18" Subwoofer DIY Options & Theorem

hi,

the first step would be to get some measuring gear, so you can see what the response looks like in your room.

at a minimum an spl meter and some test tones, or a mic/soundcard/software setup for your computer.

subs and rooms form a system. if your upper bass from a tht is slow and boomy, you likely have a large room resonance.

room resonances don't decay very fast, so notes get blended together and can definitely sound slow and or muddy.

if it is there with your tht, it will be there with most anything else in the same location.

18" drivers can be just as good as 15" drivers. the better 18" drivers are better than lower quality 15" drivers for both music as well as home theater.

the mach5 audio uxl18 would be a good choice.

how large is your room?

Some good subs in that price range are:

Mach5 IXL-18
Mach5 UXL-18
Fi Car Audio SSD-18
Dayton Reference 18

Just to name a few. I am in the same boat as you Evolvo, looking to build a really good 18" inch sub for my home theater, I am looking forward to seeing this build that you are going to do!

If you want the spike removed, try doing the measurements without the soundcard calibration.
Instead, use a loopback cable from one channel line out to one channel line in.
Be sure to apply those settings in REW.

This worked for me, and now I don't need to calibrate the soundcard because REW is doing the correction "on the fly"

If I were you, I would definitely go with an 18" driver, or possibly a pair of 15" drivers. A large, ported, 18" sub is most definitely going to go lower than your THT, but whether it will have the mid-bass output you desire, is mostly dependent on your room and listening taste. I find it strange that you say your THT goes really low but lacks mid-bass output as that is the exact opposit as what everyone else says about this horn loaded sub.

If you are truly mid bass hungry and your room is not causing a suck-out from 80-160hz, then twin 15's will likely be better than a single 18 only because (all things being equal) you will have greater efficiency. However, a driver that handles output in the <30hz area should not also be tasked to handle the 80-160hz area.

I think chest pounding bass is from 60-95 which crosses in to what I cover as the midbass band. 60-40 is hit you in the gut bass but excess 55ish Hz has caused some systems I've worked on to be muddy. 40-20 I like to call Pants flapping bass. An interesting note: the lowest octave, when played loud, tends to make my head and face hot and kind of tingly. I think this effect is due to humans fight or flight response to ULF. Anyway sorry to jack the thread, I just love bass.

-Matt

I've posted this before but seems to fall on deaf ears, so I'll take this opportunity to repeat it here, FWIW:

An analog loopback measurement of a sound card has nothing whatever to do with a frequency response measurement. Since all analog devices have DC offset blocking capacitors whose values are completely arbitrary, and thus the huge variance in roll off from one card to another, an analog loopback will show this roll off. In almost every case (100% so in all of the cases I investigated), the blocking caps are in the analog output stage.

However, when you measure the FR of your room, the signal you feed through your SC into REW never sees the analog output stage of the sound card. It goes into the SCs A/D converter and stays in the digital realm from there to REW.

That means that if your particular SC has a roll off at or above 10 Hz (which many do, or worse) in its analog output stage and you run an analog loopback from which REW makes a correction file, in most cases that correction file ends up being a boost curve to the actual A=>D signal path.

Same goes for using a generic correction file for the RS meter, and I wouldn't pay much attention to any measurement result below 10 Hz using that method to measure.

Finally, there is the noise floor of the room/measurement system, which increases as frequency decreases. To safely clear that hurdle, you have to measure at 100dB minimum, well above the level your posted graph shows. This also would entail making sure REW is at least somewhat calibrated. Since I've never used the RS meter to make a measurement, I don't know if the SPL reads while the mic is fed to REW, but if it does, used a 1000 Hz tone (where the RS meter is probably the most accurate) and set the calibration before you measure. Then, bump the sweep level to 100 dB and lose the SC correction file.

Try that and repost, if you care to and get the time.

Dave,

The analog microphone input being used is not usually flat to 2Hz and neither is the analog sound card output, usually xlr, rca or trs, into the amplifier. These are prior to the signal being converted into 1's and 0's. What am I missing here?

Not to mention the output signal from REW comes from the analog output stage of the sound card and into the device / system under test in most setups I've seen.

The sweep and sub system have nothing to do with the measurement hardware/software. You say that the SC input is not usually flat... do you have any examples? The answer is 'no' because no one knows how to measure what the A/D converter hands to REW much less actually measures it and posts that data.

Yes, there can be DC offset blocking caps in the analog input stage, but I haven't found that to be the case in my investigation of the subject in SCs, preamps, etc.

OTOH, I certainly have seen a wide variety in loopback measurements DC blocking caps values dictating the correction file curve.



The sweep generator signal and subwoofer signal chain are a different can of worms... first things first if you want an accurate measurement below 20 Hz.

If you connect the output of the sound card back to the input of the sound card and do a sweep you will measure the frequency response of the sound card. Yes, it combines the input and output stage into one measurement and you can't identify whether the input stage or output stage is responsible for the deviation from a flat response, but for the sake of measuring things with REW where you're using both the input and output stages in your measurement it's adequate to compensate for effect of the frequency response of the sound card on the final measurement.

I guess my point is that you are assuming that it is a straight A/D conversion into the speaker and back into REW. What about the analog microphone input and output stages of the soundcard? They are both being used in most cases. IOW going right through those DC blocking caps. The signal goes through the SC going both ways when doing an acoustic measurement. These input and output stages usually show some roll off which you show in your charts above. How are the measurements passing through them but not being affected by them is what I am not following. The signal is not digital the whole way.

Here are measurements of the sound card I use with REW. Left and Right channel respectively.

REW only monitors one of the channels on the input. When you switched to the right input did you also change the setting in REW?

Then explain the OP's posted response. Surely neither of you believes it is accurate? I'd be interested to know what you both believe is happening?



My sweep stays digital until the AVR converts is at the SW output. There are many ways to accomplish this. The OP said he used a Berry USB SC, which I'll bet has a digital output. If not, what is a SC with digital out, $20? How would anyone know that he didn't use the digital output of the Berry? I assumed he did.

Make that YOU can't identify which stage has the blocking caps. Of course there's a simple way to determine which and digital out keeps the output stage in digits until your SW output, which, as I said, is another story unrelated to this one.

Application of a correction table/file that doesn't apply to his exact measurement device. Outside noise during the sweep is another possibility.
*sigh* I meant it's not possible to tell solely from the measurement. I didn't say that it wasn't possible to make the determination another way. Frankly, I would expect both the input and output to have filtering to block DC on a sound card.

Here are a few suggestions:
1) Repeat the sweep with the subwoofer turned off and see what you measure at the very low end.

2) post the .cal file you're using for your meter.

3) post a graph without smoothing

4) sweep at a higher output level (higher than 70dB, like 95dB)

Your thread is not being jacked. All the info is there for you to get on track.

First, as I said, calibrate the level (see earlier post). If you don't, the dBSPL numbers on your graph are useless.

If you have a digital out option on your Berry, use it and lose the SC loopback cal file.

Not knowing what RS meter you're using and the procedure and settings is also a problem. I remember Ilkka and Ethan Winer posting this back when regarding the digital version:

More recently, notnyt said that when his digital meter was set to 'C' weight, the output jack WAS C weighted, so there is a HUGE difference in the 2 scenarios.

IOW, if your meter is NOT C weighted when set to C weight and using the output jack, using the correction file in REW would definitely give the huge boost we see down low.