Hi Pavel,
Thanks for your comments. The THD you saw was at half power. At 1W "ordinary" power its lower. Recall I wrote:
As you may have noticed I have working through variations to compare advantages and disadvantages and hopefully soon summarise options. One of my criteria is minimalism. You may have noticed my later versions use 2 MOSFET's and 4 BJTs and 2 Schottky's for the power stage. I am now working on bench testing a discrete driver stage.
Another criteria is to avoid source/emitter resistors in the power stages - including no source/emitter resistors for paralleling. I have found it is best to parallel autobias using slices like the LT1166 application note for a 350W amp. Why no source/emitter resistors in the power stages? I posted about that here https://www.diyaudio.com/forums/solid-state/355067-square-law-amp-post6219873.html
Cheers, Ian
Thanks for your comments. The THD you saw was at half power. At 1W "ordinary" power its lower. Recall I wrote:
That simulation had changes that increased the THD by 10 times - but that didn't worry me because I thought the 1W THD is good enough.
As you may have noticed I have working through variations to compare advantages and disadvantages and hopefully soon summarise options. One of my criteria is minimalism. You may have noticed my later versions use 2 MOSFET's and 4 BJTs and 2 Schottky's for the power stage. I am now working on bench testing a discrete driver stage.
Another criteria is to avoid source/emitter resistors in the power stages - including no source/emitter resistors for paralleling. I have found it is best to parallel autobias using slices like the LT1166 application note for a 350W amp. Why no source/emitter resistors in the power stages? I posted about that here https://www.diyaudio.com/forums/solid-state/355067-square-law-amp-post6219873.html
Cheers, Ian
HI Miralin,
Thanks for the suggestion. Since that post you mentioned I reported using a BD139/140 on a TO-220 Schottky (Post 138):
The BD139 is on the two Schottky's plate and the BD140 (series with the BD139) is on a non-heated plate at ambient. This halves the effective temp.co of the spreader and is just right. (With both BD139 and BD140 on the Schottky's the idle current halves! when you run at full power for 15 mins then stop the signal).
This is the best thermal arrangement I have tried so far. A driver stage is required as the input needs about 10mA at 1Vpk, and if you prefer, additional feedback to reduce distortion.
Thanks for the suggestion. Since that post you mentioned I reported using a BD139/140 on a TO-220 Schottky (Post 138):
My bench jig for Post 139 is shown below for one slice (the main heatsink with the MOSFETs is not shown but is to the right)
The BD139 is on the two Schottky's plate and the BD140 (series with the BD139) is on a non-heated plate at ambient. This halves the effective temp.co of the spreader and is just right. (With both BD139 and BD140 on the Schottky's the idle current halves! when you run at full power for 15 mins then stop the signal).
This is the best thermal arrangement I have tried so far. A driver stage is required as the input needs about 10mA at 1Vpk, and if you prefer, additional feedback to reduce distortion.
I have now bench tested one type of driver stage. I am using a "quad-core" stage (eg LT1364 but no diamond).
I have also made the driver and autobias into a subcircuit so it can be drawn like the LT1166 with drive to its Vin pin.
If you run the attached circuit you can view inside the subcircuit with Rt-clk on the AutobiasB symbol (blank area).
The driver stage when biased at 1mA (set by R1, R2 and R12,R15 of the subcircuit) there is additional voltage gain as well as current gain. Input impedance is about 10k ohms. Additional feedback is provided by the feedback pin (FB). The voltage gain into 4 ohms is 50.
It could be used as an amp, particularly if the idle current is increased closer to the optimum bias point (about 1A total) for lower distortion. But most would prefer a lower idle current -- which requires an additional gain stage and more feedback to reduce distortion.
I have also made the driver and autobias into a subcircuit so it can be drawn like the LT1166 with drive to its Vin pin.
If you run the attached circuit you can view inside the subcircuit with Rt-clk on the AutobiasB symbol (blank area).
The driver stage when biased at 1mA (set by R1, R2 and R12,R15 of the subcircuit) there is additional voltage gain as well as current gain. Input impedance is about 10k ohms. Additional feedback is provided by the feedback pin (FB). The voltage gain into 4 ohms is 50.
It could be used as an amp, particularly if the idle current is increased closer to the optimum bias point (about 1A total) for lower distortion. But most would prefer a lower idle current -- which requires an additional gain stage and more feedback to reduce distortion.
Hi Pawel,
The optimum bias is 1A total for two stages in parallel. It gives lower THD up to about 10W.
Operating at 0.5A total is a good compromise still with acceptably low distortion for typical 1W average (1W typical average for those with 90dB/W/m speakers).
Relying on THD figures is a fools game because THD does not correlate with sound quality. A recent paper backing up the noncorrelation here https://hifisonix.com/wordpress/wp-content/uploads/2017/11/Perceptual-Levels-of-distortion.pdf courtesy of 'esl 63' here https://www.diyaudio.com/community/threads/can-you-hear-10-thd.369672/post-6951246
These autobias circuits give a fast roll-off of high-order harmonics which means the THD figure (if we must use THD) does not need to be less than 0.1% to make distortion inaudible (more here). When you under-bias at say 0.5A rather than optimum of 1A the harmonics roll-off similar to cube-laws and 0.1% is OK at 1W for good sensitivity speakers. So sound wise I don't think biasing at 1A gives much advantage for the design above.
Another aspect to consider is use of current drive for moving coil drivers. Current drive offers a better listening experience than lowering these amps distortion from 0.1% to say 0.01% with standard voltage drive. Another paper on improved sound quality using current drive is here. And you may have already seen Esa's peer reviewed paper that I linked in my earlier Post 130.
Cheers,
Ian Hegglun
The optimum bias is 1A total for two stages in parallel. It gives lower THD up to about 10W.
Operating at 0.5A total is a good compromise still with acceptably low distortion for typical 1W average (1W typical average for those with 90dB/W/m speakers).
Relying on THD figures is a fools game because THD does not correlate with sound quality. A recent paper backing up the noncorrelation here https://hifisonix.com/wordpress/wp-content/uploads/2017/11/Perceptual-Levels-of-distortion.pdf courtesy of 'esl 63' here https://www.diyaudio.com/community/threads/can-you-hear-10-thd.369672/post-6951246
These autobias circuits give a fast roll-off of high-order harmonics which means the THD figure (if we must use THD) does not need to be less than 0.1% to make distortion inaudible (more here). When you under-bias at say 0.5A rather than optimum of 1A the harmonics roll-off similar to cube-laws and 0.1% is OK at 1W for good sensitivity speakers. So sound wise I don't think biasing at 1A gives much advantage for the design above.
Another aspect to consider is use of current drive for moving coil drivers. Current drive offers a better listening experience than lowering these amps distortion from 0.1% to say 0.01% with standard voltage drive. Another paper on improved sound quality using current drive is here. And you may have already seen Esa's peer reviewed paper that I linked in my earlier Post 130.
Cheers,
Ian Hegglun
Ian, current drive needs EQ according to speaker driver impedance curve otherwise the frequency response is a copy of impedance response for the moving coil driver. The other method is impedance equalization to get flat impedance. Both is uneasy and tricky. Distortion reduction may or may not happen depending on the driver used.
https://pmacura.cz/speaker_dist.htm
Hi Pavel,
Thanks for posting the link to your tweeter distortion comparisons under voltage and current drive. Well done. Very interesting.
But what is not "uneasy and tricky"?
I like a challenge. Some on diyAudio like it here because there are still some challenges.
Did you you follow the thread back 2017 using voltage drive at LF (for speaker resonance damping) and current drive above resonance (for speaker distortion reduction) here?
EG Post 8
I posted in that thread the idea of using a self-heating current sensing resistor to compensate for the voice-coil temperature upsetting the frequency response, eg Post 30 here https://www.diyaudio.com/forums/sol...rrent-drive-voltage-drive-lf-post5114877.html. It was an idea I tried 20 years earlier and mentioned in an Electronics World 1996 article https://drive.google.com/file/d/0B3...bTQ/view?resourcekey=0-Xt30BtiT8buExDhYokNt1w.
But I haven't finalised a practical version. I am stuck on transconductance amplifier designs.
Also to try is a version using an RC thermal model in the place of the self-heating sense resistor.
Also, in Post 31 using negative resistance to compensate for the voice-coil temperature - since it doesn't need an analog multiplier, so it is simpler with no added voltage offset issues etc.
So nice to hear from you.
Thanks for posting the link to your tweeter distortion comparisons under voltage and current drive. Well done. Very interesting.
But what is not "uneasy and tricky"?
I like a challenge. Some on diyAudio like it here because there are still some challenges.Did you you follow the thread back 2017 using voltage drive at LF (for speaker resonance damping) and current drive above resonance (for speaker distortion reduction) here?
EG Post 8
I posted in that thread the idea of using a self-heating current sensing resistor to compensate for the voice-coil temperature upsetting the frequency response, eg Post 30 here https://www.diyaudio.com/forums/sol...rrent-drive-voltage-drive-lf-post5114877.html. It was an idea I tried 20 years earlier and mentioned in an Electronics World 1996 article https://drive.google.com/file/d/0B3...bTQ/view?resourcekey=0-Xt30BtiT8buExDhYokNt1w.
But I haven't finalised a practical version. I am stuck on transconductance amplifier designs
.Also to try is a version using an RC thermal model in the place of the self-heating sense resistor.
Also, in Post 31 using negative resistance to compensate for the voice-coil temperature - since it doesn't need an analog multiplier, so it is simpler with no added voltage offset issues etc.
So nice to hear from you.
But what is not "uneasy and tricky"?
Hi Ian,
yes, definitely, I agree.
But more and more I turn my attention to the circuits that are practical and have the perspective of reliable circuit realization, with good results proven both by measurements and reliable, long-term function. I think that linear, electronic circuits audio has come to the point where there is no improvement in audible sonic parameters and new developments are just for the purpose (show must go on and continuous growth is requested). This is still not true in drivers (speakers) and in room acoustics. There is still a lot to improve, though the achievements in speaker design in the last decade are considerable. But I am not sure that techniques like current drive bring considerable improvement, IMO it is rather the design path as applied in Kii Three (that brings considerable improvements), which targets on biggest driver drawback and it is a rocket acceleration of low frequency distortion of smaller bass drivers with necessary large excursion at lows (current or mixed drive cannot help here), and finding the algorithm how to avoid these situations if they are not necessary. This is IMO the way to go and develop.
Last edited:
Now using MJL3281/1302 power transistors in place of IRF540/9540.
Recall Post 90 showed that this autobias loop takes MOSFETs or BJTs with little change to the bias. Here I have used the circuit of Post 139 with one pair of power transistors for an 8 ohm load. Resistors R5 and R6 are reduced from 470 ohms to 170 ohms to give sufficient base current to get 5A peak (assuming the supply voltage is sufficient) and this resistance will depend on the Beta of the transistors used. Sim's Ic and Gm plots are below:
With BJT's the idle current for optimum bias is lower than for MOSFET's. Here optimum bias is 250mA! Nice. THD at 1W is 0.15%. There is a soft clip starting from about half full power. Very nice IMO.
The driver needs to deliver about 75mA peak. Too high for a standard opamp, but several in parallel feed could do it. With BD139/140 EF driver the input impedance is about 1k. Input voltage is about 0.5V peak. Voltage gain is about 40 into 8 ohms.
I have bench tested this circuit with AC direct drive (Vin1) and R5/R6 were 330 ohms, the Betas of the power transistors were 130 and I got 4.5A peak into 8 ohms with +/-40.2V rails. This is 36V peak with 1V across two 100mR resistors (R13 and one for rail current sensing) and 3V across the power transistors. With slightly higher rail voltage and with R5 and R6 270 ohms I expect to see 5A peak into 8 ohms and with two slices 10A peak into 4 ohms (200W).
A short circuit test was done with full input swing and the heatsink temperature was monitored over time. The temperature reached 70C after a minute and then the short was removed. The amp survived. So I will add a temperature switch to disconnect the signal (or the mains) if the heatsink ever gets to 70 degrees.
Recall Post 90 showed that this autobias loop takes MOSFETs or BJTs with little change to the bias. Here I have used the circuit of Post 139 with one pair of power transistors for an 8 ohm load. Resistors R5 and R6 are reduced from 470 ohms to 170 ohms to give sufficient base current to get 5A peak (assuming the supply voltage is sufficient) and this resistance will depend on the Beta of the transistors used. Sim's Ic and Gm plots are below:
With BJT's the idle current for optimum bias is lower than for MOSFET's. Here optimum bias is 250mA! Nice
. THD at 1W is 0.15%. There is a soft clip starting from about half full power. Very nice IMO.The driver needs to deliver about 75mA peak. Too high for a standard opamp, but several in parallel feed could do it. With BD139/140 EF driver the input impedance is about 1k. Input voltage is about 0.5V peak. Voltage gain is about 40 into 8 ohms.
I have bench tested this circuit with AC direct drive (Vin1) and R5/R6 were 330 ohms, the Betas of the power transistors were 130 and I got 4.5A peak into 8 ohms with +/-40.2V rails. This is 36V peak with 1V across two 100mR resistors (R13 and one for rail current sensing) and 3V across the power transistors. With slightly higher rail voltage and with R5 and R6 270 ohms I expect to see 5A peak into 8 ohms and with two slices 10A peak into 4 ohms (200W).
A short circuit test was done with full input swing and the heatsink temperature was monitored over time. The temperature reached 70C after a minute and then the short was removed. The amp survived
. So I will add a temperature switch to disconnect the signal (or the mains) if the heatsink ever gets to 70 degrees.Hi Pawel,
Q2 was coupled to D2. Q1 at ambient. See Post 143 pic. Start current about 450mA and 400mA warmed up with no signal and 350mA just after running full power for 15 mins. It is safe thermally including a short with full swing for 1 min.
Q2 was coupled to D2. Q1 at ambient. See Post 143 pic. Start current about 450mA and 400mA warmed up with no signal and 350mA just after running full power for 15 mins. It is safe thermally including a short with full swing for 1 min.
Here's two BJT slices for 10A peak into 4 ohms for 200W in current drive.
Idle current is 520mA total. THD at 1W is 0.04% 2nd+3rd. Input is 1.2Vpk, 1.8mA for 40Vpk at 10A. Roll off -3dB is at 500kHz. The +/-9V supply is 2W total.
I think it's now a complete current drive power amp. Time to build and listen.
I think it's now a complete current drive power amp
. Time to build and listen.Now an interim listening test - one channel only. First I used parallel slice on 80V with 600mA total idle current with an 8 ohm bookshelf speaker.
Each slice has its own driver allowing TO-92's (Q3 Beta of 170 and Q4 of 200) and input impedance of each slice is 2k ohms. Driver bias is about 20mA and peak drive is about 50mA. With Beta of 120 for the NPN and 130 for the PNP and 220 ohms pull up to 9V gives 4.3A peak per slice, 9A total or 34V peak into 4 ohms.
The result was an unnaturally bright sound due to current drive of a loudspeaker intended for voltage drive. It is due to the speakers impedance versus frequency rising at the high frequency end. Adding an 8 ohm dummy load in parallel with the loudspeaker brought the the speakers frequency response close to normal (I don't have a speaker spectrum analyser to confirm how close). The music I listened to sounded clean with no noticeable distortion up to clip.
An X-Y plot for the above arrangement up to clip with a 4 ohm dummy load is shown below:
The plot is inverting with X axis of 0.2V per division and Y axis is 13V per division (my scope calibration is a bit off). The crossover regions covers about 2A or 300mV on the X-axis. This is with a 0.1 ohm degeneration resistor on each slice. When these resistors are shorted the gain doubles which means there is about 6dB of local negative feedback. This is a very low feedback amplifier yet it is surprisingly linear since the idle bias has been set to 300mA per stage for best overall linearity.
A second listening test was with voltage feedback so that the 8 ohm resistor across my speaker is not needed. I used just one slice with a 2k2 resistor in series with the input and then a 47k feedback resistor from the output to the driver input. The voltage gain becomes x21 with no load.
With an 8 ohm load the gain halves. So the output resistance of the amp is now 8 ohms. And the feedback used is still about 6dB. The sound with voltage feedback matches the above current drive amp sound with the 8 ohm in parallel with the speaker. If you need a higher gain then use a 470 ohm input resistor and a 22k feedback resistor (gain is x46 unloaded and output resistance is still 8 ohms).
The result was an unnaturally bright sound due to current drive of a loudspeaker intended for voltage drive. It is due to the speakers impedance versus frequency rising at the high frequency end. Adding an 8 ohm dummy load in parallel with the loudspeaker brought the the speakers frequency response close to normal (I don't have a speaker spectrum analyser to confirm how close). The music I listened to sounded clean with no noticeable distortion up to clip.
An X-Y plot for the above arrangement up to clip with a 4 ohm dummy load is shown below:
The plot is inverting with X axis of 0.2V per division and Y axis is 13V per division (my scope calibration is a bit off). The crossover regions covers about 2A or 300mV on the X-axis. This is with a 0.1 ohm degeneration resistor on each slice. When these resistors are shorted the gain doubles which means there is about 6dB of local negative feedback. This is a very low feedback amplifier yet it is surprisingly linear since the idle bias has been set to 300mA per stage for best overall linearity.
A second listening test was with voltage feedback so that the 8 ohm resistor across my speaker is not needed. I used just one slice with a 2k2 resistor in series with the input and then a 47k feedback resistor from the output to the driver input. The voltage gain becomes x21 with no load.
With an 8 ohm load the gain halves. So the output resistance of the amp is now 8 ohms. And the feedback used is still about 6dB. The sound with voltage feedback matches the above current drive amp sound with the 8 ohm in parallel with the speaker. If you need a higher gain then use a 470 ohm input resistor and a 22k feedback resistor (gain is x46 unloaded and output resistance is still 8 ohms).
Last edited:
Hi miralin,
Thanks for your question. The parts you mention are used to add a constant floating voltage to the Schottky voltage. The Schottky diodes give about 300mV at idle and the autobias spreader transistors need about 0.55 volts, so about 250mV needs to be added. I use constant current sources to generate this fixed floating voltage in the resistors. Prior to Post 139 I used resistors to +/-12V auxiliary rails which works fine in a simulation but not in bench testing because the MOSFETs blew when I turned off the mains (the +/-12V auxiliary rails collapsed faster than the main rails causing huge through currents in the MOSFETs). So I added the transistor CCS parts (Q5,Q6,...) to make it fail-safe at power down. No problems since then.
BTW The use of sensing diodes for the autobias loop is like the E Van Drecht patent (English translation here) such as his Fig 9c below
and using Schottky diodes like peufeu showed here with the constant floating voltage added to the Schottky voltage
but I have gone with E Van Drecht's output transistors with no drivers. Instead I use a driver stage for Vin which operates at low voltage and therefore low power and can use faster common low voltage transistors as I demonstrated in the previous post.
I found a similar idea posted by kenpeter here September 2009 but with 0.1 ohm power resistors in series with the Schottky diodes so spreader transistors don't need the floating voltage sources and associated CCS's. I was trying to avoid using any emitter/source resistors in my version because they generate unwanted higher order harmonics which necessities using lots and lots of negative feedback to eliminate them (more on that here).
I hope that helps answer your question.
Thanks for your question. The parts you mention are used to add a constant floating voltage to the Schottky voltage. The Schottky diodes give about 300mV at idle and the autobias spreader transistors need about 0.55 volts, so about 250mV needs to be added. I use constant current sources to generate this fixed floating voltage in the resistors. Prior to Post 139 I used resistors to +/-12V auxiliary rails which works fine in a simulation but not in bench testing because the MOSFETs blew when I turned off the mains (the +/-12V auxiliary rails collapsed faster than the main rails causing huge through currents in the MOSFETs). So I added the transistor CCS parts (Q5,Q6,...) to make it fail-safe at power down. No problems since then.
BTW The use of sensing diodes for the autobias loop is like the E Van Drecht patent (English translation here) such as his Fig 9c below
and using Schottky diodes like peufeu showed here with the constant floating voltage added to the Schottky voltage
but I have gone with E Van Drecht's output transistors with no drivers. Instead I use a driver stage for Vin which operates at low voltage and therefore low power and can use faster common low voltage transistors as I demonstrated in the previous post.
I found a similar idea posted by kenpeter here September 2009 but with 0.1 ohm power resistors in series with the Schottky diodes so spreader transistors don't need the floating voltage sources and associated CCS's. I was trying to avoid using any emitter/source resistors in my version because they generate unwanted higher order harmonics which necessities using lots and lots of negative feedback to eliminate them (more on that here).
I hope that helps answer your question.
This is my first 2 channel listening test using a single slice output stage and driver with 6dB of voltage feedback run on a 43V supply (with two floating supplies for 2 channels on one transformer) and two 8 ohm speakers.
The subcircuit is same as Post 154 above with the driver stage included. It gives about 20W into 8 ohms with soft clip starting at about 12W. At 10 W the simulation THD is 0.19% (file attached). The sim FFT plot is shown below at 10 watts:
Notice the harmonics rolloff at about 60dB per decade which means only the 2nd and 3rd are of any significance to what we might hear. The 2nd here is well below audibility (BTW any 2nd below 1% is inaudible). At 1W THD is 0.05%. My listening test averaging at a few watts was free from distortion. Having two channels now made speech very clear and natural and the music was very enjoyable.
The gain with 8 ohms is 12x and no load 23x. So output impedance is about 8 ohms.
The subcircuit is same as Post 154 above with the driver stage included. It gives about 20W into 8 ohms with soft clip starting at about 12W. At 10 W the simulation THD is 0.19% (file attached). The sim FFT plot is shown below at 10 watts:
Notice the harmonics rolloff at about 60dB per decade which means only the 2nd and 3rd are of any significance to what we might hear. The 2nd here is well below audibility (BTW any 2nd below 1% is inaudible). At 1W THD is 0.05%. My listening test averaging at a few watts was free from distortion. Having two channels now made speech very clear and natural and the music was very enjoyable.
The gain with 8 ohms is 12x and no load 23x. So output impedance is about 8 ohms.
Here the bias transistors of the driver are reconfigured to the diamond version to give higher input resistance (about 500k ohms). The input resistor can now be increased to 10k ohms (was 1k). The feedback loop gain is doubled (since Rin doesn't degrade the open loop gain) giving an output resistance of about 4 ohms. The bench version has a voltage gain of 16x into 8 ohms and 25x no load. A simulation file is attached.
With the extra active transistors (Q7,Q8) compensation is needed to stop oscillation when the input resistance is more than a few kohms. R9 and C5 keep the input node impedance low above 100kHz where oscillation can occur. The output impedance therefore increases at HF reaching 8 ohms at 100kHz. The amps f-3dB frequency is about 200kHz.
If you prefer an amplifier that doesn't need any compensation then the previous one with 1 k ohm input is yours. This one, with 10k ohms input resistance, is good when you use many output stages in parallel (or need the higher input resistance without an extra buffer stage).
If you prefer an amplifier that doesn't need any compensation then the previous one with 1 k ohm input is yours. This one, with 10k ohms input resistance, is good when you use many output stages in parallel (or need the higher input resistance without an extra buffer stage).
Last edited:
Hi Minek,
Thanks for asking. To make the output stage a follower you simply move the common from one side of the load to the other side. Also the subcircuit "Com" goes to "Vout" and the subcircuit has a resistor (R13) added to the splitter drive node for HF stability. Then add a HV opamp and viola.
BTW I used Subcircuit 2 (not a diamond driver) since the HF stability is better and you don't need Megohm input resistance for this opamp. Note the 0p5 feedback capacitor is the parasitic capacitance of a typical resistor. Just sims so far; I haven't bench tested this yet. Now you probably want a higher rail voltage, then parallel several stages with one opamp. You will also need the output inductor and a Zobel with real-world cables and speakers!
The LT6090-5 is the decompensated version for a minimum gain of 5. It is not a very-high slew rate opamp but is OK for 20kHz full power bandwidth. I notice very high slew rate and high voltage opamps seem fairly rare and expensive, so you will need to use a good discrete input stage/VAS to get really high slew rates. I'll leave that to yourself or anyone as I'm happy with the output stage with gain as in posts above. Also the capacitor splitting method provides speaker protection. It's simple, good sounding, robust and speaker safe -- so what more do we need for a good sounding amp? (I think many know now it's current-drive of (active) speakers with some EQ of course).
Thanks for asking. To make the output stage a follower you simply move the common from one side of the load to the other side. Also the subcircuit "Com" goes to "Vout" and the subcircuit has a resistor (R13) added to the splitter drive node for HF stability. Then add a HV opamp and viola
.The LT6090-5 is the decompensated version for a minimum gain of 5. It is not a very-high slew rate opamp but is OK for 20kHz full power bandwidth. I notice very high slew rate and high voltage opamps seem fairly rare and expensive, so you will need to use a good discrete input stage/VAS to get really high slew rates. I'll leave that to yourself or anyone as I'm happy with the output stage with gain as in posts above. Also the capacitor splitting method provides speaker protection
. It's simple, good sounding, robust and speaker safe -- so what more do we need for a good sounding amp? (I think many know now it's current-drive of (active) speakers with some EQ of course).