (seconds) = (n-1)/2Fs where n = number of taps, and Fs = sampling rate. The more taps you have, the higher the latency. The higher the sampling rate, the lower the latency. The tap count n is always a power of 2, e.g. 1024, 2048, 4096, 8192, 16384, 32768, 65536 ... etc. The sampling rate Fs is one of the common sampling rates - 44.1kHz and its multiples, or 48kHz and its multiples. How high you can actually go with n or Fs depends on the limitations of your software, processing hardware, and DAC.

You can also calculate the resolution of a FIR filter with this formula: resolution = Fs/n.

If you play a bit with the equation, you will see that if you want higher resolution, you need more taps and a lower sampling rate. But if you do that, you pay the price in latency. IMO, the lowest tap count you can get away with (while maintaining a decent amount of resolution) is 32768. The highest sampling rate supported by most hardware is 192kHz. This will give you a latency of 85ms, and a bin size of 5.9Hz. If you go up to 384kHz with a tap count of 16384, you can go down to 21ms latency but your bin size quadruples to 23.4Hz.

Note that FIR filters are capable of both linear-phase and minimum-phase DSP, so if the latency of linear-phase is too high, you can redo the filters in minimum-phase.

 

[TomSR]

You can calculate latency of a linear-phase FIR filter with this formula: latency (seconds) = (n-1)/2Fs where n = number of taps, and Fs = sampling rate. The more taps you have, the higher the latency. The higher the sampling rate, the lower the latency. The tap count n is always a power of 2, e.g. 1024, 2048, 4096, 8192, 16384, 32768, 65536 ... etc. The sampling rate Fs is one of the common sampling rates - 44.1kHz and its multiples, or 48kHz and its multiples. How high you can actually go with n or Fs depends on the limitations of your software, processing hardware, and DAC.

You can also calculate the resolution of a FIR filter with this formula: resolution = Fs/n.

If you play a bit with the equation, you will see that if you want higher resolution, you need more taps and a lower sampling rate. But if you do that, you pay the price in latency. IMO, the lowest tap count you can get away with (while maintaining a decent amount of resolution) is 32768. The highest sampling rate supported by most hardware is 192kHz. This will give you a latency of 85ms, and a bin size of 5.9Hz. If you go up to 384kHz with a tap count of 16384, you can go down to 21ms latency but your bin size quadruples to 23.4Hz.

Note that FIR filters are capable of both linear-phase and minimum-phase DSP, so if the latency of linear-phase is too high, you can redo the filters in minimum-phase.

Are there any DSP boxes that support the amount of taps and sampling rate you are talking about?

 

[ppataki]

I have considered it, only I was a bit worried about the stability of the whole setup. Once i have it set up I preferibly don't want to have to mess with it ever again so no drivers crashing or having the software forget the crossovers etc. How stable is this in your experience?

It is rock solid (once you have the setup finalized)

 

[ppataki]

IMO there is a definite audible difference between linear phase and minimum phase. Minimum-phase creates time distortion and smears sound across time. Once this is removed, the result is exceptional clarity and attack.

But I suspect the others will pounce on me for saying so because I just made a subjective statement without a blind test to back it up on a fairly controversial topic. OK I admit I don't have a blind test, but I do have linphase and minphase filters which I designed to be as close as possible to each other. I can hear a difference, and i'll stake my reputation on it If you want to read about whether phase distortion is audible ... once again, ASR delivers. Read this thread (particularly the last 10 pages) and this thread (also the last 4-5 pages).

If you want linear phase, you're going to need a lot of taps.

Anyway, enjoy your reading and check back with any questions!

I concur with that, I have the same experience - hence I always use linear phase crossover
btw. @TomSR with a PC you will have access to literally thousands of VST plugins, including linear phase ones

 

[TomSR]

I concur with that, I have the same experience - hence I always use linear phase crossover
btw. @TomSR with a PC you will have access to literally thousands of VST plugins, including linear phase ones

Will this run on something like a Raspberry PI? What else would be needed for this to work? Just an audio interface?

 

[ppataki]

Will this run on something like a Raspberry PI? What else would be needed for this to work? Just an audio interface?

No (not with the VST plugins)
You need either a Windows PC (or laptop, miniPC, etc.) or a Mac for that
+ a multichannel DAC but I think you have that already if I remember reading your previous posts correctly

 

[TomSR]

No (not with the VST plugins)
You need either a Windows PC (or laptop, miniPC, etc.) or a Mac for that
+ a multichannel DAC but I think you have that already if I remember reading your previous posts correctly

I do have pc permanently connected to the Audient EVO 16. Would i be able to use the EVO 16 both as audio interface for the pc and as DSP at the same time? What I can find on JRiver is that it's really just a media center. Is that correct? No external audio in or out?

 

[ppataki]

I do have pc permanently connected to the Audient EVO 16. Would i be able to use the EVO 16 both as audio interface for the pc and as DSP at the same time? What I can find on JRiver is that it's really just a media center. Is that correct? No external audio in or out?

You will use your EVO 16 as an audio interface (as a multichannel DAC) and then you will use your source (your PC) as the DSP
You install Jriver and use that to play your music files or stream (Tidal, etc.) leveraging its DSP engine real-time
You can use its built-in capabilities (volume, PEQ, crossover, delay, etc. etc.) and on top of that you can use 3rd party VST plugins too to have state-of-the-art DSP capability
Things like linear phase filters, AI-powered EQ, etc.
This is by far the best DSP option that one can have (IMHO)

 

[TomSR]

You will use your EVO 16 as an audio interface (as a multichannel DAC) and then you will use your source (your PC) as the DSP
You install Jriver and use that to play your music files or stream (Tidal, etc.) leveraging its DSP engine real-time
You can use its built-in capabilities (volume, PEQ, crossover, delay, etc. etc.) and on top of that you can use 3rd party VST plugins too to have state-of-the-art DSP capability
Things like linear phase filters, AI-powered EQ, etc.
This is by far the best DSP option that one can have (IMHO)

The only problem I see with this is that I won’t be playing any music files or streaming. I’ll be mixing on my live console and in reaper this audio needs to pass through the DSP. Can Jriver do that?

 

[ppataki]

The only problem I see with this is that I won’t be playing any music files or streaming. I’ll be mixing on my live console and in reaper this audio needs to pass through the DSP. Can Jriver do that?

Yes it can, it has a WDM driver that can pass any audio through its DSP engine (although this only works in Windows if I am not mistaken)

 

[Keith_W]

Are there any DSP boxes that support the amount of taps and sampling rate you are talking about?

There is only one, the DEQX Premate 8. Unfortunately it is eye wateringly expensive. You already own all the equipment that you need:

- Audient 16 Evo interface
- (presumably) an omnidirectional condenser XLR microphone + microphone stand
- PC

All that you need is software to generate filters and a convolver to host the filters for playback.

Re: software to generate filters. Take your pick: Dirac, Acourate, Audiolense, Focus Fidelity, Eclipse Audio's FIRDesigner, etc. There are free options including Denis Sbragion's DRC-FIR and REW-rePhase. I do not recommend any of the free options because you REALLY need to know what you are doing otherwise you will totally mess it up. For example, you need to manually place the impulse in the middle of the tap length when exporting filters in REW. The dialog box for .WAV export allows you to place the impulse, but the setting is easy to miss unless you know in advance that you need to set it. In short, you need to have cut your teeth learning DSP with another program before you attempt to use REW/rePhase to make linear-phase filters (but if you want minimum-phase IIR's, REW is great!).

Re: convolver. @ppataki suggested JRiver. JRiver is music playback software, and it has a lot of features including a convolver. It's not the best convolver on the market because it is lacking a few nice features. There is no quick way to load another set of filters to A-B them. You have to stop music playback, load another set of filters, and start playback again. Although JRiver can accept external digital sources (i.e. route a hardware input channel from your interface to JRiver via ASIO), its implementation is clunky and unreliable, e.g. it refuses to recognise my RME interface. I don't recommend you pass your audio from Reaper into JRiver via WDM, because it goes through the Windows mixer (Windows resamples the sound without telling you, introduces additional latency, and mixes in notification bleeps). If you want to pass audio from Reaper to JRiver, do it via ASIO or WASAPI Exclusive.

You are far better off with a dedicated convolver. The one I recommend is Accurate Sound's Hang Loose Convolver (HLC). The author @mitchco is on ASR and he's a retired recording engineer himself so it has all the pro features that you need. There are free options including CamillaDSP but Camilla's big failing is lack of ASIO support.

But please, before you spend any money, weigh up what you need. If latency is REALLY important to you, high tap count linear-phase FIR filters may not be the solution! Don't forget that each piece of software you add to your pipeline has its own buffer which will add latency. You can reduce the buffer size, but you may start getting glitches depending on how powerful your CPU is, how much RAM you have, etc. If you have a weak PC which is already doing too much, you may be forced to increase buffer size which will add latency.

 

[TomSR]

There is only one, the DEQX Premate 8. Unfortunately it is eye wateringly expensive. You already own all the equipment that you need:

- Audient 16 Evo interface
- (presumably) an omnidirectional condenser XLR microphone + microphone stand
- PC

All that you need is software to generate filters and a convolver to host the filters for playback.

Re: software to generate filters. Take your pick: Dirac, Acourate, Audiolense, Focus Fidelity, Eclipse Audio's FIRDesigner, etc. There are free options including Denis Sbragion's DRC-FIR and REW-rePhase. I do not recommend any of the free options because you REALLY need to know what you are doing otherwise you will totally mess it up. For example, you need to manually place the impulse in the middle of the tap length when exporting filters in REW. The dialog box for .WAV export allows you to place the impulse, but the setting is easy to miss unless you know in advance that you need to set it. In short, you need to have cut your teeth learning DSP with another program before you attempt to use REW/rePhase to make linear-phase filters (but if you want minimum-phase IIR's, REW is great!).

Re: convolver. @ppataki suggested JRiver. JRiver is music playback software, and it has a lot of features including a convolver. It's not the best convolver on the market because it is lacking a few nice features. There is no quick way to load another set of filters to A-B them. You have to stop music playback, load another set of filters, and start playback again. Although JRiver can accept external digital sources (i.e. route a hardware input channel from your interface to JRiver via ASIO), its implementation is clunky and unreliable, e.g. it refuses to recognise my RME interface. I don't recommend you pass your audio from Reaper into JRiver via WDM, because it goes through the Windows mixer (Windows resamples the sound without telling you, introduces additional latency, and mixes in notification bleeps). If you want to pass audio from Reaper to JRiver, do it via ASIO or WASAPI Exclusive.

You are far better off with a dedicated convolver. The one I recommend is Accurate Sound's Hang Loose Convolver (HLC). The author @mitchco is on ASR and he's a retired recording engineer himself so it has all the pro features that you need. There are free options including CamillaDSP but Camilla's big failing is lack of ASIO support.

But please, before you spend any money, weigh up what you need. If latency is REALLY important to you, high tap count linear-phase FIR filters may not be the solution! Don't forget that each piece of software you add to your pipeline has its own buffer which will add latency. You can reduce the buffer size, but you may start getting glitches depending on how powerful your CPU is, how much RAM you have, etc. If you have a weak PC which is already doing too much, you may be forced to increase buffer size which will add latency.

The PC is relatively strong running 32GB’s of ram and a Ryzen 5700X3D. So I don’t think that will be an issue.
What might be the best solution for me would be running a dsp like a 2x4HD having that do normal crossovers and then running extra software on my pc for the FIR phase correction? This way I can run the speakers and the amps standalone and also run it with extra FIR when wanted? What would your thoughts be on that?

Also what dsp is better in terms of AD conversion the hypex internal or the Mini DSP 2x4HD?

Also anyone have any thoughts on the poweramps I’ve mentioned? Any suggestions?

 

[TomSR]

Is a Mini DSP DDRC-24 not interesting?

 

[Keith_W]

The PC is relatively strong running 32GB’s of ram and a Ryzen 5700X3D. So I don’t think that will be an issue.
What might be the best solution for me would be running a dsp like a 2x4HD having that do normal crossovers and then running extra software on my pc for the FIR phase correction? This way I can run the speakers and the amps standalone and also run it with extra FIR when wanted? What would your thoughts be on that?

Of course that is possible. You would be cascading a linear-phase FIR filter (PC) with a minimum-phase IIR (2x4HD). Theoretically, linear-phase FIR filters can compensate for phase distortions caused by minimum-phase systems such as crossovers and drivers. But you REALLY need to know what you are doing. If you want to do this, I strongly recommend Acourate because it can straighten phase distortion for you with a nice and easy tool. rePhase can also do it, but the process isn't quite as easy. I am not aware of any other software that can do it. You will need to take very high quality measurements and be absolutely confident that you have captured the anechoic response before you try something like this.

Also what dsp is better in terms of AD conversion the hypex internal or the Mini DSP 2x4HD?

That is the wrong question. All modern AD conversion is effectively transparent. People can run ADC-DAC loops up to 10 times and the result is still inaudible in a blind test.

What you should be asking is "how easy is it to send output from the ADC to Reaper". I don't know what you mean by "hypex internal", but if you are referring to your interface, it should be dead easy to send signal from your Audient to Reaper. As from MiniDSP to Reaper ... well, you are in luck. I happen to have a MiniDSP SHD that I borrowed in front of me.

I can confirm that the MiniDSP ASIO driver is seen by REW, and so are all the analog inputs.

And I can also confirm that Windows can see the analog inputs as a WDM device. Mind you, I haven't actually tested the inputs by trying to record something and sending the output from MiniDSP to Reaper.

Bear in mind that if you are using ASIO with Reaper, an important limitation is that ASIO lets you use one device at a time only. This may be an issue if you are trying to take input from the MiniDSP and sending the output to your Audient DAC. Whether Reaper lets you work with more than one ASIO device - I don't know. If it doesn't, then you may need third party software like VB-Matrix.

Also anyone have any thoughts on the poweramps I’ve mentioned? Any suggestions?

No idea, sorry. I don't care about amplifiers. As long as they amplify, don't add too much "character", don't cost too much, supply enough power ... it's good enough.

If you are asking about the DSP built-in to some amps, you need to be aware that without exception all of them are minimum-phase IIR. Read that thread that I linked about understanding DSP, and pay particular attention to the fact that biquads have poor portability. You will need software that can specifically generate biquads for that model of DSP in that amp.

You asked about the MiniDSP DDRC-24. Yes of course it is interesting, but whether it is the correct product for you is another matter. What you need to do is write down a list of requirements. You have already mentioned a few: linear-phase FIR, low latency, less than 1200 Euro, reliability, ease of use. The discussion so far has shown that some of these requirements are incompatible, e.g. linear-phase + low latency. And if you want MiniDSP, you won't have enough taps for linear-phase, not by a country mile. A potential solution is FIR+IIR cascade, but it is not easy.

 

[ppataki]

JRiver via WDM, because it goes through the Windows mixer (Windows resamples the sound without telling you,

I don't think that is happening if the source app is in Exclusive mode
For example if I set Tidal to exclusive mode and play 96kHz content, then 96kHz signal will come into Jriver and if I play 44.1kHz content then 44.1kHz signal will come into Jriver, etc.

Tidal:

Jriver:

 

[TomSR]

Ok, so I'm now currently deciding between two different DSP units. I want the ability to use Dirac. So I'm deciding between a Mini DSP DDRC-24 or a Mini DSP 2x4HD. The reason why I still might want the 2x4HD over the DDRC-24 is because of the latency the DDRC-24 has. It has about 13 ms of latency with Dirac Enabled and 27 ms with it disabled. (Seems really weird to me Source: https://www.youtube.com/watch?v=22yznnaNN5I) While the mini DSP 2x4HD has a latency of about 2 ms which is a lot quicker. My thought was I can just run Dirac Live Room Correction Suite on the computer when I want better frequency response, Impulse response and corrected sound overall. But then I can also bypass it and just run the 2x4HD as correction without FIR? What is your thought on this?

Also for amplifiers I'm thinking of going with AUDIOPHONICS AP300-S500NC and the AUDIOPHONICS MPA-S125NC since I quite enjoyed the current hypex amps I had apart from their output power.

 

[Keith_W]

This is why MiniDSP annoys me, their specs page doesn't tell you what you need to know. It's a DSP product, so I want to know what kind of DSP it does and what its capabilities are.

MiniDSP 2x4HD - 32 bit / 96kHz "FIR+IIR". No mention of how many taps or biquads per channel. If you want to find out, look in the manual where you find they have thoroughly buried the info you need. Answer: 10 biquads per channel, total of 4096 taps implemented at output. FYI, 4096 taps at 96kHz is useless. Akin to painting a postage stamp with a crayon held between your toes. But at least it has some FIR capability.

MiniDSP DDRC-24 - 32 bit / 48kHz no mention of FIR but it's definitely IIR. No mention of how many biquads per channel. Again, look in the manual and you have your answer. 10 biquads per channel, no FIR.

Now it seems a bit strange to me why the DDRC-24 should have a longer latency with Dirac disabled given that it has no FIR filters. That guy in the video made no effort to investigate. He could at least have shown us the settings window to show there are no funny delays dialled in with Dirac disabled. I am not willing to accept that Dirac magically lowers the latency compared to no Dirac, because he has not proven that "no Dirac" is actually the MiniDSP in its default state.

Re: Dirac, it seems as if you want to use it to make filters for the MiniDSP, and also to make FIR filters on your PC. You will have to check with Dirac or read their purchase page very carefully whether their license allows you to do this and they aren't separate software products which means you might need to buy more than one license. With MiniDSP, you do not have to buy a Dirac license with your initial purchase, you can buy a plain MiniDSP and "upgrade" to Dirac later. I say "upgrade" because I am not convinced that Dirac does a better job than the free option (REW), but Dirac is unquestionably a better choice for beginners since it is easier to use.

 

[Digital_Thor]

How is the speaker designed? I have the Seas DXT too, and found it quite picky when it comes to baffle design.
My point is, besides lack of power for the woofer - I tried this too, the 125 plate is way too weak for this - is there anything else that sounded wrong or not right?