The Big Problem with “House Curves”

What are House Curves?

The idea behind house curves are to equalise a sound system in a way that allows the user to either “perceive” a flat response or to tune the response to the preference of the individual.

People create house curves for both Audyssey and Dirac Live systems by editing the target curve. I haven’t seen many people mess with house curves on Yamaha systems but it is also possible to do so by using a manual EQ slot and editing the PEQs.

There are many different house curves people play with: B&K, Harmon, Toole.

They all have one thing in common: they introduce a tilt in the frequency response so that bass notes play louder in volume than high-frequency tones introducing a 2, 4, 6, 8 etc dB increase in base output as compared to mid and high-frequency output.

What is a House Curve Trying to Do?

Well, our ears’ sensitivity is more sensitive to mid-range frequencies than to low frequencies. So for us to be able to hear low frequencies as loud as the midrange, they need to be played at a relatively higher volume.

What are the Issues Then?

There are quite a few. I will list them first then we can have a look at them one by one.

House curves ignore Equal-Loudness principles set out in ISO226:2003 including the fact that our ears are less sensitive to both the low AND high-end as compared to the mid-range
House curves ignore the fact that to be able to set up the correct curve, you need to understand the playback volume as our perception is volume-dependent!
House curves ignore the fact that to be able to set up the correct curve, you need to understand the original mastering volume which is content-dependent!

Issue 1: Equal-Loudness Contours (ISO226:2003)

I discussed equal-loudness contours before and in detail here. However, in summary, human hearing is not as sensitive in the higher and lower frequency ranges as it is in the mid-range. The very latest research we have available for how loudness is perceived is the international standard called ISO226:2003.

Unfortunately, unless a house curve follows the equal-loudness contours, all you are doing is loading a curve to your PREFERENCE and you won’t hear the content as it was mastered or meant to be played. While there is nothing wrong with this, as far as movie playback is concerned, you won’t hear what the director intended you to hear.

Issue 2: The Playback Volume

As the actual volume is increased or decreased, the sensitivity of our hearing changes. What this means is that loading a static curve will only work well for one particular volume level! So loading someone else’s curve without asking the question of what volume they are using the curve at is basically taking a while guess at what will work best for your setup. While it is true that most people listen to content somewhere around -10 to -25 below reference, it is still quite a range to be shooting at!

Issue 3: The Mastering Volume

Lastly, we get to the mastering volume the content was created at!

When not listening to a recorded program – let it be music, movies or TV shows – at the volume the program was recorded at or intended to be played at, the perceived loudness of the treble and base will drop off quicker than the midrange as the volume is decreased.

The one complication is that different content is mastered at different volume levels. While we have a standard for movies (85dB average with 105dB peaks – or 115dB peaks for the LFE channel), we don’t have such standards for other program material. However, we do have some best practices when applying equal-loudness contours to other content as well. For example:

Movies: 85dB as reference (this is what your AVR is calibrated to) – there is some complication with movies as the actual acoustical space also has an impact here as well as how THX and non-THX receivers handle this. (More is discussed about this in Dirac Live Perfection – coming soon to Secrets of Audyssey)
TV shows: 75dB (-10dB below reference)
Music: varying and depends on genre (between -20 and -10dB below reference dependent on dynamic range)

The Annoying Result

What this means is that your loaded curve will work great for some content at your chosen listening volume but not for other content – or the same content as you increase or decrease the volume.

What’s more, since you haven’t taken into consideration your ears’ sensitivity below and over the midrange, the program material will not have the correct tonal balance. Or some material may, while others won’t.

The Solution

The different AVRs will have different solutions. I will detail them below:

Audyssey

You need to configure Audyssey correctly and enable Dynamic EQ (and Dynamic Volume). Relevant content is linked below to get you started:

Yamaha

You will need to enable Yamaha YPAO Volume on your Yamaha receiver. Relevant content is linked below to get you started:

Dirac Live

Dirac Live does not have a Loudness Compensation solution.

This is why I spent weeks measuring Audyssey’s Dynamic EQ, Yamaha’s YPAO Volume as well as modelling ISO226:2003 exactly to enable you to configure Loudness Compensation for your Dirac Live AVR. The data is included in the SHC Dirac Live Curve Editor Excel Tool. The tool is included with the Dirac Live Perfection Guide. With this tool, you will be able to

Select the particular Loudness Compensation solution you want to emulate
Select the content you will be playing (mastering volume as the correct offset from reference)
Select your listening volume

The curve will be calculated for you to load into Dirac Live as a target curve. There are also pre-made curves that you can load up straight away.

Forget house curves! This is my recommended option.

Criticism of Equal Loudness Contours

I have found that most criticism of Equal Loudness Contours are due to a misunderstanding of the ISO226:2003 standard or due to forgetting one or more critical elements of the standard. For example, equal-loudness contours cannot be used by simply lifting them out of the standard. They need to be normalised correctly to the reference volume set as your baseline and taking both mastering and playback volume into consideration. The discussion about this is beyond this article but I discuss this in Dirac Live Perfection. The normalised data is used in the included Excel tool to create the Dirac Live target curves.

Additionally, it is true that equal-loudness contours have gone through an evolution as the standards – and our measurement methods – evolved. The current standard is the best estimation that we have – but I am sure there will be more. My opinion is that the standard works pretty well, however, you do need to take into consideration the mastering space – especially for movies that could be mastered in very well-treated and / or larger spaces than most of the home theatres. Additionally, Dolby does recommend that smaller studies are not calibrated to 85dB when mastering for the theatre but around 4dBs lower so that the master translates better to movie theatres. There is a suspicion that some of the near-field mixes were also mastered somewhat lower than 85dB for this reason.

Other criticisms seem to note that equal-loudness contours don’t relate to how “normal” sounds such as music or movies are perceived. Actually, this is a misunderstanding of multiple research findings. Perceived loudness of tones does give us an understanding for normal sounds, however, how loudness increases or decreases OVER TIME is influenced by other factors such as direction (movement away or towards), position (front versus back) and volume changes (increase or decrease of the sound volume) relative to the previous second (that is in the time domain). However, this is not what the ISO226:2003 standard is trying to tackle. The two – while closely related – impact on loudness compensation and dynamic range compression technologies differently. This might be discussed in more detail in another article.