Spectrum/RTA feature

Hi Nick,

Glad you like the new features.

Broadly speaking, the difference between a spectrum plot and an RTA plot is in how the frequency content is treated. A spectrum plot shows the level of each frequency in its analysis range, dividing the range into "bins" of equal width in Hz. For example, a 65536 length spectrum at a 48kHz sampling rate has bins that are approx 0.7Hz wide. The spectrum treats the range between 20Hz and 20.7Hz the same way it treats the range between 20,000Hz and 20,000.7Hz. That can be useful when looking at distortion components, for example, but is very different from the way we hear. Our hearing, as a very rough approximation to a very complex subject, splits the frequency range into bins whose width is bigger at high frequencies than it is at low frequencies, in fact the width is a proportion of the frequency e.g. a bin of 1Hz at 100Hz would be treated with the same weight as a bin of 10Hz at 1000Hz or 100Hz at 10,000Hz. In each of those cases the proportion is 1%. An RTA works similarly, it has bins which are a constant fraction of an octave wide. An octave is a doubling of frequency, so how many Hz are in that fraction of an octave depends on the frequency - the octave from 100Hz to 200Hz spans a lot less range than the octave from 10,000Hz to 20,000Hz. The width of the horizontal bars in the RTA plot correspond to the width of the RTA bins. If you measure a pink noise signal with an RTA what you see corresponds with how your ear perceives things, a flat line on the RTA would mean you would perceive the levels as the same across the range.

A much shorter version of that is: you will almost always want to use the RTA views and almost always with some form of pink noise signal being played through your system

The levels of the RTA depend on the octave fraction used. The smaller the octave fraction, the narrower the bins that are being analysed. The RTA shows the energy in each bin, so as the bins get narrower the energy gets lower. With RTA plots we are more interested in the shape than the level, if everything is flat then we have even reproduction across the band.

As a final tip, you can make the RTA plots look nicer by "joining the dots" of the bins rather than having bars that are the width of each bin. You control that using the option in the View settings.

What is the difference between RTA and Spectrum Analysis? In a sense, the graphs are similar yet different. I guess more to the point is, what uses would RTA have over spectrum analysis and vise versa?

Click to expand...

Spectrum analyzers are normally associated with an examination of your spectrum and it's distortion products.

With REW, there are really three ways to use it. You can play an internally generated Periodic White noise and examine the spectrum (the RTA requires Periodic Pink noise), or you can play an internally generated Sine Wave and examine the THD and THD+N, or you can simply examine the spectrum with no signal at all.

The latter is interesting because it lets you examine the room for noise energy that you may not be aware of.

Below is a pic of my office with the Spectrum analyzer running with no signal being generated.

I am using an FFT length of 65536, with an averaging of 8, and using a Hann window (standard setup).

Everything is set up as if I am going to do a measure and all the levels are set to go, but instead I simply start the Spectrum analyzer and listen to the room.

You can see my furnace is introducing some low frequency noise. If I shut it off, the noise is eliminated.

You can see I have the typical poor PC computer 60Hz hum and its related harmonics at 120Hz, 180Hz, 220Hz etc.

You can also see (amazingly), my NTSC RPTV CRT's horizontal oscillator frequency of 15.750Khz present in the room.

The TV is several rooms away and happened to be turned on, but the oscillator noise is present in my office (not that I can hear it, but there it is). Shut the TV off and it goes away.

The Spectrum analyzer can tell you a lot about your room and system before you even do a measure.

Sometimes you'll see a waterfall with a strange signal that rings out for many hundred of milliseconds. If you looked at the spectrum, you might see the source, rather than incorrectly deciding it comes from a modal resonance.... My furnace is a case in point on that issue.

You can certainly play a Sine Wave and look at the distortion products and distortion + noise results.
Here's a pic of a 1000Hz fundamental and its associated harmonics. REW gives % THD and % THD+N readouts of the fundamental plus all harmonics..

Personally, I feel it's wise to do a 1KHz simple loopback Spectrum with THD and THD+N reading of your soundcard and PC alone to establish a baseline for yourself. Then add in the system and see how much greater it is.

The RTA as you've seen is quite useful. Here's a response measure and then a quick RTA. Once this is running, I can then move the microphone around to see what different listening positions look like (all in real time). It is revealing to see the difference a few feet make when you move the mic.

The RTA also allows you to change filters and watch the RTA screen for the real time changes. It's also really useful to adjust phase on a subwoofer for the best crossover response. You simply watch the RTA screen as you dial the phase control. A lot better than taking a bunch of measures to accomplish this task.

Here's a pic of RTA and a response measurement on the same screen (measurements are selectable to be placed on the RTA screen for comparison).
I was sure not to move the mic when I took the measure and then started the RTA.

REW RTA


Then there's a feature that you may have missed that is rather cool.

Instead of using a soundcard calibration file, you can use a loopback cable on the left channel and use that as the soundcard calibration.

One of the things this new reference allows is speaker distance readout (called System Delay feature).

I can hook up a single speaker at a time and do a full range sweep and REW will tell you the distance the microphone is from the speaker tested (in msec, meters, and feet). It doesn't really apply to subwoofers given the low bandwidth of a sub, but it is quite useful for the other five or six speakers in your system. You would of course require a full range microphone such as an ECM8000 or Galaxy CM-140 to use this feature.
This is useful if you think that your simple measuring tape method to establish speaker delay trim values for your receiver isn't that accurate.

I have tested this feature and it does seem to jive with my measuring tape. That's fine, but systems may have strange delays in them that you can't account for with a tape measure. It's a useful feature.

Below is the results from a single mains speaker in my office that is exactly as indicated by REW at 10.8 feet away from the microphone.
I did a simple sweep up to 20KHz and REW shows the system delay. I could easily substitute any center or rear speaker by moving the speaker cable around and establish the acoustic distance for every speaker in my system. Now that's cool.....

brucek

Just to clarify, I would assume that means that I'd need to get splitters to split both Line In and Line Out 1/8" plugs

Click to expand...

I don't really understand. To run REW you always required splitters at the line-in and line-out plugs on the soundcard, so as to break out the left and right channels. One of those channels was always left unused in the past.

You now have the option to not use a soundcard calibration file, by connecting that unused left channel of the line-in and line-out with a cable.

Like this:



The left channel takes care of the soundcard calibration. It also is a time reference to compare the timing difference between the two channels, so that time can be translated into distance. When you run a measure sweep of any single speaker (not a sub) it will tell you how far away the microphone is from the speaker. This is the same distance information you require for distance to enter into your receiver.

I also tested this feature by looping the right channel through a BFD to determine its delay that we always assume is 1 foot of distance. Well, it's 1 foot of distance...

brucek

It appears in my situation that the RS meter could also do the channel delay feature, but I could be wrong. My speaker is 9ft away and REW says 9.05ft.

I am trying to measure misc. sounds (such as room sounds and other things) w/REW, and I am not sure if I am doing it right or not.

Like bruce wrote, I loaded the software like I was going to do a speaker measurement. But, instead of plugging the line from my soundcard out to my stereo's input, I left that disconnected so no test tone was being played through my speakers.

Just for kicks, to test it out, I hit "start measuring" and unwound a roll of duct tape in front of the mic (~6" away) to try to measure that sound. I got a spectrum that looks somewhat like what I would expect from duct tape (~75 dB peak at ~4 kHz), but it goes <0 dB starting at 70 hz and lower.

What I don't understand is why it goes <0 dB. I would understand if it was dBFS, but this is dB...I didn't think there was such a thing as negative dB.

So, I am wondering...did I set something up wrong, and if not...what's negative dB? Also, I notice that the scale on the graphs bruce shows are in dbfs, but mine are in db...how does one get the data in dbfs vs. frequency in case I wanted to do that?

Mark

ps...here's the actual plot I made from my duct tape...(smoothed in 1/3 octaves)