Ultimate Equalized V5 has been released - Page 2 - diyAudio
Go Back   Home > Forums > Commercial Sector > Vendor Forums > Vendor's Bazaar

Vendor's Bazaar Commercial Vendors large & small hawking their wares

Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Reply
 
Thread Tools Search this Thread
Old 10th January 2014, 09:31 PM   #11
diyAudio Member
 
Join Date: Feb 2002
Location: Germany
Default buffer size vs. taps?

Page 154 of the UE5 manual shows various buffer sizes and latencies but does not give the FIR filter size.

JohnK mentions here that the linear phase filter is 8192 taps but does not state the buffer size you need to set to get 8192 taps:

http://www.diyaudio.com/forums/multi...ml#post3298968

So what is the tap number per buffer size?


He also mentions that at the 5.8 Hz resolution you get with 8192 taps there may still be issues with the predicted response not matching the actual response due to issues with the windowing of the measured impulse. This may require repetitive tweaking of the target:

http://www.diyaudio.com/forums/multi...ml#post3300656

I realize that these comment might refer to v2 or v3. Are these problems resolved by now or should I expect issues trying to get linear phase from a BR or BP6 alignment tuned to 15 Hz or equalized to be flat to 15 Hz?

As an additional question: is UE5 just a streamlined subset of SE19, facilitating the work flow but omitting features, or can UE5 do things that are not possible in SE19?
  Reply With Quote
Old 10th January 2014, 10:37 PM   #12
diyAudio Member
 
Join Date: Nov 2012
Location: Melbourne
Hi Capslock,


I have tried to use the links to John's comments you have provided, but they did not work. "The webpage cannot be found" was the message I was getting.

I would prefer to read the original comments before I fully respond to your initial questions.

UE5 originated from SoundEasy, but has improved and evolved into a unique program. For instance, UE5 has "Player Mode" feature, which you would use after you are satisfied with your complete system design.
Then you switch to self-starting player mode and the program only displays small dialogue box to control volume of selected HT channels,
and shows SPL levels on all channels.

Best Regards,
Bohdan
  Reply With Quote
Old 11th January 2014, 06:23 AM   #13
diyAudio Member
 
Join Date: Feb 2002
Location: Germany
Hi Bohdan,

that was quick, thanks!

Let me try again:
rePhase, a loudspeaker phase linearization, EQ and FIR filtering tool

Quote from #51: "Boidzio UE uses partitioned, FFT convolution and is pretty fast with 8192 taps but does loose accuracy at low frequency. " - yet he does not state the UE buffer size that corresponds to 8192 taps. I think I did figure out that what you call bins is the number of samples that go into each partitioned convolution - but this is not necessarily the total length of the equivalent direct convolution FIR filter. So what I am asking is what are the FIR tap sizes associated with 1024 to 4096 bin buffers discussed on page 154 of your UE v5 manual?



The other link was:
rePhase, a loudspeaker phase linearization, EQ and FIR filtering tool

On second reading, I see I may have mixed in some info from JohnK's posts further down, such as #100, #105 and #120.

What his posts #82 and #86 really seem to say is that the 5.8 Hz resolution afforded by an 8192 tap FIR filter should be sufficient for most low order eq tasks in the 20 Hz -40 Hz region (such as rolloff correction for a dipole sub), but that UE may get the filter wrong at the very low frequency end due to windowing issues, i.e. UE will compute a filter and display its intended performance that will look fine, but the real performance of the filter may be something else, and the only recourse is to (physically?) remeasure the system performance and then iteratively tweak the target.

Did I get that right? Is this something that has been resolved since 2012? If not, is there a way to at least get rid of the physical remeasurement, such as reevaluating the calculated filter with a larger window?

Then, if this is resolved, posts #100, 105 and 120 seem to imply that there may be additional issues with the measurement process.


Finally, on SE vs. UE, I still don't get it after spending a few days browsing through either manuals and sifting through a couple of months' worth of the user group and various forum posts. I did buy and use SE a lot around v7 or 9, but never used the DE feature, and I stopped buying upgrades around v14 when I realized I was not even installing them due to being busy with job, small kids and building a house. I have been following the user group loosely, though.

The first thing I never really understood was the difference between DE and the UE feature introduced in v17. The ability to do linear phase DRC was there before UE, so once I have generated the approprate filters that will do XO, EQ and linear phase correction, what is the difference between playing them back through the DE or UE portions of SE?

Then, between SE and UE, the whole difference is that UE drops a few features, streamlines the filter generation process, has an autostart option and a player mode? Or are there other improvements such as a faster convolver?

Thanks in advance!

Last edited by capslock; 11th January 2014 at 06:28 AM.
  Reply With Quote
Old 11th January 2014, 06:43 AM   #14
diyAudio Member
 
Join Date: Feb 2002
Location: Germany
Default Sound card issues

The other thing I have been trying to understand is the current status on compatible soundcards.

1. First of all, what are the advantages of WASAPI vs. MME? Will the latencies be significantly different?

2. At this point, the Delta 1010LT is the only soundcard really known to work with the WASAPI version of UE - right?
Will the full 1010 work also? As far as I can tell, the only difference is that the AD/DA are in an outboard box.

3. The Delta 410 uses the same version of the Win7/64 driver and should be expected to work with the WASAPI version just as the 1010LT, but as of Dec 2013, there was no confirmed install of either WASAPI or MME full UE with the Delta 410, just MME lite.

4. Most of your examples involve setting the onboard ALC889 soundcard as the default playback soundcard and forcing it to resample to 96 kHz and then use an SPDIF link to one or two 1010LT. I gather this is the workaround for dealing with varying sample rates. However, how good is the resampling process that comes with Win7? Has this been tested with the ALC889? Is this done within Windows and will be as good with lesser sound chips such as the ALC592 or 892?

5. Is it reasonably safe to assume that anything that works with Win7/64 will also work with Win8.1/64? If I buy UE and dedicated player hardware, I might want to start off with Win8.1 because it has better power management.

6. The setup that might be really attractive for newcomers is to use HDMI audio and a 7.1 AVR, as it involves no non-standard hardware. There are various reports of folks doing software XO and using just the audio driver of Intel Core or the newer AMD CPUs and maybe ASIO4all but no onboard or dedicated soundcard to send the audio streams to an AVR. The only drawback seems to be that it does not work headlessly, i.e. there needs to be some sort of monitor plugged into the HDMI out of the AVR. Would this be something to look into for UE long term?


These are a lot of questions. As I have found the User Group to be frustrating when you try to recover information, I believe they are better answered here, though you might consider setting up an FAQ on your site.

Thanks again,

Eric

Last edited by capslock; 11th January 2014 at 06:55 AM.
  Reply With Quote
Old 13th January 2014, 05:49 AM   #15
diyAudio Member
 
Join Date: Nov 2012
Location: Melbourne
Hi Eric,

There are lot of questions there, perhaps we should take it step-by-step, and not necessarily in order you asked. Let’s address simple things first.

SE vs UE.

UE is a dedicated DSP program for loudspeaker filtering and equalization (and more). It has much better partitioned convolution engine, and better windowing of impulse response. There is more, but all this results in better sound quality and better DSP functionality in UE. The DE/DF features are quite simple, and will not be further developed in SoundEasy. This is because it’s now the job of the UE. On the other hand, the UE will be developed much further. UE6 will most likely have:

7.1 capabilities,
16 partition convolution engine ( I have a test code written already for it).
32k Impulse Response lengths for further improved low-frequency resolution (this is probably not necessary, but there will be this option),
Digital delays for second pair of subwoofers – the “sinks” for room mode equalization (apparently works like magic).
Data Block Buffer = 4096 as well
More threads to handle larger DSP demands.

If you have more question on this topic please keep them coming.


Low-frequency resolution.

John K. has done a great job explaining many of the details of the UE implementation.
LF resolution relates the length of Impulse Response to the sampling frequency as:

LF_resolution = Sampling_Frequency/IR_Length

Where: IR_Length = No_Of_Partitions * Data_Block_Length (or Buffer)

UE5 has 8 partitions, so if you select Data_Block_Length = 1024bins, and Sampling_Frequency = 48000Hz, then
LF_resolution = 48000/(8*1024) = 5.86Hz

If you select Data_Block_Length = 2048bins
LF_resolution = 48000/(8*2048) = 2.93Hz

If you then select No_Of_Partitions = 16
LF_resolution = 48000/(16*2048) = 1.46Hz

I hope this is getting clear now. You can do similar calculation for 96kHz sampling.



Accuracy of bass playback due to low-frequency resolution.

This issue will come up if you have sharp notches or sharp peaks in the very low frequency range. If you run with low-frequency resolution of 5.86Hz and you have a notch at exactly 3*5.86Hz=17.58Hz, this notch will have almost the correct depth. However, if you plan to have a notch at 14Hz, now the notch will be less accurate and will have reduced depth by about 3-8dB depending on the Q-factor.

Now, imagine the same requirement for 14Hz notch, but now your low-frequency resolution is 2.93Hz. This notch is very close to 5*2.93Hz=14.65Hz, and will be played much more accurately by the system with better low-frequency resolution. I have a software, that can actually model these things.

Fortunately, in real life there are not too many situations, where you would require very sharp peaks/valleys in the low-frequency range. Subwoofers have quite shallow irregularities in their LF response and rooms have quite smooth peaks there too (we do not equalize notches in the room response). Therefore, a system with 5.86Hz low-frequency resolution works very well.

You can see the EQ effect on my 18” subwoofer on page 9 in
http://www.bodziosoftware.com.au/LP_MP_Subwoofer_Tests.pdf
The response is actually measured one, and goes down to 16Hz and is very flat.

Perhaps we take a break now, and I’ll response with the sound card and PC issues comments in the next post.



Best Regards,
Bohdan
  Reply With Quote
Old 13th January 2014, 10:52 AM   #16
diyAudio Member
 
Join Date: Feb 2002
Location: Germany
Hello Bohdan,

thank you for taking the time to explain.

Is there a timeline for the release of UE6 yet? Delayed bass sinks (DBA = dual bass arrays) are the rage here in Germany in some audiophile circles.

I agree that using more partitions might be a good idea, especially as data_block_length seems to contribute a lot to latency.

According to page 154 of the UE5 manual, a buffer size of 4096 is supported, so this means a 32 k IR is already possible, albeit at 267 ms excess latency, which would potentially be nearly halved with 16 partions.


Regarding the visualization and optimization of the LF response, it seems rePhase already has an auto optimizer, see posts #94 and 96 of the thread I was referring to. Is this something that could be implemented in UE?

Regarding additional loss of LF accuracy from windowing the measurement pulse, is this an issue at all?

Looking forward to your comments on sound cards. I've am still uneasy about the default resampling to 96K. Even if Windows does this job properly, it still means everything (streaming from storage as well as playback in the 1010LT) is slaved to an oscillator on or near the onboard sound chip that is probably of very questionable jitter performance.

Thanks again

Eric
  Reply With Quote
Old 14th January 2014, 12:36 AM   #17
diyAudio Member
 
Join Date: Nov 2012
Location: Melbourne
Hi Eric,


Regarding the windowing issue in IR.

First of all, Johns comments refer to 8k long IRs, and I am not sure, if he used “wide window” too. The claim, that windowing will smooth the low-frequency response is correct. This is what would happen for say, Hanning Window. However, UE5 implements the concept of “Wide window”, explained with 4 partitions on Figure 97 in UE5 Manual. Please have a look. 75% of the window is flat, and further 10% is almost flat, therefore the “smoothing” argument is no longer important. The improvement in low-frequency accuracy is actually quite visible.

As I explained already, 8k long IR with “wide window” works very well for normal bass EQ and room EQ. If you are looking for further improved accuracy, you can use longer IRs, say 16k or 32k and “wide window”.

I am not sure how good is “optimization” in rephrase. Given the information above “optimization” is not necessary. The equalization accuracy is good, as I shown in the link in the previous post.


UE6 release.

If everything goes well, I am targeting June 2014.


Sound cards.

Currently the supported sound cards are:
Delta410 – MME mode, I do not know about WASAPI Exclusive Mode (perhaps).
Delta1010LT – MME and WASAPI Exclusive Mode.
PC Motherboards - 2in/8out in MME mode.
LynxAES16 – MME Mode and WASAPI Exclusive Mode.

I am just about to start testing Marian GmbH Trace8 sound cards. Interestingly, you mentioned you live in Germany too. Marian has a very interesting range of sound cards, and I have worked with them to fix some of the MME issues on their Trace8 sound card before. There is new driver, that promises further improvement for multi-channel operation – we’ll see. It would be fantastic, if Marian provided WASAPI Exclusive Mode drivers. Other companies have done this.

Currently, my preferred system is based on LynxAES16, described in
http://www.bodziosoftware.com.au/Hyb...gue_System.pdf
This is WASAPI in Exclusive Mode. I am actually converting my HT system with analogue amplifiers to the hybrid system with AES/EBU amplifiers from miniDSP.

An excellent, full digital system is described in
http://www.bodziosoftware.com.au/AES...kHz_System.pdf
This is also WASAPI in Exclusive Mode.


Re-sampling.

In order to keep the whole system in time alignment, there must be a single clock source in the system. So, if I play .wav file at 44.1kHz, I could simply use 44.1kHz as the clock frequency – without re-sampling, if this is your preference. Then the whole playback chain works with the same sampling frequency.

Having googled the internet, I got the impression, that jitter claims are exaggerated, and contemporary hardware uses PLL clock locking/filtering techniques, which reduce jitter by 50-100 times. I am not an expert here, so I could be wrong.


Just as a side comment, my motivation behind UE technology is the need to use a PC as audio server at the first place. I see this as an obvious need for the now and the future, so adding UE onto the same PC was just an obvious course of action.


Best Regards,
Bohdan
  Reply With Quote
Old 15th January 2014, 08:51 AM   #18
diyAudio Member
 
Join Date: Feb 2002
Location: Germany
Bohdan,

thanks for your patient explanations of the measurement issue.

A few items remain open for the soundcard thing:

- is there an inherent advantage of using WASAPI over full MME?
- I found confirmation on your home page that the Delta 410 will work with MME lite but will it really work with the full MME?
- should everything that works with Win7 also work with Win8.1?
- is it safe to assume that the 1010 will work in every way the 1010LT works?


I have two Delta410 back from Soundeasy times, one of them with external DACs because I was not happy with the performance of the on board codex. The 1010LT has similar issues and runs at about €120 used here, whereas the 1010 is not much more expensive and does have decent AD and DA chips. If I really have to buy a new card, I'd rather go for the full 1010.

I was not able to figure out what chips are used on the Trace 8, but published distortion performance is only borderline acceptable, especially if you consider its price of €300. There are consumer cards out there that can be bought for less than €50 that have much better AD/DA performance, but they can only be accessed through VST or ASIO.

I believe SE compares very favorably to other integrated packages, but not offering VST, ASIO or direct export of the filter, it will drive up the net cost to the user through its hardware requirements. Therefore I believe it will benefit from maintaining a list of supported cards and modes, with the aim of qualifying at least one low cost hardware option. If you cannot do the testing yourself, why not try to recruit beta testers?

Best regards

Eric
  Reply With Quote
Old 16th January 2014, 07:52 AM   #19
diyAudio Member
 
Join Date: Nov 2012
Location: Melbourne
Hi Eric,


1. MME has been a native Windows audio system for 10-15 years. Starting with Vista, Microsoft decided to introduce new audio system, that would allow the programmer to get very close to the sound card data buffer. This is what has been done in WASAPI Exclusive Mode, the most noticeable thing about WASAPI is reduced latency by about 30ms.

MME is very simple from programming point of view, while WASAPI is quite complex. But once you master it, then it becomes you preferred way of writing audio software for PC. You can do a lot more with it.

The difference in latency is becoming less important when you start using data block processing – which is the way to go when using “frequency domain convolution” – the FFT approach, necessary for long IRs. Blocks of data have to be read-in, processed and then send to output buffer. So, if your data block is 2048bins long and sampled at 48kHz, it will take 42.66ms just to read-in one data block. Then, you process the new block, so you can not output within the next 42.66ms. Then add some IRQ delays, then sound card buffer delays and so on. Before you know it, you have 100ms latency. Latency gets much worse for linear-phase.

Anyway, I hope you can see, that in the perfect world, we would use WASAPI, but MME is still excellent.

2. I always send out 3 versions of UE5: WASAP, MME, and MME_Lite with each order. So, you can experiment with all three versions and see which one would work with your sound card best.

3. I do not have Win8 yet, but I would assume, that audio engine in Win8 is the same as Win7. I have not seen or read anything to the contrary.

4. Several of my customers are reporting, that UE5 work fine with full Delta1010.


Finally, I fully understand, that at the end of the day, things boil down to the cost. Now, the lowest cost implementation would not require any additional soundcards. You would simply use motherboard ALCXXX codec and MME_Lite.

In the next level, you could use single Delta1010LT/Delta410 or full Delta1010.

In the top level, you would go digital with LynxAES16 and AES/EBU amplifiers. I realize, that this expensive, but the world seems to be moving to digital. My new TV has only SPDIF for audio output. My CD-player has SPDIF 2.0 output. My PayTV decoder has SPDIF output. My PC has SPDIF output, that I use for audio server playback - and so on. This is where the LynxAES16 fits perfectly.


I am considering Trace8 sound cards, because they have “balanced” inputs and outputs. This is good for getting rid of potential ground-loop and PC noises. This is typically giving you more headache than the lowest distortion figure possible.

Going fully digital is the best option, as ground-loop and PC noises do not even enter the equation.

I had reports, that ECHO sound cards work OK as well. You would have to post your question on SE User’s Group for more info.

Anyway, if the sound card driver claims to have MME or WASAPI Exclusive Mode compatibility, such sound card should work with UE. If you know of any, let me know. Perhaps there is a willing person to test it for me.

Best Regards,
Bohdan
  Reply With Quote
Old 16th January 2014, 09:07 AM   #20
diyAudio Member
 
Join Date: Feb 2002
Location: Germany
Thank you Bohdan,

so my take-away is that from a user's point of view, WASAPI and MME should not make much of a difference, and that I can progress building my new audio PC with Win8.1.

Quote:
Originally Posted by bohdan1232000 View Post
Several of my customers are reporting, that UE5 work fine with full Delta1010.[/SIZE][/FONT]

[...]

In the next level, you could use single Delta1010LT/Delta410 or full Delta1010.
I missed that part about the full 1010 being reported to work like the 1010, and not for lack of searching, so I still believe you should maintain a list of known working configurations, similar to this maybe:
USBAudioRecorderPRO

Also, your post seems to imply that the Delta410 has been shown to work with the full MME version of UE?


Best regards

Eric
  Reply With Quote

Reply


Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
Dx Turbo, the ultimate amplifier to be released before july 2013 destroyer X Solid State 10 20th June 2012 10:09 PM
'BagEnd' equalized sub... Fuzzy_Logic Multi-Way 4 19th May 2002 12:53 AM
Equalized passive Preamp. promitheus Pass Labs 2 28th June 2001 07:24 PM


New To Site? Need Help?

All times are GMT. The time now is 10:33 PM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright ©1999-2014 diyAudio

Content Relevant URLs by vBSEO 3.3.2