Thank you a lot for your answer. Please, if you can, detail a bit more about the anti-clip circuit. I am thinking of integrating that in the actual amp pcb.
So I think I'll use 2sc5200 and it's pair as they seem a bit cheaper? Are you sure there won't be any impact on the sound quality?
Regards,
Matei.
So I think I'll use 2sc5200 and it's pair as they seem a bit cheaper? Are you sure there won't be any impact on the sound quality?
Regards,
Matei.
Hi Kevin,
How amp thirsty is this design? I have a couple 55-0-55Vac 350vA transformers. Would one of these be sufficient for a stereo amp? I had initially hoped to use the transformer for an old Soundcraftsmen amp that I have but it is produces +/-98Vdc rails. I guess I'll have to stick with MOSFET's for that one.
How amp thirsty is this design? I have a couple 55-0-55Vac 350vA transformers. Would one of these be sufficient for a stereo amp? I had initially hoped to use the transformer for an old Soundcraftsmen amp that I have but it is produces +/-98Vdc rails. I guess I'll have to stick with MOSFET's for that one.
Hi Guys
A board design really depends on what heat sinks you have available to you. You end up with all the outputs in one line across the width of a low wide heatsink, i.e. one used to form the side of a chassis, or you have two rows of outputs on opposite edges of the board on a more square heat sink like Terry used for his Super amp.
Ideally there is some local filtering, or the main filters are very close to the output stage. This accommodates transients better than if the filters are far away. It is also better for many reasons to use many smaller parallel filter caps rather than one large one. The power supply is the other half of the signal path, so its frequency response and THD impact are important, especially when large currents are involved.
If you look at Rod Elliot's amp pages, he uses low filter cap values - too low for great performance. But... that fits into his design mantra of making something "good enough". He's a great guy but some of his designs are not as high fi as some are looking for here.
Like all solid-state designs, this is flexible with respect to output power. For Terry's 350VA PT, if both channels are driven each could pull 175VA, essentially delivering 175W to each speaker. A single channel driven could utilise the whole rating of the supply and push 350W into an appropriate load. For Brlmat's 250W/ch use, the PT has to be at least 500VA to support both channels - 750VA or more would be even better.
With the four pairs of outputs, you could drive 2R or 1R loads if you keep the supplies lower to keep the dissipations within the device ratings and heat sink capacity. To drive such loads at high power requires more output pairs.
All the output BJTs people use - at least the flat packages and modern numbers - have similar enough ratings to be used here. None of them cost very much, maybe $2 a piece, so skimping on outputs is not economical overall as it results in poor reliability.
As everyone says, the big costs are the heat sinks, chassis and PT(s). A dual-mono arrangement is very nice for these power levels and assures low cross-talk.
None of the devices need to be thermally bonded to other devices. The diff cascode keeps the diffs running cool and also allows the use of better devices, such as 2N5400/5401 for the whole input string of PNPs.
The Cdom compensation shown is not what I use in my amps but it is what is used in the original Blameless. An update to that uses output-inclusive compensation which transitions to normal miller at high frequencies. The result is to greatly reduce high-frequency THD. I'll post a schemo with that in it, and also the anti-clip, shortly - Monday at the latest.
For some general details about PCB design, look at my posts after page-18 on the "New Lineup Power Buffer" thread.
Have fun
Kevin O'Connor
A board design really depends on what heat sinks you have available to you. You end up with all the outputs in one line across the width of a low wide heatsink, i.e. one used to form the side of a chassis, or you have two rows of outputs on opposite edges of the board on a more square heat sink like Terry used for his Super amp.
Ideally there is some local filtering, or the main filters are very close to the output stage. This accommodates transients better than if the filters are far away. It is also better for many reasons to use many smaller parallel filter caps rather than one large one. The power supply is the other half of the signal path, so its frequency response and THD impact are important, especially when large currents are involved.
If you look at Rod Elliot's amp pages, he uses low filter cap values - too low for great performance. But... that fits into his design mantra of making something "good enough". He's a great guy but some of his designs are not as high fi as some are looking for here.
Like all solid-state designs, this is flexible with respect to output power. For Terry's 350VA PT, if both channels are driven each could pull 175VA, essentially delivering 175W to each speaker. A single channel driven could utilise the whole rating of the supply and push 350W into an appropriate load. For Brlmat's 250W/ch use, the PT has to be at least 500VA to support both channels - 750VA or more would be even better.
With the four pairs of outputs, you could drive 2R or 1R loads if you keep the supplies lower to keep the dissipations within the device ratings and heat sink capacity. To drive such loads at high power requires more output pairs.
All the output BJTs people use - at least the flat packages and modern numbers - have similar enough ratings to be used here. None of them cost very much, maybe $2 a piece, so skimping on outputs is not economical overall as it results in poor reliability.
As everyone says, the big costs are the heat sinks, chassis and PT(s). A dual-mono arrangement is very nice for these power levels and assures low cross-talk.
None of the devices need to be thermally bonded to other devices. The diff cascode keeps the diffs running cool and also allows the use of better devices, such as 2N5400/5401 for the whole input string of PNPs.
The Cdom compensation shown is not what I use in my amps but it is what is used in the original Blameless. An update to that uses output-inclusive compensation which transitions to normal miller at high frequencies. The result is to greatly reduce high-frequency THD. I'll post a schemo with that in it, and also the anti-clip, shortly - Monday at the latest.
For some general details about PCB design, look at my posts after page-18 on the "New Lineup Power Buffer" thread.
Have fun
Kevin O'Connor
Last edited:
Hi Guys
The difference between the various output devices mentioned is their power rating.
2SC5200 - 2SA1943: 150W 20mm package
2SC5242 - 2SA1962: 150W 16mm pkg
2SC3281 - 2SA1302: 150W 20mm pkg
MJL3281A - MJL1302A: 200W 20mm pkg
MJL21193 - MJL21194: 250W 20mm
For reliable high output power, it takes more 150W packages than 200W or 250W packages to dissipate the heat. Turning it around, the same number of high-wattage packages will operate with a better safety margin than low-wattage devices.
For the output inclusive compensation:
Place two 220pF caps in series instead of the single 100pF is across Q8+10
Tie a 1k from the junction of the two caps to the output node before the choke.
Have fun
Kevin O'Connor
The difference between the various output devices mentioned is their power rating.
2SC5200 - 2SA1943: 150W 20mm package
2SC5242 - 2SA1962: 150W 16mm pkg
2SC3281 - 2SA1302: 150W 20mm pkg
MJL3281A - MJL1302A: 200W 20mm pkg
MJL21193 - MJL21194: 250W 20mm
For reliable high output power, it takes more 150W packages than 200W or 250W packages to dissipate the heat. Turning it around, the same number of high-wattage packages will operate with a better safety margin than low-wattage devices.
For the output inclusive compensation:
Place two 220pF caps in series instead of the single 100pF is across Q8+10
Tie a 1k from the junction of the two caps to the output node before the choke.
Have fun
Kevin O'Connor
Struth, you rule!
I'm certanly waiting for the updated schematic and the anti - clio schematic if you are so kind to post them here. I asked about the output devices because here in romania a mjl21193 is like 6 euros... I might order them online if you say they don't need to be paired, but I have to make some research...
Also, the rads will be 3 80x65mm cpu rads per module. That will give me 240x65 effective rad. This is how it looks:
http://www.pcgarage.ro/coolere/titan/dc-k8u825x/
It's like 2 euros a piece, fan included so that'll get me to 12 euros for both channels, rad-wise.
Thank you a lot,
Matei.
I'm certanly waiting for the updated schematic and the anti - clio schematic if you are so kind to post them here. I asked about the output devices because here in romania a mjl21193 is like 6 euros... I might order them online if you say they don't need to be paired, but I have to make some research...
Also, the rads will be 3 80x65mm cpu rads per module. That will give me 240x65 effective rad. This is how it looks:
http://www.pcgarage.ro/coolere/titan/dc-k8u825x/
It's like 2 euros a piece, fan included so that'll get me to 12 euros for both channels, rad-wise.
Thank you a lot,
Matei.
Last edited:
Still4,
you are not missing anything.
230W, 1.3A @ 90Vce (for 1s), Ft>=50MHz (typ), hFE>=100 @ 8A (typ). The guaranteed minima are quite a bit lower.
Struth,
maybe you can buy those devices listed for ~$2, but most will find typical prices around double, if not triple that guide price.
You can probably buy fake 1943/5200 for ~$2 !
you are not missing anything.
230W, 1.3A @ 90Vce (for 1s), Ft>=50MHz (typ), hFE>=100 @ 8A (typ). The guaranteed minima are quite a bit lower.
Struth,
maybe you can buy those devices listed for ~$2, but most will find typical prices around double, if not triple that guide price.
You can probably buy fake 1943/5200 for ~$2 !
I remembered the rule for the number of op devices. Power to speaker times 5 or 6 shoukd be the max dissipated power per transistor times how many transistors there are. So for 250w in 8 ohms, 3 mjl21193/4 pairs should suffice. But then again 4 sc/sa pairs would suffice too. Although it would be much more reliable with 4 mjl pairs and could even drive 6 ohm load at the same power or 4 ohm load at lower power. So yeah, I'll get what I can as I understand there will be no sonic difference.
Anyway, do you guys know any reliable european store?
Anyway, do you guys know any reliable european store?
Last edited:
Profusion doesn't sound bad, although i was thinking of ordering some samples from on semi. They should be original and the risk minimal. Also cheaper...
Do you guys know what it takes to order samples from on semi ?
Also, Struth, i am still waiting for the updated schematic and the anti-clip circuitry.
Regards,
Matei.
Do you guys know what it takes to order samples from on semi ?
Also, Struth, i am still waiting for the updated schematic and the anti-clip circuitry.
Regards,
Matei.
LE: If the samples won't be shipped, i think i'll order them from Mouser. It's a big company, unlikely to sell fakes and they have pretty good prices. Also, i think i qualify for free shipping to Romania as they have, despite Digi-key, offices in Europe, their main branch being in Germany. What do you guys think about this? I know many of you order from them.
Hi Guys
Oop!
The MJL21193/94 power rating should be 200W not 250W.
The TO-3 (all metal) package allows 250W and an extra 50C of die temperature.
Which devices people gravitate towards in the TO-3PBPL (TO-264) depends on which they used in the past. The TO-3 MJ15022/23/24/25 family is essentially the same as the MJ21193/94/95/96, with the plastic packs as MJL21193/94/95/96. There are basically no electronic changes that need be made to an existing amp design to go from the TO-3 to the TO-264 - maybe some compensation for the altered layout.
Note that the TO-247 is similar to the 16mm wide TO-3P. TO-264 is similar to the 20mm wide cases.
Builders familiar with the 2SA1302/2SC3281 pair will go to the MJL1302A/3281A with ease. Since the original types are no longer in production, the other 2SA/2SC pairs are nearly identical, so an easy swap.
Back in 1992, Motorola's data book showed that they were manufacturing the 2SA1302/2SC3281 pair. For the US market, they had to repackage this and give it a "domestic" number, so the MJL1302/3281 were born. The original ratings were 150W and 200V. The latter became 230V and is now listed as 260V for the A version - only type produced now. This is what I've used in my amp designs since around '98 and they are usually about $1.95 or so.
Remember that a European vendor has to get the parts shipped in and pays duties or VAT at least, and the retail price ends up higher because if it. Many of my customers tell me that buying from North America is cheaper overall, even if the full value of the order is listed for customs. Check out Digikey and Mouser and see how the total cost compares to local.
Personally I would go with a large passive heat sink. It costs more but it is reliable. Relying on fans for all cooling invokes a stiff acoustic noise penalty and if the fans stop the basic radiator won't cut it at all.
There are many rules of thumb about how many output devices are required for a given power to the load. If the load is known and well characterised, then it is simple. Driving loud speakers is another matter. You have to design for the lowest impedance that will be typical, otherwise the protection circuitry will be cutting in and out all the time.
The no-clip circuit is attached. The signals from the diff-amp bases are buffered and then amplified by the a diff amp. The output of the diff is rectified - you can use many other types of rectifier circuit here - and integrated. The control voltage "VC" can go to either of many types of audio signal level control circuits.
Circuit-A shows what Peavey used in their CS-series amps as a part of their "Distortion Detecting technology" (DDT). A transconductance (gm) opamp is used to provide a voltage controlled resistance to divide the signal into a buffer. This is very predictable, works almost ideally, but is noisy. Any other type of gm-opamp can be used, or a VCA. A VCA requires an inverting opamp at its output, although the THAT types include an opamp.
Circuit-B shows the use of n-jfets to provide the attenuation. Like the gm-opamp, the jfets work against series resistance in the audio path. One R+jfet will likely suffice for the no-clip function, depending on how hot the signal can be from the preamp. If you want really deep muting, or maybe have to handle very large overdrive signals, then two sections can be used. For the best noise performance from the power amp, a second buffer can be added after the attenuator, so that the PA input is driven by a low impedance.
There are other types of anti-clip circuitry, some of which is much simpler than shown here but requires a different amp front end.
Have fun
Kevin O'Connor
Oop!
The MJL21193/94 power rating should be 200W not 250W.
The TO-3 (all metal) package allows 250W and an extra 50C of die temperature.
Which devices people gravitate towards in the TO-3PBPL (TO-264) depends on which they used in the past. The TO-3 MJ15022/23/24/25 family is essentially the same as the MJ21193/94/95/96, with the plastic packs as MJL21193/94/95/96. There are basically no electronic changes that need be made to an existing amp design to go from the TO-3 to the TO-264 - maybe some compensation for the altered layout.
Note that the TO-247 is similar to the 16mm wide TO-3P. TO-264 is similar to the 20mm wide cases.
Builders familiar with the 2SA1302/2SC3281 pair will go to the MJL1302A/3281A with ease. Since the original types are no longer in production, the other 2SA/2SC pairs are nearly identical, so an easy swap.
Back in 1992, Motorola's data book showed that they were manufacturing the 2SA1302/2SC3281 pair. For the US market, they had to repackage this and give it a "domestic" number, so the MJL1302/3281 were born. The original ratings were 150W and 200V. The latter became 230V and is now listed as 260V for the A version - only type produced now. This is what I've used in my amp designs since around '98 and they are usually about $1.95 or so.
Remember that a European vendor has to get the parts shipped in and pays duties or VAT at least, and the retail price ends up higher because if it. Many of my customers tell me that buying from North America is cheaper overall, even if the full value of the order is listed for customs. Check out Digikey and Mouser and see how the total cost compares to local.
Personally I would go with a large passive heat sink. It costs more but it is reliable. Relying on fans for all cooling invokes a stiff acoustic noise penalty and if the fans stop the basic radiator won't cut it at all.
There are many rules of thumb about how many output devices are required for a given power to the load. If the load is known and well characterised, then it is simple. Driving loud speakers is another matter. You have to design for the lowest impedance that will be typical, otherwise the protection circuitry will be cutting in and out all the time.
The no-clip circuit is attached. The signals from the diff-amp bases are buffered and then amplified by the a diff amp. The output of the diff is rectified - you can use many other types of rectifier circuit here - and integrated. The control voltage "VC" can go to either of many types of audio signal level control circuits.
Circuit-A shows what Peavey used in their CS-series amps as a part of their "Distortion Detecting technology" (DDT). A transconductance (gm) opamp is used to provide a voltage controlled resistance to divide the signal into a buffer. This is very predictable, works almost ideally, but is noisy. Any other type of gm-opamp can be used, or a VCA. A VCA requires an inverting opamp at its output, although the THAT types include an opamp.
Circuit-B shows the use of n-jfets to provide the attenuation. Like the gm-opamp, the jfets work against series resistance in the audio path. One R+jfet will likely suffice for the no-clip function, depending on how hot the signal can be from the preamp. If you want really deep muting, or maybe have to handle very large overdrive signals, then two sections can be used. For the best noise performance from the power amp, a second buffer can be added after the attenuator, so that the PA input is driven by a low impedance.
There are other types of anti-clip circuitry, some of which is much simpler than shown here but requires a different amp front end.
Have fun
Kevin O'Connor
Last edited:
Thank you for posting. I'll take a look at the no-clip circuitry in a minute.
You said in post 127 that i should modify the schematic a little. What you meant was that C5 should become a C5a and a C5b each of 220pf tied in series. A 1k resistor should be tied with a leg in between the C5a and C5b and the other leg just before R47, right?
Also, about the prices, this is how my cart looks at Mouser. I added some devices just to have a spare in case i fry something.
You said in post 127 that i should modify the schematic a little. What you meant was that C5 should become a C5a and a C5b each of 220pf tied in series. A 1k resistor should be tied with a leg in between the C5a and C5b and the other leg just before R47, right?
Also, about the prices, this is how my cart looks at Mouser. I added some devices just to have a spare in case i fry something.
Attachments
Last edited:
Hi Guys
Matching the output devices can make a significant difference to THD. This applies to drivers, VAS stages, input stages, etc.
The MJL21193 data sheet shows a test circuit where a complimentary EF pair produces about 0.8% THD, where the matched pair is 0.08%.
Very good performance is available from the circuit posted above without matched parts. None of the D.Self demonstrations and measurements relies on matching. I just wanted to show that for those willing to take that extra trouble, it is worth it.
Have fun
Kevin O'Connor
Matching the output devices can make a significant difference to THD. This applies to drivers, VAS stages, input stages, etc.
The MJL21193 data sheet shows a test circuit where a complimentary EF pair produces about 0.8% THD, where the matched pair is 0.08%.
Very good performance is available from the circuit posted above without matched parts. None of the D.Self demonstrations and measurements relies on matching. I just wanted to show that for those willing to take that extra trouble, it is worth it.
Have fun
Kevin O'Connor
Hi Guys
Here is the schematic with output-inclusive compensation added.
Regarding the HB: I looked at that build thread again today and see a different schemo than I'm certain I saw before. Regardless, some bits are still worth noting:
The output stage should really be a triple-EF for the high power they are targeting. Otherwise, VAS and driver THD will be much higher than they ought to be.
The offset trimpot is unnecessary and may interfere with the actual balance of the diff amp. I would leave it out.
The mounting of the output transistors is nonoptimal for a high power amp. The plastic case actually gets very hot and needs air flowing around it. The PCB is not a good heat conductor and will not cool the top of the BJT package. Manufacturers whom I respect would never mount a TO-264 like they are in the HB build. Rod Eliott has a similar project with similar transistor mounting - he should know better.
Have fun
Kevin O'Connor
Here is the schematic with output-inclusive compensation added.
Regarding the HB: I looked at that build thread again today and see a different schemo than I'm certain I saw before. Regardless, some bits are still worth noting:
The output stage should really be a triple-EF for the high power they are targeting. Otherwise, VAS and driver THD will be much higher than they ought to be.
The offset trimpot is unnecessary and may interfere with the actual balance of the diff amp. I would leave it out.
The mounting of the output transistors is nonoptimal for a high power amp. The plastic case actually gets very hot and needs air flowing around it. The PCB is not a good heat conductor and will not cool the top of the BJT package. Manufacturers whom I respect would never mount a TO-264 like they are in the HB build. Rod Eliott has a similar project with similar transistor mounting - he should know better.
Have fun
Kevin O'Connor
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.