HOLMImpulse Measurements, Need step by step dummy guide.

[Saurav]

You build an frd by pasting together several off-axis frds and specifying the angles.

That sounds like work I guess I could take some shortcuts since the midrange and woofer baffles are square, so I could assume that the vertical and horizontal measurements would be the same. That won't be true for the tweeter, and won't be true for my CNC-dream baffle.

Would you mind doing both?

OK. For the BMS, I'm assuming you want to stay with the same woofer filter, and just re-do the tweeter filter for the new driver? That would be the most apples-to-apples comparison I think.

[penngray]

That sounds like work I guess I could take some shortcuts since the midrange and woofer baffles are square, so I could assume that the vertical and horizontal measurements would be the same. That won't be true for the tweeter, and won't be true for my CNC-dream baffle.

OK. For the BMS, I'm assuming you want to stay with the same woofer filter, and just re-do the tweeter filter for the new driver? That would be the most apples-to-apples comparison I think.

Yeah, Lets keep the same woofer filter. Thx!!

[Saurav]

New measurements with 2nd box and the BMS4550 CD

I think you'll have to export those again. Looks like 'normalize to magnitude = 1' was checked, so both drivers now have the same amplitude in the impulse response:



Very different from your image:

Image not available

I'll start working with this since the tweeter filter network won't be affected by this, but new exports would help ensure that the final gain levels are right.

[penngray]

Sorry about that! I will do them again today.

[penngray]

New 4550 measurements

Here are the settings for the "Time Zero Locked" after I ran the measurement with "Detect time Zero/Causal Impulse"

Image not available

HOLM Measurements.

Image not available

Wave Export values (Just incase Im still screwing up).

Image not available

[Saurav]

That's better. I noticed that the woofer measurement seems a little different:



Red is the current measurement, blue is from the 1425 measurements, levels adjusted to get them to line up. The low end seems a little rolled off compared to the measurement with the 1425. I played with the gating to see if it would make a difference, but no luck. I'm not sure if this is a big deal, it could be the room or some other measurement issue.

Anyway, I stuck to our agreement of not changing the woofer filter. So here's a 1500 LR4 XO with the BMS:



Same deal with the shelving filter, use the closest value to the number typed in on the left. Also note the changed delay on the woofer - it would appear that the 4550's AC is further back than the 1425. Or I haven't got the response to track the LR4 curve quite right, so the delay is changing to compensate. Either way, I don't consider the woofer delay to really be part of the woofer filter (since it's compensating for the tweeter depth), so I don't think that breaks the rules

Predicted response:



That's what the optimizer thinks is best. Me, I'd drop the tweeter level maybe 0.5 to 1dB further and experiment with that. That seems to line up the tops of the peaks in the tweeter response with the woofer level, which may or may not sound more balanced.

Tweeter reversed:



Transfer functions:

[Saurav]

2kHz LR4 with the Celestion 1425:



There's a 2kHz peak in the woofer's response that's messing with this a bit. I tried notching the peaks out first, but the optimizer ends up at Q = 10, and still hunting. Try delays of 0.24 - 0.27 while checking the reverse null, the sim shows it change from a double null (where the 2kHz peak seems to add and create a peak within the null) at .24ms to a single null at .27ms.





Reverse null with 0.27ms delay:

[Saurav]

And 2kHz with the 4550, woofer filter unchanged:

[penngray]

Red is the current measurement, blue is from the 1425 measurements, levels adjusted to get them to line up. The low end seems a little rolled off compared to the measurement with the 1425. I played with the gating to see if it would make a difference, but no luck. I'm not sure if this is a big deal, it could be the room or some other measurement issue.

Thanks strange. I will try to figure that out today.

[penngray]

I checked TD12M on both speakers and they match in measurements right now. Not sure what I was doing then.

Im doing the ARTA off axis measurements for the 1425 tonight.

I didnt try any of the other crossovers yet since I have not be up in the HT room since last week.

[penngray]

Measurements with ARTA (raw, no gating)

Setup, are those values ok?

Image not available

Raw impulse

Image not available

Off axis 10 to 60 degrees.

Image not available


Off axis showing where null starts @ crossover its > 30 degrees.

Image not available


My off axis degrees are fairly rough estimates. I need to do a better job of marking the base pedistal so I know how much Im turning the stand.

[Saurav]

That looks pretty good to me. So the first one is horizontal and the second is vertical? Going up or down? Did you measure it by putting the speaker on its side and rotating it?

You can try some simple gating to clean up the curves. The yellow vertical line in the impulse response is the cursor. Clicking anywhere moves the cursor to that position. So here's what you do:

* On the top right, under Marker, click 'Del' to clear any saved information (when you first start ARTA, this should do nothing)

* Click in the impulse response window just to the left of the impulse. In your picture this would be at 6.something ms. The readout on the bottom left shows you where the cursor is (currently 5.688ms)

* Under 'Marker', click 'Set'. Now you've just set the start marker. The whole impulse response will go red, meaning there's no range that it's picking up.

* Click on the impulse response chart again, just before the wiggle at around 12ms. Doing this places the end gate. You'll see the 6ms - 12ms region go green, while the rest of the chart stays red. This shows you which region is going to be used for the FR calculations. The blip at 12ms is your first reflection, so you want the end marker to be just before that.

Once you do that, leave the markers as they are, and take all your measurements. That should clean up most of the noise in the FR curves (which are pretty readable as they are IMO).

Here's another idea, someone asked me this question in PM - why aren't we trying for a lower crossover with the BMS? I really didn't have a good answer for that one I know we've been sticking with 1500 and 2000 to have a cleaner comparison between the two. But I think eventually we should try something like 1200 on the BMS, it should be able to handle it.

[penngray]

That looks pretty good to me. So the first one is horizontal and the second is vertical? Going up or down? Did you measure it by putting the speaker on its side and rotating it?

Sorry I have not done the vertical yet, I have not figured out how to do the rotating for it.

Both are horizontal. I did the second one just to show where the null starts.

I will clean up the measurements as per your suggestions.

Yes, I would love to do the BMS around 1200.

[Saurav]

Hmm... I'm not sure there should be a null in the horizontal polars Well, let's see what the gated measurements look like, and hopefully someone else will chime in with what the polars are supposed to look like.

[penngray]

Hmm... I'm not sure there should be a null in the horizontal polars Well, let's see what the gated measurements look like, and hopefully someone else will chime in with what the polars are supposed to look like.

I was wondering about that.

I will gate the next set tonight.

[Dennis H]

Working with the cursor and marker in ARTA:
Left click sets the cursor.
Right click sets the marker.
Right double click deletes the marker.
Shift + right or left arrow moves the cursor 1 sample.
Read the cursor position and gate time below the graph in msec (samples).

Frequency resolution in Hz is the largest of:
1/(gate time in seconds) = 1000/(gate time in msec)
(sampling rate)/(sequence length)
(sampling rate)/(FFT size)

Longer gate time, longer sequence in the signal generator, and longer FFT to analyze the recorded signal, all give better LF resolution.

[penngray]

As always, thanks guys!! Its nice to finally get a complete handle on proper measurements. ARTA seems extremely easy now compared to months of fumbling around with simplistic measurements. The calibration stuff is still a nightmare and I can not get my dB axis values calibrated remotely close to real values but not a big deal (yet).

[Dennis H]

Hmmm, I just learned something from Saurav's post. I always thought you had to set the cursor before the impulse and the marker after. But his way of setting the marker before and the cursor after works too.

I still think it's better to set the cursor before the impulse if you're looking at phase. That way you can change the gate time or remove the gate without changing the cursor position and messing up the phase.

[Saurav]

Heh, I didn't know about the difference between left and right clicking Agreed about putting the cursor at the start gate to check phase. Likewise, if you're looking for the effect of room reflections, it's easier to leave the marker at the start gate and move the cursor for the end gate.

[penngray]

So do it Dennis's way?

[penngray]

Working with the cursor and marker in ARTA:
Left click sets the cursor.
Right click sets the marker.
Right double click deletes the marker.
Shift + right or left arrow moves the cursor 1 sample.
Read the cursor position and gate time below the graph in msec (samples).

Frequency resolution in Hz is the largest of:
1/(gate time in seconds) = 1000/(gate time in msec)
(sampling rate)/(sequence length)
(sampling rate)/(FFT size)

Longer gate time, longer sequence in the signal generator, and longer FFT to analyze the recorded signal, all give better LF resolution.

Should I be changing the signal generator values??

Here are the current values

Image not available

[Saurav]

So do it Dennis's way?

Either way works. The marker / right-click thing is the one you want to keep fixed, the cursor / left-click thing is the one that's easier to move.

In your case I'd probably do it 'my' way, put the marker at the start (just before the impulse) and the cursor at the end. That lets you play with moving the end gate in and out - you want to push it as far out as you can, while still having relatively clean frequency response curves. For the off-axis measurements, you're not really concerned about phase.

If you're measuring drivers for crossover work, phase is the most important thing, so there you'd have an easier time if you did it Dennis' way - put the marker at the end and leave it there, then you can play with the start gate using the cursor.

[Dennis H]

Either way works. The marker / right-click thing is the one you want to keep fixed, the cursor / left-click thing is the one that's easier to move.

In your case I'd probably do it 'my' way, put the marker at the start (just before the impulse) and the cursor at the end. That lets you play with moving the end gate in and out - you want to push it as far out as you can, while still having relatively clean frequency response curves. For the off-axis measurements, you're not really concerned about phase.

If you're measuring drivers for crossover work, phase is the most important thing, so there you'd have an easier time if you did it Dennis' way - put the marker at the end and leave it there, then you can play with the start gate using the cursor.

My mouse has a right button so I find it easy enough to move the marker (after the impulse) to change the gate. Or just double right click to remove the marker entirely and go ungated, still without moving the cursor. FWIW, that's the method he describes in the manual.

[Saurav]

LOL. Point taken.

[Dennis H]

Should I be changing the signal generator values??

Here are the current values

That looks fine although the sequence (128K) may be longer than you need, giving a frequency resolution under 1/2 Hz. Notice the time constant of over 2 sec. But better too much information than too little in the signal you record.

The FFT size (4K) is small and that would limit your LF resolution in an ungated measurement. I usually try to use as big an FFT as I can. ARTA will tell you if the FFT is too big for the number of samples it has to work with and then you can try the next size smaller.

Of course, if you gate the measurement, that's going to be what limits your LF resolution rather than the sequence length or the FFT size. A 5 msec gate means you have a 'real' data point every 200 Hz and everything in between those is a WAG by software. Gating gets rid of noise and reflections but it limits the LF resolution.

[HokieJoe]

Nothing of substance to add except to say what a great thread this is.

[coctostan]

Penngray, do you plan to add a capacitor in line for protection? If so, you might want to take impedance measurements so you can model different caps..or did you measure with a cap in line already?

I like the build. I'm thinking about trying a TD12M along with the 4550 and DE250 on the QSC horn. I might start with the 3012HO though. I will use a set of XTI1000's instead of the DCX but they have pretty much identical crossover and eq capabilities.

[penngray]

Penngray, do you plan to add a capacitor in line for protection? If so, you might want to take impedance measurements so you can model different caps..or did you measure with a cap in line already?

I like the build. I'm thinking about trying a TD12M along with the 4550 and DE250 on the QSC horn. I might start with the 3012HO though. I will use a set of XTI1000's instead of the DCX but they have pretty much identical crossover and eq capabilities.

The caps have been in place to protect the tweeters..I can not remember the values right now though (Maybe 36uF ??). I will open them up and check this weekend. There is a discussion on AVS about just adding a 4 uF cap to find out how it looks and it might make the over all crossover extremely easy. Zilch's "bridge-T" is really interesting and simple found here http://www.avsforum.com/avs-vb/showt...st18307375(ala Econo waveguide crossover stuff).



Im very interested in how those Crown amps workout, I have been considering them for awhile. I know they have decent functionality but I didnt konw they had all the functionality the DCX has.

[coctostan]

That makes sense. I should have noticed from the flat response.

The XTI (and DSI, CDi and a few others) have a pretty extensive DSP. You have the same crossovers, enough peq bands and delay. That is about all you need. It doesn't do everything the DCX does since I don't think it can split a signal 3 ways. It should be able to do everything needed for a single 2-way speaker. It is probably a slightly more expensive route at $350 each on ebay than using a DCX and a cheap multichannel amp. My room also allows for the XTIs and their fans to be in another room.

I might also try toying with an econowave derivative using their topology.

[penngray]

Sorry, I have not been able to measure again in my HT Room. I hooked up my old speakers so the family/friends/kids could enjoy some movies. I will get back to measuring this week.

I need to redo the off axis measurements to confirm that null then I need to run measurements for the new crossover settings.

I want love to see the BMS @ 1500Hz. Saurav, have you had time to build a BMS4550 crossover @ 1200Hz or 1000Hz?

[Saurav]

Nope, not yet. That's next on my list, but things have been a little crazy this past week. Hopefully in the next few days.

[penngray]

No rush at all. I know what crazy means. Im kind of stuck with a computer on my lap watching kids

[Saurav]

OK, fairly quick attempt at 1200. We really need to see how well the sims are tracking reality, especially off-axis, before we continue doing more sims

[penngray]

I will have more measurements by Thursday.

[penngray]

New measurements...

Gated Horizontal

Image not available


Gated Vertical

Image not available

On horn axis measurement

Image not available


so....I guess the fact that I did all measurements between the horn and the woofer then this is what measurement I get @ the horn axis (5" move vertically).

I think this is what I'm hearing in my tests. I believe I'm sitting about ear level with the horns.

What measurement should be building the crossovers from. Th measurements from the listening position of the speaker instead of the "generic" measurement spot (center between the drivers)?

[penngray]

1425 impedance

Image not available

1745 impedance

Image not available

Im thinking I might want to start working on using the Celestion 1745 as my "economy driver" instead of the 1425. Since it looks like it works easier below 2KHz.


I will start measuring the 4550 with XOs this weekend.

[Saurav]

so....I guess the fact that I did all measurements between the horn and the woofer then this is what measurement I get @ the woofer axis (5" move vertically).

Woofer axis or horn axis? Either you contradicted yourself or I'm confused.

Those polars don't look good IMO. So you moved the mic 5" up - how far away is it from the baffle?

On the vertical plots, did you start with the mic centered on the horn, then go up? Did you measure that by laying the speaker on its side, or moving the mic up?

[augerpro]

No idea if this will do anything or not, but I use a 16 or 8 ohm shunt resistor across the terminals of the CD. This will flatten the impedance that the amp sees and may improve any reactance the CD has with the amp and smooth perceived response. Just a thought.

[penngray]

Woofer axis or horn axis? Either you contradicted yourself or I'm confused.

Those polars don't look good IMO. So you moved the mic 5" up - how far away is it from the baffle?

On the vertical plots, did you start with the mic centered on the horn, then go up? Did you measure that by laying the speaker on its side, or moving the mic up?

Sorry about that, its a typo (trying to post in between kids screaming ), I meant the horn axis.

The measurements are still being done at approx. 1 meter.

I started the mic centered on the horn, then went up. Center of horn is about 5" higher then where I measured to build the crossovers.


btw, are the horizontal polars okay?

[penngray]

No idea if this will do anything or not, but I use a 16 or 8 ohm shunt resistor across the terminals of the CD. This will flatten the impedance that the amp sees and may improve any reactance the CD has with the amp and smooth perceived response. Just a thought.

I have read about that in another discussion. I will try to do that when I get a chance to buy some.

[penngray]

Augerpro, do you have a link that explains the shunt resistor?

[Saurav]

When 2 resistors are connected in parallel, they combine 'in inverse'. So if R1 and R2 are in parallel the final resistance R is:

1/R = 1/R1 + 1/R2

Here's a calculator:

http://www.calculatoredge.com/electronics/paralled%20resistor.htm

So let's take the 1745 impedance curve. You can see that it varies from about 5 ohms up to about 21 ohms. Now let's put a 16 ohm resistor in parallel, and calculate what the combined impedance will be. Eyeballing values from the chart, I get:

At 200 Hz, the impedance is 5 || 16, which is 4 ohms
At 1kHz, it's 21 || 16 = 9 ohms
At 2kHz it's 12 || 16 = 7 ohms
At 5kHz it's 8 || 16 = 5 ohms

As you can see, a 5 - 21 ohm swing has been brought down to 4 - 9. If you used an 8 ohm resistor, the swing would be brought down even lower, but the lowest impedance would be less than 4 ohms. In this case that might not be a problem since the tweeter won't be seeing a signal at 200Hz.

If you install Speaker Workshop (it's free, but has a pretty nasty learning curve), you can create a driver, input that impedance curve, create a network which has the driver with a resistor in parallel, and visually see the resulting impedance curve.

[augerpro]

I have read about that in another discussion. I will try to do that when I get a chance to buy some.

Where was this discussion at? Link please

[penngray]

Where was this discussion at? Link please

There was a minor point about a shunt resistor in this thread

http://www.diyaudio.com/forums/multi-way/163536-1-compression-driver-8-ohm-vs-16ohm.html

[penngray]

When 2 resistors are connected in parallel, they combine 'in inverse'. So if R1 and R2 are in parallel the final resistance R is:

1/R = 1/R1 + 1/R2

Here's a calculator:

http://www.calculatoredge.com/electronics/paralled%20resistor.htm

So let's take the 1745 impedance curve. You can see that it varies from about 5 ohms up to about 21 ohms. Now let's put a 16 ohm resistor in parallel, and calculate what the combined impedance will be. Eyeballing values from the chart, I get:

At 200 Hz, the impedance is 5 || 16, which is 4 ohms
At 1kHz, it's 21 || 16 = 9 ohms
At 2kHz it's 12 || 16 = 7 ohms
At 5kHz it's 8 || 16 = 5 ohms

As you can see, a 5 - 21 ohm swing has been brought down to 4 - 9. If you used an 8 ohm resistor, the swing would be brought down even lower, but the lowest impedance would be less than 4 ohms. In this case that might not be a problem since the tweeter won't be seeing a signal at 200Hz.

If you install Speaker Workshop (it's free, but has a pretty nasty learning curve), you can create a driver, input that impedance curve, create a network which has the driver with a resistor in parallel, and visually see the resulting impedance curve.

Thanks! That is exactly the explanation Im looking for. I like the Crossovers for dummies version Some people believe this lowers the SQ, is this just subjective audiophile stuff or is there a noticeable difference?


I have speaker workshop download but I just never spent the time to get it working.