There is a lot to be said to avoiding passive crossovers. But we also should recognise that the right crossover can actually improve the driver, like my example of a series inductor and then pick quality drivers with low inductance, then use relative linear inductance (an actual physical inductor) to suppress the potential non-linear inductance to the driver.
If I was designing an active loudspeaker I would still be using the largest series inductor I can get away with, on the midrange driver in particular. That will give the midrange amplifier used, via the series inductor, some of the same benefits that current source amplifiers give you. But then again, there are now active speakers on the market that actually use current sources on the midrange, and I say here 'current-drive' is a practical option in actives.
But going over to pre-EQ (at line-level) now introduces a whole series of new ingredients to the sauce (i.e. subjective). And some will resort to doing it digitally, so now going analog to digital, and then digital back to analog. How will that affect the overall design. And AD and DA converters working at the highest level of performance, this is not then a cheap DIY option.
Doing it cheaply like Behringer et al does not give me an appetite to do.
Just my thoughts.
If I was designing an active loudspeaker I would still be using the largest series inductor I can get away with, on the midrange driver in particular. That will give the midrange amplifier used, via the series inductor, some of the same benefits that current source amplifiers give you. But then again, there are now active speakers on the market that actually use current sources on the midrange, and I say here 'current-drive' is a practical option in actives.
But going over to pre-EQ (at line-level) now introduces a whole series of new ingredients to the sauce (i.e. subjective). And some will resort to doing it digitally, so now going analog to digital, and then digital back to analog. How will that affect the overall design. And AD and DA converters working at the highest level of performance, this is not then a cheap DIY option.
Doing it cheaply like Behringer et al does not give me an appetite to do.
Just my thoughts.
I have a custom made Jantzen inductor 18mH and DCR 3.R2 as the resistance does not need to be low and can be adjusted by the "R" value in the the LCR and that 18mH is usually enough and can be wound back if a lower value is needed. The R value is not a problem, but needs to have a substantial Wattage rating. That leaves the LCR and I can recommend a 100V Bi-Polar range that is readily available.
Here is a link for those who are interested in pro-audio variable damping amplifiers. Unfortunately in russian, so Google translate will help.
AFAIK the company moved from Luhansk to St Petersburg, maybe they still offer the amps.
https://www.ixbt.com/proaudio/sakevich-sk1200.shtml
AFAIK the company moved from Luhansk to St Petersburg, maybe they still offer the amps.
https://www.ixbt.com/proaudio/sakevich-sk1200.shtml
Interesting, but it says variable output impedance amplifiers. That is not quite the same thing. At LF, varying the output impedance simply alters the alignment of the speaker box tuning, the amplifier is neither adding or subtracting damping. I know this is a slow moving train, but slowly that understanding is emerging because it is grounded in actual physics.
I will repeat my mantra again and again: EQ (equalise) the current flat below 200 Hertz and you have effectively cancelled out the output impedance of the amplifier and that impedance no longer has control over the alignment, and it is the alignment that defines the damping. Make sure the currents at Fl, Fc and Fh are the same (assuming a sealed box) and the damping will be always be the same, even if the amplifier's output impedance is infinite. The damping is real and not as somebody suggest, mimicked. It is 100% the same. If the currents at those three frequencies (those who measure T-S Parameters will know how to find those three frequencies), stays relative to each other as SPL levels are increased and decreased, then we have locked in the damping.
Slowly but surely this will get better understood.
I will repeat my mantra again and again: EQ (equalise) the current flat below 200 Hertz and you have effectively cancelled out the output impedance of the amplifier and that impedance no longer has control over the alignment, and it is the alignment that defines the damping. Make sure the currents at Fl, Fc and Fh are the same (assuming a sealed box) and the damping will be always be the same, even if the amplifier's output impedance is infinite. The damping is real and not as somebody suggest, mimicked. It is 100% the same. If the currents at those three frequencies (those who measure T-S Parameters will know how to find those three frequencies), stays relative to each other as SPL levels are increased and decreased, then we have locked in the damping.
Slowly but surely this will get better understood.
So here's a rough concept I sketched:
There's some low-pass pre-filtering on the left, a long-tailed pair, an open-drain MOSFET and CCS output stage, and a woofer model. The woofer inductance is probably exaggerated because it uses a normal inductor instead of any special semi-inductance, so with the current feedback the gain is also exaggerated.
The feedback system has an 'inverted' notch filter. I wanted a band-pass filter, but needed something that passes DC, so ended up making a "band boost" that mostly has a gain of 1, except for the resonance, where it goes up and is somewhat adjustable. So at ~80 Hz there is more voltage feedback than elsewhere.
From 500 Hz to 20 kHz, the gain increases due to the 0.53mH speaker inductance, but this is flattened out by pre-filtering.
Bypassing the pre-filter:
It's a bit tricky so I'm open to any ideas and suggestions. High frequency noise in particular is higher than I'd like. Increasing the sense resistor (and re-balancing the feedback gains) gives a quick improvement, but the topology might need some work. The 2 feedback paths are basically fighting each other, so some first-order filter slopes might work, and trying to get the high frequencies to bypass the op-amps.
The obvious alternative is to put a notch filter on the input, but then all it does is turn down the excess gain, but not the distortion spike.
There's some low-pass pre-filtering on the left, a long-tailed pair, an open-drain MOSFET and CCS output stage, and a woofer model. The woofer inductance is probably exaggerated because it uses a normal inductor instead of any special semi-inductance, so with the current feedback the gain is also exaggerated.
The feedback system has an 'inverted' notch filter. I wanted a band-pass filter, but needed something that passes DC, so ended up making a "band boost" that mostly has a gain of 1, except for the resonance, where it goes up and is somewhat adjustable. So at ~80 Hz there is more voltage feedback than elsewhere.
From 500 Hz to 20 kHz, the gain increases due to the 0.53mH speaker inductance, but this is flattened out by pre-filtering.
Bypassing the pre-filter:
It's a bit tricky so I'm open to any ideas and suggestions. High frequency noise in particular is higher than I'd like. Increasing the sense resistor (and re-balancing the feedback gains) gives a quick improvement, but the topology might need some work. The 2 feedback paths are basically fighting each other, so some first-order filter slopes might work, and trying to get the high frequencies to bypass the op-amps.
The obvious alternative is to put a notch filter on the input, but then all it does is turn down the excess gain, but not the distortion spike.
I have always been a proponent and a user of DSP for my loudspeaker projects. Who needs those silly passive components, I thought. Along comes a tech paper by Purifi that recently caught my attention:
https://purifi-audio.com/wp-content/uploads/2022/03/220211_R05-Notchfilter.pdf
In the scenario presented, the driver (the PTT8.0X) has some 10dB tall breakup peaks in its response. The distortion of the driver has "echoes" of these peaks at lower frequencies that are N times lower where N is the order of distortion. This predominantly occurs with 3rd order distortion products, so at a frequency 3 times lower than the breakup peak, which puts the distortion peak within the passband. Bad!
I was able to chat with Lars Risbo of Purifi about the tech note until I finally wrapped my brain around the claims therein. According to Lars, distortion (broadly in the mid passband and above) is caused by modulation of the voice coil inductance due to various mechanisms. The inductance modulation within a magnetic field causes a new "distortion voltage" to be generated. This is generating various order of distortion all across the passband. This is also mentioned in this article in the section "Indefinite EMF generation due to mechanical non-idealities":
https://www.edn.com/loudspeaker-ope...y-of-current-drive-over-voltage-drive-part-2/
Also a factor is the breakup peak of the driver, which is sort of like a small frequency region of enhanced driver efficiency where more output is produced by the drive current compared to the rest of the passband. So, what is happening is that the "distortion voltage" finds a closed loop path over which it causes current flow, and this path includes the driver itself. When the distortion product falls on the breakup peak frequency, the distortion peaks just like the profile of the breakup peak following the increase in efficiency there. The way to prevent this is to reduce or block the current flow induced by the distortion voltage. In the Purifi tech note, this is accomplished via a passive (notch) network placed in series with the driver and immediately next to it without any other network in between. The series network notches out the breakup peak via a peak in impedance put in series with the driver. It is this series impedance at the breakup peak frequency that is "in the way" of the distortion voltage induced current, and so very little can flow. As a result the peak in 3rd order distortion occurring at 1/3 the frequency of the breakup peak is also reduced.
This was all new to me, and I now see some benefits from using series impedance shaping networks and filters as a way to reduce 3rd order distortion within the passband. Very interesting.
Thereis a way to do this without any passive elements, and that is to use an amplifier with high output impedance, which is what current drive produces. So I thought I would post about it here after reading Joe's post above that touches on the subject. It is something that I really did not appreciate previously. Unfortunately I believe that current drive is not possible unless the loudspeaker return is at ground potential, making amplifiers operating in bridge mode (such as many class-D amplifiers) unable to implement current drive.
Last edited:
Most implementations use a grounded current sense resistor from which the (current) feedback is taken. I'm not sure how this can work in a bridged application...
Edit, OK I did find this just now:
![bridging[1].gif](https://www.diyaudio.com/community/attachments/bridging-1-gif.1158748/ "bridging[1].gif")
But I believe this requires a dedicated design.
If you would like to present a circuit of a bridged amplifier implementing current drive, please do so. What I would like to see if how to use an existing (e.g. commercial) real world amplifier operating in bridged mode as a current source without internal modification.
Edit, OK I did find this just now:
But I believe this requires a dedicated design.
If you would like to present a circuit of a bridged amplifier implementing current drive, please do so. What I would like to see if how to use an existing (e.g. commercial) real world amplifier operating in bridged mode as a current source without internal modification.
Last edited:
Taking feedback from source resistors could be another option in a bridged design. But what's the benefit?
OK, so it's one way to skip the bulk output capacitor and still be able to use a single-ended supply.
The PSU load is more balanced, but that only seems to count for double-sided push-pull. A CCS-based class A already has a steady current draw, and it's even possible to do a "power long-tailed pair".
My earlier thinking on a mixed-mode circuit included a bridged amplifier with 'i' output on one side, and 'v' output on the other. Except that would only work one way but not the other. The current output easily uses the voltage output like a virtual ground, but the voltage output can't do the same thing because the current output acts like 100~1000 ohms and actively prevents the current from being changed by outside sources. So for the competing systems to interact with each other, overly complicated feedback circuits for frequency-dependent output impedance are still needed, as far as I can tell.
OK, so it's one way to skip the bulk output capacitor and still be able to use a single-ended supply.
The PSU load is more balanced, but that only seems to count for double-sided push-pull. A CCS-based class A already has a steady current draw, and it's even possible to do a "power long-tailed pair".
My earlier thinking on a mixed-mode circuit included a bridged amplifier with 'i' output on one side, and 'v' output on the other. Except that would only work one way but not the other. The current output easily uses the voltage output like a virtual ground, but the voltage output can't do the same thing because the current output acts like 100~1000 ohms and actively prevents the current from being changed by outside sources. So for the competing systems to interact with each other, overly complicated feedback circuits for frequency-dependent output impedance are still needed, as far as I can tell.
mixed or current or voltage mode is all possible with bridged. Don’t think to complicated. Electronics doesn’t understand what mode it works in. It is a question of feedback. Take bridged voltage amp, put opamp in front of it with your own feedback loop. Mixed or current, or negative impedance or any other feedback configuration and it will follow what you want. It will not be the fastest possible design but this is just for concept understanding that every configuration posible. CharlieLaub is showing bridged design. More or less the same i am talking about
How about drawing some more schematics. This is interesting. With so-called 'current-drive' only one side needs to be high-Z, the other can still be low-Z. Since they are in series, the sum will always be high-Z and hence a current source is not compromised.
But schematically, how would you do it. Are there several ways?
I am willing to learn from others.
But schematically, how would you do it. Are there several ways?
I am willing to learn from others.
If you mean this circuit from post 809:
https://www.diyaudio.com/community/attachments/bridging-1-gif.1158748
This is just a driver with two amps configured in a bridged operation to have the same gain. The same current flows through the driver and the sense resistor. Instead of R1 sensing the potential across the resistor referenced to ground, A3 operates as a differential amplifier. I believe the resistor values should follow these rules:
R2>>R1
R5>>R4
R3 and C are set to ensure stability and limit bandwidth.
https://www.diyaudio.com/community/attachments/bridging-1-gif.1158748
This is just a driver with two amps configured in a bridged operation to have the same gain. The same current flows through the driver and the sense resistor. Instead of R1 sensing the potential across the resistor referenced to ground, A3 operates as a differential amplifier. I believe the resistor values should follow these rules:
R2>>R1
R5>>R4
R3 and C are set to ensure stability and limit bandwidth.
It looks a bit like a Howland or improved Howland current pump as a basis. There are lots of ways to do it.
For a prototype headphone amplifier I made, the load was connected with an ordinary 3-wire phono plug: left, right, and common ground. But I wanted to avoid output capacitors (which would be leaky, therefore creating DC voltage anyway) + the convenience of a laptop-style single supply, so I used a voltage amplifier as a virtual ground for the 3rd pin, set to about 10V, and low-frequency feedback with trimmers to stabilse the DC offset of the current amplifiers on the opposite side. Bad JFET usage notwithstanding, it worked.
In that example the output transistors were configured as "open drain", raising the output impedance. But they could also use the source for output, so the starting point would be a voltage amplifier, and n-feedback from a sense resistor would raise the output impedance that way.
A big difference between between the 2 styles (open drain/collector vs open source/emitter) is that the pre-amp stages have to work differently. For open-drain the input voltage swing is low, but there's a multiplying effect on the parasitic capacitance between gate and drain, so the current handling has to be high at high frequencies. For open-source, there has to be a 'VAS' that stays linear across a very wide voltage swing, because the output transistors provide no additional voltage gain. But the source pin follows the gate with internal feedback, so the parasitic capacitance seems to have a much smaller effect.
For a prototype headphone amplifier I made, the load was connected with an ordinary 3-wire phono plug: left, right, and common ground. But I wanted to avoid output capacitors (which would be leaky, therefore creating DC voltage anyway) + the convenience of a laptop-style single supply, so I used a voltage amplifier as a virtual ground for the 3rd pin, set to about 10V, and low-frequency feedback with trimmers to stabilse the DC offset of the current amplifiers on the opposite side. Bad JFET usage notwithstanding, it worked.
In that example the output transistors were configured as "open drain", raising the output impedance. But they could also use the source for output, so the starting point would be a voltage amplifier, and n-feedback from a sense resistor would raise the output impedance that way.
A big difference between between the 2 styles (open drain/collector vs open source/emitter) is that the pre-amp stages have to work differently. For open-drain the input voltage swing is low, but there's a multiplying effect on the parasitic capacitance between gate and drain, so the current handling has to be high at high frequencies. For open-source, there has to be a 'VAS' that stays linear across a very wide voltage swing, because the output transistors provide no additional voltage gain. But the source pin follows the gate with internal feedback, so the parasitic capacitance seems to have a much smaller effect.
I am sure that A2 will not be unity gain stable!
This shares some similarities with the circuits discussed here:
https://sound-au.com/project20.htm
This shares some similarities with the circuits discussed here:
https://sound-au.com/project20.htm