I was thinking about upgrading my crossover electronics (the caps) to high quality ones, for example Jantzen Superior, but then I started thinking. I am already using HQPlayer for filters and digital room correction, why not switch DAC to a multi-channel DAC (maybe the Okto Research DAC8 Pro) and skip the speaker-crossover totally and instead drive each element directly and use an DSP crossover in HQPlayer.
Creating crossover filters in for example REW should not be a problem (I know the crossover Hz, only listen and decide which steepness sounds best).
Is this possible and a good idea? Anyone experienced with this? How much gain in sound quality? Does HQPlayer pipeline perform good synchronization between convolution to avoid phase problems?
@Magnus, your point leads me to my somewhat related question. I do have an Okto dac8pro for multichannel listening (5.2 configuration), using HQP to house my Audiolense convolution filters. I am planning to go “commando,” so to speak, and bypass my Marantz AV-8802A preamp entirely using a Coleman 7.1SW switch to change sources and RooDial via Roon for volume control. So, I will lose all bass management crossovers when I do this. (Hopefully, I will still have enough system gain without the preamp.)
I assume I will need to do high pass/low pass filters in Roon in front of HQP to accomplish the bass crossovers. Would you kindly share how you would go about doing this? It sounds as though it would also inform the OP in this thread. Thanks. JCR
But is HQPlayer suiteable for something like this, provided the crossover convolution files are made and everything else setup? Will I need some special streamer or is my current upgraded microRendu running NAA enough?
Yes, there are people using HQPlayer the way you describe.
You can also use the IIR filters provided by HQPlayer matrix processor.
Depends on the DAC you use. If you use exaSound ethernet connected DAC, it includes NAA. Otherwise with exaSound you need to use Windows or macOS computer. With Merging DACs you don’t need any NAA unless for some reason you specifically want to have such, since they are ethernet connected devices.
What is your current crossover? Is it a textbook crossover or was it made specially for your drivers?
Most loudspeaker drivers need some form of correction to become linear. Your examples with hqplayer, REW and Rephase only provide textbook filters. Changes are the end result will be worse then your current ceossover. Textbook filters hardly ever work
No, you can certainly generate corrected filters that also in addition take into account room responses.
You can run full active speaker DSP cross-overs in HQPlayer. You can run practically unlimited number of combination of convolution and parametric eq filters in HQPlayer matrix processor.
If I did this I would certainly measure the speaker with mic close, and then compensate for phase shifts and non-linearity. Then room correction on that in its own convolution and finally merge all into one convolution file/driver so 4 convolution files in total, 2 for each speaker (the speakers I use are 2-way).
@jussi_laako my current setup is Roon + HQPlayer on Windows, then stream to NAA on an upgraded microRendu, will this work with 4 channels as well provided of course that microRendu connects to a 4+ channel DAC.
I use REW to measure each driver with very close mic, outdoors when it was dead quiet (had to redo anytime a bird chirped). But all over in 1 hour, for 6 drivers. Did multiple measurements with each driver. Then use results to confirm good crossover points.
Then in listening room measure at listening position using Audiolense XO. Simply input your pre-determined XO points.
Audiolense XO spits out the mono wav correction files (impulse response) for HQplayer, which already sort out both crossover points and room correction.
130k tap FIR filters. Linear phase crossover and min phase room correction.
But you can choose whatever you like.
In HQPlayer, use matrix feature to load the mono wav impulse responses. Key part here is correct channel mapping.
I have multiple target curves (variations of Harman curve, to personal preference ) setup with different matrix “profiles” in HQPlayer.
While music is playing, using HQPDControl app on iPad to switch different profiles.
Tone controls, done the proper way.
Roon goes through HQPlayer, as does Spotify, Apple Music etc
With 100 inch screen, I use a zero latency convolver (can’t use HQPlayer here) with 65k tap min delay filters created by Audiolense - for Netflix, sport etc. No lip sync issues.
Well, this I do understand. I use Audiolense XO as well for MCH mono filters for the HQP convolution pipeline matrix. And multiple microRendus.
This is running through my Marantz preamp using the analog MCH inputs. So, of course, Audiolense XO is doing the crossovers and @jussi_laako, I indeed do not need, and have not been using, the Roon convolution engine for my Okto MCH DAC.
However, my 2.2 stereo setup, using a different USB DAC, is seen by Audiolense XO only as two full-range speakers. Therefore, my preamp has been doing the bass crossovers and that’s the functionality I lose with my preamp-bypass experiment. In HQP, I input left and right channel mono wav files generated by Audiolense XO into the convolution setup, and not the pipeline.
Might I derive the proper 2.2 crossovers from the Audiolense XO measurement made with my Okto MCH DAC in the same room? And then, can someone kindly explain how the crossovers are inputted and used in HQP?
We can move this to another thread if this is getting off topic. Thanks. JCR
Ah o.k. interesting. Applying room correction in the crossover however sounds a bit like odd practice.
I don’t have very positive experience with convolution crossover created by in room measurements or very close range measurements. The problem with close range measurments is that you have to take in account the dopple effect, the break-up behaviour and the off-axis behaviour of your drivers. The mic is at different distance for different parts of the cone. Most of the time you’ll only be measuring the output of the dustcap and the rest of the cone and surround as delayed signal. Works better fo very stiff drivers but even then, it is very tricky to get right. I did have some positive experience however by creating them from measurements in a dead room but dead room rental is expensive and the results weren’t especially better, just different, than via mathemetics combined with gated measurements and manually programmed dsp with AudioWeaver. The gated measurements however are not suitable for automated convolution. With convolution filters made from measurements you can also hear the quality of the measurement mic and preamp being used. An earthworks M30 mic with a Metric Halo audio interface made better soundig filters than with a minidsp U-mik. Bit more expensive but then again, some people spend more on a single ethernet cable. All the automated convolution methods I have tried so far were not capable or had difficulties to see the difference between difraction artifacts, breakup artifacts and direct signal. In turns out that interpretation of the measurements is still human work. Just pointing a microphone and let the computer do the rest dpesn’t lead to good resulst. Maybe I missed some developments because I lost interest in them but that’s my experience up till now. A well, Loudspeaker design is complex material, it has kept me busey for decades by now and I’m still learning. Me too thought for a while that pc power with convolution filters would be THE answer, but so far it has always ended up sounding a bit artificial and emotionless compared to my manual efforts. It ended up being just as full of compromises than other methods so you still have to choose wisely. Alltogether It takes just as much knowledge about loudspeaker design to take it a step further then average.
An Okto is off of the availability list and orders suspended for some time now. Unless you already have one or know of a second hand one available.
The Dac8 stereo orders are trickling out but not heard much about the MCH unit deliveries tho. I picked up my Stereo unit from another owner (luckily) and didn’t have to wait till the orders opened up again…when ever that might happen.
Naturally you would keep them separate as well, but just for performance reason you can merge them to get less CPU usage. The crossover filters will push the part of the room correction down to zero where its not applicable, above or below the crossover point. The sound should not be affected compared to using separate convolution files.
Are you merging the room correction part in all channels? That way it would be the same as aplying it seperatly to the whole system. Haven’t tried it that way myself but you could spare some cpu cycles by doing that indeed. Makes it a bit more complicated for tweaking afterwards but clever thinking nonetheless
When you wrote “I use Audiolense XO as well for MCH mono filters for the HQP convolution pipeline matrix.” - it’s exactly same process.
As I mentioned in my post above:
One set of impulse response files that manage both XO and room correction.
You already know how to use HQP matrix feature (including correct channel mapping - this can be a nightmare with lots of channels ! ) so you’ve already conquered most challenging part
