Jump to content

Article: Digital Signal Processing - The Ultimate Guide To High End Immersive Audio

Recommended Posts

4 minutes ago, The Computer Audiophile said:

Network DSP for all, would be great. We need to get HiFi manufacturers to embrace Ravenna. It puts everything on the network and opens up a world of possibilities. If it's good enough to capture a live orchestra, where there are no second chances, it's good enough for HiFi. 


I use Ravenna. I am sold on its benefits. For me, the main benefit is that I do not need to look for a 16 channel DAC. There are very few of those around! Instead, I can buy two 8 channel Ravenna DAC's and a microphone interface. Ravenna ties all the equipment together and tells the PC "I am a device with 16 DAC channels, 4 microphone inputs, and 8 digital inputs" (or something like that). You can put together as many channels as you want. 


The problem with Ravenna (and also Dante and AVB) is that they are pro audio standards. Not easy for us amateur hobbyists to set up. I already find my RME intimidatingly difficult, let alone Ravenna which adds the complexity of network audio and multiple modes into the mix. I keep telling myself that I am a home audio enthusiast. I am not running a broadcast studio, or routing audio in a stadium, or an airport, or any situation where network audio would be a massive advantage. But now that I am starting to run out of DAC channels, I appreciate the flexibility of Ravenna. Someone who is into immersive audio with more speakers than me would see more of a benefit. 

BTW, aren't you the guy who blew up his speakers when trying out the Merging Anubis? ;) Not easy to set up, are they. 

Link to comment
7 hours ago, The Computer Audiophile said:

Target curves and final adjustments are all about preferences. 


Welllllllllllllllllll that is a bit of a can of worms. If you read what Toole says in his book, he says that good speakers should have two properties: (1) they measure flat under anechoic conditions, (2) they have constant directivity. If you place such a speaker in a room and listen farfield, you will obtain a Harman-like curve (there is a good video by Erin on Youtube that explains why). Toole has said himself on another forum that equalizing a speaker to reproduce the Harman curve at MLP is wrong, because an on-axis correction also affects the off-axis response, which will produce reflections which are spectrally incorrect. There is no "choose your target curve based on your preference". 


Toole's motivation is to narrow the "circle of confusion" - have the studios produce music mastered through standardized sound systems, played back in our homes using speakers designed to achieve certain standards of performance. The recordings have to be mastered on systems so that they are faithful to the original sound, and have to be played back on systems that reproduce the sound of the master faithfully and accurately. Only then can we have "accurate sound". 


However, in the real world, even studios can not get something as basic as the frequency response correct. Genelec did a study using their GLM tool, which is a calibration tool for their speakers. They observed a wide variety of frequency responses in studios. And this is only for studios with Genelec speakers, who bothered to pay extra for the GLM tool. In reality the variance is probably much worse than that Genelec study. 


I would argue that this gives me license to adjust the frequency tilt as I please. For each recording, if necessary. So, like you, I have gone for a preference target. I am not an authority figure, I am merely an amateur hobbyist in an ocean of amateur hobbyists. I have no business arguing with Toole. Or Mitch, for that matter. BUT ... sometimes authority figures disagree, leaving us minnows confused. So I read what they say, try to understand their points of view, and make up my own mind. After all, the sign of an educated person is the ability to entertain contradictory points of view and weigh them up fairly. 


BTW I recently came across a new method for generating a target curve that removes the "room transfer function", restoring your speaker to a flat anechoic response. You perform a nearfield MMM of your speakers, then perform a MLP MMM. The idea is that the MLP MMM captures the "speaker + room" response, and the nearfield is speaker only. If you subtract the nearfield MMM from the MLP MMM, you will obtain the "room transfer function". If you set this as your target curve, it will correct the nearfield response to flat. I have tried this, and it works for frequencies above transition (i.e. 4x Schroder, about 440Hz in my room). If you want to learn more, google "Magic Beans room correction joe n tell". You will find some videos. 

Link to comment

If you look further in that ASR thread you linked, you will see my experiments with Magic Beans. I do not agree with Toole's opinion, because I don't think he understood the objective of Magic Beans. Toole is famously against custom target curves that replicate the Harman curve. In that same thread he said it is not a Harman "target", it is a Harman "result", i.e. what the speaker naturally reproduces at the MLP if the speaker was designed to be flat under anechoic conditions. The idea of Magic Beans is to restore the anechoic performance of the speaker by removing the "room transfer function". It is not about artificially creating a Harman target. And even if it was, I have no objection to it, provided that speaker directivity is constant and it does not mess up the off-axis response. 


Toole also thinks that this is some kind of money making gimmick. This is not true, anybody who has DSP software which has the ability to manipulate curves is able to copy the workflow and implement Magic Beans correction for free. I posted an Acourate workflow in the same thread. That experiment did not cost me a cent. 


In any case, I have said upthread that I have no business arguing with Toole. I personally think he is incorrect on this point, but you should remember that my opinion carries no weight. I am an amateur hobbyist. He is Toole. Having said that, I am under no obligation to do everything he says with my own system. 


Unfortunately, my speakers are not a candidate for Magic Beans correction. This is because they have mixed driver types, with horns >500Hz and normal woofers/subs below. I have measurements that demonstrate that the output from the horns does not decrease in volume with distance, whereas the woofers/subs decrease as expected. This produces an upward tilting curve at MLP, which is the opposite of what nearly every other speaker type produces. The Magic Beans correction actually does correct the nearfield response to flat (indicating that I followed the procedure correctly), but it produces an upward tilting response at MLP. After I gave my feedback to joentell, he implemented a "directivity detect" feature in his app. 


There is a detailed analysis of my speakers on ASR under my system thread. It does not behave like a normal speaker, which is why the DSP solution is different to a normal speaker. 

Link to comment

I should point out that Toole thinks that we should correct below Schroder. Above Schroder, he says it is OK to apply broad tone controls, but he thinks that applying correction to a target is the wrong approach. He is essentially arguing against more granular correction - after all, a "broad tone control" is the same as a very loosely applied target curve. When you make your correction, it is trivial to select how granular you want your correction to be. You can make it really hug the target curve, or you could make it sort of "trend" towards the target curve (the Toole approved approach), or anything in between. The granularity of the correction is a continuous variable. Since it is so easy to do, you can make a range of filters and listen to which one you think sounds best.


I don't have a firm opinion on this one, I do both. I can flip between the filters with ease, a push on the button on my convolver and I have a new filter loaded. And if you use HLC, the music doesn't even stop. 


Of course, making the on-axis response look perfect might mess up the off-axis response. So I regularly check what is going on off-axis as well. Same procedure, 45 deg off axis, and your choice of SPS or MMM. 


And BTW, I wouldn't dismiss the Magic Beans idea. Do you understand how it works? Those videos don't do a very good job explaining it, you really have to watch a few videos and aggregate all the information together to understand what the intention is. I posted a summary of my findings in that ASR thread, and joentell thought that it was an accurate description of the method. 

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

  • Create New...