Amplifier Group Delay Affecting Measurement Accuracy

I recently bought some new speakers and decided to run frequency response plots on them. The plots were showing an 18-22dB level drop off from 11KHz to 20KHz. (See screenshot) I had run plots full range with my previous 4-way speakers over the past 12 years with with no high frequency rolloff. I experimented with 256K,512K and 1 MB sweeps with 48K and 96K bit rates with no change except for a few minor bumps in the high frequency range.. I then hooked up a current probe and voltage probe to the 465B scope along with an AC meter. All inputs to the to the speaker looked fine and the Radio shack spl meter did not show any high frequency speaker dropoff. I was using the REW Generator for steady state measurements. I have a Motu 2 audio input with one output going the the system preamp. I hooked up the second output to the scope that was measuring the signal into the speaker. I was able to measure the speaker amp input delay as approximately 15uS at 10KHz and 25uS at 20KHz. The speaker input delay is enough to cause the 18-22dB drop (as 25uS will put the speaker input 180° out of phase @20kHz with the output of the second Motu 2 output. The reason for the lack of dropoff in the old speakers is the RTR-6 tweeters were driven by a direct coupled tube amp without group delay.

I am surmising that the amplifier group delay amount increases as frequency increases. As the mic distance is an unknown variable I assume REW measures the first few cycles to get a time constant and keeps it during the sweep

I have some questions:

1. Is this group delay normal in most amplifiers? Mine is a Parasound A-21 400 watt

2. Is there anyway to compensate for this group delay issue in REW? I read the manual but did not see any mention of group delay.

Hi John, Thx for the quick response. I have an amplifier with none/minimal group delay that measures high frequency speaker signals with no problems. Measuring the the new speaker with a sound level meter using the signal generator shows the speaker is not drastically rolling off the high frequencies as per the screenshot. Doesn't the REW sw lock on the initial signal when it reaches the measuring microphone?

How do I do this? That can be dealt with by using the acoustic timing reference to synchronise the capture.

Thx for your input. As it rules out amplifier/signal chain variation it leaves the size of the speaker/transducer size as the main variable. The new speaker has at least 16 times the radiating area of the of the RTR tweeters that measured flat. The speaker has 7 vertical columns segmented in a 45° total angle. AS the column width is approximately 7 time the 20KHz wavelength. I expect the polar response to be quite narrow. I do get flat response above 10KHz using stepped sine waves and measuring with sound level meter. I do plan to run some more swept tests to see if I can resolve high frequency dropoff.