The interesting fact I uncovered is the very primitive current limiting of the 303 helps double the effective useful range. By the time things are getting serrious even alcohol should not keep the truth away. My 7 amps should offer at least 10 usable amps.
I have two spare holes on the heat sinks , three in fact. If I use the 1000 VA transformer and some Motorola TO220 drivers I can go that far. The VAS turned up to 6 mA ( 2SD756 if so ). I would fit the 1000 VA first and would hope that is a big step forward. Next the voltage raised ( 80 ) as it reduces Cob a bit. After that the extra 2 devices. Then the drivers. Finally the VAS. Each piece logged as to how it sounds
Here is a little speculation I did ages ago when the 2000 VA 0-55 0-55 turned up, I have 2 . Doubtless unworkable or something upside down? All this is to say in the least number of practical steps to how it might work. Remember that SOA if very lucky would allow 9 amps sine wave. The cascode is a happy accident. The MPSA 42 is a very low gain device. The BC337 or BC547C a high gain device. The overall FT might still be that of MPSA42. I can use better if I choose. The cascode is set high to avoid Early effect. In reality Early effect might be rather nice. I have seen a small resistor emitter to collector on old designs. 33R for example. That swamps the Re of 6R. The long tail pair 10 R is to remind me to try a value.
The idea of the double VAS is to have like a bicycle two feet on the peddles. That is to source and sink equally , how much current comes later. The 27K 3 nF is to remind me to adjust the loading for least 50 kHz distortion. BCV 61 will outperform any home made current mirror in terms of balance. It's 20 V rating is OK as 2V is what it will see. 2 V is usefully away from the 0.55V sometimes seen. If the BCV swamping resitors set at 0.2 V and Vce 0.2 V there is still 1.6 V of headroom . Although current a driven VAS I think that is wise. Much like a Peter Walker design. Alas BCV 62 would be useless for current mirror No 2.
Note I use power transistors as drivers. The reason is they will be very high gain if this low in current and can cope with 2 watts free air if given some PCB copper. The 4R7 is to stop punch through. I suspect 1 R would be OK.
As I throw away a few volts to the cascode I could equal that to the long tail. If so a simple resistor should be OK and might even be nice. The 68 R RC filter with I guess 2200 uF could have R increased. The LTP CCS is just to say do something.
The outputs can be Exicon 10N/P20 ( 25 ) and VAS at 7mA ( 100 R 18K and 4n7 ). If so Vbe MPSA42 becomes a single 470R. That would be a mildly improved HH-1200. There would be no driver or pre driver and no thermal tracking.
I have two spare holes on the heat sinks , three in fact. If I use the 1000 VA transformer and some Motorola TO220 drivers I can go that far. The VAS turned up to 6 mA ( 2SD756 if so ). I would fit the 1000 VA first and would hope that is a big step forward. Next the voltage raised ( 80 ) as it reduces Cob a bit. After that the extra 2 devices. Then the drivers. Finally the VAS. Each piece logged as to how it sounds
Here is a little speculation I did ages ago when the 2000 VA 0-55 0-55 turned up, I have 2 . Doubtless unworkable or something upside down? All this is to say in the least number of practical steps to how it might work. Remember that SOA if very lucky would allow 9 amps sine wave. The cascode is a happy accident. The MPSA 42 is a very low gain device. The BC337 or BC547C a high gain device. The overall FT might still be that of MPSA42. I can use better if I choose. The cascode is set high to avoid Early effect. In reality Early effect might be rather nice. I have seen a small resistor emitter to collector on old designs. 33R for example. That swamps the Re of 6R. The long tail pair 10 R is to remind me to try a value.
The idea of the double VAS is to have like a bicycle two feet on the peddles. That is to source and sink equally , how much current comes later. The 27K 3 nF is to remind me to adjust the loading for least 50 kHz distortion. BCV 61 will outperform any home made current mirror in terms of balance. It's 20 V rating is OK as 2V is what it will see. 2 V is usefully away from the 0.55V sometimes seen. If the BCV swamping resitors set at 0.2 V and Vce 0.2 V there is still 1.6 V of headroom . Although current a driven VAS I think that is wise. Much like a Peter Walker design. Alas BCV 62 would be useless for current mirror No 2.
Note I use power transistors as drivers. The reason is they will be very high gain if this low in current and can cope with 2 watts free air if given some PCB copper. The 4R7 is to stop punch through. I suspect 1 R would be OK.
As I throw away a few volts to the cascode I could equal that to the long tail. If so a simple resistor should be OK and might even be nice. The 68 R RC filter with I guess 2200 uF could have R increased. The LTP CCS is just to say do something.
The outputs can be Exicon 10N/P20 ( 25 ) and VAS at 7mA ( 100 R 18K and 4n7 ). If so Vbe MPSA42 becomes a single 470R. That would be a mildly improved HH-1200. There would be no driver or pre driver and no thermal tracking.
Last edited:
Nigel? Are you all right? You are scaring the hell out of me! I counted (gulp!) 23 transistors there, must be your life's first! 
A suggestion, if I may - instead of 1N4148, use BAV20, MUCH faster.
Bypass all signal path caps with polypropylene smaller value caps.
Get that darn inductor at the output out of the way, it's ruining your damping factor, and besides, you will not need it, only amps with insufficent power supplies actually need it.
A suggestion, if I may - instead of 1N4148, use BAV20, MUCH faster.
Bypass all signal path caps with polypropylene smaller value caps.
Get that darn inductor at the output out of the way, it's ruining your damping factor, and besides, you will not need it, only amps with insufficent power supplies actually need it.
IN4148 from memeory is not too bad. It is any diode really.
I could loose two transitors and be OK. I must not loose the BCV 61 as that allows some slewing without too much DC offset. The 2 x 2SA1085 is very helpful. MPSA 42/92 could be 2N5551/5401. BC337-40 could be BC109C if my brothers stock has some. The 2200 uF tells me to think hard. The idea is industrial grade PSU and high grade local.
SOA might only be 9 amps total giving 18 amps possible. If MOS FET it might rise 50%. 18 amps 2 R is 650W. Seems about right. Ron would be about 0R4 with 6 FET.
Input cap is 10 uF poly. The feedback one could have 1 uF . The gain of the amp is 76. That is a CD player can drive it. A gain of 48 seems ideal if 47K 1K . It could be gain of 20 ( VAS comp? ). That is 47K and 2K4. That would allow feeding signal into input minus. Any op amp should drive that. My experiments suggest it is a waste of time. One day it might not be.
We could loose 7 devices and have a better HH-1200. I suspect it might just be the better amp. For now that is not important. Been there, done that. What I have done is graft a Hitachi onto a Quad via HH. The HH was used by the BBC as a do everything amp. A Range Rover before they were posh.
Brief specifications HH.
Power Output Per Channel At Clip
VX150 80W (4 Ohm) 60W (8 Ohm) 170W (8 Ohm Bridged Mono)
VX200 100W (4 Ohm) 65W (8 Ohm) 205W (8 Ohm Bridged Mono)
VX300 155W (4 Ohm) 100W (8 Ohm) 315W (8 Ohm Bridged Mono)
VX450 240W (4 Ohm) 150W (8 Ohm) 500W (8 Ohm Bridged Mono)
VX600 325W (4 Ohm) 200W (8 Ohm) 650W (8 Ohm Bridged Mono)
VX900 475W (4 Ohm) 285W (8 Ohm) 965W (8 Ohm Bridged Mono)
VX1200 630W (4 Ohm) 375W (8 Ohm) 1300W (8 Ohm Bridged Mono)
T.H.D. <0.03% (<0.08% for VX1200) at rated o.p. 20H-20KHz
I.M.D. <0.03%
Inputs XLR and Jack. Electroic Balance, transformer balanced (optional) or unbalanced selected by switch.
Damping Factor >300 at 100Hz. 8 Ohm load.
Hum and Noise -100dB (20Hz - 20KHz)
Slew Rate 60V/uS
Outputs Binding posts (VX200 - VX600 also fitted with XLRs)
Protection Short circuit, open circuit, load missmatch, thermal and DC.
Cooling VX150-VX300 Convection cooled. VX450-VX1200 2 speed fan*.
*runs silently on slow speed and rarly runs full speed.
Rack Size and Weight
VX150 1U 6.3Kg
VX200 2U 9.6Kg
VX300 2U 11.5Kg
VX450 2U 16Kg
VX600 2U 16.3Kg
VX900 3U 22.6Kg
VX1200 3U 24.8Kg
I could loose two transitors and be OK. I must not loose the BCV 61 as that allows some slewing without too much DC offset. The 2 x 2SA1085 is very helpful. MPSA 42/92 could be 2N5551/5401. BC337-40 could be BC109C if my brothers stock has some. The 2200 uF tells me to think hard. The idea is industrial grade PSU and high grade local.
SOA might only be 9 amps total giving 18 amps possible. If MOS FET it might rise 50%. 18 amps 2 R is 650W. Seems about right. Ron would be about 0R4 with 6 FET.
Input cap is 10 uF poly. The feedback one could have 1 uF . The gain of the amp is 76. That is a CD player can drive it. A gain of 48 seems ideal if 47K 1K . It could be gain of 20 ( VAS comp? ). That is 47K and 2K4. That would allow feeding signal into input minus. Any op amp should drive that. My experiments suggest it is a waste of time. One day it might not be.
We could loose 7 devices and have a better HH-1200. I suspect it might just be the better amp. For now that is not important. Been there, done that. What I have done is graft a Hitachi onto a Quad via HH. The HH was used by the BBC as a do everything amp. A Range Rover before they were posh.
Brief specifications HH.
Power Output Per Channel At Clip
VX150 80W (4 Ohm) 60W (8 Ohm) 170W (8 Ohm Bridged Mono)
VX200 100W (4 Ohm) 65W (8 Ohm) 205W (8 Ohm Bridged Mono)
VX300 155W (4 Ohm) 100W (8 Ohm) 315W (8 Ohm Bridged Mono)
VX450 240W (4 Ohm) 150W (8 Ohm) 500W (8 Ohm Bridged Mono)
VX600 325W (4 Ohm) 200W (8 Ohm) 650W (8 Ohm Bridged Mono)
VX900 475W (4 Ohm) 285W (8 Ohm) 965W (8 Ohm Bridged Mono)
VX1200 630W (4 Ohm) 375W (8 Ohm) 1300W (8 Ohm Bridged Mono)
T.H.D. <0.03% (<0.08% for VX1200) at rated o.p. 20H-20KHz
I.M.D. <0.03%
Inputs XLR and Jack. Electroic Balance, transformer balanced (optional) or unbalanced selected by switch.
Damping Factor >300 at 100Hz. 8 Ohm load.
Hum and Noise -100dB (20Hz - 20KHz)
Slew Rate 60V/uS
Outputs Binding posts (VX200 - VX600 also fitted with XLRs)
Protection Short circuit, open circuit, load missmatch, thermal and DC.
Cooling VX150-VX300 Convection cooled. VX450-VX1200 2 speed fan*.
*runs silently on slow speed and rarly runs full speed.
Rack Size and Weight
VX150 1U 6.3Kg
VX200 2U 9.6Kg
VX300 2U 11.5Kg
VX450 2U 16Kg
VX600 2U 16.3Kg
VX900 3U 22.6Kg
VX1200 3U 24.8Kg
Last edited:
DIY Micro SIM ! It's rare we can score over the modern world. Even Nano!
https://www.youtube.com/watch?v=1u5-s63Dc2U
https://www.youtube.com/watch?v=5qktcs8rRjE
https://www.youtube.com/watch?v=1u5-s63Dc2U
https://www.youtube.com/watch?v=5qktcs8rRjE
Last edited:
Can be done this way. The voltage doublers allow low ripple and higher power cheaply. No real need for a tail CCS. This is just specultion and can be bettered. VAS can be 10 mA. With MOS FET's slew rate can be reduced as output device is fast. 300 mA to output devices = 50 watts of static heat.
Add zener to rails. The resistors in the doubler were from another project and not a tested value. It was to nudge me in the right direction.
Would help if VAS caps were in the right place and applies to the other diagram. Collector 2 to base 1 seems best. The split to output is OK.
Last edited:
The interesting fact I uncovered is the very primitive current limiting of the 303 helps double the effective useful range. By the time things are getting serrious even alcohol should not keep the truth away. My 7 amps should offer at least 10 usable amps.
I have two spare holes on the heat sinks , three in fact. If I use the 1000 VA transformer and some Motorola TO220 drivers I can go that far. The VAS turned up to 6 mA ( 2SD756 if so ). I would fit the 1000 VA first and would hope that is a big step forward. Next the voltage raised ( 80 ) as it reduces Cob a bit. After that the extra 2 devices. Then the drivers. Finally the VAS. Each piece logged as to how it sounds
Here is a little speculation I did ages ago when the 2000 VA 0-55 0-55 turned up, I have 2 . Doubtless unworkable or something upside down? All this is to say in the least number of practical steps to how it might work. Remember that SOA if very lucky would allow 9 amps sine wave. The cascode is a happy accident. The MPSA 42 is a very low gain device. The BC337 or BC547C a high gain device. The overall FT might still be that of MPSA42. I can use better if I choose. The cascode is set high to avoid Early effect. In reality Early effect might be rather nice. I have seen a small resistor emitter to collector on old designs. 33R for example. That swamps the Re of 6R. The long tail pair 10 R is to remind me to try a value.
The idea of the double VAS is to have like a bicycle two feet on the peddles. That is to source and sink equally , how much current comes later. The 27K 3 nF is to remind me to adjust the loading for least 50 kHz distortion. BCV 61 will outperform any home made current mirror in terms of balance. It's 20 V rating is OK as 2V is what it will see. 2 V is usefully away from the 0.55V sometimes seen. If the BCV swamping resitors set at 0.2 V and Vce 0.2 V there is still 1.6 V of headroom . Although current a driven VAS I think that is wise. Much like a Peter Walker design. Alas BCV 62 would be useless for current mirror No 2.
Note I use power transistors as drivers. The reason is they will be very high gain if this low in current and can cope with 2 watts free air if given some PCB copper. The 4R7 is to stop punch through. I suspect 1 R would be OK.
As I throw away a few volts to the cascode I could equal that to the long tail. If so a simple resistor should be OK and might even be nice. The 68 R RC filter with I guess 2200 uF could have R increased. The LTP CCS is just to say do something.
The outputs can be Exicon 10N/P20 ( 25 ) and VAS at 7mA ( 100 R 18K and 4n7 ). If so Vbe MPSA42 becomes a single 470R. That would be a mildly improved HH-1200. There would be no driver or pre driver and no thermal tracking.
Nigel,
On paper this is an interesting blend of the old - the Quad 303 output stage with a modern front end. Your work on the output stage I can follow as you are sticking with the spirit of the original design - using power output devices as drivers which should avoid any thermal issues with multiple output transistors.
With regard to the second differential VAS stage you have a lot of transistor junctions in there with a current mirror load. While the loading of the output stage should not burden the operation of that stage, it is one that is dynamic in nature - quite unlike a constant current source which is commonly used to have a reasonably constant current flow through the temperature sensing voltage multiplier.
Triple output stages are less easy to stabilize than normal two stage arrangements and one would not want to feed them with a fluctuating reference voltage. There may be some point I have missed that would benefit my understanding in this regard. A vague thought in this regard was the use of Thermaltrak devices for the driver stage instead of the 2SA, 2SC ones.
My idea is the output tripple of the Quad or Crimson type rough speaking is like one FET. The speed is slightly less so the BCV 61 will ensure the slewing is a bit faster as very little is lost on the input stage if so. If you like we not only have 2 feet on the peddles now, we have two strong feet. The double VAS for a small phase complexity has almost double the slew rate and is symetrical, A single VAS might have to be 4 times more able to do the same basic job. That brings doubts as to real stability. The analogy is the one legged cyclist even with physical devellopement can not push and pull on the one peddle so well. Worse is that a VAS has no momentum to get it along.
The HH was about 600 watts 2 R and 60 V/uS. For an amp with no input stage current mirror that is not bad. If my conjectire is right FET's do not require very fast drivers as the loop is maintained much better due to the fast output device. Bipolars when compounded to get high current gain are like cast iron pisons. 1920's engines as a sales gimmick had 4 valves per cylinder. Alas with iron pistons it was to fool the public that it was an aero engine. A lady was looking at her life on TV and saying tick to all the good things ( Miranda Hart ). We engineers do the same. In life it might not make for happiness and guess what, in engineering also. 6 V /uS will do 100 watts 8R as long as no stupidly high signal is passed and FET outputs. The Hypex sets this as about 50 kHz so I think that's about where I would. The Hypex is a very fast amp in the class D sense. The fact it has to limit the audio does not make for a dull sounding amp. Conjecture. It is much faster sounding as it is switching faster? Conjecture 2. It sounds brighter as some well controlled fizz has sub harmonics ? I have measured them so they exist, they should if the 1820's conjecture of Fourier is correct. Like Pioneer Legato Linear that reused aliasing products. Alas into non musical HF. They used non white light if you like to light a picture. What it did prove is extended HF has something we can hear. A digital source can be 22.05 kHz if standard CD. However it can switch in the 100's of MHz. Thus RC filtered and digital are not exactly the same. The reason I say this is do not like spoken English assume a right or wrong way. Shakespere used the common man to voice ideas. I suspect Mr S was one although uncommon in ability?
Quad used a lot of compensation. It is a bit optimistic to think this doodle will work. In cost terms it is starting to beat the MOS FET and it should not be drastically worse. The ultra HF distortion might be less good. As tripples ( NPN ) work well is Bryston and Crimson and Quad uses a very simple VAS we might infer this can work. We then hope to reduce the problems to VAS compenstaion alone. That way although we loose the purity of the Quad signal path we gain in being able to see exactly what we did.
A speculation. When my brother was around he said the Ft of transistors is less good than the common base specification which is seldom given. Lets put MPSA42 at 50 MHz. As it is common base if a cascode could Ft be greater ? This is implied in many 1970's text books. It seems to me that the cascode can be almost something for nothing if so. Who cares we loose 5 V. Forces us to build a propper PSU then.
The double VAS current mirror is to get as much as possible for few components. Hitachi proved it to work very well. It does attempt the impossible. With all respect to Douglas Self I think he sometimes goes into the same region as correct English. That is to say somethings are so important as to stop a person reading if one of the Cardinal sins of electronics is commited. Design is like a war. One sometimes goes to the region of moral doubt.
There is a better quadruple Widmer current mirror . It looses an exra volt I guess. If MPSA 92 it would be low cost. It is two standard ones where the diode side swaps over and they are in series. Origin is LM741 I think although I don't think used there. Vastly easier than the tripple transisor one.
The HH was about 600 watts 2 R and 60 V/uS. For an amp with no input stage current mirror that is not bad. If my conjectire is right FET's do not require very fast drivers as the loop is maintained much better due to the fast output device. Bipolars when compounded to get high current gain are like cast iron pisons. 1920's engines as a sales gimmick had 4 valves per cylinder. Alas with iron pistons it was to fool the public that it was an aero engine. A lady was looking at her life on TV and saying tick to all the good things ( Miranda Hart ). We engineers do the same. In life it might not make for happiness and guess what, in engineering also. 6 V /uS will do 100 watts 8R as long as no stupidly high signal is passed and FET outputs. The Hypex sets this as about 50 kHz so I think that's about where I would. The Hypex is a very fast amp in the class D sense. The fact it has to limit the audio does not make for a dull sounding amp. Conjecture. It is much faster sounding as it is switching faster? Conjecture 2. It sounds brighter as some well controlled fizz has sub harmonics ? I have measured them so they exist, they should if the 1820's conjecture of Fourier is correct. Like Pioneer Legato Linear that reused aliasing products. Alas into non musical HF. They used non white light if you like to light a picture. What it did prove is extended HF has something we can hear. A digital source can be 22.05 kHz if standard CD. However it can switch in the 100's of MHz. Thus RC filtered and digital are not exactly the same. The reason I say this is do not like spoken English assume a right or wrong way. Shakespere used the common man to voice ideas. I suspect Mr S was one although uncommon in ability?
Quad used a lot of compensation. It is a bit optimistic to think this doodle will work. In cost terms it is starting to beat the MOS FET and it should not be drastically worse. The ultra HF distortion might be less good. As tripples ( NPN ) work well is Bryston and Crimson and Quad uses a very simple VAS we might infer this can work. We then hope to reduce the problems to VAS compenstaion alone. That way although we loose the purity of the Quad signal path we gain in being able to see exactly what we did.
A speculation. When my brother was around he said the Ft of transistors is less good than the common base specification which is seldom given. Lets put MPSA42 at 50 MHz. As it is common base if a cascode could Ft be greater ? This is implied in many 1970's text books. It seems to me that the cascode can be almost something for nothing if so. Who cares we loose 5 V. Forces us to build a propper PSU then.
The double VAS current mirror is to get as much as possible for few components. Hitachi proved it to work very well. It does attempt the impossible. With all respect to Douglas Self I think he sometimes goes into the same region as correct English. That is to say somethings are so important as to stop a person reading if one of the Cardinal sins of electronics is commited. Design is like a war. One sometimes goes to the region of moral doubt.
There is a better quadruple Widmer current mirror . It looses an exra volt I guess. If MPSA 92 it would be low cost. It is two standard ones where the diode side swaps over and they are in series. Origin is LM741 I think although I don't think used there. Vastly easier than the tripple transisor one.
Nige, you are comparing apples with watermelons here. MOSFET output devices do have their clear advantages, such as requiring minimum drive current, so a BJT pair placed before them sually serves more as a buffer than anything else, it's there more to offload the VAS than to actually drive the MOSFETs. and yet, their great advantage of a tremenou frequeny response is also their greatest problem, as it makes the succeptible to oscillation much more than a standard BJT power device with an Ft of say 6 MHz.
You know what I'm about to say now, but I'll say it anyway - no MOSFET amp I have ever heard could do drums as faithfully as BJTs, period. Once they can do that, I'll gladly use them.
What I do objket to is your failure to use a JFET input. Demian taught me a thing or two about them, the main one being that JFETs are gluttons for current, so where for a BJT diff pair you'd use up to say 1.5 mA oer device, a JFET will work MUCH better at say 5 mA per device. The only problem is voltage, they are usually limited to 40V at best, so an 2SK170 input stage needs a cascode with say 2SC2240 low noise Toshiba BJTs.
As for the slew rate, it is usually determind by the Miller cap compensation on the VAS and the current fed to the input stage. Even assuming a relatively high cap value of say 33 pF, if the current i 5 mA, that's still 151 V/uS. That will get you incredible bandwidth, however assuming you run your VAS at something decent, say 10...15 mA of bias current. That will get most transistors well into their most linear service range, I would suggest 2SC3503/2SA1381.
You know what I'm about to say now, but I'll say it anyway - no MOSFET amp I have ever heard could do drums as faithfully as BJTs, period. Once they can do that, I'll gladly use them.
What I do objket to is your failure to use a JFET input. Demian taught me a thing or two about them, the main one being that JFETs are gluttons for current, so where for a BJT diff pair you'd use up to say 1.5 mA oer device, a JFET will work MUCH better at say 5 mA per device. The only problem is voltage, they are usually limited to 40V at best, so an 2SK170 input stage needs a cascode with say 2SC2240 low noise Toshiba BJTs.
As for the slew rate, it is usually determind by the Miller cap compensation on the VAS and the current fed to the input stage. Even assuming a relatively high cap value of say 33 pF, if the current i 5 mA, that's still 151 V/uS. That will get you incredible bandwidth, however assuming you run your VAS at something decent, say 10...15 mA of bias current. That will get most transistors well into their most linear service range, I would suggest 2SC3503/2SA1381.
I am an absolute idiot regarding booze, since I use it in next to nothng quantities, once in two months or so of a bit of sherry or Cointreau (orange brandy from where else but France), but even I can actually tell apart the taste of Grant's whiskey.
You take the booze, I'll settle for haggis myself. Finally made it to Edinburgh in 2009 and LOVED it. It had been one of my preferred destinations ever since I left UK, I've been wanting to go to Scotland. Even my cell phone calls me with the tune "Scotland, The Brave".
I felt like the interviewer in the Monty Python Ann Elk sketch over the response to my last post.
Last edited:
Well, a lot of doodles do actually work, but how is another matter altogether. Thank you for a stable amp with a slew rate of 20 V/uS, in this day and age. Mind you, I'm not a slew rate addict, but I do think it's symptomatic that fast a,ps tend to sound best to me. I think a nominally 100W/8 Ohms amp should have a slew rate of at least 100 V/uS because I know that's easily possible if you set out for it to be so.
Just to quickly remind one and all that Otala and Lohstroh got 100 V/uS from a nominally 25/50W into 8/4 Ohms amp in 1973.
Just to quickly remind one and all that Otala and Lohstroh got 100 V/uS from a nominally 25/50W into 8/4 Ohms amp in 1973.
Dejan, that "requiring minimum drive current" is not true in the real world of the circuit, which maybe is the reason for the MOSFET amps not doing it for you. At the moment I'm playing with a high power, low distortion amplier with MOSFET OS in the simulator, and, unfortunately, they do require high gate current - this is a combination of the MOSFET's parasitic capacitances, and effective feedback driving them hard to linearise crossover distortion.
As it is, at the moment I've cut back a bit on the idle current of the drivers, which has reduced the performance above the audio bandwidth - but I didn't want to go silly on the heat in the driver devices.
Last edited:
I'm responding to the beginning of this thread, because I didn't feel like reading 1700 pages of posts.
But to me, knowing the limitations of measurements is important, as well as knowing how these measurements are performed. Ever noticed that those great distortion numbers from the amps of the 70's never mentioned what frequency and power level the distortion was measured at? Almost any amp can have great distortion when driven to a 1V output and at 1KHz and driving an 8Ω resistive load. Specmanship plays a huge part. Very few manufacturers will tell you their piece of equipment sounds or measures anything less than perfect. So every(or at least most) amp out there has great specs, and then we hear differences in amplifiers with the same or similar specs, and we conclude that good sound can't be measured.
THD measurements mask frequency response, phase shift, spurious response and things like TIM distortion. Any measurement that involves Fourier analysis can misrepresent things that are aperiodic in nature, which is like 60% of what music is.
So it may not be that these things cannot be measured, but that some of the ways we measure these things don't cover everything that our ears can hear.
But to me, knowing the limitations of measurements is important, as well as knowing how these measurements are performed. Ever noticed that those great distortion numbers from the amps of the 70's never mentioned what frequency and power level the distortion was measured at? Almost any amp can have great distortion when driven to a 1V output and at 1KHz and driving an 8Ω resistive load. Specmanship plays a huge part. Very few manufacturers will tell you their piece of equipment sounds or measures anything less than perfect. So every(or at least most) amp out there has great specs, and then we hear differences in amplifiers with the same or similar specs, and we conclude that good sound can't be measured.
THD measurements mask frequency response, phase shift, spurious response and things like TIM distortion. Any measurement that involves Fourier analysis can misrepresent things that are aperiodic in nature, which is like 60% of what music is.
So it may not be that these things cannot be measured, but that some of the ways we measure these things don't cover everything that our ears can hear.
Last edited:
Are you using "stopper" resistors near the MOSFET gates - by way of making a low-pass filter.
- Status
- Not open for further replies.