The entry level 2x4 miniDSP has a 24bit 48kHz DAC and a 28bit/56bit DSP engine. Assuming typical 9dB of boost required for a heavy LT, that is same as 3 bits less resolution (2^3 = 9dB) or 21 bits of total resolution on the DAC. I think that is still sufficient for handling typical 16bit resolution of CD quality?
This is an interesting point guys - in order to flatten the responses on the horns - I have mainly used attenuation.
Does anyone know how the minidsp performs this attenuation (up to -18 dB) in PEQ?
I am using a minidsp 4x10 HD and feeding it a digital signal.
Is there loss of resolution in the same way?
If this attenuation in the PEQ is then combined with the volume attenuation - could this become significant?
Does anyone know how the minidsp performs this attenuation (up to -18 dB) in PEQ?
I am using a minidsp 4x10 HD and feeding it a digital signal.
Is there loss of resolution in the same way?
If this attenuation in the PEQ is then combined with the volume attenuation - could this become significant?
Last edited:
Hear you krivium and xrk971 and agree : )
Took a check my own DSP what electric signal it sends to DACs and can see would benefit power amp for woofer had some make up gain or more sensitive input, or even better XO in 80Hz area as bushmeister's new plans. DSP is JRiver and internal run 64bit plus DACs are 120dB capable so there is some DR to do with.
Signal sent to DSP in first picture is normal pink noise so to see electrical speaker correction and scale steps 3dB, second picture is acoustic output on axis at one point in space, and third picture is same pink noise seen on another computer without any filters.
Took a check my own DSP what electric signal it sends to DACs and can see would benefit power amp for woofer had some make up gain or more sensitive input, or even better XO in 80Hz area as bushmeister's new plans. DSP is JRiver and internal run 64bit plus DACs are 120dB capable so there is some DR to do with.
Signal sent to DSP in first picture is normal pink noise so to see electrical speaker correction and scale steps 3dB, second picture is acoustic output on axis at one point in space, and third picture is same pink noise seen on another computer without any filters.
Attachments
Last edited:
BYRTT, you just need more woofers 
....
The example above uses 21 dB between highest and lowest peak. Maximising the gain as to (just) not clip the signal and you're still left with about 16 bit resolution. Gain structure is important though. You don't want to spill bits unnecessary.
That's some difference in top and bottom though. You have almost as much difference as I have. I use up 24 dB between the lowest attenuation and largest peak. But that does make it flat from 17 Hz to 17 KHz. My DAC does 20 bits according to a test on Stereophile. Which means I theoretically still have 16 bit left, as long as I adjust the gain structure properly.
My room does worse than that though. Or my ears.... and even my speakers! I bet I don't get to enjoy 96 dB of resolution...
....The example above uses 21 dB between highest and lowest peak. Maximising the gain as to (just) not clip the signal and you're still left with about 16 bit resolution. Gain structure is important though. You don't want to spill bits unnecessary.
That's some difference in top and bottom though. You have almost as much difference as I have. I use up 24 dB between the lowest attenuation and largest peak. But that does make it flat from 17 Hz to 17 KHz. My DAC does 20 bits according to a test on Stereophile. Which means I theoretically still have 16 bit left, as long as I adjust the gain structure properly.
My room does worse than that though
. Or my ears.... and even my speakers! I bet I don't get to enjoy 96 dB of resolution...
Last edited:
My guess is if you only attenuate, you're throwing away bits if you can't adjust the gain upwards. The room that's left in the MiniDSP to be able to boost will be unused and the (reserved) bits are wasted, unused as well... Unless you can up the internal gain. But upping that gain would still be boost
.I don't know if I buy the boost vs cut with one being better than the other. I understand where that might come from but it all depends on how the separate components are used and what they are exactly.
In the end there's a curve that works, in order to maximise the gain structure you don't want to have any more room above that curve than to leave the maximum peak signal unclipped. Just enough not to clip your DAC, making sure your Amp can handle that peak input.
For me there's nothing wrong with boost, as long as you're not trying to boost a null
. It's nothing to be scared about when you're working with digital signals. As long as you understand what you can and cannot do with boost.Upping the internal gain and then using attenuate only is resulting in the same net curve as boosting half and attenuating the other half of the signal.
Only choosing to attenuate on a device that has an equal space reserved for boost is like I described above, you're actually wasting bits...
Thanks structure analyze wesayso, agree with woofers it will improve gain structure for all DAC outputs, hard thing is probably in home enviroment measure down there to get midbass verse sub slopes perfect.
Think will drop that 10 incher then and get 22w's in duty with 10F's in a enclosure with nice 3 inch rounded sides as on Grimm LS1 and then add a pro 15 incher down at floor level.
it will improve gain structure for all DAC outputs, hard thing is probably in home enviroment measure down there to get midbass verse sub slopes perfect.Think will drop that 10 incher then and get 22w's in duty with 10F's in a enclosure with nice 3 inch rounded sides as on Grimm LS1 and then add a pro 15 incher down at floor level.
... and what happens if by unluck your eq algorythm just 'create' a peak output not present in initial source signal Ronald (which is likely to happen much often than you could think using digital frequency modification)? Well you don't throw bits out doing an eq: you are eqing a signal! You are not targeting a level, you modify a frequency response... This is not excactly the same as using an attenuator for level control.
About the audibility of cuts versus boost in eq this is phase related. This is one of the reason passive analog eq are still highly praised by sound engineer around the world.
And i think your way of thinking is flawed about the boost in digital: even if you boost (using floating point for dynamic) you end up with an attenuation not to overload your signal path. In the end this is equal from a (digital) gain structure, except you gained headroom when you cut and are doing the frequency response minimizing audibility of phase.
Sorry to Hijack the thread...
Last edited:
... and what happens if by unluck your eq algorythm just 'create' a peak output not present in initial source signal Ronald? Well you don't throw bits out doing an eq: you are eqing a signal! You are not targeting a level, you modify a frequency response... This is not excactly the same as using an attenuator for level control.
If the device you use has enough head room to boost without clipping you're leaving that space unused. Why would an EQ algorithm create a peak not present in the original source... I don't follow you here. Are we not responsible for the EQ we apply? I certainly don't get any nasty surprises.
There's more to the story of coarse, how much headroom is needed and how does your device of choice handle that headroom.
If You EQ with your eyes on minimum phase behaviour the EQ boost is going to fix that phase(*), not make it worse, if you are doing your job right, that's the whole point of EQ-ing to begin with. You just shouldn't want to blindly EQ non minimum phase behaviour. Like I said: know what to boost and what not to boost.
The exception being to create some kind of house curve, but for those you are not using narrow Q boosts.
(*) for a certain point in space and time
That's why I specifically said that you need the room to play the loudest peak you want to play. Nothing flawed there i.m.h.o, if a MiniDSP allows for 18 dB boost and you're only using cuts you waste 18 dB worth of bits, if the device doesn't clip with the use of that 18 dB full on boost. That would depend on what the maker of that device intended. Does it have headroom above that reserved boost region.
That would depend largely on how the device is designed. I know how this works in JRiver, and if you look at it form that point of view, I attenuate everything because I create the needed headroom by lowering the internal volume in my case.
I make sure I don't clip my signal when it's maxed out. But You have to know the device you're using to know if you indeed use all those bits. Wasting headroom is a waste of bits. Having headroom is a necessity
.
Last edited:
Because digital is not as direct as analog for this kind of treatment. I will overly simplify but when we talked about digital attenuation we talked about some not finished division (i don't know how to say it in english so: sometimes you'll have rsults of calcul of this kind:1.222556664488946323166877etc,etc...) at one point you have to stop the calcul and then you have an error. For eq it can result in artefacts: in some peaks not presents in original signal. If eq algorythm rely only on gain set in the eq to perform leveling this is not taken into account and you end up with a peak which was not present in initial source program.
It was one of the reason early digital (pro) gear had a 'sound' (not really pleasant) it was not totally mastered and users wasn't aware about that kind of things, and 16bits being 16bits we tryied most of the time to use fullrange of scale... result: nasty 'digital' cold sound at the (digital) source not DAC oddities.
When you choose to use cut (for digital eq) you gain additional headroom because even a cut can produce a peak, but in this case you have margin left.
I agree about that. But you are not wasting headroom when you don't use boost: you have 'additional' headroom allowed in case of you need a boost. That is the reason for increased bit (48 bit fixed, 32 floating point,...) for treatments in software or harware dsp (being summing in DAW, eq, comp,...) this give you digital margin. This is the cushion of signal path in Bob Katz own terms, if you know what i mean.
Last edited:
I see your point Ronald, but i have difficulty to see an eq like something to 'fix' phase. You know it's a professional habits: for me this is the tool to fix FR issues. And for fixing phase i'll look at FIR... But that's me.
Ok. I would like to say all of you X,Bushmeister,etc... how i'm impressed by the work you've done with this project. It's really tempting to try same thing... So much temptations so little time (and money!).
Which is why you need headroom. But wasted headroom is still wasting bits, you're not using them. Ever!
But the key point is knowing how the device is constructed. MiniDSP can boost. But at what point is it going to clip. Can you use that boost?
If you can, without the risk of clipping the signal... you may call it headroom in case you need a boost... that doesn't make sense if you don't want to boost.
In the end we are playing songs. If you setup the system to play the most demanding track, it has more headroom on every other song played. That's why I like JRiver. They set an arbitrary headroom needed to play music.
You can analyse the library, set the volume levelling option and every other song that's less demanding has more headroom than that most demanding track. It's a volume control based on the average listening level according to the R128 algorithm. So I don't have to touch a dial and my songs with big dynamic range sound as loud as the ones that are crushed. The crushed songs have huge headroom, they won't clip. Because on average their peak values are much lower. The high DR dynamic songs determine the headroom that's needed. The amp sets the desired SPL level.
I do get your point about calculation error though. But it's unlikely I'm close to the headroom set. Just know where it is and what it is, gain structure.
Thanks Krivium. You should make one - come join the fun. The XT1464 is probably available in France as well as the LTH142. I like how Bushmeister's turned out though as his design had exceedingly low disortion. It may be all that epoxy putty, sorbothane, felt, tar, rock wool. You name it, he added it to make the cabinet and horn silent.
Well that don't make sense if you don't make boost or other treatments you are right.
BUT as you are doing treatment (this is the 'raison d'etre' of a loudspeaker management dsp unit) and developper of the product don't know what final user is going to do with it it give you this additional headroom in order to perform treatments.
I have the same 'waste' when i record a performance in stereo in my DAW: it allow for 256 tracks to be recorded and mixed real time including plug ins (realtime too) and a summing buss running 48bit float calcul... Is it needed for 2 tracks? No. Have the summing buss the capacity to cope with 24db peaks (past the theorical odbfs!): yes. Is it needed? no. Do i push fader the fader of the 2 track up to 18db because the soft can cope with it and won't 'waste' DR: no. It's not needed.
The point about that is once you use a processor you are not only 'playing song' as artefact from treatments can happen between your player and your dac: you have to take that into account as i do when i use an eq or comp on some tracks in a mix.
When you say that minidsp allow for +18db of boost it's like saying my car is able to go 220km/h because the speedmeter indicate that number at the end of the range.
For the minidsp i don't know if the DR allowed for boost is ok for a 0dbfr peak +18db, and will probably never know as i don't plan to use one.
I hope for users this is the case but i doubt given the price the unit is sold and what i've seen with digital pro gear 100x the price.For the user that want to know i think minidsp have a forum and they can asnwer (or not). As for my car... maybe running downhill with a hurricane pushing the back of the car... i could approach 180km/h!
Most users don't know about those issues that is the reason i give the advice that when you have choice between boost/cut choose the cut option if you want clean sure output.
I know you know your gain structure very well Ronald and most (if not all) of choice you make are well thoughts.
I don't want to pollute the thread more about that subject, so if you want to still debate about that i'll happily do in pm or by mail.
I'll probably do in the future. You opened new ways for the synergy approach for the one like me which are poor in woodworking skills.
I am a little out of my depth here - but I know in the specific case of the minidsp 4x10 HD, when feeding it a digital signal (as I do via AES/EBU input), there is a master digital input attenuation option.
You are advised to cut this digital input signal by the same amount as your 'maximum' boost level. So if I boost a peak 18dB, I cut the input digital signal the same amount to avoid 'digital clipping'.
i.e. If I input a badly compressed song, and look at the digital input level - often it peaks at 0dBFS, so this provides no headroom at all to boost further within the DSP.
(Of course for a lot of dynamic uncompressed recordings it is rare that the digital input ever hits 0dBFS).
So to prevent signal saturation and loss of fidelity, I will always cut the input digital signal by the amount of the maximum total boost in my PEQs.
All of the above is only for digital input of course! Please correct me, if I am wrong.
You are advised to cut this digital input signal by the same amount as your 'maximum' boost level. So if I boost a peak 18dB, I cut the input digital signal the same amount to avoid 'digital clipping'.
i.e. If I input a badly compressed song, and look at the digital input level - often it peaks at 0dBFS, so this provides no headroom at all to boost further within the DSP.
(Of course for a lot of dynamic uncompressed recordings it is rare that the digital input ever hits 0dBFS).
So to prevent signal saturation and loss of fidelity, I will always cut the input digital signal by the amount of the maximum total boost in my PEQs.
All of the above is only for digital input of course! Please correct me, if I am wrong.
Last edited:
When you say that minidsp allow for +18db of boost it's like saying my car is able to go 220km/h because the speedmeter indicate that number at the end of the range.
For the minidsp i don't know if the DR allowed for boost is ok for a 0dbfr peak +18db, and will probably never know as i don't plan to use one.
I hope for users this is the case but i doubt given the price the unit is sold and what i've seen with digital pro gear 100x the price.For the user that want to know i think minidsp have a forum and they can asnwer (or not). As for my car... maybe running downhill with a hurricane pushing the back of the car... i could approach 180km/h!
Most users don't know about those issues that is the reason i give the advice that when you have choice between boost/cut choose the cut option if you want clean sure output.
I know you know your gain structure very well Ronald and most (if not all) of choice you make are well thoughts.
I don't want to pollute the thread more about that subject, so if you want to still debate about that i'll happily do in pm or by mail.
An here we come to the point of my ramblings... I don't know what MiniDSP allows boost wise, and I don't know if there's headroom above that. But if I were using such a unit: I would want to know!
For the record, my car should actually run about the same speed that's noted on it's speedo.
But before I put any trust in the speedo I'd check if the diameter of tyre I use is the one that should be on there. That's what I'm saying. I'd even confirm using other means.Here's a picture to show what car that would be:
Speedo goes to 240 km/h. Original top speed was recorded as 245 km/h for this car. I may have lost a few horses over time, but changing my exhaust should make up for that with about ~20 more horses than the original one. As a licenced mechanic I make sure it's engine is in the best condition it can be.
I'll stop here too. I couldn't drag this any further OT than I already did but hope I achieve what I want to achieve here. Don't put your trust in anything, check it! And after that double check! If you don't know what the devices you use are doing you are playing in the dark. So don't assume based on some random advice. Check your own gain structure. Are you save? Do you know why? Is there any headroom? (there should be)
Can your amp's input handle the max signal the DAC produces? Don't gamble here... get a grip on it. After that, and only then, you can make your (informed) choice on the best way to handle the situation.
Does that make sense? That's my advise here.
This is totally right for me.
And give a clue about the 'headroom' allowed for treatment (in digital in the core of minidsp). The digital signal path (gain structure) is probably done this way:
digital input receiver (max input 0dbfs) in 24bit fixed > minispd core: (24bit converted to 56 bit fixed or 32 bit floating point) > digital level control > processing (probably some headroom allowed maybe 6 or 12db past 0dbfs taking into account possible new peaks created by treatments) > digital output or dac (dithered to 24 or 16 bit fixed) max output 0dbfs> analog out.
All treatments done in digital in the core of minidsp won't harm soundquality as long as you don't allow for digital clipping, so as long as you use the digital input attenuator already mentionned to counteract a boost (in reasonnable range which -18db is): you can't ear digital level attenuation when performed in 56 fixed point/32 bit floating* so you won't loss quality doing that and keep allowed headroom in digital gain structure into minidsp.
*well some people could probably hear a degradation in soundquality but they are exception (batman or something)...
Well in fact when mastering is done usually max peak output allowed is -0.1dbfs. You probably won't notice them on your peak meter for high DR materials (because consumer's one are not as sensible as the one used in proworld (ballistic is more 'relaxed' for consumer's one) but this is possible for peaks of this amplitude to be presents in non highly compressed source.
At the end this is not a problem if your gain structure is well respected (this is the point about Ronald comments and i agree with him about that) and you have safe margin.
If you wan't to be sure about peaks in tracks you'll have to perform an analysis about them looking for them ( Soundforge, Wavelabs or other editors software package give this option including rms levels,etc,etc...). Most of the time, non highly compressed material will hit max peak (-0.1dbfs) more often that you'll have thoughs but as they stay peaks (so don't stay high level for long period) you won't notice.
Don't be surprised to find tracks that exceed 0 dbfs. It shouldn't happen but it does. More on modern day recordings than the older ones, but exceptions can be found there as well.
Here's JRiver's analyser on a couple of tracks. So even with all the care in the world, it's already clipped. I whish it was an exception.
Do note that in this above example, the sample remains under that 0 dbfs number. But the R128 algorithm reports clipping.
I have "a thing" for "devices" that actually are able to do what they claim to do.
Here's JRiver's analyser on a couple of tracks. So even with all the care in the world, it's already clipped. I whish it was an exception.
Do note that in this above example, the sample remains under that 0 dbfs number. But the R128 algorithm reports clipping.
I agree!
Nice car! Won't publish pictures of mine because situation i described is true!
I have "a thing" for "devices" that actually are able to do what they claim to do
.Attachments
Last edited: