It is difficult to say anything in particular about the OLTF without the phase plot in plus and minus 180 degrees. The shape of the loop itself looks ok.
If the phase and minus 135- 140 degree points in the phase plots can be established, closed loop and the the amount of feedback possible can be guesstimated, just like Bolstert has shown.
If the phase and minus 135- 140 degree points in the phase plots can be established, closed loop and the the amount of feedback possible can be guesstimated, just like Bolstert has shown.
I´m not sure how to fix the phase Error in the measurement with REW. . I use loopback for timing reference.
I traced your curves and overlaid it with a 4th order LR HP at 12.5 Hz in pinkish. What puzzles me completely is how an Open Baffle Woofer can show a 4th order roll off. With Current Drive you boost Fo to a 6-12dB peak, but a 4th order roll off is still somewhat unusual.
I calibrated the soundcard with loopback, used L channel output as timing source and made a splitter so that L and R input get the same signal.Then i run the signal trough the MFB circuit inkluding the current feedback amplifier to the speaker and in to accelerometer. This signal is INVERTED, so I applied Inverting function in REW. See the difference between the phase with and without inverting the phase.
But still i think there is a timing issue...
How do you secure absolute accurate phase information in REW?
(and another issue at 200-230Hz broken voice coil? playing discrete tones gives rattle and it is impossible to find where? I have to try another woofer later)
Inverted phase:
Not inverted phase:
But still i think there is a timing issue...
How do you secure absolute accurate phase information in REW?
(and another issue at 200-230Hz broken voice coil? playing discrete tones gives rattle and it is impossible to find where? I have to try another woofer later)
Inverted phase:
Not inverted phase:
The 4the order roll off below 10Hz is still somewhat strange and has i.m.o high impact on the phase rapidly approaching +180 degrees. This might wel be one of the classic piezo accelerometer attachment issues that plagues MFB in real life situations. In a closed box, your Fbox would be higher, say 50Hz, leaving more headroom for feedback. This is the drawback of the very low Fo in your OB application.
With a notch @14.1Hz, 12dB depth and a Q =2.0 I massaged the loopshape into some 12 dB feedback.
With a notch @14.1Hz, 12dB depth and a Q =2.0 I massaged the loopshape into some 12 dB feedback.
@esl 63. Do you still need a simulation for the electrical part of the schematic?
I used 700pF and 500Meg for the sensor capacitance and the internal ACH01 resistor.
If someone knows the exact values please let us know.
And a bit of oftop : I read the thread about active bass absorption in HiFiForum.nu forum.
Respect for the effort! If I can steal some time I want to try a combination between active absobtion and MFB.
I like the idea of super-tiny bass absorber. 🙂
I used 700pF and 500Meg for the sensor capacitance and the internal ACH01 resistor.
If someone knows the exact values please let us know.
And a bit of oftop : I read the thread about active bass absorption in HiFiForum.nu forum.
Respect for the effort! If I can steal some time I want to try a combination between active absobtion and MFB.
I like the idea of super-tiny bass absorber. 🙂
Active absorbtion is really simple... microphone close to the speaker, and try to minimize phase shift, crank up volume until it oscillates and then down with the amplification until it is stable. I made some mistakes in that old design, and will correct it. Maybe put a thread here at DiyAudio. I think Im on my way to get the measurements with REW working after some struggeling. Recapped the soundcard, electrolytics in signal path,, >15 years old and dry. deoxidized the connectors, resoldered the PCB where connectors is mounted (micro cracks?) and added caps in the power supply.
It went much more silent, no simulation needed, for the moment. Thanks!
It went much more silent, no simulation needed, for the moment. Thanks!
Made tons (>300) of distortion measurements without seeing any improvements with MFB, hours become days...
Tried to figure out what was wrong, apparently my MFB CFB loop was stable with 14dB feedback, and since i had to crank up the volume with 14dB to get same volume as with the MFB connected i knew that the accelerometer was sending an inverted feedback signal and of course harmonics... so what was happening why was there 1.3% distortion with and without MFB? Only component that is in common is the amplifier.
I measured the amplifier distortion including the input stage/filter on piratelogic card over a fixed 10ohm resistor 0.011%..
Yeah.. one more component is in common, the UMIC microphone.
Today I hardwired the soundcard into the PXE output and run some measurement:
First one is without MFB measured on PXE output of EVE 2020 MFB board from Pirate logic, see schematic in earlier post. 0.49% THD
Now we get 0.18% THD... with the UMIC there would still have been 1.3% and I would work in the darkness with bad input data... The sound level from the microphone was less than 100dB so i guess it is damaged? Anyone have any idea what´s wrong?
Anyway.. now I will go back to the optimizing phase of the MFB loop and phase margin work.
Next thing is to make an OLTF plot from the PXE output.
Tried to figure out what was wrong, apparently my MFB CFB loop was stable with 14dB feedback, and since i had to crank up the volume with 14dB to get same volume as with the MFB connected i knew that the accelerometer was sending an inverted feedback signal and of course harmonics... so what was happening why was there 1.3% distortion with and without MFB? Only component that is in common is the amplifier.
I measured the amplifier distortion including the input stage/filter on piratelogic card over a fixed 10ohm resistor 0.011%..
Yeah.. one more component is in common, the UMIC microphone.
Today I hardwired the soundcard into the PXE output and run some measurement:
First one is without MFB measured on PXE output of EVE 2020 MFB board from Pirate logic, see schematic in earlier post. 0.49% THD
Now we get 0.18% THD... with the UMIC there would still have been 1.3% and I would work in the darkness with bad input data... The sound level from the microphone was less than 100dB so i guess it is damaged? Anyone have any idea what´s wrong?
Anyway.. now I will go back to the optimizing phase of the MFB loop and phase margin work.
Next thing is to make an OLTF plot from the PXE output.
OLTF Measurements is WORKING!!
There is two methods to measure the open loop transfer function:
Acoustically with a microphone, you need two measurements, one without feedback "A" and one without feedback "B"
You use the aritmetic functions in REW A-B and then take A-B result and divide by B, so (A-B) / B
Timing is wery important so that the phase information is perfect! A 15" woofer and a lot of low pass filtering in your MFB circuit makes the high frequency timing signal too weak! I added a tweeter in parallel with woofer with a switch so just after the acoustic timing chirp signal is heard, the tweeter is disconnected. Not to interfere with the CFB amplifier... A normal voltage amplifier will not care, but remember that the speaker is in the feedback loop of a CFB! Now the acoustic timing works perfect!
Electrical, by connecting the signal in and out somewhere in the feedback loop see post 4 on the schematic where it says "OPEN LOOP" there is a jumper.
When you do the electrical OLTF measurement remember to Invert the phase, to get it right! (Not necessary on the acoustical measurements)
Here is the two measurements and they match very well!!
Acoustic OLTF:
And below the electric OLTF:
As you can see there is quite good phase margin at low frequency BUT there is an oscillation of 0.5Hz! so I will try to make C8 a little smaller i have 10uF today.. next oscillation is at 2-3 kHz and I will decrease C12 a little. When you can trust the measurements you are much better out.
So lets see tomorrows update with new C8 and C12! So we can get a larger phase margin and with that more feedback!
There is two methods to measure the open loop transfer function:
Acoustically with a microphone, you need two measurements, one without feedback "A" and one without feedback "B"
You use the aritmetic functions in REW A-B and then take A-B result and divide by B, so (A-B) / B
Timing is wery important so that the phase information is perfect! A 15" woofer and a lot of low pass filtering in your MFB circuit makes the high frequency timing signal too weak! I added a tweeter in parallel with woofer with a switch so just after the acoustic timing chirp signal is heard, the tweeter is disconnected. Not to interfere with the CFB amplifier... A normal voltage amplifier will not care, but remember that the speaker is in the feedback loop of a CFB! Now the acoustic timing works perfect!
Electrical, by connecting the signal in and out somewhere in the feedback loop see post 4 on the schematic where it says "OPEN LOOP" there is a jumper.
When you do the electrical OLTF measurement remember to Invert the phase, to get it right! (Not necessary on the acoustical measurements)
Here is the two measurements and they match very well!!
Acoustic OLTF:
And below the electric OLTF:
As you can see there is quite good phase margin at low frequency BUT there is an oscillation of 0.5Hz! so I will try to make C8 a little smaller i have 10uF today.. next oscillation is at 2-3 kHz and I will decrease C12 a little. When you can trust the measurements you are much better out.
So lets see tomorrows update with new C8 and C12! So we can get a larger phase margin and with that more feedback!
I will show two OLTF measurements and the difference between changing C12 from 22n to 2,2n... I can not get enough phase margin at 2-3kHz
0dB is at 80dB line!! And the phase margin is how many degrees you are from 180deg at the point where amplitude is crossing 0dB line (80dB)
First is C12 22nF:
And next is C12 2n2.
How can I increase the phase margin at 3kHz?
I tried with a Zobel link on the driver (since I use current feedback and the Inductance of the driver increases the impedance at high frequency.
I tried to add a physical resistor of 10 Ohms in series with the driver but it did not change the phase margin either.
C16 was also changed to 3,3 nF but it went worse.
Low frequency stability seems OK though!
First of all: how do you know there is an oscillation at 0.5 Hz? Your graphs run until 2 Hz only, as far as I can see.
Regarding your 3 kHz issue: how about notching that peak out?
Looking at the graphs: i.m.o. the entire 550 Hz -5 kHz range should be notched out and flattened in some way or other.
Regarding your 3 kHz issue: how about notching that peak out?
Looking at the graphs: i.m.o. the entire 550 Hz -5 kHz range should be notched out and flattened in some way or other.
When I increase the loop gain I see the membrane start to swing gently in and out with approx 0.5 Hz.
I agree that the amplitudes between 300Hz and up could be notched away, but I have limited knowledge on how to achieve that without introoducing phase shift that decreases the stability I need between 20Hz -150Hz.
I can use IC2a and IC2B as components for playing with extra filters.
Thanks a Lot Boden!!
https://rmsacoustics.nl/papers/whitepaperMFBdesign.pdf
On page 22, Robert-H Munnig Schmidt shows that he reach up to 30dB of gain with the same accelerometer and some clever filtering.
How is that possible?
I agree that the amplitudes between 300Hz and up could be notched away, but I have limited knowledge on how to achieve that without introoducing phase shift that decreases the stability I need between 20Hz -150Hz.
I can use IC2a and IC2B as components for playing with extra filters.
Thanks a Lot Boden!!
https://rmsacoustics.nl/papers/whitepaperMFBdesign.pdf
On page 22, Robert-H Munnig Schmidt shows that he reach up to 30dB of gain with the same accelerometer and some clever filtering.
How is that possible?
This is the shape of the loopfilter RMS uses. As you can see this is not a monotonous transfer function, but an optimally shaped complex filter, digitally achieved with the ADUA 1777 AD/DA processor. Analogue you can achieve the same result by using 4 or 5 gyrator circuits.
Well I am satisfied if I can go from current 12 db to 18dB.
6dB extra...
30 dB is a "dream"... but I have to accept my own limitations here.
6dB extra...
30 dB is a "dream"... but I have to accept my own limitations here.
Blocks for possible analogue inmplementation: : 1 Linkwitz transform circuit, 4 gyrators, 1 Lowpass 2e order.
Using VituixCAD to trace your plots, I imported into spreadsheet to see what could be done to increase gain.
The phase does match the magnitude response (ie minimum phase), so I think you have worked the problems out of your OLTF measurement technique.
With one 1st order LP filter and 3 or 4 Param-EQs you could increase gain to > 20dB...at least at HF.
I am skeptical of the measurements or stability at LF...the response should continue to fall with decreasing frequency below resonance with at least of slope of -12dB/oct transitioning to steeper slopes depending on the size of coupling capacitors. Is it possible your current feedback amplifier is marginally stable at LF? Also, the array of sharp peaks and dips starting at 200 Hz is troubling. Typically you should only have to deal with one major peak from the ACH-01 mount.
I'd recommend getting a set of 4 measurements to better understand what you are dealing with at LF and HF.
Measurements should be taken with no circuitry between your computer and amplifier.
With standard voltage source amplifier:
1) measure response with near field microphone
2) measure response with ACH-01 using soundcard microphone input directly (see Fig. 2 in attached *.pdf)
With your current feedback amplifier setup:
3) measure response with near field microphone
4) measure response with ACH-01 using soundcard microphone input directly (see Fig. 2 in attached *.pdf)
This will hopefully help isolate the source of the LF and HF anomalies seen in the OLTF.
What about the drive signal to the driver? Life is nicer working with electricity rather than air (yecch, has water, carbon pollution, no fun). If the MFB is doing what you are paying for, the drive signal should be distorted. As in days of yore, I'd scope it to see how the driver signal goes weird under the influence of MFB as the cone motion increases. (Take care with the alligator clips because you may have a complicated ground reference.)
The biggest glory of MFB is "fast bass". We've been living with subwoofer distortion forever and not all that irritating. But why is there no attention to quantifying (or at least displaying) transient goodness?
Why is anybody fussing about MFB north of maybe 150 Hz? Not a path leading to happiness working with ordinary cones and magnet structures.
While appealing to get lots of feedback operating, do the math. An amp needs a great deal of feedback to get down to a respectable level. But reducing a sub from the usual 10-20% distortion to 5% should be just fine with modest (safe) amount of feedback.
The biggest glory of MFB is "fast bass". We've been living with subwoofer distortion forever and not all that irritating. But why is there no attention to quantifying (or at least displaying) transient goodness?
Why is anybody fussing about MFB north of maybe 150 Hz? Not a path leading to happiness working with ordinary cones and magnet structures.
While appealing to get lots of feedback operating, do the math. An amp needs a great deal of feedback to get down to a respectable level. But reducing a sub from the usual 10-20% distortion to 5% should be just fine with modest (safe) amount of feedback.