Hi Guys
Informed caution is good; suspicion is negative.
I've used the MJL1302-3281s since Motorola released them. They licenced the manufacture of the 2SA1302-2SC3281 and the drivers 2SA1306-2SC3298, which were highly complementary devices, much more so than the MJs and MJEs Motorola had been making up till then. The US market in particular would not accept the Japanese part numbers, so Motorloa recast them as MJLs. Some of the processing methods spilled over in the adaption of the MJ1502x series into flat packages, but those remained as low-Ft parts.
The MJL1302A-3281A are within 10% match just by design. If you bother to match them closer THD will go down dramatically.
In an amplifier designed for a 20kHz bandwidth, one would expect that a "slow" 4MHz device should be fast enough - and it is in the right circuit. It has 200 times the desired bandwidth in effect, and is still 40 times faster than a 100kHz bandwidth. In most circuits, the base capacitance is bootstrapped and is nonproblematic. It is only when the voltage across the device is low that the capacitance multiplies and the circuit runs into drive problems. The capacitance curves have shown this since transistor data was first published, so again, we see a false economy of having too-low rails for a given power output.
The craziness of testing with 100kHz+ signals is of no purpose but ego stroking. No audio source has content much above 20kHz and the profile of frequencies in nature and in music follow that pattern, with a sort of natural rolloff through the mids to treble - not that the amp should have this property. Depending on which math you wish to believe, Bruno says TIM lives outside the feedback loop and is entirely mitigated by the use of degeneration of the input stage. band-limiting the input signal also prevents TIM provided THD20 is low for the amp. So, even here there is a difference of design protocols: one where degeneration is a given and one where it is not. Others think the amp has to have a zillion vols-per-microsecond slew rate to be "acceptable". Again, no music signal has ever been measured that a 20V/usec amp cannot handle. Band-limiting the input has many other virtues as those here and elsewhere have noted.
I have no emotional investment in the Thompson-Lin circuit as it is not what I prefer. I've built several over the years without issue, without base-stops et al as said above. The beastie I tested recently to prove the above circuit works fine. There is cross-rail high-frequency decoupling per Self and 820uF per rail of local storage. I lay my amps out so the output stage is very close to its power supply filters and I use distributed filtering. No problems so far.
Have fun
Informed caution is good; suspicion is negative.
I've used the MJL1302-3281s since Motorola released them. They licenced the manufacture of the 2SA1302-2SC3281 and the drivers 2SA1306-2SC3298, which were highly complementary devices, much more so than the MJs and MJEs Motorola had been making up till then. The US market in particular would not accept the Japanese part numbers, so Motorloa recast them as MJLs. Some of the processing methods spilled over in the adaption of the MJ1502x series into flat packages, but those remained as low-Ft parts.
The MJL1302A-3281A are within 10% match just by design. If you bother to match them closer THD will go down dramatically.
In an amplifier designed for a 20kHz bandwidth, one would expect that a "slow" 4MHz device should be fast enough - and it is in the right circuit. It has 200 times the desired bandwidth in effect, and is still 40 times faster than a 100kHz bandwidth. In most circuits, the base capacitance is bootstrapped and is nonproblematic. It is only when the voltage across the device is low that the capacitance multiplies and the circuit runs into drive problems. The capacitance curves have shown this since transistor data was first published, so again, we see a false economy of having too-low rails for a given power output.
The craziness of testing with 100kHz+ signals is of no purpose but ego stroking. No audio source has content much above 20kHz and the profile of frequencies in nature and in music follow that pattern, with a sort of natural rolloff through the mids to treble - not that the amp should have this property. Depending on which math you wish to believe, Bruno says TIM lives outside the feedback loop and is entirely mitigated by the use of degeneration of the input stage. band-limiting the input signal also prevents TIM provided THD20 is low for the amp. So, even here there is a difference of design protocols: one where degeneration is a given and one where it is not. Others think the amp has to have a zillion vols-per-microsecond slew rate to be "acceptable". Again, no music signal has ever been measured that a 20V/usec amp cannot handle. Band-limiting the input has many other virtues as those here and elsewhere have noted.
I have no emotional investment in the Thompson-Lin circuit as it is not what I prefer. I've built several over the years without issue, without base-stops et al as said above. The beastie I tested recently to prove the above circuit works fine. There is cross-rail high-frequency decoupling per Self and 820uF per rail of local storage. I lay my amps out so the output stage is very close to its power supply filters and I use distributed filtering. No problems so far.
Have fun
Interesting comments Kevin.
You seem to prefer fully symmetrical topologies in general, Bryston's design philosophy. Some of their designs, if I'm not mistaken, forgo LTP degeneration, presumably to increase loop gain at the expense of IP linearity, which even in an undegenerated form is already more linear than the rest of the amplifier and not a major source of distortion.
Can you comment on some of the issues that need to be considered and your recommendations for band limiting the input, particularly when an undegenerated bipolar input pair is used in the design? Do you think this approach has merit, if we are trying to hit a loop gain target while keeping the number of gain stages to a minimum, which introduce extra phase shift that usually needs to be compensated for by slowing down the front-end, which is in itself detrimental to performance.
I have had good success with very simple designs that rely on fast silicon and a minimum number of gain devices, but are able to achieve quite high ULGF intercepts, seemingly as a consequence. I've often wondered whether with more advanced and optimised compensation it could achieve performance comparable to more complex circuits.
You seem to prefer fully symmetrical topologies in general, Bryston's design philosophy. Some of their designs, if I'm not mistaken, forgo LTP degeneration, presumably to increase loop gain at the expense of IP linearity, which even in an undegenerated form is already more linear than the rest of the amplifier and not a major source of distortion.
Can you comment on some of the issues that need to be considered and your recommendations for band limiting the input, particularly when an undegenerated bipolar input pair is used in the design? Do you think this approach has merit, if we are trying to hit a loop gain target while keeping the number of gain stages to a minimum, which introduce extra phase shift that usually needs to be compensated for by slowing down the front-end, which is in itself detrimental to performance.
I have had good success with very simple designs that rely on fast silicon and a minimum number of gain devices, but are able to achieve quite high ULGF intercepts, seemingly as a consequence. I've often wondered whether with more advanced and optimised compensation it could achieve performance comparable to more complex circuits.
Hi Ranchu32
None of Bryston's designs use degeneration on the diff amps, nor base-stops on the outputs. I do prefer fully symmetric designs and the designs by Chris Russell and Stuart Taylor are ones I admire. Bryston's amps are on the verge of being unmeasurable by an AP - their preamps are already there. Their approach is to my way of thinking superior to most.
There do not have to be a lot of stages to have enough gain to have powerful feedback. Nor do you have to linearise every stage as Self shows to achieve low THD - at least not in the ways that he shows. Self admits to not liking symmetric circuits, yet made a reasonable effort to show what they can do. However, he did not use BJT types that people who do like symmetrics use, so crippled the result to some extent. For that matter, he has yet to update the nested feedback figure with corrected values after several editions of his book. Could be just extremely busy but looks a lot like a tinge of bias as he denounces nested feedback, too.
I believe there are some aspects of audible performance that we have not yet determined or know how to measure or simulate. When you sim a Bryston circuit the results can be disappointing, yet an AP has difficulty to measure it.
My recommendation for bandwidth limiting is no different than most people inasmuch as a 100kHz bandwidth is sufficient to allow through any musical harmonic that we may be able to detect (even beyond our hearing). Higher than this is just noise, so why amplify it?
Many PAs have an input conditioning circuit, such as balanced input adaptor, where the bandwidth limiting should be placed. The input of the PA itself does not have to be affected and can now be driven by a known low impedance. The input circuit can be given a small amount of gain, allowing the PA gain to be dialled back further reducing THD.
There are lots of ways to achieve low THD, as the many innovative designs here show. Some are not so low THD but have interesting ideas for investigation. The input stage on my high power amp uses double cascodes - Bob thought that was frightful - but they work great.
have fun
None of Bryston's designs use degeneration on the diff amps, nor base-stops on the outputs. I do prefer fully symmetric designs and the designs by Chris Russell and Stuart Taylor are ones I admire. Bryston's amps are on the verge of being unmeasurable by an AP - their preamps are already there. Their approach is to my way of thinking superior to most.
There do not have to be a lot of stages to have enough gain to have powerful feedback. Nor do you have to linearise every stage as Self shows to achieve low THD - at least not in the ways that he shows. Self admits to not liking symmetric circuits, yet made a reasonable effort to show what they can do. However, he did not use BJT types that people who do like symmetrics use, so crippled the result to some extent. For that matter, he has yet to update the nested feedback figure with corrected values after several editions of his book. Could be just extremely busy but looks a lot like a tinge of bias as he denounces nested feedback, too.
I believe there are some aspects of audible performance that we have not yet determined or know how to measure or simulate. When you sim a Bryston circuit the results can be disappointing, yet an AP has difficulty to measure it.
My recommendation for bandwidth limiting is no different than most people inasmuch as a 100kHz bandwidth is sufficient to allow through any musical harmonic that we may be able to detect (even beyond our hearing). Higher than this is just noise, so why amplify it?
Many PAs have an input conditioning circuit, such as balanced input adaptor, where the bandwidth limiting should be placed. The input of the PA itself does not have to be affected and can now be driven by a known low impedance. The input circuit can be given a small amount of gain, allowing the PA gain to be dialled back further reducing THD.
There are lots of ways to achieve low THD, as the many innovative designs here show. Some are not so low THD but have interesting ideas for investigation. The input stage on my high power amp uses double cascodes - Bob thought that was frightful - but they work great.
have fun
Hi Terry
I haven't built it on this board BUT i made a front-end board and used it with an existing output stage I have. That output stage is identical to this one in all the details objected to above. No problems at 400W+ output into 8R or into 4R with 100V rails and a hefty heat sink. For continuous use into 4R it would be good to reduce the rails or add more outputs. My board has eight pairs, so no worries there.
The PCB house I use charges about 170 for a prototype run, which I won't spend on a circuit that is not my fave. I have boards for the fully symmetric stuff including the large EF3 output stage. I don't use the discount board places as I would not trust them with designs I want to retain.
Have fun
I haven't built it on this board BUT i made a front-end board and used it with an existing output stage I have. That output stage is identical to this one in all the details objected to above. No problems at 400W+ output into 8R or into 4R with 100V rails and a hefty heat sink. For continuous use into 4R it would be good to reduce the rails or add more outputs. My board has eight pairs, so no worries there.
The PCB house I use charges about 170 for a prototype run, which I won't spend on a circuit that is not my fave. I have boards for the fully symmetric stuff including the large EF3 output stage. I don't use the discount board places as I would not trust them with designs I want to retain.
Have fun
Maaan, it's been a long time, but i've been so so busy.
Kevin, thank you for still working on this schematic.
Regarding the boards, some chinese factory can make three boards with shipping via China Post for 45 dollars
Size: 190.50*82.55MM 3pcs
Material/ Thickness: FR-4 / 1.6MM
Layer: 2
Copper thickness: 1/1OZ
Surface treatment: HASL
Soldermask: Green
Silkscreen: White
I suppose they are ok as they have good feedback.
You can also choose DHL for 10 more dollars i think.
By the way, I almost finished my control, protection and monitoring project I started for a certificate in Informatics. I will make a thread for it in the next two months i think. It uses an arduino(actual management) and an atmega8a(outputting video) and a portable dvd player with AV input. It implements AC monitoring, Soft start, delayed speaker coupling, DC protection, current monitoring and overcurrent protection (and i think it is fast enough to be considered a short circuit protection), temperature monitoring and pwm fan driving. I hope you guys will like it
Kevin, thank you for still working on this schematic.
Regarding the boards, some chinese factory can make three boards with shipping via China Post for 45 dollars
Size: 190.50*82.55MM 3pcs
Material/ Thickness: FR-4 / 1.6MM
Layer: 2
Copper thickness: 1/1OZ
Surface treatment: HASL
Soldermask: Green
Silkscreen: White
I suppose they are ok as they have good feedback.
You can also choose DHL for 10 more dollars i think.
By the way, I almost finished my control, protection and monitoring project I started for a certificate in Informatics. I will make a thread for it in the next two months i think. It uses an arduino(actual management) and an atmega8a(outputting video) and a portable dvd player with AV input. It implements AC monitoring, Soft start, delayed speaker coupling, DC protection, current monitoring and overcurrent protection (and i think it is fast enough to be considered a short circuit protection), temperature monitoring and pwm fan driving. I hope you guys will like it
Hi Guys
Brimat: NEVER use 1oz copper for a power amp or any circuit where current is high. The layout is designed for 2oz copper.
2oz copper used to be the default but these days you have to specify it to get it. It usually does not cost any more than 1oz.
A friend used a cheapo PCB house and got some bad boards. I'll stick with who I know.
Your project sounds interesting. people have been connecting computers to tube amps and solid-state amps for years. I prefer linear as it can be repaired or modified and kept working for decades; digital stuff goes out of fashion so quickly it can be difficult to maintain some of it over the long term.
Have fun
Brimat: NEVER use 1oz copper for a power amp or any circuit where current is high. The layout is designed for 2oz copper.
2oz copper used to be the default but these days you have to specify it to get it. It usually does not cost any more than 1oz.
A friend used a cheapo PCB house and got some bad boards. I'll stick with who I know.
Your project sounds interesting. people have been connecting computers to tube amps and solid-state amps for years. I prefer linear as it can be repaired or modified and kept working for decades; digital stuff goes out of fashion so quickly it can be difficult to maintain some of it over the long term.
Have fun
Hi Guys
One slip of the finger and several minutes of typing are gone. Ironically, it was to save the text before posting.
Anyhoo...
The short form of what I had written is simply that there are many ways to physically construct a circuit. The layout above is only one, and it is the intuitive first form being a unitised board supporting the entire circuit. This has construction compromises attached.
For example, one can solder the MJLs from the underside and mount the board parallel to the flat side of the heat sink, necessitating a mass of aluminium that can fit into 3U or taller. This is not necessarily good for the solder connections on the PCB but it frees up floor space within the chassis to allow dual PTs. It also allows one to wire things as they wish as noneof the audio connections are on the board, and thus not in any prescribed position that may be difficult to accommodate by a person of average metal working skill.
Alternatively, one could have the PCB perpendicular to the heat sink, allowing a shorter heat sink that might not be able to dissipate all the heat generated depending on the supply voltages, idle settings, and usage.If the design is being used for lower power, then the heat management issue is averted and the lower profile justified. The PCB also eats up floor space for supply components.
There are good reasons to split the circuit into two boards or more, to optimise the flow of currents, the proximity to audio connections, and to facilitate the use of available heat sinks, chassis, etc. These would be second or third form layouts that one comes to when considering more facets of the total package, and is what I prefer for my own amps. The single board approach is convenient when someone else is expected to acquire the board for themselves. Imposing the purchase of two or more boards seems presumptuous.
Have fun
One slip of the finger and several minutes of typing are gone. Ironically, it was to save the text before posting.
Anyhoo...
The short form of what I had written is simply that there are many ways to physically construct a circuit. The layout above is only one, and it is the intuitive first form being a unitised board supporting the entire circuit. This has construction compromises attached.
For example, one can solder the MJLs from the underside and mount the board parallel to the flat side of the heat sink, necessitating a mass of aluminium that can fit into 3U or taller. This is not necessarily good for the solder connections on the PCB but it frees up floor space within the chassis to allow dual PTs. It also allows one to wire things as they wish as noneof the audio connections are on the board, and thus not in any prescribed position that may be difficult to accommodate by a person of average metal working skill.
Alternatively, one could have the PCB perpendicular to the heat sink, allowing a shorter heat sink that might not be able to dissipate all the heat generated depending on the supply voltages, idle settings, and usage.If the design is being used for lower power, then the heat management issue is averted and the lower profile justified. The PCB also eats up floor space for supply components.
There are good reasons to split the circuit into two boards or more, to optimise the flow of currents, the proximity to audio connections, and to facilitate the use of available heat sinks, chassis, etc. These would be second or third form layouts that one comes to when considering more facets of the total package, and is what I prefer for my own amps. The single board approach is convenient when someone else is expected to acquire the board for themselves. Imposing the purchase of two or more boards seems presumptuous.
Have fun
Don't you hate that when it happens!
It's happened to me on more than one occasion, and after a long post I usually just give up. Wish there were a way that forum posts could be saved somehow online in a small, temp disk buffer assigned to you when you logon - that you could easily recall from the user CP. But, alas I don't how to do it if there is. Very frustrating.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.