Quote:
This is quite a spread in frequency response. Given the SPL levels you are using for the measurements, I think you have way too much room interaction with your speakers.

I would suggest setting the meter at 80db @ 1k and placing it in your listening position. Also, consider using the "C" weighting and "slow" settings.

I think 24db swings in the mid band can be improved upon with speaker positioning. Would you comment on the size of your room and your speaker positioning?

I did use the C weighting but I used the fast response instead of slow. I will run again with new settings.

Also, I didn't use any of the of adjustments per frequencies that are out there for the radio shack analog meter. I guess I will do that with the orginal and new measurements. I doubt the digital is programmed to compensate.

I'll post on room specifications later but basically it is an average size livingroom that breaks at a 90 degree angle into the dining room and kitchen area. The speakers are located along the short wall before entering the dining room area. Listening position is a chaise longue with the back approx. 11 feet from the speakers. The custom cabinets are on the floor approx. 3 feet apart. The Sansui speakers are on endtables with the bottom of the cabinet approx. 2 feet off the floor and 6 feet apart.

Hz Correction
10 20.5
12.5 16.5
16 11.5
20 7.5
25 5
31.5 3
40 2.5
50 1.5
63 1.5
80 1.5
100 2
125 0.5
160 -0.5
200 -0.5
250 0.5
315 -0.5
400 0
500 -0.5
630 0
800 0
1000 0
1250 0
1600 -0.5
2000 -1.5
2500 -1.5
3150 -1.5
4000 -2
5000 -2
6300 -2
8000 -2
10000 -1
12500 0.5
16000 0
20000 1

I just realized that calibrating a RS SPL meter or one of its knockoffs to 70dB @80 on the meter using a pink noise track and then running a log sweep tone track (20 to 20k) and manage to keep the entire track within the setting of the meter @70 should be good enough for this meter. Doing this I notice that I have a coloration problem between 80 and 90 Hz with the Realistic custom cabinets which spike up to 87dB. I verified this by going through the individual frequency test tones to pinpoint what frequencies were the problem. For the rest of the track the entire thing is between 60 and 80dB with 70dB as the midrange. The Sansui SP2500 manages to get through the entire track without going below 60dB or spiking above 80dB. Playing all four speaker results with the same spiking at 80 to 90 Hz. Correct the coloration at the low frequency range with the Realistic custom cabinets and I should be good to go. Not great but just good. Of course, I can't pick up any measurement below 25Hz and above 16kHz as described above, which is sort of pointless since it is a mid-fi system at best. BTW, these Sansui SP2500 are really cool loudspeakers. Especially when playing vinyl. I got a pair in mint condition on Ebay for a steal. I think these are the only loudspeakers Sansui put any effort into that came out with positive results.

Thanks for the input. I'll do some experimenting!

I have become a little better at this. I have also included the use of a better amplifier (HK3380). The TEAC AG640 was loud and that is just about all I can say about that piece of equipment.

The digital meter's advertised accuracy is +/- 2dB so I see no reason for using the various correction charts that are found plentiful on the Internet to compensate for the meter's known nonflat measurements. Instead, as a compromise between the measured SPL readings and the generic correction charts, I use an average of the current actual reading with the previous actual reading to be charted for frequency response averaged SPL. There is so much controversy about which Radio Shack SPL meter correction chart is best that averaging for this test is more practical due to the difference between each manufactured meter, frequency data spacing, and natural hearing.

After I had my data sets completed I calculated the average deviation as well as standard deviation for 160Hz through 16kHz based on the average SPL. Average deviation is a compromise between the standard deviation and a more acceptable weighted standard deviation since there are no acceptable weights for the Radio Shack SPL meter that can be weighted with each frequency tone measured. This is important to counterbalance data frequency spacing with natural hearing. I prefer the average deviation over the standard deviation because it is more of a natural deviation.

http://1982celica.illusionfxnet.com/Charts/SPLCharts3.html

Again, I'm experimenting and any input would be informative.

Eventually, I would like to move on to using the PC to plot a log sweep but that is later and not within the scope of practical and basic use of the RS digital SPL meter, which I am presently using it.

The only thing I have done as far as software is concerned is read up on the various programs. Seems to me the more variables you place between yourself and the meter the more likely the error rate will only increase giving false results. I haven't' actually used anything as of yet. As far as the RS meter is concerned it is a good tool to verify noticeable audio peaks and dips in a room and measuring basic frequency response. Generally, the flatter the response the better. I think Stereophile Magazine favorable reviews average about 91% for loudspeakers with a frequency response deviation of less than 3.0dB and about 63% for 3.0-5.0dB. A loudspeaker should have less than 5.0dB to begin with before considering things like room acoustics and so forth. I'm basically just reiterating what the editor of the magazine has published in the past concerning loudspeaker measurement. Especially, frequency response measurements. The flatter the response the more likelihood it will receive a favorable review upon further testing.

I forgot to mention that those percentages I was using earlier were for deviation for mid and treble response.

Also, here is a software program that details the importance of proper set up to measure response.

http://www.pensa.fr/freqresplot/indexe.htm

I know, it's a French website.

I have compared the results of using the digital meter and creating a manual frequency response chart with one created using a real time analyzer and found no material differences for deviation. I think the digital meter is more than appropriate for home use.

http://rssplmeter.1982celica.illusionfxnet.com/Sansui%20SP2500%20Charts.html

I used True Audio's RTA (TrueRTA) at its highest resolution for the comparison above. It's nice to have if you have it but I don't think it is a "must have". Also, if you don't have the software set up correctly your data is going to be flawed. For general home use I don't recommend wasting money on a RTA if all you need is reassurance that what you're hearing measures up. No pun intended. I also saved some money running the RTA by cheating a little by using it under WINE in SuSE Linux and using the Kmix software with the operating system. Otherwise, in addition to the RTA software you also will need a good sound card and maybe a mixer rather than the usual integrated type as well as a measuring mic other than the RS digital meter's. I haven't even mentioned the various cables you may need.

Thanks. I'll look into that one as well.

Quote:
Please critique my loudspeaker graphs. This is the first time I have run tests like this and I may need to run them again under other conditions. I realize this is ballpark measurements due to the measuring instrument but any feedback would be informative on the subject. Thanks.

http://1982celica.illusionfxnet.com/Charts/SPLCharts.html

Well, this is sort of weird. the above link is obsolete now. I determined that the Qtc of the custom Realistic cabinets was way to high, like over 1.12. This was causing a major spike around 100Hz and making the bass sound like one note below about 60Hz, so to speak. Anyway, I spent most of the day screwing around with calculations and how to compensate this high Qtc and out of whack F3 and Fb. I ended up tuning the box to a lower frequency at Fs and then added a smaller diameter port hole in the rear of the cabinet. This smoothed things out in the lower bass region somewhat. So, I took the RTA and calculated the average deviation for 50Hz-20Khz using a log sweep as well as a pink noise track. At 1/24 octave on the RTA this came to 251 different frequencies for each test. The average deviation for the log sweep and pink noise was 5.07 and 3.79 dB, respectively. I then got the average deviation for the Sansui SP2500, which was already tested using the pink noise that was 4.43 dB. I then averaged all three deviations and it came to 4.43 dB, the same as the Sansui alone. This is really strange. When I play all four loudspeakers at the same time I can't differentiate between the two sets any more. In other words, combined, all four sound like one set of two loudspeakers. This phenomenon was not occurring before all this messing around with the custom Realistic cabinets. Now, is this just bullshit or did I stumble on something here with all my dicking around? Separately, these two sets of loudspeakers sound different because one set is on the floor and other set is on stands. But together they sound as one. Very strange.