do you know how big that psu would be? especially if the idea of providing separate supplies for the buffers, but even without it would be HUUUUUGE.
i know this is a wish list and all and in other builds i would possibly agree, i think we should stick to more realistic requests that actually might happen. doing something like salas shunts really should be left up to you imo, psrr of this amp is very very large; spending huge amounts of space and producing lots of heat, adding costs of massive heatsinks (where to put them?) and considerably larger transformers to create a salas shunt that can produce 2 amps seems ....
. i'm in favor of having a somewhat improved power supply onboard just for good measure, but nothing like that, nothing that would cause the board to be 10x larger, short traces compromised and cost several times more to build.
not for my board, even though i think the salas shunts i have are excellent, they are often impractically large and for a headamp that otherwise takes up a few square inches, it just seems like folly to me, especially considering cmrr of the lme49990 is 137db and psrr is 145db. the buffers are not as flash, but considering the measured performance of the first version, that cant be too bad either.
adding the salas shunts would also increase the level of experience needed to build it considerably.
i know this is a wish list and all and in other builds i would possibly agree, i think we should stick to more realistic requests that actually might happen. doing something like salas shunts really should be left up to you imo, psrr of this amp is very very large; spending huge amounts of space and producing lots of heat, adding costs of massive heatsinks (where to put them?) and considerably larger transformers to create a salas shunt that can produce 2 amps seems ....
. i'm in favor of having a somewhat improved power supply onboard just for good measure, but nothing like that, nothing that would cause the board to be 10x larger, short traces compromised and cost several times more to build.not for my board, even though i think the salas shunts i have are excellent, they are often impractically large and for a headamp that otherwise takes up a few square inches, it just seems like folly to me, especially considering cmrr of the lme49990 is 137db and psrr is 145db. the buffers are not as flash, but considering the measured performance of the first version, that cant be too bad either.
adding the salas shunts would also increase the level of experience needed to build it considerably.
its possible to make very high performance shunt regs with another lme49600/10 and a few other parts. shunts can be quite small, but unless a complete redesign of the salas shunt is done, using smd parts and probably not even then, it cannot be even average size and the current required is not in its usual range, to supply 1a for one side for example, you would generally shunt half that again and many people go even further. meaning that the power supply could be burning something like 50W
sorry, leaving now. dont want to seem controlling. that isnt my intention, i just think that we should be treating this as what it is, an opportunity to upgrade to an already successful and excellent design, not asking owen to design an entirely new amp and maybe not even have it sound as good.
sorry, leaving now. dont want to seem controlling. that isnt my intention, i just think that we should be treating this as what it is, an opportunity to upgrade to an already successful and excellent design, not asking owen to design an entirely new amp and maybe not even have it sound as good.
Last edited:
Hi Owen,
Glad to see you're doing well. This is what I'd be interested in:
1. Full kit preferred, but wouldn't mind bare boards
2. No PSU, there are plenty of excellent offerings already (Twisted Pear LCDPS/LCBPS/Placid, AMB σ22, etc.)
3. Prefer SE, but would also be interested in a balanced version if offered
4. 2 (or if only boards were offered, more than 2)
Thanks!
Glad to see you're doing well. This is what I'd be interested in:
1. Full kit preferred, but wouldn't mind bare boards
2. No PSU, there are plenty of excellent offerings already (Twisted Pear LCDPS/LCBPS/Placid, AMB σ22, etc.)
3. Prefer SE, but would also be interested in a balanced version if offered
4. 2 (or if only boards were offered, more than 2)
Thanks!
opc,
I'll increase my interest to 2 bare boards.
Also like qusp's idea of a population-optional se/balanced input converter. I'm planning on running mine SE input initially, just grounding an input and taking the small harmonics hit, with the plan of building/buying a balanced DAC at some point down the road. A converter would likely do a better job than just grounding an input.
I'll increase my interest to 2 bare boards.
Also like qusp's idea of a population-optional se/balanced input converter. I'm planning on running mine SE input initially, just grounding an input and taking the small harmonics hit, with the plan of building/buying a balanced DAC at some point down the road. A converter would likely do a better job than just grounding an input.
Hi Owen, I was hoping you'd do another run, and signed up to diyAudio just in case you did! Thanks for that!
1. Do you prefer a board only, or a full kit?
- Full kit would be preferable
2. Would you rather have a PSU included or not?
- PSU included
3. Would you be interested in a fully balanced version (bridged output with two LME49990) or the existing version?
- Single ended preferably, but I'd still be interested if it was fully balanced.
4. How many units are you interested in?
- Two please.
Thanks again!
1. Do you prefer a board only, or a full kit?
- Full kit would be preferable
2. Would you rather have a PSU included or not?
- PSU included
3. Would you be interested in a fully balanced version (bridged output with two LME49990) or the existing version?
- Single ended preferably, but I'd still be interested if it was fully balanced.
4. How many units are you interested in?
- Two please.
Thanks again!
opc - a couple of ideas on the output buffers and power supply, FWIW, since it is looking like balanced output is popular and might happen, which would mean physical places on the board for 4 LME49600/610s. Still thinking up ways to wedge in paralleled buffers on the single ended version, lol. 
A fully balanced-output version would take 4 LME49600/610s, 2 per channel bridged - but so would a paralleled-output-buffer version of the existing single ended output board. Maybe it would be possible to do some clever (population-optional) design and using short wire jumpers to configure the board for either case.
case 1. Populate all 4 LME49600/610s and jumper/configure for fully balanced (bridged) output, along with populating the additional parts needed before the buffers.
case 2. Populate all 4 LME49600/610s and jumper/configure for paralleled (2 per channel) singled ended operation. Don't populate any additional parts needed for the fully balanced output. Thermal output should be roughly the same as 1 buffer since the power dissipation is now divided among the two. Buffer input capacitance is doubled, though, as was mentioned in that LME49600 thread.
case 3. Populate only 2 LME49600/610s, one per channel, and jumper/configure for singled ended operation. Don't populate any additional parts needed for the fully balanced output. This case then reduces to the existing board configuration.
For the power supply: whether there is a power supply on board or not, provide a jumper between the output buffer power supply rail and the LME49990s rail for each rail polarity so the buffers could optionally be fed from a separate supply, like qusp mentioned. Then the buffers could be separately fed with that +/-22V (for the LME49610s), or just the same +/- 15 for the LME49600 but from a different supply if someone really wants to go nuts with psu separation - like me.
case 1. rail jumpers connected - buffers and op amps fed from same psu, as in the existing board.
case 2. rail jumpers removed, op amps fed from one supply (could be the on-board supply, if there is an on-board supply), buffers fed from a different (off-board) supply (and possibly different voltage). Current of that supply could also be picked to match either one buffer per channel or two, as per above.
One downside in all the jumpers is increased trace length and inductance, of course, but you wind up with a board that is configurable to be more things to more people. Since everything is surface mount I'm assuming any jumpers would be short surface jumpers between adjacent pads.
A fully balanced-output version would take 4 LME49600/610s, 2 per channel bridged - but so would a paralleled-output-buffer version of the existing single ended output board. Maybe it would be possible to do some clever (population-optional) design and using short wire jumpers to configure the board for either case.
case 1. Populate all 4 LME49600/610s and jumper/configure for fully balanced (bridged) output, along with populating the additional parts needed before the buffers.
case 2. Populate all 4 LME49600/610s and jumper/configure for paralleled (2 per channel) singled ended operation. Don't populate any additional parts needed for the fully balanced output. Thermal output should be roughly the same as 1 buffer since the power dissipation is now divided among the two. Buffer input capacitance is doubled, though, as was mentioned in that LME49600 thread.
case 3. Populate only 2 LME49600/610s, one per channel, and jumper/configure for singled ended operation. Don't populate any additional parts needed for the fully balanced output. This case then reduces to the existing board configuration.
For the power supply: whether there is a power supply on board or not, provide a jumper between the output buffer power supply rail and the LME49990s rail for each rail polarity so the buffers could optionally be fed from a separate supply, like qusp mentioned. Then the buffers could be separately fed with that +/-22V (for the LME49610s), or just the same +/- 15 for the LME49600 but from a different supply if someone really wants to go nuts with psu separation - like me.
case 1. rail jumpers connected - buffers and op amps fed from same psu, as in the existing board.
case 2. rail jumpers removed, op amps fed from one supply (could be the on-board supply, if there is an on-board supply), buffers fed from a different (off-board) supply (and possibly different voltage). Current of that supply could also be picked to match either one buffer per channel or two, as per above.
One downside in all the jumpers is increased trace length and inductance, of course, but you wind up with a board that is configurable to be more things to more people. Since everything is surface mount I'm assuming any jumpers would be short surface jumpers between adjacent pads.
Last edited:
hmmm interesting idea, but there would have to be an extra summing amp as in the current SE design, in the way of the signal path for the balanced output and all the power and feedback parts associated with it. both layouts would be compromised, well actually no, only the balanced one would be. i dont see an easy way to have a proper instrumentation amp as for the SE version, while not compromising the layout for the balanced version. with a 4 layer pcb it might be easier, as perhaps you could have the optional parts on the underside, with 2 layer this would break up the ground fill, but with 4 layer perhaps it could be kept to a minimum.
do remember part of the beauty of the current design was the compact high performance build with very short traces. we really need to make sure we dont ruin that with over complicating it. i'm guilty there too perhaps
the buffers basically operate in class A by design, so i wouldnt think 4 units would be the same power consumption and dissipation as 2. i could be wrong there, but i would think that due to that regardless of whether the load demands double the current it will still be provided
perhaps a rerun of the se board as well?
do remember part of the beauty of the current design was the compact high performance build with very short traces. we really need to make sure we dont ruin that with over complicating it. i'm guilty there too perhaps
the buffers basically operate in class A by design, so i wouldnt think 4 units would be the same power consumption and dissipation as 2. i could be wrong there, but i would think that due to that regardless of whether the load demands double the current it will still be provided
perhaps a rerun of the se board as well?
Hi Guys,
I've gone back through the last few pages and compiled the attached list based on people's requests.
It looks like there's overwhelming support for a kit version, so I will be offering both a kit option, and a bare board option.
As for amp style, I'm going to spend the day laying out two more versions. I will make the following available:
1. Balance input to SE output - the same as the one previously offered
2. SE input to SE output - one LME49990 and one LME49600 per channel
3. Balanced input to Balanced output - Still up in the air - see below
For the balanced input to output I'm debating between using one OPA1632 per channel with a pair of LME49600 in bridged output, or using a pair of LME49990 each driving one LME49600 in bridge output. If anyone has any suggestions on which they would prefer (and why) then please let me know.
As for parallel output devices, this isn't going to happen. I've tried it, and it's not as good as a single device. The only reason I can see for doing it would be to deliver more current to a very low impedance load, which is not the aim of this headphone amplifier. In bridged mode, this amp will be able to swing 80V p-p with 22V rails, and drive over 250mA of current. I've never heard of a pair of headphones that require more than that, which makes parallel output devices a waste of time, money, and sound quality. More is NOT always better! Build one, and you'll see what I mean.
The other aspect that is still up in the air is the PSU. Since the desire for a PSU is pretty much split, I don't think I'm going to include it on the board. I will, however, include an optional off-board PSU. If anyone has legitimate suggestions for improvements to the PSU then let me know and I'll look into them. I really believe that an over the top supply is a true waste in this application though. It's just not needed thanks to the natural PSRR of the amp.
I'll post the two new layouts later today.
Cheers,
Owen
I've gone back through the last few pages and compiled the attached list based on people's requests.
It looks like there's overwhelming support for a kit version, so I will be offering both a kit option, and a bare board option.
As for amp style, I'm going to spend the day laying out two more versions. I will make the following available:
1. Balance input to SE output - the same as the one previously offered
2. SE input to SE output - one LME49990 and one LME49600 per channel
3. Balanced input to Balanced output - Still up in the air - see below
For the balanced input to output I'm debating between using one OPA1632 per channel with a pair of LME49600 in bridged output, or using a pair of LME49990 each driving one LME49600 in bridge output. If anyone has any suggestions on which they would prefer (and why) then please let me know.
As for parallel output devices, this isn't going to happen. I've tried it, and it's not as good as a single device. The only reason I can see for doing it would be to deliver more current to a very low impedance load, which is not the aim of this headphone amplifier. In bridged mode, this amp will be able to swing 80V p-p with 22V rails, and drive over 250mA of current. I've never heard of a pair of headphones that require more than that, which makes parallel output devices a waste of time, money, and sound quality. More is NOT always better! Build one, and you'll see what I mean.
The other aspect that is still up in the air is the PSU. Since the desire for a PSU is pretty much split, I don't think I'm going to include it on the board. I will, however, include an optional off-board PSU. If anyone has legitimate suggestions for improvements to the PSU then let me know and I'll look into them. I really believe that an over the top supply is a true waste in this application though. It's just not needed thanks to the natural PSRR of the amp.
I'll post the two new layouts later today.
Cheers,
Owen
i have to think on that, the susy chips are superb, but i have an abundance of designs using them already. if it happens, i have a stock of them, so thats a plus, but i will be using ths4131 just to be different haha, it was the original one that caused the commotion afterall.
i'm fine with either though really as it really is a superb chip, then again so is the lme49990; its really lineball for me. i might have to do a quick little experiment and add some buffers to my portable dac which uses the dgn powerpad version. which would be the same signal path exactly but cheating a bit by using zfoil smd feedback/filter resistors
for the psu i recommend looking at the lt1764a and only the lt1764a; using 2 connected in series for a bipolar supply, comes in the same type of package as the buffer and is quite simply the highest performance ic reg i have ever used or seen reports of. low low noise and speed off the charts. if the current needed was lower i would recommend the lme49600/10 plus low noise reference and unity gain stable opamp, as described in the buffer datasheet app note; ive tried this and it works a treat. could maybe use that for the input opamp supplies and something beefier for the output as the buffer is of course limited to 250ma
unfortunately linear have not supplied a partner negative reg that is quite up to the quality of the lt1764a and lt1963a, the closest is lt1185. that would require people know what they are doing with hooking up signal input and would only be able to be used on the balanced version, so hmm, would have to use lt1764a/lt1963a and lt1185 which is a much better option than lt317/337 (which themselves are mismatched in quality) i dont think its enough to not consider them, still leagues ahead imo
i'm fine with either though really as it really is a superb chip, then again so is the lme49990; its really lineball for me. i might have to do a quick little experiment and add some buffers to my portable dac which uses the dgn powerpad version. which would be the same signal path exactly but cheating a bit by using zfoil smd feedback/filter resistors
for the psu i recommend looking at the lt1764a and only the lt1764a; using 2 connected in series for a bipolar supply, comes in the same type of package as the buffer and is quite simply the highest performance ic reg i have ever used or seen reports of. low low noise and speed off the charts. if the current needed was lower i would recommend the lme49600/10 plus low noise reference and unity gain stable opamp, as described in the buffer datasheet app note; ive tried this and it works a treat. could maybe use that for the input opamp supplies and something beefier for the output as the buffer is of course limited to 250ma
unfortunately linear have not supplied a partner negative reg that is quite up to the quality of the lt1764a and lt1963a, the closest is lt1185. that would require people know what they are doing with hooking up signal input and would only be able to be used on the balanced version, so hmm, would have to use lt1764a/lt1963a and lt1185 which is a much better option than lt317/337 (which themselves are mismatched in quality) i dont think its enough to not consider them, still leagues ahead imo
Last edited:
2 balanced pcbs for me and maybe matching semis depending on what they are. definitely agree with no psu on board now i think about it, as i could try one of these out in my portable sabre as a amp module, as it will probably make a decent enough iv stage with buffer as well. will see what the psu is before i make that decision
are you really scared of balancing your headphones or something? you know i can help with that right
Ha! qusp - hey thanks for the offer. I may take you up on that sometime. I do think balanced is better but right now its just a lack of anything balanced to power. My Shure SRH840s are SE. For my next headphone purchase I will probably give balanced a good hard look. One of these days I may do some surgery on the Shures. Actually it may not be that hard now that I think about it a bit, if they are running separate grounds down the headphone cable. Maybe I will decide to get one of opc's bal->bal cards too before it is all over.
My next "balanced" thing will probably be buy/build a DAC. I've learned a lot from your balanced DAC comments on what to go looking for.
Thanks again!
qusp - hey thanks for the offer. I may take you up on that sometime. I do think balanced is better but right now its just a lack of anything balanced to power. My Shure SRH840s are SE. For my next headphone purchase I will probably give balanced a good hard look. One of these days I may do some surgery on the Shures. Actually it may not be that hard now that I think about it a bit, if they are running separate grounds down the headphone cable. Maybe I will decide to get one of opc's bal->bal cards too before it is all over. My next "balanced" thing will probably be buy/build a DAC. I've learned a lot from your balanced DAC comments on what to go looking for.
Thanks again!
Last edited: