PLParEQ1 and PLParEQ4

[fritzman]

I simply hate statements like "plugin XYZ uses 3 times more CPU than plugin YZX and that indicates, that it is much better",

But you're not saying that this was my statement, right? Because it wasn't.

I'm don't want to sound jealously and i maybe a little bit biased due to my own work, but if everyone tells them that there EQ is great, they maybe blind to see everything which is still worse here.

Christian, there are several people who are telling the dev of PLParEQ that it has tons of bugs. Besides of that it sounds great, what's wrong with that?

I was in a similar situation.
Another thing which arouses me is, that all these papers are not very intuitive to read. What do they say at all? At least it should be something like our EQ is good, because of all these technical items build in. And it's only fair to measure them. And in some points they simply were wrong. At least i felt the need to show them.

Regards,

Christian

Yes, the papers should be correct. My bias is simply towards good sound ATM and not towards papers, poor GUIs, measurements and such. And towards a bug free version of this EQ.


Best wishes, FRitz

[duncanparsons]

I think there is both a technical and aesthetic middle ground here.

Technically, if something says 'I can cut your signal at xHz' you do expect it do just that. If the controls don't allow you to attain xHz, then there is little value in positing such a claim; if xHz can apparently be attained, but the output doesn't match the claim, then again we loose value.

However, a couple of Hz in a mix is difficult to discern, so maybe a suggesting the frequency rather than using the exact freq is probably acceptable. Outside of a mix, this can be more worrying, I know a number of people with perfect pitch who can tell A4 detuned by a couple of Hz.

In my plugs, my filters always show '~xHz', since I prefer to merely suggest where we are headed, and leaves me a little room for error (and you should use your ears anyway!)

Christian has done valuable work in examining the output of the plug. I know he has tested many other peoples plugs, as well as his own work, and he has exacting standards. Him testing this plug isn't at all out of the ordinary; and him testing public claims is something that everyone should do if they have the tools available.
However, there is a lot to be said to taking something at face value, and just using it - after all no one is asking you to like it and if you do all the better.

Aesthetically, it does all come down to taste. If Christian doesn't like it, that's up to him; if someone else does, fab. The most salient part of my psychology degree was given in one of my earliest lectures - the concpet of 'individual differences' - basically since we are all different (cue 'Life of Brian'..), we will all respond to different stimuli in a different way, tho' patterns of behaviour may emerge. We won't all like the same EQ, and we will all have different reasons for that dis/like.

In the end, so long as the plug isn't damagingly buggy, and does what it claims to the best of it's ability, then it will probably keep the users who like it.

For myself, at the end of the day, I have all the EQ's I currently need to restore lost dynamics, so another one doesn't particularly catch my eye..



DSP

[fritzman]

Hi Duncan,

just out of interest: What was your point now to "dislike" or not check ot whatever it is that you have against this plugin? Not defending it in any way, as you said and I totally agree, in the end it's all subjective. I'm just curious.


Best wishes, FRitz

[dbmcclain]

Christian is correct that neither we, nor anyone else, can bend the laws of physics. We approached this problem from a purely mathematical viewpoint. The off-frequency measurements you are seeing are likely due to innaccuracies in the bilinear transform mapping from the s-plane to the Z-domain.

The size of the buffers has primarily to do with the level of accuracy achieved in frequency resolution and also helps to decorrelate the ends of each block to help bury time aliasing. The use of 8-fold overlap is the biggest reason for the high quality, and secondarily, the use of the composite Windowing function.

That windowing function has the property that, applied twice in succession, enables perfect reconstitution at 50% overlap. Hence, if you have 50% overlap perfection, you have arbitrary overlap as perfect also -- viz.. take each pair of windows at 50% and they are perfect, all other offsets can be treated similarly, so the resulting sum of all is likewise perfect reconstruction.

When upsampling to high internal SR, you need long blocks to gain frequency resolution at the very very low frequencies. This is the same problem faced by any FFT-based filtering package. Your frequency resolution is determined solely by the duration of your windowed buffers.

The fact that you don't measure 20 dB gain at 20 Hz is likely due also to an insufficient buffer size. Viz... 20 Hz, at Q=20, means your bandwidth is only 1 Hz. Hence, you need a buffer that represents 1 sec of sound. Our largest buffers are only 100-200ms of sound, so you can see that we have a gross mismatch at these extreme settings.

Once could use a larger buffer, at the expense of memory, and higher throughput latency.

But look, after all is said and done, I told you exactly how I did all of this on the web page. You are free to replicate and improve on these results.

Cheers,

- DM

[dbmcclain]

...on the implementation side of the plugin, the knobs interface has inherent inaccuracies -- the frequency and Q knobs are really manipulating logarithmic quantities and the resolution shown on screen is a rounded value.

Accuracies in reproduction are painstakingly produced by using 64-bit floats throughout, with 80-bit intermediate values, and then taking great care to dither back to 32-bit floats (24-bit mantissas).

A better GUI would allow you to specify exactly what frequency you want. But as an instrument for music, and with relatively broad Q peaks, it is the ear that primarily dictates how it is used, not the math.

- DM

[duncanparsons]

Hi Duncan,

just out of interest: What was your point now to "dislike" or not check ot whatever it is that you have against this plugin? Not defending it in any way, as you said and I totally agree, in the end it's all subjective. I'm just curious.


Best wishes, FRitz

hours in the day, basically. If I'm going to chack it out, I want to be able to do it justice, rather than a quick 'oh, that's nice' test. Being a developer myself, I respect the work that others put in, and if I can't give it the time it deserves, I would rather not waste the dev's time; and I would prefer people to do the same for me (as they have done).

It looks nice enough, but what with a full time job, 3 children and a long commute to and from work each day, my musical hobby has to contract sometimes..

ATB
DSP

[dbmcclain]

I am interested in learning about why Nuendo and Wavelab have so much trouble with the plugin.

Changing sample rates is treated in exactly the same way as changing Quality level. If you can change quality without problems then using changing sample rates should be no problem either.

The problems we uncovered yesterday had to do with the way that the plugin DLL is used with multiple instantiations. Windows is now smart enough to make only one copy of the DLL resident, and so any global variables will get stepped on by the multiple instances. We backed up and made sure that the only shared globals are now read-only tables, e.g., the dither table -- which BTW, is now a Gaussian dither with 1/3rd bit STDEV, contrary to the web page which states it is Triangular.

The crashes we were all experiencing had to do with former global buffers being resized by other instances.

What remains are difficulties with the host environment in the face of rather large memory block allocation requests. We found that performing any allocations on the rendering threads could spell hazard. So we make sure that no memory allocation occurs on the rendering thread, and all allocations occur on the user (GUI level) thread.

I still have problems switching Quality levels on Sonar 5 Pro, but not on T2 or Cubase SL (1.06). I can only surmise that Sonar, and perhaps Wavelab and Nuendo, are trying to capture memory requests to help control their global environment, and they don't like being hit by large changes, especially during rendering.

But if anyone can run a tool like Plugin Consultant and log the results to a text file, I would like to see what is happening in the troubled environments. Until that time, since I have neither Wavelab nor Nuendo, we are shooting in the dark.

Heh... in fact one of the Nuendo users who complained of Side processing dropouts and crashes yesterday, admitted last night that he was using an old P3 laptop. That may be the source of the problem...

Cheers,

- DM

[bmanic]

That was me. But i don't use nuendo. I use fl studio. I've tried the plugin today on my main system. Same problems. Side not working at all and crashes host at Q5.

- bManic

[fritzman]

I am interested in learning about why Nuendo and Wavelab have so much trouble with the plugin.

...

But if anyone can run a tool like Plugin Consultant and log the results to a text file, I would like to see what is happening in the troubled environments. Until that time, since I have neither Wavelab nor Nuendo, we are shooting in the dark.

Cheers,

- DM

Hi David,

what exactly do I have to do? I've got plugin consultant and I've got the latest version of PLParEQ4. I ran it but where is the log? Sorry, maybe that's faster than reading all the manual...


Best wishes, FRitz

[fritzman]

Ah, think I've got it. Sending email ...

[Sascha Franck]

For myself, at the end of the day, I have all the EQ's I currently need to restore lost dynamics [...]

I wonder if anybody else noticed that...

[bmanic]

For myself, at the end of the day, I have all the EQ's I currently need to restore lost dynamics [...]

I wonder if anybody else noticed that...

Jep.

[Sascha Franck]

Jp.

Err... yes.

[fritzman]

For myself, at the end of the day, I have all the EQ's I currently need to restore lost dynamics [...]

I wonder if anybody else noticed that...

Eh, no. But funny ...

[JonnySun 2.0]

@dbmcclain: I cannot understand why you didn't speak about public beta testing. That's what you are doing here.
What would be the reason to buy a good sounding eq-plug for much money if the "appetizer" are full of bugs?

So, if you want to sell a copy of System 1200 or the 10-band-version your task is to test it on ALL SYSTEMS and HOSTS.
I am a Cubase SX 3-User, it's almost useless to test it in Cubase SL 1.06 That's not an reference, actually.
I am very disappointed, downloaded version 1.16 and it still doesn't work.