Digital or Analog?
So has anyone compared their PC digital crossover to a purely active analog one like say the Marchand Electronics Linkwitz Riley type?
I'd be interested to know if anyone has experience with
-DCX2496
-Windows PC
-Linux PC Brutefir
-DEQX
-crossover in speaker (what designmeasurement tool do you use?)
-active line level crossover (type of opamps used?)
I'm running into some limitations with my PC crossover. While it is a step up from the Behringer DCX2496, the latency on my PC high enough that I couldnt run a DVD movie through it. I am starting to see that as a serious limitation.
The DEQX claims 4 msec's or less latency using "long FIR convolution". That would make it useable for movies and stereo.
Could an active line level crossover compete with the digital crossover in detail and transparency? With limited digital headroom on modern recordings recorded at -3db could it actually sound inferior?
How can digital crossovers not throw out bits so that they can do EQ boosting? If you didnt you'd have digital clipping wouldnt you? So even the mighty DEQX must reduce the input digitally by something like 10db to have the headroom to boost and cut frequencies.
So has anyone compared their PC digital crossover to a purely active analog one like say the Marchand Electronics Linkwitz Riley type?
I'd be interested to know if anyone has experience with
-DCX2496
-Windows PC
-Linux PC Brutefir
-DEQX
-crossover in speaker (what designmeasurement tool do you use?)
-active line level crossover (type of opamps used?)
I'm running into some limitations with my PC crossover. While it is a step up from the Behringer DCX2496, the latency on my PC high enough that I couldnt run a DVD movie through it. I am starting to see that as a serious limitation.
The DEQX claims 4 msec's or less latency using "long FIR convolution". That would make it useable for movies and stereo.
Could an active line level crossover compete with the digital crossover in detail and transparency? With limited digital headroom on modern recordings recorded at -3db could it actually sound inferior?
How can digital crossovers not throw out bits so that they can do EQ boosting? If you didnt you'd have digital clipping wouldnt you? So even the mighty DEQX must reduce the input digitally by something like 10db to have the headroom to boost and cut frequencies.
Re: Digital or Analog?
I never worked with PC crossovers but I do have Marchand crossover with updated BB opamps. I also have DCX2496 heavily modified (Lundahls transformers and XBosoz after DA converters).
http://www.diyaudio.com/forums/showthread.php?s=&threadid=76095&highlight=
I also have passive xover in speakers that I could easily turn on and off just for the purpose of comparing to different set up.
http://www.diyaudio.com/forums/showthread.php?s=&threadid=21523&highlight=
My preference would be the least amount of electronic in the chain. I use to listen straight signal from very good DA converter to passive volume to amp and through passive xover in speakers. That gives the ultimate clarity. But... Active crossover gives more precise control particularly over delay and phase alignment and obviously more options. DCX was great improvement over Marchand xover specifically in the region of noise and clarity. Unmodded DCX was not option for me so I went to do all of modification. In the first place I replaced complete output circuitry with Lundahl transformers and that was the ultimate sound for me. Unfortunately not all amps have enough gain so I added Xbosoz preamps in the output. Still very good and transparent.
I have DEQ in front of DCX and that gives much more freedom regarding digital signal. I do not have to do much of equilazation so I do not know what to tell you what are the limits.
When I did test with passive xover in the speakers vs. digital xover there is not much difference after the fact that there is attenuation due to elements in the path of the signal. In some set ups that maybe is not big deal but in mine it is. I am using low powered 2A3 amp for the mid driver. Active set up allows me to use different amps: tube, class A for tweeters and class AB for the bass and that makes big difference. Since all amps and speakers are with different powers and sensitivity I level that with various gains in preamp sections after DA converters in DCX. My 6 channel balanced volume control is between XBosoz preamps and amps. That way I avoid distortion in the DA section of DCX.
All in all I prefer digital signal vs. analog xover or passive. Still very good set up is possible with careful design on passive xover, but that also very much depend on complexity of speaker design.
I did have a chance to listen DEQX in very good system, but I didn't like the sound of opamps output in it. The ultimate thing would be option with digital outs but I haven't had a chance to listen to it. I am very confident that modified output DCX could be compared to DEQX and that there is not much difference between them - certainly not equal to the price difference. The biggest difference is in output section and I would strongly advise anyone to completely abandon opamps in output of DCX.
So much from me, I hope this gives you some of the info. I am sure others will fill you in areas that I have no experience.
Stay well
AR2
Daveis said:
I never worked with PC crossovers but I do have Marchand crossover with updated BB opamps. I also have DCX2496 heavily modified (Lundahls transformers and XBosoz after DA converters).
http://www.diyaudio.com/forums/showthread.php?s=&threadid=76095&highlight=
I also have passive xover in speakers that I could easily turn on and off just for the purpose of comparing to different set up.
http://www.diyaudio.com/forums/showthread.php?s=&threadid=21523&highlight=
My preference would be the least amount of electronic in the chain. I use to listen straight signal from very good DA converter to passive volume to amp and through passive xover in speakers. That gives the ultimate clarity. But... Active crossover gives more precise control particularly over delay and phase alignment and obviously more options. DCX was great improvement over Marchand xover specifically in the region of noise and clarity. Unmodded DCX was not option for me so I went to do all of modification. In the first place I replaced complete output circuitry with Lundahl transformers and that was the ultimate sound for me. Unfortunately not all amps have enough gain so I added Xbosoz preamps in the output. Still very good and transparent.
I have DEQ in front of DCX and that gives much more freedom regarding digital signal. I do not have to do much of equilazation so I do not know what to tell you what are the limits.
When I did test with passive xover in the speakers vs. digital xover there is not much difference after the fact that there is attenuation due to elements in the path of the signal. In some set ups that maybe is not big deal but in mine it is. I am using low powered 2A3 amp for the mid driver. Active set up allows me to use different amps: tube, class A for tweeters and class AB for the bass and that makes big difference. Since all amps and speakers are with different powers and sensitivity I level that with various gains in preamp sections after DA converters in DCX. My 6 channel balanced volume control is between XBosoz preamps and amps. That way I avoid distortion in the DA section of DCX.
All in all I prefer digital signal vs. analog xover or passive. Still very good set up is possible with careful design on passive xover, but that also very much depend on complexity of speaker design.
I did have a chance to listen DEQX in very good system, but I didn't like the sound of opamps output in it. The ultimate thing would be option with digital outs but I haven't had a chance to listen to it. I am very confident that modified output DCX could be compared to DEQX and that there is not much difference between them - certainly not equal to the price difference. The biggest difference is in output section and I would strongly advise anyone to completely abandon opamps in output of DCX.
So much from me, I hope this gives you some of the info. I am sure others will fill you in areas that I have no experience.
Stay well
AR2
Re: Digital or Analog?
I have programmed a PC crossover and using my means I cannot imagine how you can preserve the SQ and have it in sync. It is over 100 ms in my case. But I am sure it is not difficult at all to delay the pciture, is it ?Daveis said:
Maybe I do not understand sth here but for me it this reasoning is so much simplified that is rather far from the truth. The actual bits are in the time domain whereas the bass boost is in the freq. I have experienced sth like 1 bit loss (I had to divide it by 2 not to get an overflow) when I equalized heavily (over 6 dB the under 100 Hz area) . By ears it sounds much much more dynamic than the analog counterpart. But this has rather little to do with the bits I think ... But it is only my observations not scientific research.Daveis said:
Can someone confirm that lowering the volume digitally by 6 db is the equivalent of throwing out the LSB(least significant bit) of the 16-bit signal? In other words, for each bit you are reducing the sound level by what db?
In retrospect, my Panasonic XA-SR70 receivers do the first 20db or so of volume control in the digital domain. I dont think it harms the sound appreciably.
Pawelp, what programming language and sound libraries did you use for your crossover?
In retrospect, my Panasonic XA-SR70 receivers do the first 20db or so of volume control in the digital domain. I dont think it harms the sound appreciably.
Pawelp, what programming language and sound libraries did you use for your crossover?
To the question phrased this way the answer must be yes. I wanted to just point out you are talking about lowering volume in the actual time domain.Daveis said:
And are we really loosing sth ? We can do it the other way round - we can say that we reduce the outstanding mid and/or high frequencies not boost bass. Would it mean that we gain bits ? The thing for me is really not about loosing bits - since it is in the frequency domain we can only talk about flattening the freq response. Actually what happens is that the sound gets really "fast", punchy and dynamic.
The real issue though is how the woofer handles the additional excursion ...
C++
Daveis said:
The error in this view is that nobody that is the least bit concerned about sound quality uses 16 bit words for anything but the input from the CD. When you have a 24 bit output word size, the bits aren't 'lost', but are rather just shifted down into the lower bits. As long as the D/A stage has sufficient resolution, you aren't losing anything at all.
*IF* you only used 16 bit words at the output, then you'd 'lose' 1 bit for each 6dB reduction. This is not a great way to do things, but if you dither the result properly I suspect that it would actually sound pretty decent. without dither though, it would be noticably inferior to other approaches on a decent system.
Actually, the Pannys use an odd scheme. I *think* that from a reading of -20 or so up to 0, they actually BOOST the signal digitally. From -20 down another 12 or 18 dB depending on model, they drop the power supply voltage. Only below this point do they use digital attenuation. I played around with this as I am running my Panny XR25 off batteries for the output supply. In the range where the Panny uses voltage to control volume, in my setup it does nothing. However, below this range it works fine.
dwk123:
My current music starts off as 16-bit CD and is processed at 64-bits (double precision floating point?) by Voxengo VST plugins. Then it is probably dithered to 16-bit output. Voxengo claims that all internal processing is at 64-bit.
If I understand you correctly, if I converted the 16-bit to say 24/96 and then did EQ/crossover with 64-bit precision and then outputted through a 24/96 DAC I'd have superior sound.
What I'm trying to wrap my head around is that to do EQ boosting I've found that I need to lower the volume of music digitally by 10db or so. That way boosting frequencies doesnt cause them to clip. The writers of the Voxengo plugin tell me that conversion to 24-bits doesnt help. I suspect that's because all their math is done internally at 64-bits and only at the end dithered to 16-bit.
Just for fun I may try a Behringer SRC2496 (if I can find one) to see if I can hear a difference.
----
The question I asked Voxengo:
I've noticed that many modern CD's are recorded at -3db. When I try to do EQ with boost I create digital clipping. Any way to have music I play through GlissEQ automatically lower the volume so that it doesnt clip? Would scaling 16-bit audio to somewhat less than 24-bit's increase the digital headroom for doing EQ?
----
Their reply:
I suggest you to use limiter on output - for example, Voxengo Elephant. Increasing bit-depth from 16 to 24 won't help here.
-----
I think their answer tells me that they are doing math at high precision without truncation at intermediate steps. If I feed the output of their EQ plugin into their limiter, it should automatically limit some of my CD's that are recorded with compressed dynamic range. My understanding is that in my current VST plugin host that sound can be processed at any arithmetic precision internally, but that sound is converted to 32-bit float when it goes from plugin to plugin. That is still a bit higher than 24-bit.
On another DIYAudio thread it seemed someone said that the Behringer DCX2496 looks at your min/max EQ settings and automatically scales your audio so that it wont clip.
My current music starts off as 16-bit CD and is processed at 64-bits (double precision floating point?) by Voxengo VST plugins. Then it is probably dithered to 16-bit output. Voxengo claims that all internal processing is at 64-bit.
If I understand you correctly, if I converted the 16-bit to say 24/96 and then did EQ/crossover with 64-bit precision and then outputted through a 24/96 DAC I'd have superior sound.
What I'm trying to wrap my head around is that to do EQ boosting I've found that I need to lower the volume of music digitally by 10db or so. That way boosting frequencies doesnt cause them to clip. The writers of the Voxengo plugin tell me that conversion to 24-bits doesnt help. I suspect that's because all their math is done internally at 64-bits and only at the end dithered to 16-bit.
Just for fun I may try a Behringer SRC2496 (if I can find one) to see if I can hear a difference.
----
The question I asked Voxengo:
I've noticed that many modern CD's are recorded at -3db. When I try to do EQ with boost I create digital clipping. Any way to have music I play through GlissEQ automatically lower the volume so that it doesnt clip? Would scaling 16-bit audio to somewhat less than 24-bit's increase the digital headroom for doing EQ?
----
Their reply:
I suggest you to use limiter on output - for example, Voxengo Elephant. Increasing bit-depth from 16 to 24 won't help here.
-----
I think their answer tells me that they are doing math at high precision without truncation at intermediate steps. If I feed the output of their EQ plugin into their limiter, it should automatically limit some of my CD's that are recorded with compressed dynamic range. My understanding is that in my current VST plugin host that sound can be processed at any arithmetic precision internally, but that sound is converted to 32-bit float when it goes from plugin to plugin. That is still a bit higher than 24-bit.
On another DIYAudio thread it seemed someone said that the Behringer DCX2496 looks at your min/max EQ settings and automatically scales your audio so that it wont clip.
Is this thinking correct.
You take 16-bits. To lower volume you have to shift/divide bits. No matter how you do that you are throwing out data from the source.
Now take that 16-bit signal and shift/multiply it into 24-bit data. If you reduce the volume of that 6-12db you are just shifting 1-2 bits over. You havent lost any of the original 16 bits until you shift/divide the 8-bits you gained out.
This is why digital volume controls work to a point. The Pansonic receivers use it to good effect.
So I really should try to find a VST plugin to convert from 16-to-24 bits? Or failing that get a Behringer SRC2496 to do the conversion in hardware instead?
I looked at the DBX processor... Yikes it's $3750.. as bad as the DEQX. I'm still happy with me PC and DCX2496.
You take 16-bits. To lower volume you have to shift/divide bits. No matter how you do that you are throwing out data from the source.
Now take that 16-bit signal and shift/multiply it into 24-bit data. If you reduce the volume of that 6-12db you are just shifting 1-2 bits over. You havent lost any of the original 16 bits until you shift/divide the 8-bits you gained out.
This is why digital volume controls work to a point. The Pansonic receivers use it to good effect.
So I really should try to find a VST plugin to convert from 16-to-24 bits? Or failing that get a Behringer SRC2496 to do the conversion in hardware instead?
I looked at the DBX processor... Yikes it's $3750.. as bad as the DEQX. I'm still happy with me PC and DCX2496.
Back to the 'how to boost a signal without losing bits' question...
If the signal is below 0dB at that frequency where the boost is occurring then boosting at that frequency will not clip the signal at that frequency.
However, if the overall signal has now increased above the 0dB limit then clipping will occur.
to improve S/N on my DCX2496 I have run the inputs (digital) at -1dB and boosted the crossovered outputs up to +6 or +8dB. Each output still does not clip with a 0dB input, because each output is only reproducing a small part of the total spectral signal.
If the signal is below 0dB at that frequency where the boost is occurring then boosting at that frequency will not clip the signal at that frequency.
However, if the overall signal has now increased above the 0dB limit then clipping will occur.
to improve S/N on my DCX2496 I have run the inputs (digital) at -1dB and boosted the crossovered outputs up to +6 or +8dB. Each output still does not clip with a 0dB input, because each output is only reproducing a small part of the total spectral signal.
I know this is an old thread but I wondered if there is anything available off the shelf nowadays that answers the original question . could you use for instance an optical splitter and 3 ultra curve pros and just eq all the frequencies you didn't want to go to each particular driver down to zero so for instance the midrange driver was only getting midrange frequencies and so on . 3 dacs with remote volume control afterwards and your there ?????
MiniDSP expensive ? 
Imagine my goals are the same and for a 3-way active setup with 1 stereo DAC/way I'd need to take 3 from the commercial DAC I chose recently and it costs more than $2k each. Furthermore I also had to spit out 3 differently filtered I2S digital signals to feed the DACs - shall we take 3 Raspberry Pi-s with a decent reclocker or maybe shall I build a PC with 3x Pink Faun PCIe cards and hope that Foobar supports all 3 of them simultaneously ?
This is hell of a money OR it's very complicated with commercial stereo audiophile grade DACs - or you stay on the cheapo side from eBay with little cheap DAC modules, USB and then you might end up with more compromise due to low quality parts than having an "ordinary" stereo setup while having a traditional active crossover instead of filtering in digital domain. But I have no experience with both of these so I don't have a comparison either, just thinking loud.
Today I still see digital-input-capable DSPs the only way to achieve filtering in digital domain with the (hopefully) precise math in the background while keeping cost and complexity on a reasonably low level.
Using a classic computer for filtering: I'm afraid of crashes, not always by hardware but by software / OS: years ago in some cases my sound card had a terrible loud signal stuck into a loop on all frequencies and this huge buzz was kept on 'til I shut the PC off by hand. Now when the same happens and your separate DAC suddenly decodes and sends a huge bass tone to your tweeter's amp, you're doomed and goodbye expensive tweeters. Oh, you gonna protect it with a passive element, a capacitor, just in case such thing happened ? -> you're losing the pure passive-element-free signal path (at least from amp to driver) so you lost one of the most interesting aspects of active amplification.
Imagine my goals are the same and for a 3-way active setup with 1 stereo DAC/way I'd need to take 3 from the commercial DAC I chose recently and it costs more than $2k each. Furthermore I also had to spit out 3 differently filtered I2S digital signals to feed the DACs - shall we take 3 Raspberry Pi-s with a decent reclocker or maybe shall I build a PC with 3x Pink Faun PCIe cards and hope that Foobar supports all 3 of them simultaneously ?
This is hell of a money OR it's very complicated with commercial stereo audiophile grade DACs - or you stay on the cheapo side from eBay with little cheap DAC modules, USB and then you might end up with more compromise due to low quality parts than having an "ordinary" stereo setup while having a traditional active crossover instead of filtering in digital domain. But I have no experience with both of these so I don't have a comparison either, just thinking loud.
Today I still see digital-input-capable DSPs the only way to achieve filtering in digital domain with the (hopefully) precise math in the background while keeping cost and complexity on a reasonably low level.
Using a classic computer for filtering: I'm afraid of crashes, not always by hardware but by software / OS: years ago in some cases my sound card had a terrible loud signal stuck into a loop on all frequencies and this huge buzz was kept on 'til I shut the PC off by hand. Now when the same happens and your separate DAC suddenly decodes and sends a huge bass tone to your tweeter's amp, you're doomed and goodbye expensive tweeters. Oh, you gonna protect it with a passive element, a capacitor, just in case such thing happened ? -> you're losing the pure passive-element-free signal path (at least from amp to driver) so you lost one of the most interesting aspects of active amplification.
Having sad this and gone through tons of videos of Mr. AIX aka. Dr. Mark Waldrep, master material quality and our hearing capability I think 96/24 is already enough for ANYBODY wanting a decent high-end sound, especially if I'm a PCM guy and don't believe in DSD ('cause that's the case). Just spare for a miniDSP with digital input, no I2S here as far as I know but still the best solution today I think.
Nice accurately clocked separate I2S connections (per way) would be the top of the mountain in reality, possibly with PS Audio compatible pinuot via HDMI balanced connection, into 6 ultra-high-res mono DACs and then proper playback software on computer side but hey, it's 2018 only, we still have plenty of room to advance.
Nice accurately clocked separate I2S connections (per way) would be the top of the mountain in reality, possibly with PS Audio compatible pinuot via HDMI balanced connection, into 6 ultra-high-res mono DACs and then proper playback software on computer side but hey, it's 2018 only, we still have plenty of room to advance.
hahah you're right I need to keep looking for another option . I'll look again at mini dsp . once you start thinking of buying multiple dacs etc the cost adds up quickly probly more cost effective to buy some high quality studio monitors like Neumann 310
I found a nice sounding dac with remote control preamp in the Yamaha dsp a3090 I paid £60 so thought about multiple of these as you could always insert a better dad in the signal path and still take advantage of the remote volume control if you wanted to .
kind regards James
I found a nice sounding dac with remote control preamp in the Yamaha dsp a3090 I paid £60 so thought about multiple of these as you could always insert a better dad in the signal path and still take advantage of the remote volume control if you wanted to .
kind regards James
+1 to lot's of good miniDSP offerings.
The 2x4HD is a great way to wade in, IMO....it even has a bit of FIR capability.
The nanoDIGI is a 2x8, digitial in / digital out, for $170....... if you are really wanting to supply your own DACs.
I use a set of 4 openDRC-DI's that feed a proaudio stagebox with 8 balanced DACs, but that's a pretty expensive solution.
I havent been able to find an truly affordable high quality commercial solution.
Best so far is a Linea Research ASC-48 (or Danley SC-48 which is the same unit)...but it too, is $$$.
Yeah that's pretty good idea, try it and inform us here if you like the sound or not.
At the moment I'm still thinking which way to go. Get a mini DSP from a friend and try it with its digital input and then several crossover models inside.. or try the good-old-classic 3 way active filter opamp circuitry built into the preamp or something completely different.. not sure yet.
But for the very beginning a good signal generator and a good measuring microphone will do the trick and tell me if I can omit the DSP or I will need it 'cause if I can omit it, I'll use something like this (resized for my crossover points) and then use the expensive DAC which also employs a built-in volume control on digital level (so I don't need to use an attenuator in the signal path at all). Not sure yet. Anyway, the measurements and my ears will tell me, what I need. I would suggest the same for anybody who wants a decent sounding system ('cause you know.. there're DSP'd systems which sound very nice on paper but in reality they're just way too lifeless as some people told me - I think it's about taste)..
At the moment I'm still thinking which way to go. Get a mini DSP from a friend and try it with its digital input and then several crossover models inside.. or try the good-old-classic 3 way active filter opamp circuitry built into the preamp or something completely different.. not sure yet.
But for the very beginning a good signal generator and a good measuring microphone will do the trick and tell me if I can omit the DSP or I will need it 'cause if I can omit it, I'll use something like this (resized for my crossover points) and then use the expensive DAC which also employs a built-in volume control on digital level (so I don't need to use an attenuator in the signal path at all). Not sure yet. Anyway, the measurements and my ears will tell me, what I need. I would suggest the same for anybody who wants a decent sounding system ('cause you know.. there're DSP'd systems which sound very nice on paper but in reality they're just way too lifeless as some people told me - I think it's about taste)..
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.