|
|
Post by marcl on Sept 8, 2024 10:12:40 GMT -5
The drawing on the right does look like a more concise wiring scheme. If it was all on one side it would be easy the other way, IF I mounted the connector facing downward so wires only had to go up and not backtrack.. Having the connector facing up was done to facilitate testing where I'll be disconnecting the servo and drive wires at various times. In the end it is a very clean implementation with wires neatly going through the rubber grommets and all dressed on one side of the back. I'll use the nylon sleeve to dress them better after testing is done. |
|
|
|
Post by marcl on Sept 8, 2024 12:12:05 GMT -5
Started the bench test. So far so good! I'm using a miniDSP2x4HD as audio interface for my laptop, and measuring with REW and UMIK-1 at one meter. Driving both L/R inputs to the plate amp at a test level of 85dbC. Patio door is open to reduce axial resonance in that direction. I'll coagulate and disambiguify results later ... but if you look close you can see the astounding difference in impulse response between the dashed line with no servo, and the solid line with servo.  |
|
ttocs
Global Moderator 
I always have a wonderful time, wherever I am, whomever I'm with. (Elwood P Dowd)
Posts: 8,168
|
Post by ttocs on Sept 8, 2024 13:38:01 GMT -5
... but if you look close you can see the astounding difference in impulse response between the dashed line with no servo, and the solid line with servo. That's a great demonstration of what servo can do! I've not seen another comparison using Impulse, which is a great way to show what's going on. Usually, folks just show frequency response which doesn't completely show what's going on. |
|
|
|
Post by leonski on Sept 8, 2024 14:45:32 GMT -5
Any way to do measurements OUTDOORS? With a large enough area, you could simulate anachoic space....Raise the subs UP as much as practical.....
I'm certain that any neighbors within earshot would be impressed and want to help!
|
|
|
|
Post by marcl on Sept 8, 2024 15:01:31 GMT -5
Okay, I will likely do some more testing including the outdoor ground plane test (no room resonances) but this pretty much tells the story. Test Conditions: - Room: 15'x11'x8'
- Speaker position: 6' from front wall. 4' from left side wall, centered 4' above the floor
- Mic position: 3' in front of speaker
- miniDSP2x4HD as audio interface feeding both L/R inputs of the plate amp
- Test Level: 75dbC
First, note the response in the Room Simulation showing the major axial resonance at 37Hz. This is evident especially in the response with Servo disconnected.  Next the Frequency Response. I tested various combinations and decided that I would leave the extension filter at 14Hz. I did the sweep with Servo disconnected (dashed pink) and then connected the Servo and tested Hi, Med, Lo Damping (pink, green, blue respectively). I was surprised to see the effect the Servo had on the resonance peak. In fact I raised the level of the measurements with Servo to get a more realistic comparison. As you can see less damping results in somewhat higher output below 100Hz. Not as much difference between Hi and Med as there is between Med and Lo. Attachment DeletedAnd the very important Impulse Response! Similar results as Frequency Response. Has anyone seen a woofer - let alone a sub - respond like this?   I also tried enabling the 10Hz Rumble Filter. The effect on Frequency Response was okay ... rolling off below 10Hz without sacrificing response down to 16Hz. BUT ... it really screwed up the Impulse Response, so I won't be using the Rumble Filter. The only other controls allow you to select 20Hz or 28Hz as the extension filter rather than 14Hz. They allow the response to roll off as you would expect. I see no reason to set it at anything other than 14Hz. And here's THD for no Servo and Hi Servo (bright pink, dark pink respectively). The Servo trace is a little lower below 40Hz and a little higher above 40Hz .... but overall values are mostly under 2% and no higher than 3% ... which is MUCH lower than my Outlaw subs ever measured.  All of this is really better than I would have hoped for especially the response below 20Hz! |
|
|
|
Post by marcl on Sept 9, 2024 12:09:14 GMT -5
After talking with ttocs this morning I reran the measurements just with the Servo/Damping connected or disconnected (there is no bypass switch on the amp ... just Hi, Mid, Lo). Basically the same results but with levels more normalized and controlled. The issue with levels is that if I have Servo/Damping connected and set the level with pink noise to 75dbC, when I disconnect the Servo/Damping the level goes up to 92dbC. I don't know why this is the case. But for measurement purposes I used 75dbC for both sets of measurements and adjusted the level in the miniDSP. Red - Servo/Damping OFF Pink - Servo/Damping On Blue - Servo/Damping On + 10Hz Rumble Filter Looking at the frequency response and distortion I'd be inclined to leave the Rumble Filter on. But I don't like what it does to the impulse response.      |
|
KeithL
Administrator
Posts: 10,274
|
Post by KeithL on Sept 9, 2024 13:16:07 GMT -5
I can actually make a guess about your level discrepancy. In general most "servo" things work by applying negative feedback. And, if you "switch off" or "reduce" negative feedback, without changing anythi8ng else, then the gain goes up. (I would guess that, when you switch it on, the feedback that applies the "servo correction" is just added to whatever NFB is already there.) (But, of course, it should be easy enough to add some attenuation somewhere to kill off that extra gain.) After talking with ttocs this morning I reran the measurements just with the Servo/Damping connected or disconnected (there is no bypass switch on the amp ... just Hi, Mid, Lo). Basically the same results but with levels more normalized and controlled. The issue with levels is that if I have Servo/Damping connected and set the level with pink noise to 75dbC, when I disconnect the Servo/Damping the level goes up to 92dbC. I don't know why this is the case. But for measurement purposes I used 75dbC for both sets of measurements and adjusted the level in the miniDSP. Red - Servo/Damping OFF Pink - Servo/Damping On Blue - Servo/Damping On + 10Hz Rumble Filter Looking at the frequency response and distortion I'd be inclined to leave the Rumble Filter on. But I don't like what it does to the impulse response. View AttachmentView AttachmentView AttachmentView AttachmentView Attachment |
|
KeithL
Administrator
Posts: 10,274
|
Post by KeithL on Sept 9, 2024 13:21:38 GMT -5
Considering the wavelengths of the sound a subwoofer will be making I don't see any issue with "dimensional resonances"... So what you would really be talking about would be "plain old vibration" or some cabinet resonance excited by it. Your cabinets are quite heavy, and heavily braced, so I doubt this would be much of an issue. Dynamat is mostly used to prevent back wave sound from the drivers from bouncing around inside the cabinet... Or from making the cabinet walls vibrate... Neither of which should be a serious issue for you.  (If anything maybe consider adding some black putty along the seams where the panels meet... but I doubt it would make any difference. ) I would be looking at some sort of modern solid state accelerometer... Most ceramic phono cartridges actually work by the pressure on the stylus FLEXING a strip of piezoelectric material. (So they rely on applying sideways force to the strip while the other end is held firmly in place... and they have very little range of motion.) Another commonly used strategy is to use a dual-voice-coil driver... And use the second voice coil as a "pickup"... (Some custom drivers have a specially designed "sensor coil".) Otherwise note that the sensor MUST be connected to the center of the driver. ALL woofer cones flex and resonate to some degree. The center is going to be the best place to actually model the position of the voice coil and so of the cone overall. (And you can glue a sensor to the center of a solid dust cap without affecting balance.) If you really want to have "fun" with modeling and programming... You can also use the back-EMF of the driver's main voice coil to "track what it's doing". But that gets really complex... Another really interesting option would be to use a tiny microphone... (The new solid state microphones used in things like smart phones and Alexa's are really tiny and very rugged.) A microphone measures the movement of air relative to the microphone... If you were to mount a tiny microphone on the cone, with the microphone "moving and detecting the air pressure in front of the cone"... IN THEORY the microphone would be giving you a really accurate indication of the actual power being coupled to the air at that point... (It's telling you the actual air pressure at that point at the front surface of the cone.) But, since this would vary across the cone, and such complications, it would be really tricky... and I've never heard of anyone doing it. But, of course, don't forget that you have to get that signal back to your electronics... (So an extra set of flex cables... ) Or... for a real science-fiction idea... How about a self-powered piezoelectric sensor... With an electronic package that digitizes the readings... And sends them back through the drive wires that power the voice coil. (You could modulate them on a high frequency carrier... like "Ethernet over power" works.) And then, of course, there's LASER interferometry... which works non-contact...  Another interesting thought is this... With an open baffle sub the actual audio signal has a very long wavelength... Therefore, having a sensor a few inches away from the surface of the driver might actually give you BETTER information about the audio wavefront itself. So how about mounting a microphone on the baffle, to the side, several inches behind the driver. Or even on a support strut, going across the cabinet, directly behind, and five or six inches away from, the center of the driver... This would show you what the actual waveform IN THE AIR looked like. (Remember that the goal is to accurately reproduce the audio waveform... and NOT specifically the movement of the driver.) And, since it's only a tiny fraction of a wavelength away from the driver... You should be able to correct the driver to compensate for variations in the air-borne wavefront... Hmmmm... that might actually work pretty well at very low frequencies. Wow this is quite a bit of interesting stuff (as usual). I'm going to read again in detail after I finish driving the last four bolts to mount the third driver. Just to give you some info on this specific build though .... the drivers are GRResearch 12" 16ohm servo speakers with indeed a servo coil as well as the drive coil. They will be driven by a Rythmik servo amp. Both amp and speaker are specifically made for open baffle use, with the amp incorporating the required shelf filter (internally) in addition to the user adjustable filter and damping controls. But the specific conversation about resonance and damping is not about the driver itself, it's about the cabinet and the debate over whether the recommended Norez material does anything if applied to the inner surfaces of each square cavity. So what I want to know is ... IS there a cabinet resonance? IS it significant in a range that will be excited by the drivers (16-120Hz)? And IF so ... can Norez or some other material like Dynamat significantly dampen that resonance? My hypothesis - partially confirmed by the tapping and calculation exercises earlier in this thread - is that there is no significant resonance. And if there was, the foam part of Norez would do nothing below 500Hz so it would be a matter of the Norez damping layer vs Dynamat or MLV. I think the only way to tell from here is the ground plane measurement of the three drivers playing in the cabinet and measured from 2m away. That's probably a couple days away. View Attachment |
|
|
|
Post by marcl on Sept 9, 2024 14:50:47 GMT -5
I can actually make a guess about your level discrepancy. In general most "servo" things work by applying negative feedback. And, if you "switch off" or "reduce" negative feedback, without changing anythi8ng else, then the gain goes up. (I would guess that, when you switch it on, the feedback that applies the "servo correction" is just added to whatever NFB is already there.) (But, of course, it should be easy enough to add some attenuation somewhere to kill off that extra gain.) After talking with ttocs this morning I reran the measurements just with the Servo/Damping connected or disconnected (there is no bypass switch on the amp ... just Hi, Mid, Lo). Basically the same results but with levels more normalized and controlled. The issue with levels is that if I have Servo/Damping connected and set the level with pink noise to 75dbC, when I disconnect the Servo/Damping the level goes up to 92dbC. I don't know why this is the case. But for measurement purposes I used 75dbC for both sets of measurements and adjusted the level in the miniDSP. Red - Servo/Damping OFF Pink - Servo/Damping On Blue - Servo/Damping On + 10Hz Rumble Filter Looking at the frequency response and distortion I'd be inclined to leave the Rumble Filter on. But I don't like what it does to the impulse response. View AttachmentView AttachmentView AttachmentView AttachmentView AttachmentThanks Keith! Yes my other friends from EE school were saying he same thing and one postulated that 15-20db of negative feedback seemed reasonable and that would account for the difference. I don't know if it makes a difference in the way the servo circuit works, but there is no bypass switch so the only way I can test without the servo engaged is to pull off the sensor wires from the speakers. So you're saying with no feedback, the servo would increase the gain? |
|
|
|
Post by marcl on Sept 9, 2024 15:03:19 GMT -5
Considering the wavelengths of the sound a subwoofer will be making I don't see any issue with "dimensional resonances"... So what you would really be talking about would be "plain old vibration" or some cabinet resonance excited by it. Your cabinets are quite heavy, and heavily braced, so I doubt this would be much of an issue. Dynamat is mostly used to prevent back wave sound from the drivers from bouncing around inside the cabinet... Or from making the cabinet walls vibrate... Neither of which should be a serious issue for you. (If anything maybe consider adding some black putty along the seams where the panels meet... but I doubt it would make any difference. ) Wow this is quite a bit of interesting stuff (as usual). I'm going to read again in detail after I finish driving the last four bolts to mount the third driver. Just to give you some info on this specific build though .... the drivers are GRResearch 12" 16ohm servo speakers with indeed a servo coil as well as the drive coil. They will be driven by a Rythmik servo amp. Both amp and speaker are specifically made for open baffle use, with the amp incorporating the required shelf filter (internally) in addition to the user adjustable filter and damping controls. But the specific conversation about resonance and damping is not about the driver itself, it's about the cabinet and the debate over whether the recommended Norez material does anything if applied to the inner surfaces of each square cavity. So what I want to know is ... IS there a cabinet resonance? IS it significant in a range that will be excited by the drivers (16-120Hz)? And IF so ... can Norez or some other material like Dynamat significantly dampen that resonance? My hypothesis - partially confirmed by the tapping and calculation exercises earlier in this thread - is that there is no significant resonance. And if there was, the foam part of Norez would do nothing below 500Hz so it would be a matter of the Norez damping layer vs Dynamat or MLV. I think the only way to tell from here is the ground plane measurement of the three drivers playing in the cabinet and measured from 2m away. That's probably a couple days away. View AttachmentAgreed. Most everyone who builds these things and takes a picture has the Norez in there. But I think it's just very easy to convince most people that foam and asphalt will do something .... and so it's just Danny doing marketing ... very successfully! The off-axis ground plane test will tell me a lot. But here are two measurements taken with servo on Hi and 14Hz extension filter. Darker pink the speaker is on a stand 3ft off the floor. Brighter pink it's in the same spot but on the floor. Measured 3ft away. Interesting how abruptly the response changes right below 40Hz which is the fundamental resonance of the length mode of the room. Beyond that ...100Hz is the next significant peak, and that is the fundamental resonance of the width mode of the room.  |
|
|
|
Post by marcl on Sept 9, 2024 15:05:12 GMT -5
I can actually make a guess about your level discrepancy. In general most "servo" things work by applying negative feedback. And, if you "switch off" or "reduce" negative feedback, without changing anythi8ng else, then the gain goes up. (I would guess that, when you switch it on, the feedback that applies the "servo correction" is just added to whatever NFB is already there.) (But, of course, it should be easy enough to add some attenuation somewhere to kill off that extra gain.) After talking with ttocs this morning I reran the measurements just with the Servo/Damping connected or disconnected (there is no bypass switch on the amp ... just Hi, Mid, Lo). Basically the same results but with levels more normalized and controlled. The issue with levels is that if I have Servo/Damping connected and set the level with pink noise to 75dbC, when I disconnect the Servo/Damping the level goes up to 92dbC. I don't know why this is the case. But for measurement purposes I used 75dbC for both sets of measurements and adjusted the level in the miniDSP. Red - Servo/Damping OFF Pink - Servo/Damping On Blue - Servo/Damping On + 10Hz Rumble Filter Looking at the frequency response and distortion I'd be inclined to leave the Rumble Filter on. But I don't like what it does to the impulse response. View AttachmentView AttachmentView AttachmentView AttachmentView AttachmentAnd this just in from Brian Ding at Rythmik: "Servo feedback is “negative” feedback. So if one disconnects the servo feedback, the “gain” of the amp will be higher. This is the concept of “open loop gain” vs “close loop gain”. If disconnecting servo does not give you any difference, then that servo feedback is a gimmick. Without servo, the damping of the system is also very LOW, or the Q value is going to be very high. In your REW, you will find a huge peak at around 20hz." |
|
KeithL
Administrator
Posts: 10,274
|
Post by KeithL on Sept 9, 2024 16:57:13 GMT -5
That's going to depend on exactly how they implemented the servo circuitry. From the way you describe it " the only way I can test without the servo engaged is to pull off the sensor wires from the speakers"... When you do that you are basically removing the input from the servo circuitry... But the servo circuitry itself is still "connected and idling"... In general ALL modern amplifiers work by way of negative feedback. At a very simple level, the description of a modern amplifier would be very similar to the description of a servo. And you can think of the way negative feedback controls the gain of an amplifier as "a servo control on the output"... (Some fraction of the output is compared to the input and the output is then "corrected as necessary" so that they match.) With a normal amplifier there is one, or sometimes more, electrical paths from the output to the input, by which that NFB is applied... What they've done is to add another path that includes the driver itself by way of that sense coil... So now, at least to some degree, the motion of the driver itself is being compared to the input signal, and a correction signal is being generated... And the simplest way to apply that would result in a lowering of the overall gain when the correction signal is applied... (In general you don't ever use positive feedback because that generally results in instability - a really loud buzzer.) I can actually make a guess about your level discrepancy. In general most "servo" things work by applying negative feedback. And, if you "switch off" or "reduce" negative feedback, without changing anythi8ng else, then the gain goes up. (I would guess that, when you switch it on, the feedback that applies the "servo correction" is just added to whatever NFB is already there.) (But, of course, it should be easy enough to add some attenuation somewhere to kill off that extra gain.) Thanks Keith! Yes my other friends from EE school were saying he same thing and one postulated that 15-20db of negative feedback seemed reasonable and that would account for the difference. I don't know if it makes a difference in the way the servo circuit works, but there is no bypass switch so the only way I can test without the servo engaged is to pull off the sensor wires from the speakers. So you're saying with no feedback, the servo would increase the gain? |
|
|
|
Post by leonski on Sept 9, 2024 23:08:51 GMT -5
Can the enclosure be thought of as a very short tube? Any advantage to making the shelves Wider?
|
|
|
|
Post by marcl on Sept 10, 2024 2:06:22 GMT -5
Can the enclosure be thought of as a very short tube? Any advantage to making the shelves Wider? Interesting leonski ... always "thinking outside the box"  So in the sense that a ported sub works that way with a tuned tube port .... yes, theoretically. But once the "port" becomes the size of one side of the enclosure, and at the wavelengths involved here (9ft to 70ft) ... no. You won't get a pipe organ-like note playing when the air is pushed out of the "tube" by the speaker. At least not at a frequency or amplitude that you could hear. "Port chuffing" is a thing with ported subs though, because the pressure is pretty high and the tube relatively narrow. That's why they taper and round the exit of the tube. But ... the size of the "shelves" is actually the baffle size. And with an open baffle design the baffle size is very important. Ideally - for the best low bass extension - you would have no shelves ... just a really big flat surface extending several feet in all directions around the driver. That being impractical, we fold the baffle into various shapes like the W or the slot designs ... or the H-frame used here. So if the shelf was wider or deeper it would create an effective baffle size that is bigger. That would lower the rolloff frequency of the design. The current effective baffle size is 28". |
|
|
|
Post by marcl on Sept 10, 2024 2:20:38 GMT -5
So for those following along ... I want to emphasize something from a couple posts up regarding how the servo works and its effect on the measurement results.
Because the servo function can't actually be defeated, it's always trying to work even if the sense coil is disconnected from the amp. The result of this is that with this amp it's impossible to get an accurate measurement of the sub without servo control to compare the response with servo control. With the sense coil disconnected the feedback loop in the amp is compelled to increase gain, causing exaggerated overall amplitude and very high peaks at the resonances of the room.
I think it's still valid for comparing impulse response, but the impulse response without servo control probably looks a bit worse than it would be with a non-servo amp. So ... when I do the ground plane test outdoors this week, I intend to include a test with a regular non-servo amp for the purpose of comparing the impulse response. But even that won't be completely apples:apples because the regular amp won't have the shelving filter required to compensate for the dipole cancelation.
BUT ... since I'm driving the amp from a miniDSP, I can add a low shelf to the output and adjust its parameters until the output of the amp matches the shape of the output of the servo amp. I believe the shelf should start around 100Hz and with a slope of around 6db/octave .... then I'll tweak from there. So then I'll know the answer to the question of what the shelving filter that Rythmik adds to the OB version of the A370 amp actually looks like.
|
|
|
|
Post by marcl on Sept 10, 2024 8:39:53 GMT -5
LAWNMOWERS! 
|
|
|
|
Post by marcl on Sept 11, 2024 16:17:30 GMT -5
Among the reasons to use an open baffle dipole sub is that the room resonances are not engaged as much as a conventional sealed or ported sub. With sealed or ported the pressure wave is omnidirectional, bouncing off all walls, ceiling and floor. And those resonances can add up either constructively or destructively. With the open baffle dipole the room modes to the sides and top and bottom of the sub are not engaged as much, due to dipole cancelation. Here is an illustration. I measured at 3ft using a regular amplifier with no servo and no shelving filters to compensate for the dipole cancelation. I had the sub pointed in the direction of the 15ft length of the room (orange trace). Then I turned the sub 90 degrees so it was playing across the 11ft width of the room (green trace). Resonances of the length mode are 37.5, 75, 112.5, 150. For the width modes 51, 102, 153. The ceiling is 8ft, and its modes are 70.3 and 140.6. You can see the modes illustrated at the bottom of the graph and they correspond to the trace colors.  Indeed, with the speaker pointed in the length orientation there is a peak at 37.5 , but not at 51. With it pointed in the width orientation the peak is at 51 but not 37.5. Beyond that the modes start to stack and at the microphone can add up either positive or negative. As we would expect, the floor/ceiling mode is the same for both orientations but there's actually a cancellation in both measurements around 140. So then I did a series of measurements as close as I dared to the cone (given the long travel at low frequencies). Here again with the regular amp with no servo and no shelving filter compensation, you see a very flat response from this speaker right down to just below 30Hz, where it rolls off 12db/octave. And this is as expected because the Fs of the driver is ... 28.3Hz!   Next week sometime when I finally pull the Outlaw subs out of the listening room and install the dipoles, I will put the 12" Outlaw sub in the den in the exact same spot and measure to see how it excites the room modes. SCIENCE! |
|
|
|
Post by marcl on Sept 12, 2024 13:33:57 GMT -5
The proverbial "natural response of the speaker" ... well ... this is as close to anechoic as I'll ever get. Ground plane test at 6ft. I'm in the dipole null with my house 20ft behind me. To the left and right there's virtually nothing close enough to reflect back.  16-300Hz, 10Hz rumble filter on; Hi, Med, Lo damping settings (light, med, dark blue). No smoothing at all. THD no more than 1% @40hz. THD peaks at 16%, 13%, 11% @16hz for Lo, Med Hi damping respectively.   I'm done with the theory, now to build up the other one and get them in the room! |
|
ttocs
Global Moderator
I always have a wonderful time, wherever I am, whomever I'm with. (Elwood P Dowd)
Posts: 8,168
|
Post by ttocs on Sept 12, 2024 13:51:40 GMT -5
The proverbial "natural response of the speaker" ... well ... this is as close to anechoic as I'll ever get. Ground plane test at 6ft. I'm in the dipole null with my house 20ft behind me. To the left and right there's virtually nothing close enough to reflect back. View Attachment16-300Hz, 10Hz rumble filter on; Hi, Med, Lo damping settings (light, med, dark blue). No smoothing at all. THD no more than 1% @40hz. THD peaks at 16%, 13%, 11% @16hz for Lo, Med Hi damping respectively. View AttachmentView AttachmentI'm done with the theory, now to build up the other one and get them in the room! This proves it. We should bring our systems outside for the best frequency response! No correction needed! |
|
|
|
Post by marcl on Sept 12, 2024 13:57:29 GMT -5
The proverbial "natural response of the speaker" ... well ... this is as close to anechoic as I'll ever get. Ground plane test at 6ft. I'm in the dipole null with my house 20ft behind me. To the left and right there's virtually nothing close enough to reflect back. View Attachment16-300Hz, 10Hz rumble filter on; Hi, Med, Lo damping settings (light, med, dark blue). No smoothing at all. THD no more than 1% @40hz. THD peaks at 16%, 13%, 11% @16hz for Lo, Med Hi damping respectively. View AttachmentView AttachmentI'm done with the theory, now to build up the other one and get them in the room! This proves it. We should bring our systems outside for the best frequency response! No correction needed! But what's really wild is that measured from 3ft in an untreated 11x15x8ft room with no correction other than what's built into the amp for the dipole ... it looks like this!  |
|
