Woofer problem in a 3-way design

[Martyn]

Hi, Folks.

I have a 3-way project that I started a couple of years ago and then put on the back burner. I've now resurrected it, but this time using a miniDSP and active crossovers. So far, I've built a test-box to prove the concept. If it works, I'll build a pair of fancy cabinets and I'll document the whole thing at that time. Unfortunately, I have a problem with the woofers...

The problem is that the bass has some "fuzzy" distortion. It also doesn't play as low as I expected, but that's secondary at the moment. The bass sounds like the player of a string bass isn't clamping down on the strings tightly enough, or that the skins on a drum are loose. It's not "listenable" as it is.

The design uses a pair of Peerless HDS 830883 6.5" woofers connected in parallel. As you can see in the photograph, the box is lined with a combination of acoustic convoluted foam and rigid fibreglass. The midrange is sealed in a pyramid-shaped enclosure. The space outside the apex of the pyramid connects the lower four levels to the tweeter level at the top for a total of 40 to 45 litres. The second level up houses a rear-facing 3" diameter port about 7" long flared at one end.

I have run the woofers in parallel and individually, so it's not a faulty woofer - it would have to be both, which seems unlikely. I can mute the mid and tweet, which means that I can eliminate those, as well as any interaction they might be having with the woofers. I have also eliminated the amplifier. I'm crossing to the mid at 325 Hz with a LR 12dB slope (I've tried 24 dB as well). I have also tried a LR 48dB at 35 Hz in case there's anything nasty happening low down. The box is 3/4" MDF with braces dadoed in. The baffle is held on with 14 woodscrews and is sealed with draft-excluder - there's nothing loose. The woofers have six mounting holes; I'm using woodscrews in just three of them, but they are tight and there really is little or no vibration to be felt.

I'm not very knowledgeable when it comes to the electrical side of speakers, so I'm hoping that the problem is something basic that I've overlooked. In any event, I need some help!

Anyone have any ideas?

Thanks in advance,

Martyn

[Ray_D]

Questions

You said miniDSP and active crossovers. Isn't the miniDSP the crossover? It sounds similar to driver mounting issues I have had. Why are you only using three screws one the woofers? You could try running a frequency sweep to see if you can excite a resonance which you might hear or feel with a finger on the driver frame(s). How hard are you driving them? Is the problem evident at all drive levels?

[Martyn]

Yes, the miniDSP provides active crossovers. I was labouring the point for those who aren't familiar with miniDSP products. Three screws: they clamp the drivers to the baffle quite solidly and this is just a test box...the "real thing" will use machine screws and inserts. Done lots of frequency sweeps, but there's almost no vibration. I'm not driving them hard - at least I don't think so, but it's hard to know what's going on when you're testing in your living room. There's certainly not much excursion. I should really do some tests outside, but the weather's not conducive at the moment (plus I'm still learning how to use the miniDSP).

[Matt M]

Hi Martyn,

You need measurements. The less knowledgeable you are, the more.

I think TMWW tower speakers are awesome, but also a real PITA to do right.

What are the other drivers you are using?

-Matt

[Martyn]

Hi, Matt.

Yes, but the less knowledgeable one is, the harder the measuring becomes. Learning how to use the miniDSP and Room EQ is enough for now! The tweeter is an OW2 and the mid is a PHL 1120.

[Matt M]

This is a sensible combination of drivers. This might result in a really nice loudspeaker. That's the good news.

The bad news: it doesn't matter if you have an active or crossover or a passive one - someone has to design the crossover.
It's not as simple as adjusting for driver SPL and setting crossover points. It really isn't. This isn't car hifi.
And I am not talking about high ambitions here, just the basics you would expect from a retail speaker that cost a few hundred bucks.

The very minimum effort is to trace the manufacturer's SPL plots. Run those through a cabinet simulation - you need to know the effect of the baffle. And then try to design filters that mimic a proven design, that is comparable, e.g. "Visaton Atlantis" would be a good start.
I think I could prepare you some data to work with over the next days, if you want to give it a try.

-Matt

[Martyn]

This is my first experience with a miniDSP product, which I'm pairing up with Room EQ Wizard for measuring the drivers individually in the cabinet and then together. First time around, I constructed filters individually and got the speaker to sound quite good - except for this woofer distortion. I habitually start with Unibox to figure out my box design. I've just started playing with the Auto EQ feature in REW to generate biquad coefficients for each driver, but haven't had time to get very far as yet. If we get some decent weather, I might try a ground plane measurement for the woofers, but I don't have a laptop so I'll have to drag a lot of gear outside.

It seems to me that with multichannel DSP crossovers and filters, things like baffle step become irrelevant once I'm measuring the assembled speaker. All I have to do is get the gating right, construct a reasonably level response, filter out the bumps, and then tweak the voicing. I'm over-simplifying, but isn't this correct in principle?

[TEK]

I think I would have disconnected the mid and tweeter.
After that I would have connected a regular amp/preamp and tried a sweep tone from 15-1000hz (or something like that).
That should get you a OK sweep. If the noice is gone - you know there is something going on with your minidsp.
If you still habe the same issue it has to be the drivers or the box. From what you are saying, the box is the most likely source if it is not he minidsp...

[Matt M]

Okay, so you have a microphone. Good.

It seems to me that with multichannel DSP crossovers and filters, things like baffle step become irrelevant once I'm measuring the assembled speaker. All I have to do is get the gating right, construct a reasonably level response, filter out the bumps, and then tweak the voicing. I'm over-simplifying, but isn't this correct in principle?

I would state it differently: it is irrelevant wether you compensate baffle-step with a DSP or a passive crossover. The process is the same: you try to fit a specific target function. Knowing the target is the tricky part.

Your woofer tuning looks good from specs btw.
Did you do a near-field measurement of the port to check if it is working as intended?

Regarding your initial problem: try crossing lower (200Hz) if your midrange is fine with this. Or try crossing higher (>550Hz).
If you look at other builds you won't find many that do TMWW with lower crossover frequency of 350 Hz.
E.g. Monitor Audio Platinum Floorstander or the wavecor ardents cross >500Hz. Many of the 800 B&W TMWW-type models too.
The Mini-Statements went 3,5-ways to make it work at 400Hz. I must admit, I do not understand the problem fully myself: something regarding power response and floor bounce.

It would be nice if you could provide more information upfront, pictures, measurements.

Btw. a good way to get a feel what is happening at the listening position is what I call the (lazy man's) "stereophile listening place measurement":
measure listening pos, 18" to the right, to the left, 10" up, 10" down. Use 1/6 octave smoothing and average those five measurements.

Do the same measurement for a pair of reference speakers and compare the results.
-Matt

[Martyn]

There's no obvious audible distortion when doing a sweep of just a woofer, but it's very obvious when playing music through just a woofer. I've done a measurement with the mic two diameters from the woofer, but even so the gating isn't effective for eliminating first reflections at those frequencies. I could try a measure just in front of the woofer cone, but I don't know how informative that would be.

[Martyn]

Good idea with the port...I haven't tried that, although I have tried two quite different port lengths (I found significantly different T-S parameters quoted by different sources. I'm now using those from the Tymphany website).

Just to be clear, I'm running as a TM(WW), i.e. three channels. However, I do have the option of running four channels and controlling the woofers individually. Might be fun to try, although I should probably walk before I can run. Nonetheless, the distortion is present if I drive just one woofer, so they can't be interfering with each other,

I'll play around a bit more and then post some pics

[xandresen]

I use a miniDSP 4x10 for a 3way crossover.

I have run into the problem of distortion several times - in each case it has been the sound levels versus the miniDSP.

If you overdrive the input of the miniDSP, it clips the waveform and produces considerable amounts of distortion.

If your input signal and the crossover design asks the miniDSP to produce more output voltage then it is capable of,
again it clips the waveform and lots of distortion.

I find it's easy to get the levels wrong when starting out with a new design.

a miniDSP has switches on the pc board to allow you some adjustment of max input/output level.

[Martyn]

Thanks for the tip...I suspect that's what has happened (see my next message).

[Martyn]

*sigh* I was going to post a trace, but I seem to have forgotten how to do it! I managed it for my original post, but now...

I'll put it down to a "senior moment" and try again tomorrow.

[Martyn]

I have played a little more with Auto EQ in REW. When I replaced my hand-constructed filters with the Auto EQ biquads, the bass cleaned up! This is still early days in that my use of Auto EQ was just a first pass and not sophisticated by any stretch of the imagination, but the speaker sounds much better and it's actually looking quite promising. I was going to attach a trace of the woofers and also of the port, but I'm going to have to revisit that first. I think I mentioned that different sources quoted different T-S parameters for these woofers...well, the ones from Tymphany's website are telling me that I need a shorter port (a lot shorter), so I'm going to do that next. I'll also try to do a ground plane measurement before much longer, although I have a busy week ahead already.

With any luck, I'll be able to commit to a final cabinet design fairly soon and get working on it in the new year. I'll post a woofer and port trace as soon as I've modified the port.

My sincere thanks to everyone who has responded thus far,

Martyn


Hmm...I see that the traces have attached themselves anyway.

[Dave Bullet]

Both channels / both speakers or just one / one side?

[Martyn]

There is only one cabinet - it's just a test box. I've learned the hard way not to build ambitious designs until the drivers and their configuration have been proven!

[Martyn]

A week later...

Over the course of the last week, I've spent a lot of time on this - much more than I intended. I have done literally hundreds of sweeps: near field, far field, individual drivers and in combination. I even managed to drag the gear outside for some free field measurements.

I eventually settled on crossing first at 600 Hz with an LR fourth order, and again at 2400 Hz. I found that Room EQ Wizard's Auto EQ feature sometimes works very well, but not always. Ultimately I ran an Auto EQ on the overall response curve and then used the resulting data to hand-build the corresponding filters in the miniDSP. Even so, I still had to tweak them further.

The image below shows the three response curves measured individually at one metre and gated accordingly. They have been shaped by the crossovers, but have not been filtered further and are unsmoothed:

The second image shows the combined response curve when the mid has been padded down 4 dB, all the filters are in place, and it has 1/6 octave smoothing:

I then moved the speaker to its typical location and measured again, this time from the listening position about nine feet away and with the microphone at ear level and point straight ahead (i.e. not directly at the speaker). The difference was dramatic:

I then re-worked all the filters to produce a more or less acceptable shape (1/6 octave smoothing). I didn't put too much effort into this, because by now I'm quite confident that the miniDSP will be capable of producing a good result with the final cabinet:

Unfortunately , I still have a problem with the bass. I'm sure that my original distortion problem was caused by my lack of familiarity with the miniDSP, but now I have an apparent lack of bass. It appears to be rolling off quite early, even when the speaker is positioned only a couple of feet out from a corner of the room. I should mention that no filters have been applied to the woofers.

The mid and tweeter sound quite good, so I think this will be a good speaker if I can find out what's happened to the base. I've checked the polarities, which are fine. The 3" diameter port is flared at one end and 11cm long (per Unibox). Perhaps I've over-done the acoustic foam and glassfibre?

Any other suggestions?

[TEK]

From your description it seems like your theoretical model tells you that the base should mot fall off as early as it dors.
Could you share the expected base respons?
My initial thought is that this is how those drivers delivers in the given box. If so you would have ti use eq to lift the lowe end or take down the mod, tweeter and the upper part of the base driver, resulting in the whole speaker to be a heavier load.
Just some ideas from the top of my head - but I have to mention that I have no experience in speaker design - just in building. So be very critical to anything I says...

[Face]

Bypass the MiniDSP and measure the woofer section again. If it continues to roll off that early, something is amiss with the box or tuning.
Can you post up your box sim?

Sent from my SM-G935V using Tapatalk

[Martyn]

This is my Unibox model (is this what is meant by a bass simulation?):



I'm expecting an F3 of around 40 Hz and an F10 of around 33 Hz.

Bypass the miniDSP: do you mean bypassing the crossovers and filters in the miniDSP? If I do that, I get the same woofer response as is shown in the first image in my previous post, except that the response isn't rolled off by the 600 Hz crossover. The lower roll-off is much the same. I can probably bypass the miniDSP completely if that's what you mean...I'll have to think about that further.

[Dave Bullet]

Most systems need 3 - 5 dB of baffle step compensation. Can you export the transfer function? Here's mine for my 3 ways. Ignore the absolute amplitude (Y axis) values - it is the relativity that counts.

In the attached graph - see how the bass (green) is approx 3 dB higher than the red (mid). Also attached is my BDS simlation - showing the bafflestep rise of the woofer (pink) and midrange (yellow). This takes into account the baffle dimensions (mine is a truncated triangle). You can estimate your bafflestep F3 frequency (that is -3dB from 0dB level where all frequencies will radiate off the baffle and not wrap around via 115 / your baffle width in metres).

In anycase, you can see the pink and yellow lines rising. This combined with the drivers natural behaviours, lead me to specify the attached electrical transfer function.

I'm using a passive crossover - but the concept is the same. You are aiming for the same acoustic output levels and acoustic roll offs.

Can you extract your electrical transfer function from mini-DSP? We should see the midrange being more attenuated then the bass as mine shows, and depending on where you crossover, this will be apparent in the transition between drivers (if in the middle of bafflestep) or moreso on one driver or another (i.e. an applied slope)

[Martyn]

Here's the woofer response without going through the miniDSP:



The red line is both woofers measured midway between them, and the green line is just the top woofer measured on axis. The mic was 40" away in both cases. The responses are gated at 3.5 ms and otherwise unsmoothed. Good idea to eliminate the miniDSP as the cause of the problem, which I think we have now done.


Dave, these are the individual responses:



They were taken outside with the mic at 20" on axis to each driver in each case. No gating, but 1/3 octave smoothing. Note that the tweeter was shot 5 dB higher than the others in this instance. Is this what you wanted to see?

The next image shows the individual responses with their active crossovers in place (LR24 at 600 and 2400 HZ), but with no smoothing or filters. They were measured indoors at 40" and gated at 3.5 ms.



Everything shows a pretty consistent roll-off below 200 or 300 Hz...

[Bear]

Let me ask what could seem like a stupid question: When you were designing your box, did you exclude the volume of the internal bracing and the volume of the port (using the OD, not the ID)? Your model uses a 41L box volume, and the picture of the box in the original post doesn't look like a 40 liter box to me (e.g., a 9"x24"x12" outside dimension box using 0.75" sheet goods will gross about 29L before subtracting the port, the bracing, the driver volume, and any other acoustically opaque materials). If you combine baffle step and a too-small box, then the significantly diminished bass would be expected. Excessive excurion would then generate a fair bit of distortion, even before hitting mechanical limits.

Another possibility is a problem in the relevant amplifier (i.e., downstream from the MiniDSP). Are you keeping the amp constant between the various tests?

[Martyn]

I have a fairly detailed spreadsheet for calculating the volume. Overall, the box is 40" x 9" x 14" for a gross internal volume of 59 litres. After subtracting the volume of the mid enclosure, braces, drivers, the port, and an allowance for crossovers (which I now won't use), I get 43.5 litres. So if anything I'm actually a couple of litres oversize. It's easier to add a bag of sand to an oversize box than it is to add volume to an undersize box, but it was a good question.

The amplifier is an equally good question. I keep a record of the "volume" setting with each test and try to use the same setting every time if I can. This doesn't always work out, but at least I have a record and can allow for it. Of course, when all drivers are running, the volume knob is out of the equation. I'm using an old Arcam multi-channel AVR for the amplification. There's a potential risk that the outputs have been conditioned according to some HT requirement, but I have ensured that all speaker outputs have been set to the same levels, i.e. "large" with zero compensation and delay. There's a similar risk for the inputs, but the only way I can get separate channels of amplification for multiple channels of input is to use the DVD-A/SACD analogue inputs. I believe that the multichannel analogue inputs to the AVR pass through "unconditioned" except for the volume control. Nonetheless, I use a "front speaker" input for the woofer channel just in case. My recent test that took the miniDSP out of the equation involved using a TOSLINK input to the AVR instead, which therefore eliminated the possibility that the DVD-A/SACD inputs were conditioning the signals somehow.

[Juhazi]

Bass drivers might be faulty units, but most likely you have too much eq/gain in low end. But also you shoud systematisize your approach like this:

1. Connect minidsp-amps to drivers but don't set any eqs, xo or anything else. Set the speaker on a stand in the middle of the room, pillows on the floor. Turn down wife and kids, AC, dishwasher, etc.
- measure each drivers's "raw" response with same signal level, at 40-60cm distance, mic height between mid and tweeter, use 12 ms IR gating, no smoothing
- measure bass drivers, both connected, also in nearfield say 10cm from baffle in midpoint of them, use 60ms IR gating
- look at responses and set eq for each driver (bass as one unit) - don't do anything that raises gain over zero line in minidsp graphics!

2. measure again the equalized drivers at higher spl ( calibrated mic should show at leas 100dB at 60cm) level and look at distortion
- set spl levels same by reducing gain of those units that play at higher spl! (typically mid needs - 6-12dB)
- preferably use a turntable under the speaker and measure also off-axis responses 10¤ steps up to at least 60¤, preferably up to 180¤

3. set crossovers
-WM around 400-500Hz (depends on your baffle width and distortion in the lower end of the midrange)
- MT around 2500Hz (depends on your mid's directivity and tweeter's low end distortion)

4. set polarities according to xo type
- with LR2 you must invert the polarity of the midrange

5. measure summed responses of WM and MT
- measure also with midrange in different polarity and check summation
- set some delay for typically the higher-playing unit until you get good summation and deep xo null with wrong polarity of mid
- check also step response to fine-tune delay

6. measure all 3 units simultaneously
- set more negative gain for the midrange! With LR2 it's response gets boost from both the W and the T
- listen to it with music at normal distance, you can now also do a listening spot room response measurement with 500ms gating
- bass response obviously is'nt right, set even more negative gain for mid and tweeter

7. measure bass separately in many different ways - floor plane at various distance, listening spot, speaker at different distance from walls
- this is the most nasty and difficult task, even small changes with eq can alter the sound very much!
- also listen to music with total system after every change . Repeat the next day etc.

8. Enjoy sweet music!


Greetings,
Juhazi

I started learning this in 2013 https://www.diyaudio.com/forums/mult...ml#post4231049

[Martyn]

Juhazi, than you for such a detailed approach. I think I've done most of that, although not in such a well-structured way. I spent the first two or three days just learning how to use the miniDSP and Room EQ Wizard, although there are still a few things I'm struggling with. It's good to have a disciplined method to follow.

I stand the speaker on a box that puts the tweeter the same distance from the ceiling as the top woofer is from the floor and then use a single gate time for all measurements. For the whole speaker, I typically measure at 1 metre on the midrange axis. I have also measured each driver on its own axis at 50 cm and also near field inside and outside. Background noise in the room is usually below 50 dB. I haven't summed the WM and MT combinations in pairs. Nonetheless, I will work through your process carefully, paying particular attention to the SPLs.

What is bothering me most is that I have this low end roll-off whatever I do...even if I measure without any EQ, regardless of whether I measure inside or outside, with or without other drivers, and so on. Until I can get the low end to come up, there doesn't seem to be much point in doing anything else. I have another pair of woofers (for the second speaker of the pair eventually) so I could try those. It seems unlikely that both of the current drivers are damaged, but I'm getting to the point where almost anything is worth a try.

Here's a question for any Room EQ Wizard users: under the Target Settings tab of the EQ section, I set the speaker type to "none" and then set the LF Rise and HF Fall slopes to zero so that is no roll-off at either end. REW defaults to 75 dB for the "Target Level", but I have almost never been close to this when I click "Set target level" - REW will nearly always set a much lower level. If I try to force a higher target level, REW will warn me that this is more than 10 dB away from the recommended level. Sometimes I get a message telling me that some high percentage of my data points are outside the recommended range. How can I rectify this? I don't see how this could be the cause of my problem, but I mention it just in case. It's also irritating!

[Juhazi]

Hi,
I have never used REW autocal.

About bass response measured this way - it should not be straight down low! When speaker is in middle of a room and on a stand, bass doesn't get the boundary reinforcement that it has at normal position on the floor and near wall(s). But anyway your measurements show excaggerated downfall...

http://arqen.com/acoustics-101/speaker-placement-boundary-interference/

By the way my procedure comes from John Reekie. You can mostly skip the open baffle/dipole things

https://www.hifizine.com/2010/12/prototyping-4-way-open-baffle-speaker-with-the-minidsp-2x4/

https://www.hifizine.com/2011/03/refining-a-4-way-open-baffle-speaker-minidsp-2x4/

https://www.hifizine.com/2011/09/prototyping-dipole-bass-system/

[Martyn]

Thank you...good links.

[Bear]

Bass drivers are often run through a "rub and buzz" set of tests for many of the reasons this thread highlights. We haven't used the term yet, but that's basically what people have been working you through.

Unibox gives you a predicted anechoic response. With a vented box like yours and a tuning like yours -- assuming there is not an implementation problem -- you should see a 24dB/octave rolloff below the tuning frequency (if you align the Fb and F3 in Unibox reasonably closely). This can vary based upon the realized Qtc, but set that aside completely.

In the real world, on finite-sized baffles, you will see about a 6dB rolloff for baffle step loss, plus some additional artifacts from edge diffraction. Jeff Bagby's Diffraction and Boundary Simulator spreadsheet can help you simulate what to expect in real-world implementations. Attached is a picture of the baffle step response for a project I'm working on (welcome to analysis paralysis!!). You can offset much of the baffle step loss by floor-loading your woofers. After accounting for driver roll-off, baffle step, and diffraction effects, then you would add the room reinforcement.



In a typical room with typical placement near walls and corners (i.e., non-Audiophile optimized), you can easily see +6 - 12dB/octave of gain going from 100Hz down to 20Hz. Since there's no such thing as a free lunch, there is often one or more room cancellation modes that create large dips/suckouts in the 100 - 200Hz range. If you look at some of JonMarsh's measurements of his various works, you can see these artifacts quite clearly. They aren't there because Jon doesn't know what he's doing...

So, what do I think I see? You have a really early roll-off for your woofers -- but it is a relatively gentle roll-off (-4 to -5dB/octave from 200Hz on down; is your port blocked?). That's likely a result of baffle step plus a few other things since it doesn't stop at just a -6dB cut. Baffle step plus a room cancellation would give you the roll-off from ~250Hz, but there would also be significant reinforcement in the low bass that manifests itself in the measurements.

Without auditing your spreadsheet, I'll trust that your volume calculation is correct. However, unless there is a lot of open volume behind the mid and tweeter chambers that isn't visible in your one photo, I'm a bit skeptical that you've got 40+ liters. A test of your volume calculations and implementation would be to pull all of the egg crate and yellow foam out of the woofer volume and see if you saw an increase in the bass extension. Yes, that's changing two variables. However, I'm still a bit stuck on the implementation questions.

Similar to Juhazi's list, here are some things to check (some of which you've done):
1) Rub and Buzz - test each woofer individually. Use one and only one amp and channel for testing -- you need to control the variables closely. Run sweeps, run chirps, grab your favorite bass workout tracks (but not too low!) and run each driver through its paces listening for artifacts like the one you have identified. Don't use the miniDSP for this. Also note, bass drivers that aren't broken-in can sound "thin" for a while until the suspensions loosen up.

2) Like #1, but now wire the drivers together and run a shortened version of #1. You are listening for phase issues between the drivers and their interaction with the room. Issues here should be a bit more subtle, but we're tying to troubleshoot each piece of the assembly through the process.

3) Check your mid!! "Strangled midrange" is a common problem when the mid enclosure isn't well-suited to the midrange in question. If the problem hasn't presented itself in the woofers when you were testing them individually, then the midrange is the next logical place to look. It should be broken-in as well. Various vocals and choral music help here, but start with the test tones. No low bass unless you like ordering new drivers and starting over. :-)

4) Add-in the miniDSP (see Juhazi's list!). Start with just the minimalist crossovers and definitely make sure that you aren't inadvertently causing the unit to clip the inbound signal. I would steer well away from LR2 slopes at this point, and focus on LR4 and BW3. When you've fully ruled-out physical/mechanical issues, then worry about implementation.

- Play your test tracks at various volume levels to see if the problem is related to changes in volume. Identify when it is just barely noticeable and when it becomes objectionable. Since you're Canadian and identified the issue in bass guitar, I assume that you have everything that Rush has ever recorded to test variations on the problem. Just ignore Geddy's early vocals. :-D

- See if you can isolate the problem to a set of frequencies to match specific test tones against. If so, play with the crossover points and slopes to see how it affects your perception of the problem. Measurement can help here (e.g., IMD bursts)

- You may have a faulty miniDSP or some other quantization problem. Consider upgrading the the 2x4 HD. It's just a much, much better processor, but obviously US$200. An inexpensive class D desktop amp with a gain/volume control can help isolate your existing preamp and amplifier from the equation (assuming that the new amp isn't faulty itself, of course).


5) Start adding your EQ filters, one at a time or in small groups (if you can't isolate them). Start with the ones that make the biggest changes (especially high Q / high gain). Test each one individually first, then build the stack going from most aggressive to least aggressive. You are testing to see if the problem is a quantization issue that is driven by an outright math problem or that your unit starts running out of processing power. This is where the 2x4HD can help, among its many talents.

That's probably $0.025 worth of advice, but that's about all that I've got. I don't have the expertise that others have, so I'll step back to the shadows.

[Ray_D]

I suggest you take all of the stuffing out of your enclosure and make new measurements. I believe stuffing for woofers is overrated. The pictures do not show the openings in the braces. Are you sure they are not too restrictive?

[xandresen]

One more thing to watch out for

When we gate a measurement, the gating establishes a frequency limit below which accuracy gradually vanishes.

For a 3.5msec gate, the lower accuracy limit is about 1/3.5ms => 286 Hz.

Some of your problem with bass response comes from the gating. The error would vary with frequency and where in the room you made the measurement.


You also provide an un-gated woofer response. I don't see any room resonances in it. I have never seen a living room measurement that didn't show pronounced room resonances below 200 Hz. ??
We don't want to have the software equalize the room at the same time we are trying to design/measure a loudspeaker

[Martyn]

Bear, thanks for taking the time to write such a detailed post. There's a lot to get through and it will take me some time...I'm already being quizzed along the lines of "What do you have planned for tomorrow, dear?". I can respond to a few points directly:

Here's the drawing for my box...it's quite accurate except for the small circles on the side panel. The gross volume below the lower midrange brace is about 40 litres, so add the space around the mid and above the tweeter and subtract all the "allowances" and we get back to 43 litres or so. I'll pull out most of the glass and foam as you and Ray have both suggested.



My miniDSP is the 4x10HD since I will need six channels for a pair of 3-ways. I have the sensitivity jumpers and gain switches set to give me zero gain overall in the miniDSP. I have bypassed the miniDSP as another member suggested and thus eliminated it as the source of the problem. I have also used different channels in my AVR in order to eliminate that. However, I haven't tried using one and only one channel for all single driver measurements...that's an elegant thing to do. I wouldn't expect to see a difference in this instance, but I'll try it anyway.

I'm not a Canadian by birth and was never a fan of Rush during my mis-spent youth, but I was referring to an acoustic double bass anyway! I think my original distortion problem has been resolved...it's this early roll-off of the bass that's the problem. In a listening test beside my usual 3-way speaker, the mid/tweet in my test box is very good, but the low end is sorely lacking for any track that has significant mid/low content. I start by matching the SPLs of the two speakers, but some tracks will sound much quieter because the woofers are rolling off over such a wide range.

Back at it...

[Martyn]

Good points. I used the bounce calculator at mehlau.net with the speaker on a box about 40% into the room and rotated slightly off the centreline of the room. The ungated responses were taken outside my house which is on a hill. The speaker was on its side with the centres of the drivers about a metre off the ground and with the mic 50 cm away on axis. The outdoor ungated and indoor gated responses have a similar shape, as do the near field and far field responses. Even allowing for a beginner's experimental errors, the woofers are clearly lacking in a side-by-side listening test with my usual speakers, so something is amiss somewhere.

[Face]

I suggest you take all of the stuffing out of your enclosure and make new measurements. I believe stuffing for woofers is overrated. The pictures do not show the openings in the braces. Are you sure they are not too restrictive?

Agreed, ported enclosures only need the walls lined.

Sent from my SM-G935V using Tapatalk

[Bear]

Note that the 4x10HD (172MHz DSP of some form that can't do FIR filters) isn't the same processor as the 2x4HD (400MHz SHARC that can do FIR filters).

[Juhazi]

So, with 4x10HD you can play with dip switches inside (two sets), they help to get you suitable gain to match input voltage and output voltage. I had to wiggle with those too, but I don't remember which settings I have. https://www.minidsp.com/forum/hardwa...-9vrms-default

Basically I believe that with bass you might get clipping in digital domain, in minidsp.

How high is your source's output volume usually set, as 0-10 scale? Do you use source or power amp as volume control? Some people recommend feeding minidsp with maximal signal level and gain/volume pot of the power amplifier. I use my HT receiver as source and it's volume adjust, then analog RCA to minidsp 4x10HD and true power amps with fixed gain. Minidsp's main volume at input is set to 0dB. I have never noticed excesss noise or sound deterioration at almost silent volume level.

But I can also induce overdrive/clipping this way (in minidsp input) - and then I get same level of H2 and H3 throughout the passband when doing a sweep (at too high level).

If minidsp eq settings go too high, over safety margin, signal starts clipping at only in those specific frequency ranges.

I think that most problems with dsps come from these gain match problems and from using too much positive eq gain. The original unbalanced Minidsp 2x4 (without HD) stands only 0.9V input voltage, and that is also default setting of 4x10HD. The 2x4HD clips at 2,4V input.

Edit: the unerlined was wrong, my memory failed. Both 2x4 and 2x4HD have jumpers for input. Output of 2x4 has 0.9Vrms max and 2x4HD has 2V max. 4x10HD has dips in/out 0.9 or 2Vrms (all as unbalanced) When input is set to 0.9V it is too easy to make it clipping! But for me it is necessary to gets spl level to match with HTamps other 3 channels! Then spl is around 105dB at 3 meters and that is more than enough for me.

[Martyn]

Very interesting. I have the miniDSP DIP switches set to 2V RMS to give me zero gain unbalanced (which seems logical in my simple world). However, I'm using a Squeezebox as my source and do not attenuate its output. I use an AVR after the miniDSP as an amp/preamp and volume control, and I have the volume control set around its usual level, so this seemed reasonable too. Yesterday I started afresh and have got as far as a response curve that's within +/- 2.5 dB on axis without setting any filters - with the exception of my usual woofer roll-off (I was quite pleased with this). My next step was going to be the removal of stuffing from the box, but flipping the DIP switches is such an easy thing to try that I'll do that first.

[Martyn]

Up-Date - Progress at Last

It feels like I'm getting close.

I made a fresh start yesterday and followed many of Juhazi's and Bear's suggestions. I started by taking individual responses for each driver from 50cm on each axis gated at 4.7ms. I used the same amplification channel at the same SPL in my AVR for each test. I then switched to a different channel and did them again. Good news: all measurements were very consistent.

I then repeated the tests while bypassing the miniDSP. This meant that I had to use an optical input to my AVR and use either the "Satellite" or "Tape" selectors. Again, the curves were very consistent, but were about 4 dB lower. Since the gain in the miniDSP was set to zero, I have to assume that the 4 dB difference is due to the different route through the AVR.

Back at the miniDSP, I then repeated the tests with the gain DIP switches in their default positions. This produced the same curves as previously, but with the expected 6 dB drop. Seeing no obvious benefit, I set them back at zero gain since that seems more logical to me and also gives me the volume control setting that I'm used to on the AVR.

With the mid measuring hot, I reduced its gain by 10 dB in the miniDSP, turned the AVR up by 10 DSP, and shot everything again. This time I left the mic on the midrange axis for all measurements. I then worked on my crossover points and measured the drivers in pairs (WM and MT). When all three were looking good, I inverted the mid to determine and optimise the delays.



At this point I had a promising response curve which was +/- 2dB from 200Hz to 10kHz. I now removed a fair amount of glassfibre and foam from the woofer box. I left 2" of glass on the back wall and on the top and bottom panels, but removed everything else. I measured the response curve again and could detect no visible difference. I did some final EQ tweaking on the response curve which smoothed it out quite well, although it was still around +/- 2dB. With 1/3 octave smoothing, this drops to +/- 1dB. Here's the EQ'd but unsmoothed curve:



Here's a distortion curve (fundamental and THD), but I must admit that I don't really know whether this is good or bad:



I did some quick listening tests with the speaker on its box in the middle of the room, and have to say that it's sounding very much better. I ended up using the same crossover points and slopes as previously, but with the midrange down 10dB versus 4dB previously and with quite different delays.

If I get time tomorrow, I'm going to put my "daily driver" on the box in the middle of the room and do a quick measure just out of interest, then I'll run them side by side in the listening location.

[scottvalentin]

Excellent perseverance and in the end you have some nice drivers so will be able to end up with a very nice speaker.

For the MiniDSP, I had issues if my source volume (in my case PC) was turned up all the way. I ended up with fuzzy, distorted bass as you describe. So I ended up running my PC at 50% volume and turning up the AVR volume to compensate. I was going from PC-MiniDSP-AVR multichannel input for 4 channels for a 2-way stereo pair.

One thing I haven't seen is a nearfield measurement of your woofers with no gating. All of your measurements have the bass rolling off above 200hz, but this is likely the result of gating at 3-5 ms. This is why your graphs show no bass.

For distortion, leave the boxes for the 2nd harmonic and 3rd harmonic to display. Your total distortion looks pretty good in the sweep above. You would be looking for any spikes in third order (3rd harmonic) distortion. These spikes would tell you of a resonance or where a driver is playing in a range where it isn't happy (eg tweeter crossed too low).

[Juhazi]

Glad that you made it work better! I forgot to mention, that minidsp control software has input level meters (bars) - you should check input level there during sweeps or when playing music .

Distortion /THD) curve looks like showing only ambient noise in you room. Play the speaker at much higher spl like 100dB at 50cm. Toggle also Noise floor to see ambient noise level!

[Martyn]

Thanks for your comments, Scott. I have three or four Squeezeboxes scattered around the house and run them unattenuated as a matter of course with no discernible issues. I use the SB volume control on the remote for fine tuning if one track is louder than another. I suspect that a PC environment is rather different. I particularly like the Squeezebox Touch - they come up on the used market from time to time for under $100.

I did some near field measurements last week. Here's the woofer ungated and unsmoothed:



Thanks for the distortion tips. Here's the plot from yesterday showing the 2nd and 3rd harmonics and noise floor:



Juhazi, I'm using the mic at 50 cm from the driver with the AVR volume at 60 dB (which I suppose really means 40 dB of attenuation. Even at this level, I'm quite close to triggering a feedback loop, so I think I'm as loud as I can go. The distortion curves are almost entirely below the noise floor, which I assume is good. Thanks for the tip on the miniDSP input level meters...one of those things that's obvious once it's been pointed out to you!

In summary, although the low end is now sounding a lot better, I still can't explain the roll-off below 200 Hz. Today I'm hoping to do a comparison with my "daily driver". The latter uses a 10" woofer in a sealed box. It's very crisp, but sometimes I wish that it went a bit lower. The double Peerless woofers in my test box were supposed to get me a bit lower in a slimmer box with only a small loss in crispness. I'm not confident that this is going to happen. I might have to make a difficult decision between accepting the shortfall or abandoning the Peerless in favour of a pair of Dayton RS225 drivers. Unfortunately, these are larger diameter and will drive up the size of my final design box significantly. This will strain an important element of my design goals while giving me a volume that's far too big.

[Juhazi]

Your 50cm distance response looks quite normal regarding bass. Nearfield measurement with long gating shows how reflex port tuning works. You should measure also at port's mouth. Nearfield is useless for distortion because mic gets saturated too easily.

So, now that your speaker and measuring gear are obviously functioning well, you can take measurement of a single speaker in it's normal position and mic at your listening spot. This is called room response. However, if you sit close to (back)wall, take mic a little forward. Do a slow sweep (512) and set IR to 500ms and smoothing to 1/3 octave or psychoacoustic. Now you will see the sound balance and how the low end works in your room. But you will also see lots of reflections and modes in the room. Move the mic and measure at least from 5 different positions, you can hold the mic in your hand. Then overlay responses and toggle "Average responses" too!

Distortion plot of REW shows a rather nice spl response. You really must play the speaker at higher level, but without clipping mic input (100dB at 1 meter at least, with calibrated mic). Anyway, to see real distortion, you must see H2 and H3 well above noise floor! Typically distortion is highest down low and around crossovers. You must check that furniture and windows are not rattling!

If you detect a peak in distortion, use generator and play sinewave at that frequency. Walk around the room and close to the speaker to find out where that noise/distortion comes. It should be quite easy to detect, usually a peak is from a loose driver or wire rattling or something in the room resonating. High H4/H5 might be box leaking etc.

[xandresen]

Looking at the near-field frequency response of the woofer at 6mm, ungated, I wonder if your microphone pre-amp has a filter turned on to block bass below 100Hz?

thoughts as a fellow miniDSP 4x10 user
The frequency response of the drivers is affected by the selections we make in three places.
I find it easy to focus on either Input or Output EQ and forget what I set in the others!

1. Input page EQ
I do my baffle correction here since Input EQ affects all drivers.
For the slope correction:
EQ type = High_Shelf, Frequency = 250, Gain= -4dB, Q = 0.5
this works well for me with a 10.5" wide cabinet in a 14x28 foot living room.
Some people might increase the gain to -4.5 or -5 dB. -6 dB is too much bass.

2. Output page EQ and Output Page Filters
make sure you Bypassed the HighPass filter in the bass driver's Filters section.
an LR filter is 6dB down at turnover so one set at 35 Hz would drop that
frequency by 6 dB. Turn the HighPass's Bypass on and off while playing
bassy music so you can hear what you are missing.

[Martyn]

It's possible that that near field response was shot outside...my notes aren't quite clear and my memory is worse! I was learning a lot of new things at the same time, so I can't be certain. I'm using the miniDSP UMIK USB microphone, so it's powered over USB from my PC. Other than this, I don't really use PC audio very much, so I don't know much about what goes on in Windows 7 and sound cards.

Interesting about doing some of your speaker EQ on the Input page...I assumed that that was for doing a room EQ for each channel. Yes, one of the first things I did was to clear the default filters.

Another post will follow soon...