john k, is your impression about the miniDSP based on their precessing power or the already available plugins?
the DDX technology takes up to 10 convolution filters per channel, now i don't actually get the meaning of it, indeed it does not sound easy at all, but what would be the power/complexity needed by an hardware implementation of the resulting filters from soundEasy?
the DDX technology takes up to 10 convolution filters per channel, now i don't actually get the meaning of it, indeed it does not sound easy at all, but what would be the power/complexity needed by an hardware implementation of the resulting filters from soundEasy?
I believe that would be HOLM Acoustic's DSPreLab. Only problem is you need a DSPre to go with it and I've never been able to get Thomas (owner of Holm) to quote on either a three or four way version. In addition to Allocator and miniDSP, Motu and Steinberg are pretty close as well. If I didn't have a requirement for linear phase I'd look to a Motu Ultralite in standalone mode.
The recommended sound card list has been updated to have some connection with reality? Last I checked the listed cards had all been off the market for quite some time. Also, Bohdan's made a number of statements about the code that lead me to conclude SoundEasy's audio core is probably not implemented very well. That's consistent with the poor UI design and its general unreliablity, so I see the need for a recommended sound card list as more of a symptom than a solution. We are, after all, talking about software whose unit tests are so badly written (or perhaps nonexistent) it's unable to do a simple frequency sweep reliably. And one whose author has so little passion for quality he doesn't bother to reply to the threads in his own forum even after key supporters of SoundEasy have reproed a problem.
So you can see why I'd be sceptical of any DSP module from Bohdan. Even if the control UI's stable and easy to use I wouldn't trust the filter synthesis. The great thing about stuff like Allocator is you can grab the demo and measure the results at zero cost---I modeled and measured several crossovers and EQ configurations in Allocator before comitting to it. That approach obviously doesn't work for hardware, but unless I see a value proposition comparable to Allocator or Motu it's a non-starter for me.
Please don't think I'm singling Bohdan out here; you can check the miniDSP forum here on DIY and see my post were I told them essentially the same thing about feature requirements. Also, in addition to testing the Allocator and PLParEX3 demos, I verified HOLMImpulse with measurement equipment before adopting it. If SoundEasy works for you and cuibono, that's great. But for me the observed reliability is not viable when combined with the lack of support. Hence I cannot, in good faith, be as positive about it as you two are.
I am certainly not a dsp hardware expert. Bohdan has done all the coding of the algorithms used in SE. I have provide guidance and theory. The biggest stumbling point seems to be that to do the type of filtering we do doesn't fit the standard IIR, FIR algorithms commonly used in audio processing. I could not tell you how much processing power is required. I just don't know off hand. It's not on my side of the equation. How this started out was basically I asked Bohdan, "Can you do this?" and he took a look and it evolved into the UE. Anything that might develop with miniDSP would require new plug-ins. We may still do something with them but I don't think it will be the UE. I really don't want to say anything more at this time re miniDSP.
My understanding is that Bohdan will be updating his PC and sound devices this year so things may change. Anyway, all I know is that all the sound functions (measurement, dsp, etc) work very well and are stable on my PCs so I can't really comment. Zaph Audio uses SE for his designand measurements, so I guess he has had success with it to. The only problem I have with the SE dsp crossover emulation is the PC nice. I have one PC that is very quiet and one that is a little too noisy to suit me. But I never have any problems with filter emulation or stability. I have run the filters for days without problems, PC related or otherwise.
Anyway, if we can get this out of the PC I think it will be very different than SE. And I suspect it will be extended to 4-way systems.
Indeed, one of the main reasons I bought SoundEasy was because you and Zaph use it professionally. Unfortunately I wasn't able to replicate your results and, since my 32 bit XP laptop runs everything else just fine, the problem's plainly SoundEasy. If a DSP unit's going to compete effectively with Allocator, Holm, Motu it has to be radically different from SoundEasy. You obviously know your business and Bohdan better than I do but, ya know, the fact you've a business selling a separate user's guide would seem to say something about your business partner's ability to deliver quality software. Best of luck mitigating that execution risk.
Please understand that SE is 100% Bohdan's baby and business. I only have input in the form of ideas and I try to test it as thoroughly as possible. I do this as part of the DIY community.
I wrote the guide because the user's manual is rather complex and I felt that SE, despite what some say about the learning curve, is about the best bang for the buck out there. The manual is more of a reference to what SE can do rather than a "how to". With the guide I tried to make the code more accessible to the first time user. It is a completely independent effort. Bohdan hs been gracious enough to recommend it to purchasers of SE because of all the effort I have provided to improving the features in SE over the years.
We shall see what happen moving forward with the UE technology. I have wanted to get the SE dsp technology out of the PC for years. I hope it will finally happen.
Understood. Anyway, the unwindowed, zero padded FFT used in v16 and earlier is not particularly sophisiticated. You don't mention what the implementation in v17 is but I'd be curious to know how well it handles phase correction and where its IMD floor is.
Can you explain? In both theory and measurement I've found even order LR and odd order Butterworth sum flat in both warped phase and linear phase. Also, so far as I know, Bessel filters don't sum flat regardless of whether they're warped or linear phase.
I'd encourage you to spend some time with Allocator. Not only is its combination of textbook and measurement more flexible than v17's measurement based approach (at least as I understand its description) Allocator's also easier to use.
Your comments would indicate that you do not understand what the UE in SoundEasy does and how it operates. I can understand that if you don't have V17 and have not played with it. There is a lot more to the dsp filtering that is discussed inthe SE manual. No big deal. I downloaded the Allocator demo and played with it today. It is certainly more complex to set up than the UE in SE.
As for the Butterworth and LR flatness there is a difference in how the two codes operate. The Allocator sets up a conventional speaker and the linearizes the phase of the system. This may or may not result in linear phase response of the individual band passes which compose the system.
The UE linearizes the phase of each band pass so the linear phase system is always the sum of linear phase band passes.
With LR acoustic targets the Allocator and UE will yield the same result, a flat linear phase system with linear phase band pass components. In the case of odd order Buterworth targets, the Allocator will yield a flat, linear phase system but the band pass responses will not be linear phase and the polar response will remain that of a typical odd order crossover.
The UE set to linear phase when using Butterworth targets would have a 3dB hump in the response at the crossover point which is correctable with a notch filter and the polar response would be like the LR family. With the UE the tilt of the polar response can be controlled by adding delay to one (or more) pass bands.
Anyway, here is an example of how it handles phase. This is a band pass response with 30 Hz B2 HP and 300 Hz LR6 LP cut offs.
An externally hosted image should be here but it was not working when we last tested it.
There's abundant evidence SoundEasy usability is poor for most folks and, so far, you're the only one I know who's asserted Allocator more difficult to use than SoundEasy. So I'm curious why you say that. Different folks definitely put concepts together differently and find their own paths through software, so measuring software usability's more of a probability distribution thing than something deterministic. It's always interesting to look at the thinking that puts folks in different parts of the distribution.
Not a big deal for those with project cash to burn and spare time to set up the measurements. Thanks for sharing the implementation details; while it's possible to correctly guess how Bohdan coded the feature the text itself is ambiguous (for the record, Jan had to document essentially the same thing in Thuneau's forum as the Allocator manual's even less specific). It's been my experience the even order LR alignments in Allocator sound slightly better to my ear than the odd order Butterworths. Disentangling the filter implementation and radiation pattern effects from what might be varying with the drivers as the crossover changes is tricky and doing the measurements to determine what's going on hasn't made its way to the top of my priority stack yet. It would certainly be interesting to do a blind A/B of Allocator and SoundEasy against a few targets.
And your reply would indicate you see purchasing SoundEasy v17 and measuring the phase response of UE's outputs prerequisite for general discussion about filter properties.
Not a big deal for those with project cash to burn and spare time to set up the measurements. Thanks for sharing the implementation details; while it's possible to correctly guess how Bohdan coded the feature the text itself is ambiguous (for the record, Jan had to document essentially the same thing in Thuneau's forum as the Allocator manual's even less specific). It's been my experience the even order LR alignments in Allocator sound slightly better to my ear than the odd order Butterworths. Disentangling the filter implementation and radiation pattern effects from what might be varying with the drivers as the crossover changes is tricky and doing the measurements to determine what's going on hasn't made its way to the top of my priority stack yet. It would certainly be interesting to do a blind A/B of Allocator and SoundEasy against a few targets.A) No, not at all. You inquired about the UE's ability to correct phase. So I responded by way of an example.
B) I can only say that you are entitled to your option based on your experience but that I would disagree with your generalization of "most folks". I have no doubt that some users find the general and extensive features in SoundEasy difficult to master. However, what ever the usability of SE is in general doesn't necessarily carry over to the UE. The UE is pretty much a stand alone section of the code. So let's compare apples to apples; setting up the Allocator vs the SE UE. I'll use a single pass band for example.
With the Allocator you must go to the Allocator screen and set up the crossover for the system in the form of specifying the crossover frequencies and order. This sets the acoustic targets. Next, the driver responses must be loaded. At that point the user can view the SPL response compared to the Allocator target. Then comes matching the SPL response to the target. For the high pass cut off this requires manual manipulation of up to 3 stages of cascaded 2nd order and 1 stage of 1st order, and adjusting the cut off frequency of each and the Q of each 2nd order stages. It is purely trial and error. For the low pass cut of the same manual manipulation is required. Not that this is difficult, but that is what is required. After that there is the possibility of manually adjusting between 1 to 4 stages of parametric equalization by setting the center frequency, gain/cut and Q of each stage. Lastly, there may be the need to adjust high or low shelf equalization, again, setting frequency and gain. This is all done manually by the user. The level of agreement to the target is a function of this manipulation and the patients of the user.
Doing the same thing in the UE requires loading the driver responses; and setting the band pass acoustic target type, order, and hi/low cutoffs. A button is clicked to display the target and the unfiltered driver SPL data. Not really any different than the Allocator at this point. Then the frequency limits over which the response is to match the acoustic are specified (these typically would be an octave or more above/below the band pass cut offs). A button is clicked to automatically calculate and display the filter transfer function required to "equalize" the driver SPL to the targeted acoustic response. Clicking a third button displays the final acoustic response compared to the target. There is no manual adjustment required, no guessing what filter Q might work best, the required filter transfer function is computed automatically. This can be accomplished in the UE because, as you put it, the dsp processing used is "not particularly sophisiticated" (I would beg to differ).
Thanks for clarifying the scope is one specific task.
Granted, there's certainly a set of users who'd be happy to click a button and think their speaker+crossover is now a perfect allpass. That set probably also overlaps significantly with folks who aren't well aware of acoustic measurement challenges and hence are at risk of plugging in bad data. So, as is so often the case with SoundEasy, I think the vision behind the feature's great but from what I know at the moment it's hard for me to get behind its implementation.
I agree the dial in process could be less tedious than Allocator makes it. DSPreLab and SigmaDSP are a couple examples of a lighter weight EQ workflow.
The desired transfer function is computed only if the desired measurement is loaded. For every set of polars I've seen none of the particular measurements is really what you want EQ for. So, unless UE has the ability to adjust the synthesized filter (I didn't see anything in the manual) that tedious but straightforward dialing in of EQ in Allocator's less hassle than SoundEasy. Because, with SoundEasy, it would seem you have to go tweak the data to import to get it to do what you want. Or is chapter 19 missing some features?
Granted, there's certainly a set of users who'd be happy to click a button and think their speaker+crossover is now a perfect allpass. That set probably also overlaps significantly with folks who aren't well aware of acoustic measurement challenges and hence are at risk of plugging in bad data. So, as is so often the case with SoundEasy, I think the vision behind the feature's great but from what I know at the moment it's hard for me to get behind its implementation.
I agree the dial in process could be less tedious than Allocator makes it. DSPreLab and SigmaDSP are a couple examples of a lighter weight EQ workflow.
So what's v17's output IMD level after the FFT blocks are recombined and how does it do the recombination? At least for v16 the way the manual read was the blocks were unwindowed and unstitched with only zero padding to mitigate discontinuities. Since it took me about 15 minutes to get Allocator up and running the first time I used it I gave up on spending the hours upon hours needed to fight through SoundEasy v16 to set up a crossover to measure. Hence I never measured v16's distortion. But from the docs I'd guess maybe 1%.
Not everything is addressed in chapter 19. There are also some equalization functions that can be used to shape the system response; shelving functions and PEQ. These are applied to the system overall, not the individual pass bands. Thus if you have a mid/tweeter x-o of 3k Hz and want to add a BBC dip you can just use the PEQ to place a dip in the system response. This would correctly modify both the mid and tweeter filters. With the Allocator you would have to apply the eq separately to each.
While what you say here is correct, iit really isn't a SE issue at all. It applies to any and all speaker design software. You can not design a good speaker, passive, active, digital or analog if the SPL data you are working with is bad. GI/GO
I had not measured this before so I made some measurements the other day. The results I got were THD and IMD were at the residual of the sound card which was less the -85dB below the input level . I.e. not an issue.
Latency: With phase correction latency has to be a little higher than desired because the impulse must be shifted tot he right on the time axis to make it causal. I believe that currently the latency is some where between 150 and 200 msec. Bohdan was playing around with that and how many partitions were used. It's not optimized (minimized) at this point.
Cool; could be a viable feature. Just need a demo version to try it out.
I'm unsure either approach is correct or wrong. They've different tradeoffs and both are arguably optimal within the context of the surrounding design they're used in.
For couple of years now I have been running v15 and v16 on a regular basis on a dedicated, quiet PC. It's pretty easy these days to acquire an older PC with WinXP and make it quiet if it's not. Other than issues of software upgrade popups that overlay the SE graphics causing a stutter when they appear, both versions were essentially 100% stable for literally weeks at a time.
There is an issue with the source, though. I do not play wav files, I have a separate dedicated music PC that I would like to run into the digital input on the SE machine. As it is, I use the output of an A/D to the analog inputs on the SE machine with SE at 100% output to an MSB 8-channel preamp. Remove control from my listening chair.
I'd love to avoid the extra D/A - A/D stages, though.
Dave
Dave, you will have to instruct me on how the make a PC quiet. Ibought a deicated XP machine for just that use and, unfortunately, it is a little noiser than my work horse, though a lot faster.
Are you sure the SPDIF input doesn't work? I never tried it but I see it listed in the preferences screen for the Delat 410.
I don't have any idea how old is your computer but these instructions could help to suppress noisy elements in a computer
- Processor cooler : replace it by a more efficient and quiet. A zalman for example.
- Box fan : Add a fan speed controller (5..12V ) or replace by a quieter or suppress it
- Power supply fan : buy a new power supply with a large fan (120mm), or without fan or change the fan by one with rolling ball.
- Hard disk : buy a new one, the newer disks are very quiet. The quieter are the ssd disks
.- The fan of the graphics card or motherboard : suppress them but, but you must add a box fan. It also exists quieter small fan. Some motherboards integrate acpi controller and this could be reduce noise but with XP if not active during installation, useless to activate on the bios.
The problem with fan, smaller they are, faster they run and noisy they are. You should keep speed of the fan below 2000rpm. In general your computer becomes a little more hot after the transformation but temperature variations and noise must be not very large (10°C for the temperature)
Happy to oblige! I made my pc silent a while ago, and love it very much.
John, check out this website, they do good work making pc's quieter.
I've had a couple of issue though - but I'll start by saying that having a quiet pc is really awesome. Before it was such an annoyance, and now it is soooo quiet. I just replaced all the fans, and the hard drive with products that got high scores from the above website.
So my issues are related to two of the three fans I replaced. One fan was on the processor - I got a fancy new heatsink and fan (which was too big for my enclosure, btw), but the fan is never recognized by my motherboard. It runs fine, but I always get an error at startup, and I can't fix it in the BIOS. The other issue was replacing the fan in the PSU - I always have to jumpstart it. Whenever I start the machine up, the fan needs to be 'flicked' to get going. I guess it isn't getting enough voltage? Either way, they are small issues, and I wish they weren't there, but having a silent pc is really awesome.
John, check out this website, they do good work making pc's quieter.
I've had a couple of issue though - but I'll start by saying that having a quiet pc is really awesome. Before it was such an annoyance, and now it is soooo quiet. I just replaced all the fans, and the hard drive with products that got high scores from the above website.
So my issues are related to two of the three fans I replaced. One fan was on the processor - I got a fancy new heatsink and fan (which was too big for my enclosure, btw), but the fan is never recognized by my motherboard. It runs fine, but I always get an error at startup, and I can't fix it in the BIOS. The other issue was replacing the fan in the PSU - I always have to jumpstart it. Whenever I start the machine up, the fan needs to be 'flicked' to get going. I guess it isn't getting enough voltage? Either way, they are small issues, and I wish they weren't there, but having a silent pc is really awesome.
I think you can have lower RPM with PWM modulation if it's supported by the mainboard (and fan IIRC).
I have pretty big and slow fans (CPU undervolted with a 12cm fan below 1000RPM, PSU is 12cm, so as slow as it gets) but it's still too loud when I turn the volume lower.
The best you can do is to get slower parts and undervolt them in order to use fanless designs. There are fanless CPU heatsinks, PSUs and graphics cards.
As for integration and stability:
I hope it helps that Linux and Windows7 both have completely new software audio where you can insert third party filters etc regardless of soundcard.
For example I played around with Pulseaudio and got a pretty stable (and unoptimized) 4-way crossover with LADSPA filters.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.