Harman/Kardon Onyx Studio 8 vs Onyx Studio 9

Aug 26, 2026 03:44 PM By Bashar

120

Its said “Be careful what you wish for, it may come true” :) I was criticizing some of the calculations and tests before but with new test bench, things getting even more complicated :( Now there are several different results and calculations with test v0.8, v1.0 and v1.0.1 in site and very hard to compare different speakers with different test bench results. And unfortunately so many inconsistent test results in new test bench. I have few points to mention for future: 1)There are some problems in tests of the speakers which have “room correction feature” as its commented here about Onyx 8 and 9. Its worth to test all speakers (with this feature) again after you place them in test point and use them for awhile before starting the test. (As an example, I use HK GoPlay Play3 time to time and its previous Std Error figures were 2.15 and reduced to 3.26 with new test bench which seems not so reliable) 2)New Loudness tests and weight of Max SPL seems a bit over rated especially for daily use. Let’s take Tribit Stormbox Blast for instance. It has 101.3dB Max SPL and 4.11 DRC Max. HK GoPlay3 has 98.3 Max SPL and 2.98 DRC Max. Tribit has 8.4 Loudness rating while HK Go Play has 7.8. In daily life most of the people would prefer to listen music in 98.3dB with 2.98dB DRC, rather than 101.3db with 4.11dB DRC. For future, maybe better to test DRC@90dB (like in Soundbars) and increase its weight more than 10% in Loudness ratings. In real life usage, there is not a huge difference between 98.3dB and 101.3dB because most of the time people listen music between 80 to 95 decibels. That’s why in reality, the difference between 85db and 88dB is not the same as 99dB and 102dB.

Thanks for the feedbacks Bashar. Normally all 1.0 speakers should have been brought up to 1.01, and the full batch of retest is completed, that is 96 speakers! Is there a specific model that you see still on 0.8 that is of interest to you?

For automatic room correction, normally we play Pink Noise for at least 10 minutes before measuring the speakers. We will later in the year update the test bench so hopefully we can meet more of your expectations. That said 1.1 dB difference in error to a target will always be within tolerance of speaker measurements in a semi reflective room. We can make it better but it cannot be precise to 1 dB, I would not be honest to promise that.

For the Loudness, It is also planned to update this test down the line. I think I andwered to an other comment, but the main difference you see is because we measure at 1m instead of 2.

In all cases, we’ll take good notes and hopefully will adress your concerns.

Best.

Its said “Be careful what you wish for, it may come true” :) I was criticizing some of the calculations and tests before but with new test bench, things getting even more complicated :( Now there are several different results and calculations with test v0.8, v1.0 and v1.0.1 in site and very hard to compare different speakers with different test bench results. And unfortunately so many inconsistent test results in new test bench. I have few points to mention for future: 1)There are some problems in tests of the speakers which have “room correction feature” as its commented here about Onyx 8 and 9. Its worth to test all speakers (with this feature) again after you place them in test point and use them for awhile before starting the test. (As an example, I use HK GoPlay Play3 time to time and its previous Std Error figures were 2.15 and reduced to 3.26 with new test bench which seems not so reliable) 2)New Loudness tests and weight of Max SPL seems a bit over rated especially for daily use. Let’s take Tribit Stormbox Blast for instance. It has 101.3dB Max SPL and 4.11 DRC Max. HK GoPlay3 has 98.3 Max SPL and 2.98 DRC Max. Tribit has 8.4 Loudness rating while HK Go Play has 7.8. In daily life most of the people would prefer to listen music in 98.3dB with 2.98dB DRC, rather than 101.3db with 4.11dB DRC. For future, maybe better to test DRC@90dB (like in Soundbars) and increase its weight more than 10% in Loudness ratings. In real life usage, there is not a huge difference between 98.3dB and 101.3dB because most of the time people listen music between 80 to 95 decibels. That’s why in reality, the difference between 85db and 88dB is not the same as 99dB and 102dB.

Aug 11, 2026 07:23 PM By RtingsUser9403519

3

It’s difficult to believe that the bass extension was reduced from 46hz to 109hz. I don’t think I really ‘believe’ 46hz (it is the result of an intermittent peak), but I don’t think I believe 109hz either.

I agree, with you that the new version don’t reproduce a whole octave lower. To be honest, there are no standardized calculation of what constitute low frequency extension. Our old measurements had quite severe room interractions in the bass that made it difficult to trust assessment of extension. In the new test bench we have been able to comppensate for most of these but in the case of speakers with room correction,which the Onyx 9 have we measure at 1m to see the behaviour, basically, not removing the room modes since the speaker should compensate for them. It is possible that your experience at home could differ.

As for The LFE calculation, the shape of the response for this specific speaker may push the extension above what would be subjectively audible. For full transparency, we calculate the extension at the point in the roll off where the Bass response drop to lower than 70 dBSPL (or 6 dB below our reference Level). In some regards, it is fair to say that this speaker “tricked” our algorithm, since as you can see on the graph, it does fall below at 109 Hz, but then there’s a low frequency bump at 53 Hz, which altough technically in the rolling off part of the response still gives significant energy below our published Low Frequency extension.

I would say that you are right to not trust a single number like that, the whole curve tells a different story. It is challanging to have a number all would agree with when the roll off in the response is not a smooth continuous slope.

Hope that help and thanks for flagging this. Frequency response should not be consodered an exact universal metric. There are some quirks and it can be seen in looking at the individual responses (the grey traces), there is already large difference depending on the measurement microphone positions (we use a 15 microphones array).

It’s difficult to believe that the bass extension was reduced from 46hz to 109hz. I don’t think I really ‘believe’ 46hz (it is the result of an intermittent peak), but I don’t think I believe 109hz either.