GRIMM Audio LS1c & SB1 DSP Speaker Review

[Pearljam5000]

I'm not. I've got technical questions right from the beginning, and I was hoping for some anwers. But it is always about the bass (see the bigger Revels for comparison) and the price. I understand the urge to question the price. I would rather ignore it (way beyond my paygrade). But the technicalities are in my range, so to say.

Then ask again maybe someone would answer
I just don't see how it contradicts other comments here

 

[Frank Dernie]

I wouldn't compare the Genelec with it's LF Box though because Genelec on it's own is already reaching deeper than the Grimm with LF Box according to the anechoic measurements, and the Genelec will also benefit by room reinforcement likewise.

Suit yourself, my reading of the data shows the Grimm goes lower and will need less manipulation in room, but keep those knickers twisting away

 

[gnarly]

This is wrong, me thinks:
"At first it seems most straightforward to measure the sound with a microphone. Indeed this can be done, however… a microphone will measure all sounds and not only the sound from the loudspeaker, like people talking, noises of cats and dogs, but also reflected sound via walls. Especially the latter are a problem as they come later and cause the feedback controller to correct something that is no longer there, a typical cause for instability. One may think that mounting the microphone inside the enclosure would solve this. Unfortunately however, ..."

In feedback there is a clean signal and a difference, the error. The f/b will attenuate the error. Outside noises, not contained in the signal, are corrected until they vanish, not beyond! In consequence the outside noise will be canccelled by the speaker, it acts like a 'black hole' for them, suck in and don't let it back out.

Meyersound found the same thing Grimm did...that servo control in their turn of the century X-10 studio monitor, reacted to sound beyond that of the direct signal.
Meyer said it caused unnatural sounds in recording studios. For example, a studio door being opened or closed was said to cause a lot of trouble.

The cancellations you envision require first the time it takes for the outside noise source to reach the sub. Noise which passes by ears on the way to the servo sub. The servo sub then reacts to produce it's own counter to the direct noise, making a second unwanted source of direct noise that's delayed in time vs the original noise.
The direct noise and servo reply cannot cancel each other out in the room at listening positions, being time separated. Bad ju-ju.

 

[Lars Risbo]

Boundary aka ground plane measurements:

The boundary acts as a mirror that generates a phantom mirror speaker on the other side of the boundary. This doubles the SPL generally, ie adds 6dB.

If we place the mic on the boundary (eg floor) then we have symmetry between the speaker and its mirror phantom speaker. This means again we are free of any cancellations due to the boundary.

However, if the speaker is placed straight up then a close boundary mic is off the speaker axis and the measured response will reflect that, ie droop with rising frequency. this droop will follow the baffle step initially as the speaker goes from omni to forward firing.

We can avoid the off axis droop by tilting the speaker downwards to get the mic on the speaker axis. With a preferred mic distance of at least 3 times the longest box dimension (and 2m in 2034 std) we will typically have to lift the speaker off the ground too. This allows an near perfect on axis measurement and the ground plane adds 6dB across all frequencies up to some frequency where the finite mic to boundary distance causes cancellations.

If we do this outdoors on a parking lot (no ceiling) and houses and cars are far away then we have a near anechoic measurement (just with 6dB gain).

Again for the speaker standing straight and the mic on the ground we get a droop versus frequency since the mic is off axis at higher frequencies. The closer the mic the more off axis it is.

Now if we lift the mic to eg the tweeter axis hight then we still get the plus 6dB at low frequencies but soon we have a near infinite notch due to the cancellation caused by the ground reflection and we have a series of periodic notches becoming less deep and lower peaks in between. This reflects that the ground reflection comes off axis and is attenuated as the frequency goes up. This gives a shelving down response with periodic notches. The shelving again follow the baffle step. A larger baffle moves the shelving corners down and the sound arriving to the listener is less influenced by the floor and other boundaries.

 

[sigbergaudio]

I will be measuring our Sarannas (that go pretty deep) in an anechoic chamber again on the 18th of June if anyone has any bass-in-an-anechoic-chamber theory they want me to test.

 

[Lars Risbo]

even very large anechoic chambers are only good down to about 50Hz

 

[sigbergaudio]

even very large anechoic chambers are only good down to about 50Hz

Yes, I'm not sure if anything is actually perfectly accurate / gives the right information this low. But we do have a comparison of this chamber (Seas) and an NFS for what it's worth. They're similar down to 50hz as you say. At 30hz the chamber is about 3dB lower than the NFS. At 25hz it's about 6dB lower than the NFS.

So if we assume the NFS is accurate, the chamber probably shows about 6dB too low at 20-25hz at least for that specific speaker that was used for the comparison.

When I've used it with subwoofers in the past it felt like it was even worse than that at least compared to a ground plane measurement. But I guess that makes sense.

So if a ground plane measurement is 6dB above a "true" or correct anechoic measurement (aka NFS if we trust that), that means the chamber is possibly up to 12dB off compared to a ground plane measurement, which I think is perhaps the best way we have to do a subwoofer/deep bass measurement.

 

[amirm]

even very large anechoic chambers are only good down to about 50Hz

Hence the reason Klippel advertises NFS even for anechoic chambers to get response down to 20 Hz.

 

[amirm]

So if we assume the NFS is accurate,

There is nothing to assume. Since bass frequencies are almost omni, it takes just a few low order basis functions to describe it. This means NFS computes this in its sleep.

There was an early issue where this assumption was taken too far when the bass drivers/ports were way below the tweeter but this was sorted out long time ago. And was not a weakness of the algorithm itself.

From memory, the NRC anechoic chamber loses that aspect at around 120 Hz. The Harman one is around 80 Hz. Calibrations are used to compensate but that creates error similar to the issue above for (old) NFS. That is, where the bass is coming from impacts the calibration.

The wedges you need to diffuse/absorb 20 Hz need to be massive and you are just not going to find that in any anechoic chamber.

 

[Lars Risbo]

NFS is of course not perfect. Nor is the LSI. Would be good to document its limitations. Probably better than many anechoic chambers.

 

[Lars Risbo]

There is nothing to assume. Since bass frequencies are almost omni, it takes just a few low order basis functions to describe it. This means NFS computes this in its sleep.

true the evaluation is simple but the identification of these few coefficients is numerically very sensitive.

 

[amirm]

NFS is of course not perfect. Nor is the LSI. Would be good to document its limitations. Probably better than many anechoic chambers.

We have done a number of comparisons against anechoic chambers to give us confidence as far as results (in good hands):

Neumann KH150 DSP: https://www.audiosciencereview.com/forum/index.php?threads/neumann-kh-150-monitor-review.39922/
NFS results:

And company's anechoic chamber measurement:

It doesn't get any better than this. I have done the same comparison with Genelec and results are nearly as perfect of a match.

 

[amirm]

true the evaluation is simple but the identification of these few coefficients is numerically very sensitive.

You obviously have to know what you are doing. Fortunately NFS self-checks its computed response against actual response and plots an error chart. Here are those results for Grimm LS1c:

As you see, before the tweeter takes over, the error is well below 1%. The loss of accuracy as you go toward 20Hz is actually not an NFS error but caused by the reduced output of the speaker.

The enemy of NS is complexity of sound field which occurs at higher frequencies. Here, the LS1c performs exceptionally well with the response being smooth and again, below 1% computed error. Had that gone up, I could have increased the measurement points to bring the error back down.

As an extreme example of that kind of error is the Magnepan LRS:

I always look at this graph first and will re-measure if needed. This occurs very rarely. But even when it does, the impact on on-axis tends to be small. Here is the Magenpan again:

In contrast, the computed vs predicted for LS1c land on top of each other (red and green):

We can also see the room impact as seen in the dashed blue. That is subtracted from the red, giving you the smooth response I posted in the review.

So unlike an anechoic chamber, we are not running blind here. We know the limitations of the system.

 

[amirm]

Adding on, the major advantage of an anechoic chamber is speed. If you own one, you can perform in a few minutes what takes a few hours with Klippel NFS. This is assuming you don't need the high resolution (1 degree) of Klippel NFS and can live with CE-2034 spec.

Klippel NFS is however superior to outdoor measurements due to avoidance of temperature gradients causing phase shifts and such.

 

[thewas]

We have done a number of comparisons against anechoic chambers to give us confidence as far as results (in good hands):

Neumann KH150 DSP: https://www.audiosciencereview.com/forum/index.php?threads/neumann-kh-150-monitor-review.39922/
NFS results:

And company's anechoic chamber measurement:

It doesn't get any better than this. I have done the same comparison with Genelec and results are nearly as perfect of a match.

From what I know Neumann also doesn't have or use a huge anechoic chamber for the lower bass region but the typical methods like outside or ground plane measurement and more recent even the NFS service from Klippel and I wouldn't be surprised if it would be very different for Genelec. Still like you say the NFS has long proven its reliability also in that region compared to other methods.

 

[a4eaudio]

even very large anechoic chambers are only good down to about 50Hz

Have you worked with the Mic-in-the-box method?
Information seems to be pretty limited on the "typical" forums, but it is well-enough known to be included in REW and VituixCAD.

 

[Heinrich]

Meyersound found the same thing Grimm did...that servo control in their turn of the century X-10 studio monitor, reacted to sound beyond that of the direct signal.
Meyer said it caused unnatural sounds ...
The cancellations you envision require first the time it takes for the outside noise source to reach the sub. Noise which passes by ears ...

Won't recap feedback theory - I really tried to boil it down, but envisioned that it won't be appreciated here. Only one point. As I said, the system's performance is limited by the microphone. If it starts to compress its output due to overload, the system as a whole gets into run-away mode, the limited output capability of the speaker (as the actuator) completes the mess..

Otherwise, your take on why it won't work is flawed. The room is not part of the feedback loop; one can formulate a theoretical contribution, though, so please go ahead. I speak maths pretty well (better than English, actually) ;-)

 

[Heinrich]

Have you worked with the Mic-in-the-box method?
Information seems to be pretty limited on the "typical" forums, but it is well-enough known to be included in REW and VituixCAD.

On bad bass (courtesy of working review index):

https://www.audiosciencereview.com/forum/index.php?attachments/revel-f328be-cea-2034-spinorama-frequency-response-measurements-png.92621/

and

https://www.audiosciencereview.com/forum/index.php?attachments/revel-f208-twoer-speaker-spinorama-cea2034-frequency-response-audio-measurements-png.130626/

Compare to Grimm's - the Revels are not known for being bass shy.

Regarding the ripples in midband, I withdraw my question. I don't ask to switch positions of tweeter and bass anymore. It would have helped with a listener standing up, walking around, see ceiling bounce. I can't afford the Grimm's anyway. Nice testbed.

 

[NTK]

Won't recap feedback theory - I really tried to boil it down, but envisioned that it won't be appreciated here. Only one point. As I said, the system's performance is limited by the microphone. If it starts to compress its output due to overload, the system as a whole gets into run-away mode, the limited output capability of the speaker (as the actuator) completes the mess..

Then the mic is not fit for the job. Use a better mic. How is it not predicted during the design process?

Otherwise, your take on why it won't work is flawed. The room is not part of the feedback loop; one can formulate a theoretical contribution, though, so please go ahead. I speak maths pretty well (better than English, actually) ;-)

Then please explain what the subwoofer is supposed to do, when the disturbance (e.g. door slam) was already heard by the listener before it was picked up by the mic (since the listener sits in between the door and the subwoofer).

The subwoofer must do nothing in response to the door slam, because it can't reverse history. So please explain how the feedback control system figures it out.

 

[kimmosto]

NFS results looks systematically unreliable below 40 Hz if some LF radiators such as ports are in rear panel. For example, DI could jump to -5 dB without any possibilities to do that in real life.