Using REW to Plot Crossover Response Curves?

[Megastat]

REW's impedance test function works really well for me. Using the jig diagram proved easy and the results appear very accurate.

What I'm wondering is if a way exists in which I could send a sweep to the input of a speaker crossover and plot the response curves of the outputs?

If there is a way to configure a jig to do this, it would be incredibly useful for generating accurate digital clones of passive crossover response slopes. Drivers would have to be connected for the crossover points to be valid, and it would be one output at a time.

Cheers

 

[JLM1948]

That is quite feasible.
You have to send the output of the souncard to a power amp that drives the speaker; I would recommend being very cautious about levels, it's so easy to fry a tweeter! Typically, speaker measurements are carried out at 2.83V RMS (1W nominal into 8 ohms); for frequency response measurements, level is not critical.
Then you need to pick up the signal at the driver's terminals. That's the part that must be thoroughfully thought.
The signal must be attenuated in order to not saturate the soundcard's input. The degree of attenuation depends very much on the characteristics of te soundcard. A typical SoundBlaster card has microphone level inputs that saturate with only a few millivolts, whilst semi-pro soundcards have line level inputs that accept levels up to several volts.
This information is paramount for "designing" the attenuator. Said attenuator is going to be of teh "L" type, which consists in just two resistors or alternatively a potentiometer.
A L-type attenuator can be calculated on-line. The sum of the values of the resistors (or the potentiometer's nominal value) must be significantly higher than the speaker's nominal value. Typically 10kohms is satisfactory, but 5k and 20k also. It is not critical.
OTOH, the value must be in the same order of value or lower than the souncard's input impedance (generally a few kilohms).
Now, one must identify precisely what terminal of the driver is the cold (ground) and what is "hot".
Connecting the ground terminal of the sound card to the "hot" speaker terminal results in injecting current in the soundcard's ground circuit, which can damage it seriously.
For a minimum risk of error, the soundcard ground terminal should be connected to the cold terminal of the speaker, and the souncard's input should be connected to the driver's "hot". In the absence of certainty, probing the two driver's terminals should show one with no signal and the other active, which obviously the correct one.
Beware that some speakers do not confirm to this simple scheme.
For example speakers with time-correction APF's (All Pass Filter) have signal on both sides of the driver(s).
Another case is for amplifiers with BTL outputs; thses amps have signal on both terminals. I would recommend not using that type of amp for measurements.
In both cases, the measurement must be taken differentially, which is possible if the soundcard has balanced inputs. In that case, the attenuator would be a U-type.

 

[sm52]

Megastat,
A crossover simulator isn't what you want?

 

[JLM1948]

A crossover simulator isn't what you want?

I don't think so. He needs to emulate the response of an existing x-over. After taht, he may enter the results as target responses, that the simulator would try to achieve.

 

[sam_adams]

You can measure the electrical response of the crossover by simply using the output of the soundcard. Connect the soundcard output to the terminals of the speaker and the input of the soundcard to the particular driver's terminals that you are going to measure. Run a sweep on each driver and then use the Trace Arithmetic function of the All SPL tab to combine the responses. Depending on the type of crossover that you are measuring, you may have to invert the polarity of one or more of the measurements to get the correct combined response curve.

A two way, 2nd order crossover with the tweeter polarity inverted is shown below:

Red curve is the woofer. Green is the tweeter. Orange is the combined response. The polarity of the tweeter response was inverted before the the responses were combined as that is the way the tweeter is connected in the system.

All the measurements above were taken with the drivers in a test enclosure of the same dimensions as the final enclosure with provisions for the test leads to be present outside the enclosure.

Acoustic responses may be measured with the drivers in the enclosure by disconnecting the driver(s) that are not being measured and substituting either a straight resistive load for the disconnected driver or by calculating and constructing a simulated equivalent speaker load for the disconnected driver(s). A simulated load would have to be outside the enclosure—the number of capacitors, inductors, and resistors of the simulated speaker will be quite bulky—and provisions would have to be made to have it connected properly for testing. Remember, you will need some sort of timing reference to make this work. A brief bit of creative thought on how to make that work is all that is required.

 

[Megastat]

Then you need to pick up the signal at the driver's terminals. That's the part that must be thoroughfully thought.

Hi JLM1948, I
typically use an amplifier with an attenuator, or the headphone output of my soundcard to drive the loads. If it's as simple as sending a stepped sine sweep from one channel of the soundcard's outputs to the crossover input and having the output from the crossover's tweeter output sent into one channel of mic input, that would be great.

What I'm not sure of, is if this type of frequency measurement needs to be connected to compare the signal out to signal in (like the im-pedance jig).

Megastat,
A crossover simulator isn't what you want?

I have modelling software, but the goal is to plot the electrical curves of the high and low pass sections without disassembling the crossover to measure the individual component values.

 

[JLM1948]

Hi JLM1948, I
typically use an amplifier with an attenuator, or the headphone output of my soundcard to drive the loads.

The main issue there is that most soundcards cannot drive appropriately the speaker. A speaker is supposed to receive signal from a very low source impedance, which is seldom the case with sondcards. the result is that the measured curves are not the same as those when driven from a power amp.

What I'm not sure of, is if this type of frequency measurement needs to be connected to compare the signal out to signal in (like the im-pedance jig).

No. You don't need to compare, since the signal that is sent to the speaker is flat (unless the amp has a non-flat frequency response).

I have modelling software, but the goal is to plot the electrical curves of the high and low pass sections without disassembling the crossover to measure the individual component values.

that's exactly what you get with the suggested method.

 

[JLM1948]

You can measure the electrical response of the crossover by simply using the output of the soundcard. Connect the soundcard output to the terminals of the speaker and the input of the soundcard to the particular driver's terminals that you are going to measure. Run a sweep on each driver and then use the Trace Arithmetic function of the All SPL tab to combine the responses. Depending on the type of crossover that you are measuring, you may have to invert the polarity of one or more of the measurements to get the correct combined response curve.

there are two major flaws in your method:

• The soundcard has a source impedance that is significantly higher (10-20 ohms) than the source impedance of a power amp (0.01-0.1 ohm). The transfer function will be significantly different
• Using arithmetic sum to compute the overall response is bound to give errors at the x-over point, even when taking in account possible polarity inversion. In addition, it does not take into account the difference of sensitivity between drivers.

 

[mboxler]

Hi

First post here

Recently discovered this way to test the voltage across a crossover output. The following plot is one of many that I don't understand. It's a high pass filter that should look like the LTspice plot, but for some reason, the plot gets fuzzy until around 80db. Is there a setting I'm missing?

Line out-->Amplifier-->Crossover input-->Crossover output (with driver attached)-->Line in

Thanks in advance!

Mike

 

[JLM1948]

Well, it looks like there is a notch in the response at about 5k, on which LTspice and REW agree.
The fuzziness is just noise. I suggest you do the same measurement with the input signal muted (but with the amp and soundcard connected). You'll see the same fuzziness.

 

[mboxler]

Thanks. I assumed it was noise and considered increasing the signal, but worried about damaging the soundcard.
It's a steep filter and I'll just assume everything from the noise and below is okay.

 

[JLM1948]

Yes, you must know that noise, after AD conversion and FFT results in bumpy response instead of a nice continuous fraph.

 

[mboxler]

I have a separate issue. Klipsch used autotransformers in his early crossovers to attenuate the voltages to drivers. When I ran some plots on a crossover, the voltages seemed a little off. I finally simplified my test. I ran a plot of the source voltage, then ran a plot off the -3db tap of the autotransformer. The voltage across the -3db tap ended up being -2.8db.

When I used a DMM, the difference between the source voltage and the -3db tap measured -3.05db.

That's quite a difference. Is this normal, or is there something I'm missing?

Thanks, Mike

 

[JLM1948]

When I used a DMM, the difference between the source voltage and the -3db tap measured -3.05db.

That's quite a difference. Is this normal, or is there something I'm missing?

Thanks, Mike

0.3dB is a minor error, about 3%. Do you know what your DMM's precision is? And its frequency response.
REW is not exempt of error either. What soundcard are you using? Significant error can come from the reference point of the measurement setup.

 

[mboxler]

0.3dB is a minor error, about 3%. Do you know what your DMM's precision is? And its frequency response.
REW is not exempt of error either. What soundcard are you using? Significant error can come from the reference point of the measurement setup.

The soundcard is the audio jacks on my ASUS motherboard, but I may have found the issue. The level setting on my Line In was set to 50. When I moved it up to 100 my plots improved dramatically. I assume that digital reduction(?) caused the issue?

 

[JLM1948]

I assume that digital reduction(?) caused the issue?

Veru likely.
Another possible cause of errors is the unbalanced connections. Ground current circulation can offset measurements.