Ar303 info.

[AR303 and classic 18 user]

In addition to the measurments of the AR classic 18 I have been involved with, I have also been taking some measuremnts for my AR303s. As with the classic 18 the measuremsnts presented are for the raw drivers in the enclosure without crossover parts. (I thought I would post the results as a way to give back for all the great information I have received from this site.) Attached are the results. Here are a few notes on the setup. Its proably not ideal but works well enough for my purposes...

The mike used was a radio shack analog sound meter that was calibrated to the radio shack spec in the manual.

SPL data is relative only and was not calibrated to an absoulute level (same voltage used for each driver).

All are measurements taken at 1.0 V rather than standard 2.83 V. (Did not want to risk tweeter...)

Soundeasy MLS measurement system with realtek 2 channel sound card.

The data is near field and far field data (1 meter) merged together to form the complete response.

[AR303 and classic 18 user]

Well I finally managed to get through the process of building the response from the raw driver SPL data. To be clear about what is shown, a computer program is generating the transfer functions for the crossovers and applying that to the measured in cabinet SPL data shown in the previous post.

The first attached image shows the schematic as entered into the program. I used the 303-a schematic as a basis which can be found in the library section of this website.

A couple of things about the SPL data plots:

1)Overall the technique used to measure the data seems in general to be working. In particular when the crossover is applied to the SPL data for the woofer and the midrange something resembling the specified response is calculated.

2)The measured phase is consistent with the published phase and the computer program predicts that the tweeter needs to be hooked up with its positive lead attached to the negative terminal. (Drivers with reversed polarity are blue.)

3)The sore thumb here is the high frequency response. The tweeter SPL appears to be too high.

Things to look into:

1)Revisit the microphone calibration file being used to measure the response.

2)Revisit the measurement conditions perhaps the assumptions that the conditions were the same for each driver was not correct.

3)Revisit the merge process between the near field and far field data, i.e. what determines what the best merge frequency should be and should priority be given to the far field or near field data?

But even with the offset in the high frequency response I think one begins to get some sense of how the crossover interacts with each speaker to produce the total system response.

[AR303 and classic 18 user]

I had some time to verify the calibration I was using. Although not strictly AR related, I thought it would important for people to be aware of the calibration since it is critical to the final result measurements. I went through the process of verifying the calibration of my RS meter using the stereophile review data as a reference. Although not strictly a true calibration, the resulting validation calibration should allow for a determination if the original calibration file I am using for my meter is reasonable or not. (I don’t have the equipment to do a full calibration.)

To start out, I total took system measurements for my 303 speaker system. These measurements were performed at 1 m and 1 volt. The raw impulse data was saved so that the various calibration data could be applied to generate the various system responses. I started with calibration data in the RS 2050 manual, made, a system frequency response curve and compared it to the stereophile reference. This data is plotted in Fig 1. The match isn’t outstanding but I think one can see that there are major features in both the user 303 and the reference 303 that correlate like the dip at 1 kHz and the peak at 6 kHz. (Looking at the reference stereophile data it can be seen that the response of the AR 303 is very flat. )

With this first iteration complete, I calculated the difference between my system response and the reference system response (Fig 1. green plot). I then corrected the radio shack calibration by the same amount, generated a new system response with the corrected calibration and repeated this process for several iterations. The results for the final iteration are plotted in Fig 2.

Since it is a different set of speakers, I will only use this calibration to compare against the calibration I had used previously to generate the plots in the first post. This original calibration file was from two data sets that are commonly found on the internet that over time I had merged together. (Only from 1000 Hz above is being validated and below 1000 Hz remains as is.) The results are graphed in Fig3 which shows the various calibration data. Over all, the validation matches the “internet calibration” quite well and I think this shows that the original calibration set is reasonable. Most of the differences are probably due to the fact that two different 303 speaker systems were used for measurements. However, there is a dip at ~4 kHz that the RS manual data shows that the internet calibration does not. Interestingly, the validation also has a large discrepancy at ~4 kHz here that might actually suggest a correction could be applied to the original calibration file?

I think the conclusion that can be drawn here is that the raised tweeter response is not caused by a faulty calibration file. I will need to keep looking for the source of the raised tweeter response from the second post.