Tuesday, March 8, 2016

LM-1C The Process

Here I'll quickly go over the process I took in designing the LM-1C.

As usual, my tools of choice are OmniMic and DATS V2 plus XSim.

Cabinet Selection

I'm not making my own cabinets, but instead selecting among available off-the-shelf cabinets from Dayton. I would like to say that I had simulated the bass response in a tool, then went looking for the right cabinet, but in fact that happened late in the process.

The truth is I doubled the cabinet volume for the LM-1 (0.281 cubic feet x 2 = 0.562) and looking for the closest appropriate cabinet.  This turned out to be the MTM-0.75.  Sounded too big, but in fact according to WinISD was too small. Anyway, using WinISD I plugged in the values for the Peerless 830991, and selected 2 drivers, vented.  WinISD wants a 1.4 cubic foot cabinet. Far too large for the location I'd like to put it in, so I retried with the 0.75, and set the port diameter to 1.5" (0.0381m).
 This raises the -3db to 64 from 50 Hz.  Since we're using this with a subwoofer, and a typical center channel speaker is cut off around 80 Hz, this is kind of an irrelevant compromise.  There's also a slight rise (0.5 dB) centered around 160 Hz.

Overall a very nice compromise.  Setting the port count to 2, WinISD recommends a 2 7/8" long port.

So, to summarize:

  • Cabinet Volume: 0.763 cubic feet / 21.6 liters
  • Ports: 2 x 1.5" diameter and 2 7/8" long
  • f3 = 64 Hz. 
You could also use the  Dayton 1 cubic foot cabinet, and get around 55 Hz.  Not worth the size. If you really wanted a high output center, you should probably select the larger 830990 Peerless drivers, justifying a larger cabinet.

Cabinet Preparation

Driver Location

Center the tweeter about 2 5/8" below the top edge of the baffle. The woofers are vertically centered on the baffle 5 7/16" away from the center.  Sorry for the odd measurements, truth is I did the driver locations in metric. 5 1/2" away from the center is probably close enough too. 


Next, is to route the driver holes and mount the drivers and ports. Here's what it will look like:

I have surface mounted the woofers, but inset the tweeter.  For the tweeter I set my router depth to 0.2" and first routed a 4 1/8" outer rim, then set the jig to 3 1/2" That provided me with the inner and outer flange rings.  Then select something in the middle to route out the remaining ring.


Since I knew this would be my center, after testing I hot glued both of the crossover boards to the rear, bottom panel.  You want to keep the two boards apart, and out of the way of direct radiation from the woofers. 


By feeling the cabinet while watching movies I discovered that the front chamber is pretty vibrant ( literally! ) so I added some thick Sonic Barrier.  Multi-layer works best.  It's also good to put the crossover parts out of the way of the direct reflection from the drivers, and put a small piece directly behind the tweeter. Of course the tweeter itself doesn't create much standing waves in the cabinets but by adding padding around the tweeter we minimize the energy that hits the tweeter housing, and, by extension, the opportunity for that energy to re-radiate or distort the tweeter.

I half-filled the rear chamber with Acousta-Stuf and it does in fact reduce some left-over boominess. Place it so it's tucked up against the rear cabinet wall, and around the ports.  Ideally you want to leave the ports clear all the way to the drivers.

Felting Up 

I added a 4" wide sheet of PSA black felt underneath the tweeter to minimize diffraction effects. It's one of my favorite tweaks.  It's of course optional, but I encourage it, and it's cheap. Cut the circle out roughly, apply to the panel, and then use a sharp, clean razor to gently saw through the felt at the tweeter's flange cut out. 


Measurements are the step that must be taken before crossover design can begin accurately. We must know the driver impedance, frequency response and acoustic offsets in order to begin.  For the LM-1C I've done this for you, but you may be interested in doing this yourself.

If measuring your own design, connect long test leads to the tweeter and run them out through one of the ports.  Do the same to 1 of the woofers, but jumper the two woofers so they are electrically in parallel.

Driver Impedance

If you haven't already, measure the tweeter impedance (i.e. Z).

Measure the woofer impedance in the cabinet. This can be done anywhere, so you can move the speaker to your PC if you need to.  This will give you the combined effect of the woffer and cabinet and port(s).

Drive Frequency Response

Put the speaker where you will use it. Follow the procedure for calculating driver distances, explained in a previous post. For now test the woofers together. For the woofer especially this is not a step that can be skipped.  In place the woofer, cabinet, baffle step, room and furniture will all come into play.

You might be tempted to take one of two alternative approaches in regards to the woofer testing:

Route 1 Woofer and Measure

Won't work because the box volume is 2x what you are simulating.  See what happens in WinISD with this approach.

Route 2 Woofers, measure 1

Nope, the second woofer will act as a passive radiator and screw with the impedance and FR.

There is an approach which would be equivalent:

Create half sized cabinet with 1 port

This is the only far-field approach which will yield the right results. However, it requires you actually build a test cabinet.

Hacking the files

Because we are using 2 drivers in a single cabinet measurements get a little tricky.

You can take one of two routes now.  You can pretend you measured a single woofer and create a single woofer in XSim which has the combined measurements or you can hack the data you gathered and in XSim to get data for 2 separate woofers.  The advantage of the second option is that you can try different things like a 2.5 way design if you want to or re-use the data to simulate a TW configuration in smaller cabinets.  A very useful thing if you are building a 5.1 system but only the center is dual-woofer.

Anyway, to estimate single driver data from this MTM cabinet you will need to do the following:

Hack the ZMA files

The driver impedance will change from the spec once it's in a cabinet, especially if you are using a ported configuration.  This is why to be accurate it's important to measure the impedance of the driver in place and not rely solely on published specifications.  However, it's not absolutely necessary. We can estimate that we only care about the impedance at the top end near the crossover point, and use the published specs, or vendor ZMA files.

The impedance of the individual woofers will be 2x the measured, but the frequency and the phase will be the same.  So open up your ZMA file, plop it into your favorite spreadsheet, including Google Sheets, and copy the three columns.  They are not labelled however they are:

  • Frequency
  • Impedance
  • Phase angle

In your copy, double the measured impedance values. Save it as the new, single driver impedance.

Hack the FRD files

The FRD files are similar, but the center column is amplitude. You'll want to do the same as you did above, but subtract (not divide!) 6dB from the amplitude. Save it and you have a new FRD file with the contributions of a single woofer.

Test your results

If you chose the dual simulated woofer approach you need to test that you hacked everything correctly.  For this we'll start creating the crossover.  Put down three drivers. Ignore S1.  Wire S2 and S3 in parallel:

Now, examine the Frequency Response and Impedance graphs.  They should match exactly the original measurements in SPL or Ohms.  They should not match the hacked files. The final result should be 6dB higher in FR and half the Z of the hack.

With either approach, you should be starting with exactly the same FR and Z as measured. Only then can you start the crossover design.

Now import the tweeter FR and Z (impedance) files, and start cooking.

Using other Tools

One thing I cannot stress enough is that this process is only as simple as it is thanks to OmniMic's automatic blending feature.  It automatically blends the gated high frequency response with the non-gated low frequency response.  If you use REW then you end up doing a lot of work that is not needed to get the right measurements before you can import them into XSim (or whatever your crossover design program is). Also, we don't need to calculate room gain and add it, blah blah blah. Our measurement and source of truth is the speaker and room itself. This process is faster and more accurate and less error prone by the nature of it's simplicity.

Lastly, don't trust most speaker maker's specs, especially the T/S measurements. The cost is no indicator of likelihood of accuracy.  ScanSpeak is notoriously off spec for instance, while the Focal and Peerless woofers I've tested are pretty close.  Buy a driver first and measure it before making cabinet selections.

LM-1C In Room Measurements

Free Field 

This first measurement I took last. I placed the LM-1C on a Hsu subwoofer and measured a foot away. Very smooth response:

In the bass 80 Hz is about -4 dB, but please note the extension to 20kHz! This is a perfect HT speaker! Especially for the outlay. I strongly suspect on a dedicated stand it would be even better. The curve is so close to ideal I won't bother with EQ here.

On Entertainment Center

Jaggedness below 500 Hz attributable to the usual room/speaker interactions, however the upper shape of the response is pretty much on target.  One good thing though is that the sensitivity at 1 kHz is actually a little better than predicted, and the response from around 600 Hz to 1 kHz is much smoother than predicted, a good thing. Plot was taken with 1/24th octave smoothing.

I also took the liberty of running some distortion measurements at 1 watt. The measurements past 800 Hz are all below 0.35% HD.  Most excellent values for the spend.

I may in fact have benefitted somewhat from the bass coil having unusually high resistance.  I used a pair of coils to make it, and the total coil impedance ended up being around 1.2 Ohms, yielding a much smoother mid-bass response. Never stop looking for serendipity!

Saturday, March 5, 2016

Introducing the LM-1C

The LM-1C speakers are part of the Leach Memorial speaker project, started here.

The 1C's are tonally balanced to be placed as shown, on an entertainment center. On or in an entertainment center however is never ideal so despite my best efforts, modest equalization will be required for them to sound their best. If stand mounting, EQ is strongly suggested to make up for the minimal baffle-step compensation built in.
Originally I was going to make the LM-1 and then add a center/D'Appolito-like speaker for the center or for those who wanted more output and clarity, but I made it first because I really needed a center channel more than I needed main or surround speakers. However, the qualities of a good center channel like this is exactly the same as for any other speaker in a home theater or stereo system. You may use the LM-1C as a center channel with four LM-1's or you can make all of your home theater speakers LM-1C.  The choice is yours.  There's nothing particularly "center" or "home theater" about the LM-1C.

The LM-1C is MUCH better free standing than the LM-1 however. The LM-1 does better in a bookshelf or on top of a desk or entertainment center. So this is one reason for choosing the LM-1C over the LM-1. 

Now, I did promise to post the design and details of the LM-1, but the 1C is a separate project.  It doesn't mean I won't post the schematic.  Just I won't be as diligent. The picture is the tweeter crossover prototype which I have built from parts I already had lying around. You'll need another small board for the woofer.

Instead of wire ties and anchors you can just hot glue instead. Wire ties are best for tweaking and experimenting.

As you can see, the LM-1C has the same driver complement as the LM-1, just doubled up on the woofers, and the ports are not the same either. One thing I've noticed so far is that the speaker needs to be tilted up to give the correct tonal balance. 

Here is the schematic.  I was using actual measured values for the coils, based on the parts I had on hand.

I used the  Dayton 0.75 cubic foot MTM in Cherry along with a pair of 1.5" ports cut to 2 7/8".



Here we go, the "money shot." Prices are for the individual speaker

  • Cabinet: $149
  • Drivers: $95
  • Crossover: About $75 and up. 
  • Miscellaneous: Speaker connectors, 2x 1 1/2" ports
Approximately $350 assembled.


Update March, 2015:
I took measurements on top of a sub instead of on an entertainment center. Results were MUCH better than below, so I strongly encourage you to try this speaker out in free space on a stand if possible. Of course, if you can't, your receiver/processor's normal room EQ should handle the changes needed automatically.

Original article:
After assembling a speaker it's always a good idea to compare the finished impedance plot to the predicted plot.  This is a quick way to tell if you messed something up.

Here you can see I got it pretty much spot on.  The minimum impedance is around 4.3 Ohms at 195 Hz with an electrical phase near zero (not shown). If you are using any sort of solid state amplifier this will be a very easy load.  Some tube amplifiers may not be happy with the impedance peak though, so consider adding a 100 Ohm resistor across the speaker inputs if you are using a tube amp.  Also System7 over at DIYaudio likes to do this just to prevent the chances of ultrasonic oscillations.  May be worth trying.

Of course, what everyone wants to look at is the FR.  Sadly I only have the predicted so far.  I just don't have the time / energy to measure it again, especially since I'm still moving it around.  But for now, here is the FR as predicted by XSim.  As before, this should be far-field responses on the cabinet, as opposed to quasi-anechoic.  Scale starts at 20Hz.

Cheating with DSP

So this is all well and good, but putting a speaker on top of or in a TV stand is rarely the best sounding location even if it is the most convenient in the case of the center channel. At the end of the day I cheated a little by adding parametric filtering via miniDSP:

Center Channel EQ

Sadly I have lost the "final" results, but as you can see the EQ required was pretty modest, and mostly involved broad cuts between 70 and 250 Hz with a couple of boosts. I deliberately limited the EQ to be between 30 Hz and 500 Hz to minimize stress on what really is a small speaker, and to avoid over-EQing what is otherwise a pretty well balanced design.

The bosts at 62  and 300 Hz are 2 and 3 dB respectively, while the 3 deepest cuts average around 6 dB. Overall a good power balance, especially with the lowest change being a significant cut around 40 Hz. This should not stress the amplifier or speakers extraordinarily, and indeed when playing Pirates of the Caribbean and Battle Los Angeles at theater levels from 9' away there was no audible strain at all. In addition to this EQ, I relied on the Oppo 103's integrated bass management to provide high pass filtering at 80 Hz.

Of course, this EQ is very much situation dependent! I measured the speaker after I had placed the TV behind it, and then used OmniMic to create my filters, so I would not encourage anyone to take this chart and attempt to apply it to their own build.

What I mean to show is that a little judicious EQ can help improve the LM-1C or really any center channel speaker. Subjectively the thing that improved most was reducing boominess in the dialogue, which is usually in the 100-200 Hz range. Reducing that made dialogue much clearer and the center much more transparent.