What voltage at the gate should I aim at with the voltage divider? (Standard circuit with 38v rails and 200/400 resistors)
My FETs bias up at about 5.9V each on the gates. This can vary somewhat, but adjusting for 5.9V will get you in the ballpark.
The actual adjustment should be done to obtain 1/2 AVCC at the input though.
If your DAC is running 3.5V AVCC then you should adjust each leg of the NTD1 to get 1.75VDC at the source of the FET.
Once adjusted, plug in the DAC and adjust it again. It will drift over time, but once set at operating temp it will stay.
Mke sure to double check the polarity on that source voltage. It must be +1.75VDC relative to GND. I have accidentally set this voltage to -1.75 on several occasions because I was holding the probes in the wrong orientation and rushing :S
Regards,
Owen
I have just finished my NTD1. There's some things to finish, of course, and something to rework but the first impression is very good. I made it for sacd listening mostly as I have another dac for cd and hi-res listening.
The heatsink under the board is small so I need to change it for a bigger one. I had to lower the voltage to be able to touch it for a few seconds.
The sound is very good. I listened to Tchaikovsky Nutcracker sacd rip today. That was a pleasure.
P.S. To adjust to 1,75V is quite difficult I should say. The measuring is always around it.
The heatsink under the board is small so I need to change it for a bigger one. I had to lower the voltage to be able to touch it for a few seconds.
The sound is very good. I listened to Tchaikovsky Nutcracker sacd rip today. That was a pleasure.
P.S. To adjust to 1,75V is quite difficult I should say. The measuring is always around it.
Last edited:
Excellent work zz1969!!
I think this is the first picture of the new board I have seen 🙂
Indeed that heatsink is a bit small as you've already discovered. It's worth finding a bigger on so that you can get the rails back to full voltage.
As for the adjustment to 1.75V, I should clarify that this is not particularly critical. There is only a very small difference in performance (almost not measurable) between 0V and 1.75V, so being even a decent amount off is not going to impact performance to any degree. The critical thing here is to not let this voltage exceed AVCC, or go negative, since that does have a big impact on distortion (and could potentially damage the DAC).
If I ever do another version I might consider adding a resistor on the lower side of the pot to provide an offset and decrease the sensitivity of the pot to the bias adjustment as it is a bit picky at the moment.
Excellent work on the build, and glad to hear you're enjoying the sound!
Regards,
Owen
I think this is the first picture of the new board I have seen 🙂
Indeed that heatsink is a bit small as you've already discovered. It's worth finding a bigger on so that you can get the rails back to full voltage.
As for the adjustment to 1.75V, I should clarify that this is not particularly critical. There is only a very small difference in performance (almost not measurable) between 0V and 1.75V, so being even a decent amount off is not going to impact performance to any degree. The critical thing here is to not let this voltage exceed AVCC, or go negative, since that does have a big impact on distortion (and could potentially damage the DAC).
If I ever do another version I might consider adding a resistor on the lower side of the pot to provide an offset and decrease the sensitivity of the pot to the bias adjustment as it is a bit picky at the moment.
Excellent work on the build, and glad to hear you're enjoying the sound!
Regards,
Owen
Wow, thank you Owen for such a detailed and thorough response! Really appreciate you taking the time to work through my brain dump and provide advice. 🙂
It's taken me a bit to digest everything and come to a few conclusions based on your ideas and suggestions, but I think I'm almost there:
I've reached out to Riedon, Caddock, Vishay, and Texas Components to see if they can provide custom value D-Pak style power resistors with similar or better characteristics than those of the PFS35 series.
43R, 47R, 50R, 91R, 100R, and 110R are the closest PFS35 standard values, and even then only the 100R and 50R are non special-order (high min quantity) from DigiKey.
Any suggestions for other resistor options if none of the above suppliers can provide an exact match at a reasonable price?
I'm assuming it would be best to run a pair of LT3042 supplies per channel (four rails, two boards), similar to how the NTD1 PSU is set up.
Are you suggesting that the BAL-BAL should be powered from the same supplies as the NTD1, or that an additional two LT3042 boards should be used for the BAL-BAL?
If the BAL-BAL was powered by the NTD1 supplies, I'd guess that the LME49724 would be required, since the 18V rails would be above the 16V max of the OPA1632.
From purely a performance perspective, which configuration would you recommend?
I've decided to take the BAL-BAL approach for single ended conversion over transformers, so now I just need to figure out how many LT3042 boards I need to finalize my order. 🙂
I'm up for going DC coupled. I've seen it mentioned a few times throughout the various NTD1 threads and I'd like to give it a shot since I'm already this close.
Thinking I'd use a SMT jumper where C2, C3, C22, and C23 would go on the board. Any particular type of jumper you'd recommend?
Any recommendations on a relay muting circuit? Leaning towards the AMB ε12 since there's a board available. There's also the ESP Project 104 schematic.
Do you have a recommended value for R5, R6, R15, and R16 in this case? My guess was 9 or 10k, assuming the pot stays at 5k.
Haha, I guess I should have expected that. My assumption is that the SSLV shunt regs may not be the best choice for the NTD1 or BAL-BAL in this case due to the high current.
I am curious how the LT3042 compares to the Salas Reflektor-D when used on DAC or XO boards.
It's taken me a bit to digest everything and come to a few conclusions based on your ideas and suggestions, but I think I'm almost there:
I've reached out to Riedon, Caddock, Vishay, and Texas Components to see if they can provide custom value D-Pak style power resistors with similar or better characteristics than those of the PFS35 series.
43R, 47R, 50R, 91R, 100R, and 110R are the closest PFS35 standard values, and even then only the 100R and 50R are non special-order (high min quantity) from DigiKey.
Any suggestions for other resistor options if none of the above suppliers can provide an exact match at a reasonable price?
I'm assuming it would be best to run a pair of LT3042 supplies per channel (four rails, two boards), similar to how the NTD1 PSU is set up.
Are you suggesting that the BAL-BAL should be powered from the same supplies as the NTD1, or that an additional two LT3042 boards should be used for the BAL-BAL?
If the BAL-BAL was powered by the NTD1 supplies, I'd guess that the LME49724 would be required, since the 18V rails would be above the 16V max of the OPA1632.
From purely a performance perspective, which configuration would you recommend?
I've decided to take the BAL-BAL approach for single ended conversion over transformers, so now I just need to figure out how many LT3042 boards I need to finalize my order. 🙂
I'm up for going DC coupled. I've seen it mentioned a few times throughout the various NTD1 threads and I'd like to give it a shot since I'm already this close.
Thinking I'd use a SMT jumper where C2, C3, C22, and C23 would go on the board. Any particular type of jumper you'd recommend?
Any recommendations on a relay muting circuit? Leaning towards the AMB ε12 since there's a board available. There's also the ESP Project 104 schematic.
Do you have a recommended value for R5, R6, R15, and R16 in this case? My guess was 9 or 10k, assuming the pot stays at 5k.
Haha, I guess I should have expected that. My assumption is that the SSLV shunt regs may not be the best choice for the NTD1 or BAL-BAL in this case due to the high current.
I am curious how the LT3042 compares to the Salas Reflektor-D when used on DAC or XO boards.
output filter
Does anyone ever consider adding an output filter? I know many feel it is unnecessary, but I am a little uneasy about passing along a high degree of ultrasonic noise. Seems like it would be easy to add a couple of small film caps to G on each channel anyway. I assume that one could figure 200R for determining the corner frequency, right?
Does anyone ever consider adding an output filter? I know many feel it is unnecessary, but I am a little uneasy about passing along a high degree of ultrasonic noise. Seems like it would be easy to add a couple of small film caps to G on each channel anyway. I assume that one could figure 200R for determining the corner frequency, right?
Quick update on the resistor front:
Caddock has a 100 unit minimum for custom values of the MP725 series.
Texas Components can produce custom values without minimums, but they don't have an exact match of the PFS35 specs. Their closest model is the TX220S in a TO220 surface mount package.
Their resistors use Foil technology, and thus only have a power rating of 10 watts max when chassis mounted. This is quite a bit lower than the 35 watt rating of the PFS35.
If the NTD1 has 200mA going through each leg in this configuration, then using P=I^2*R, the power through each 45R and 98R resistor is 1.8W and 3.92W, respectively.
Assuming that's the correct method of applying the formula here, it seems like it may be possible to use the TC resistors?
Owen recommended using two transformers on post #750, But after looking at other fet based circuits, why can't we use a single 2x32v transformer to the L and R +-?
Is there any disadvantages?
Is there any disadvantages?
If...
You are using OPC's power supply, you have a true dual mono supply. Theoretically, the advantage of a dual mono supply should be greater stereo separation, and hence better soundstaging. Having separate transformers for each channel extends the dual mono nature, and agin, theoretically should result in the best performance. If you are going to the extent of having the NTD1 as your output stage, I see no reason to compromise things by using a single transformer for power, why not go all the way and keep things dual mono.
You are using OPC's power supply, you have a true dual mono supply. Theoretically, the advantage of a dual mono supply should be greater stereo separation, and hence better soundstaging. Having separate transformers for each channel extends the dual mono nature, and agin, theoretically should result in the best performance. If you are going to the extent of having the NTD1 as your output stage, I see no reason to compromise things by using a single transformer for power, why not go all the way and keep things dual mono.
Just to clarify. There is only three connection points for power per channel on OPC supply.
So the options would be 16x16v centre tapped or 32v for +- with no ground and then dual secondary windings or dual transformers.
I understand the L-R separation but am unsure if there is any audible - measureable benefit.
On the 9018 DAC I have separate regulators on the AVCC but not separate supply's
So at what point are there diminishing returns?
So the options would be 16x16v centre tapped or 32v for +- with no ground and then dual secondary windings or dual transformers.
I understand the L-R separation but am unsure if there is any audible - measureable benefit.
On the 9018 DAC I have separate regulators on the AVCC but not separate supply's
So at what point are there diminishing returns?
yep...
I guess that would be a hard question to answer for sure, without doing an involved A/B listening test. But, the transformers recommended by Owen are not expensive, so having two is not really any problem for me, I figure, why worry about it and just go dual mono. Unless of course you also have a chassis which will not fit everything, I guess that could be a problem.
I guess that would be a hard question to answer for sure, without doing an involved A/B listening test. But, the transformers recommended by Owen are not expensive, so having two is not really any problem for me, I figure, why worry about it and just go dual mono. Unless of course you also have a chassis which will not fit everything, I guess that could be a problem.
My mistake, the supply is not designed with center tapped input.
The Wire - All Boards and Kits Explained Here! - diyAudio
The Wire - All Boards and Kits Explained Here! - diyAudio
So I was looking at the NTD BOM for it, trying to figure out the end cost.
I would rather source my own parts, lets me pick the parts, and leave out parts I tend not to use, like connectors (I'll just solder things in place).
I noticed you can only buy the 180 ohm power resistors in qty of 500 from Digikey, kind of a problem since I don't need that many 😉
Saw some other people have had the same problem, earlier in this thread.
Looking at alternatives, I noticed in the NTD picture, caddock mp725 were used, at 200 and 400 ohm values. But I also saw earlier that Owen said the NTD sounds better when the resistors are lower value, I guess that's why the PL uses 180 and 390 ohm.
Can I go lower? 150 ohm mp725 resistors are available. I guess I would lose some gain, but hopefully I'll have enough gain in my system to make up for it.
Or, should I stick with MP725's in 200 and 400? I can find those.
Thanks
Randy
I would rather source my own parts, lets me pick the parts, and leave out parts I tend not to use, like connectors (I'll just solder things in place).
I noticed you can only buy the 180 ohm power resistors in qty of 500 from Digikey, kind of a problem since I don't need that many 😉
Saw some other people have had the same problem, earlier in this thread.
Looking at alternatives, I noticed in the NTD picture, caddock mp725 were used, at 200 and 400 ohm values. But I also saw earlier that Owen said the NTD sounds better when the resistors are lower value, I guess that's why the PL uses 180 and 390 ohm.
Can I go lower? 150 ohm mp725 resistors are available. I guess I would lose some gain, but hopefully I'll have enough gain in my system to make up for it.
Or, should I stick with MP725's in 200 and 400? I can find those.
Thanks
Randy
The Wire Headphone Amp: Balanced versus Single-Ended?
Thinking about my first headphone amp build but I have to say I am thoroughly confused regarding this. Are there truly any benefits to balanced headphones?
The contribution from Elias Gwinn repeated on this site is interesting.
I'd like to hear the views of Owen and others. TIA
Thinking about my first headphone amp build but I have to say I am thoroughly confused regarding this. Are there truly any benefits to balanced headphones?
The contribution from Elias Gwinn repeated on this site is interesting.
I'd like to hear the views of Owen and others. TIA
Hi SGK,
I'll pull the talking points straight from the thread you refer to:
This doesn't sound like an answer from an engineer. It sounds like fast talking around a perfectly valid point. The cable doesn't need to be long, or even of any significant resistance. Any amount of shared return causes crosstalk, and it doesn't take much. Just the 3-pin 1/4" jack and a few cm of wire inside the amp is all it takes to decimate crosstalk. Let's look at the numbers:
Consider a pair of 16 ohm IEMs. How much common ground resistance is required for a rather poor crosstalk figure of -80dB? The answer is a mere 1.6 milliohms! That's the resistance of a 3cm long 22AWG wire. The contact resistance of a very good Neutrik headphone jack is in the neighborhood of 5 milliohms. Let's assume a headphone jack with a very optimistic 3cm GND return. That's a best case shared resistance of 6.6 milliohms which gives us a crosstalk value of -67dB. That's pretty sad considering the DAC is capable of better than 130dB of channel separation.
How much resistance is required to do our DAC justice? You'd need to dip below 5 u-ohms, and you'll need liquid nitrogen and directly soldered connections to get that low.
Or... you just used a balanced system and the insurmountable issue above simply disappears when you eliminate the shared return.
In addition to the above, the statement that "most headphones do not share a common conductor" is completely incorrect. Sitting beside me is a pair of AKG studio headphones, that, like many of AKG's studio headphones, uses a detachable 3-pin mini XLR. That means the GND return is shared through the entire length of the cable, and I would consider these to be "of decent quality".
Finally, saying "most headphone cables are effectively balanced inherently" is total nonsense. This one doesn't even need an explanation. Just because you have two wires in your hand doesn't mean you have a balanced connection.
Although 55kHz isn't really that impressive, I generally agree that the added slew rate is not a significant advantage here. This has more to do with the design of the amplifier than with the differential output. If slew rate is all important, then you can easily design a single ended amp with an extremely high slew rate.
Again, as with slew rate, this is more of an issue with the amplifier design than it is with the differential output. The Wire BAL-BAL has an output impedance of less than 7 milliohms from DC to 20kHz, and although this is indeed twice the output impedance of the SE-SE variant, it's totally irrelevant in the grand scheme of things. Contact resistance and a few feet of headphone cable will increase this by 10-fold, making the amplifier's contribution negligible.
I also take serious issue with the statement "increased distortion... and altered frequency response. This is common knowledge for bridged amplifiers."
This is not "common knowledge" because its completely false. Distortion will decrease in this arrangement (cancellation of 2nd harmonic) and there is no impact on the frequency response of the system.
This is correct, but I'm not sure why this is being discussed. If anything, it's an argument in favor of a fully differential system where the transducer at the end of the chain can provide perfect CMRR. A transducer connected in a single-ended manner cannot do this as it has one lead with signal, and the other referenced to ground.
This is the exact same statement as #3. No sure why it's stated twice. See my answer above.
In all honestly, I think both connection types will provide a fantastic listening experience, and I really don't have a dog in this fight since I will happily offer you either an SE-SE or a BAL-BAL depending on your taste or requirements. What I do take issue with is trying to argue against what is unquestionably a technically superior solution in a balanced headphone connection just because you don't offer a product with a balanced output.
Regards,
Owen
I'll pull the talking points straight from the thread you refer to:
This doesn't sound like an answer from an engineer. It sounds like fast talking around a perfectly valid point. The cable doesn't need to be long, or even of any significant resistance. Any amount of shared return causes crosstalk, and it doesn't take much. Just the 3-pin 1/4" jack and a few cm of wire inside the amp is all it takes to decimate crosstalk. Let's look at the numbers:
Consider a pair of 16 ohm IEMs. How much common ground resistance is required for a rather poor crosstalk figure of -80dB? The answer is a mere 1.6 milliohms! That's the resistance of a 3cm long 22AWG wire. The contact resistance of a very good Neutrik headphone jack is in the neighborhood of 5 milliohms. Let's assume a headphone jack with a very optimistic 3cm GND return. That's a best case shared resistance of 6.6 milliohms which gives us a crosstalk value of -67dB. That's pretty sad considering the DAC is capable of better than 130dB of channel separation.
How much resistance is required to do our DAC justice? You'd need to dip below 5 u-ohms, and you'll need liquid nitrogen and directly soldered connections to get that low.
Or... you just used a balanced system and the insurmountable issue above simply disappears when you eliminate the shared return.
In addition to the above, the statement that "most headphones do not share a common conductor" is completely incorrect. Sitting beside me is a pair of AKG studio headphones, that, like many of AKG's studio headphones, uses a detachable 3-pin mini XLR. That means the GND return is shared through the entire length of the cable, and I would consider these to be "of decent quality".
Finally, saying "most headphone cables are effectively balanced inherently" is total nonsense. This one doesn't even need an explanation. Just because you have two wires in your hand doesn't mean you have a balanced connection.
Although 55kHz isn't really that impressive, I generally agree that the added slew rate is not a significant advantage here. This has more to do with the design of the amplifier than with the differential output. If slew rate is all important, then you can easily design a single ended amp with an extremely high slew rate.
Again, as with slew rate, this is more of an issue with the amplifier design than it is with the differential output. The Wire BAL-BAL has an output impedance of less than 7 milliohms from DC to 20kHz, and although this is indeed twice the output impedance of the SE-SE variant, it's totally irrelevant in the grand scheme of things. Contact resistance and a few feet of headphone cable will increase this by 10-fold, making the amplifier's contribution negligible.
I also take serious issue with the statement "increased distortion... and altered frequency response. This is common knowledge for bridged amplifiers."
This is not "common knowledge" because its completely false. Distortion will decrease in this arrangement (cancellation of 2nd harmonic) and there is no impact on the frequency response of the system.
This is correct, but I'm not sure why this is being discussed. If anything, it's an argument in favor of a fully differential system where the transducer at the end of the chain can provide perfect CMRR. A transducer connected in a single-ended manner cannot do this as it has one lead with signal, and the other referenced to ground.
This is the exact same statement as #3. No sure why it's stated twice. See my answer above.
In all honestly, I think both connection types will provide a fantastic listening experience, and I really don't have a dog in this fight since I will happily offer you either an SE-SE or a BAL-BAL depending on your taste or requirements. What I do take issue with is trying to argue against what is unquestionably a technically superior solution in a balanced headphone connection just because you don't offer a product with a balanced output.
Regards,
Owen
It is indeed normal... The DAC itself will dissipate about 30W with normal 38V rails. The supply will dissipate another 10-20W depending on the input voltage, and the transformers will dissipate a few watts each depending on the VA rating. It's perfectly fine for the heatsink to sit at 50C or 60C, but what you don't want is for the DAC IC itself to run above 50C.
My heatsink measures 21cm x 15cm x 5cm and runs at 45C. This keeps the ESS DAC itself at about 41C.
This kind of performance from such a simple circuit doesn't come for free! You pay for it in heat...
Regards,
Owen