I've been on and off about this design because it looks like the only design that can reach acceptable (to my pickiness) THD+N performance for super low impedance headphones and IEMs at the upper frequencies. In other words, if you pick up a pair of 8 or 12 Ohm IEMs, then performance around 20kHz is pretty bad on almost everything. Even the O2 looks like it'll be out of spec for my pickiness if I got a pair of K3003's. This one looks like it'll be perfect even though it's very much overkill for such in terms of power ability.
What I want to do is use this design behind an AD797 buffer circuit.
What I want to do is use this design behind an AD797 buffer circuit.
the wire pwns it, this is just the datasheet (well without the multiloop FB) and a pretty average pcb. nothing specifically wrong with that, but singling it out for uber high performance seems a bit strange considering afaik there is no empirical data. check the plots on the wire, 20khz is a walk in the park, as is driving 2 ohm loads in the form of the LPUHP, so i gather it would be the same with the other amps in the family since they share pretty much everything
Last edited:
The user guide for the build (http://www.ti.com/lit/an/snaa052/snaa052.pdf) has a lot of measurements. I can only infer 8 Ohm performance, but with a 9-15V supply it looks like it'll be under 0.003% THD+N for the entire listening range at what I believe are normal listening levels. The cool thing is that if I had a second gain setting and plugged in my standard headphones with a 15V supply, I'd get 50x better performance. Not bad. I'll have to go back and look at the wire again. It's definately smaller.
yes but those measurements are taken with a properly laid out pcb, you cannot infer it with the ebay one. like i said anyway the wire pwns it hehe even just the THD+N 2nd harmonic of the first gen bal-se boards is -120db/0.0001%, let alone the latest if you want to look these are even better. then 3rd is -125 and the noise floor -145. perfectly linear and these numbers dont really change driving 2, 4 , 8 , 32, 300ohms
Well, no servo like the reference design. I remember reading someone post on this forum that they had some noise with the ebay board, but it could have been caused by poor wiring.
I etched some boards along the lines of the reference design, but with the option to switch the buffer outside the feedback loop with a jumper just to see what it sounds like. That was over a year ago, but still haven't found the motivation to actually populate the board.
I wonder what resistor value they used to split the shared inputs to the pair of LME49600. On OPC's multi paralleled LME49600 he didn't use any.
Can this board be all that bad? It says they've sold 118 of them. I seem to recall seeing it or something from this seller a year or so ago.
Do you really need a board though? This one's not ground planed. You could just use a bit of Vero type board and solder the parts wired point to point. That gets rid of any adverse effect from the copper tracks and reduces the number of solder joints in the signal part and elsewhere.
Can this board be all that bad? It says they've sold 118 of them. I seem to recall seeing it or something from this seller a year or so ago.
Do you really need a board though? This one's not ground planed. You could just use a bit of Vero type board and solder the parts wired point to point. That gets rid of any adverse effect from the copper tracks and reduces the number of solder joints in the signal part and elsewhere.
Last edited:
"Can this board be all that bad? It says they've sold 118 of them. I seem to recall seeing it or something from this seller a year or so ago."
i wanna compare with The Wire but i just missed the wire GB
"Do you really need a board though? This one's not ground planed. You could just use a bit of Vero type board and solder the parts wired point to point. That gets rid of any adverse effect from the copper tracks and reduces the number of solder joints in the signal part and elsewhere."[/QUOTE]
Actually, no, i found this ebay pcb n its seem copy from the wire design
i wanna compare with The Wire but i just missed the wire GB
"Do you really need a board though? This one's not ground planed. You could just use a bit of Vero type board and solder the parts wired point to point. That gets rid of any adverse effect from the copper tracks and reduces the number of solder joints in the signal part and elsewhere."[/QUOTE]
Actually, no, i found this ebay pcb n its seem copy from the wire design
Actually, no, i found this ebay pcb n its seem copy from the wire design[/QUOTE]
its nothing like the wire, just a similarity (not even the same) in the 2 chips used. the ebay board is a slightly lacking copy of the datasheet without the servo, with no ground plane and no proper decoupling. it uses the metal can chips, which although nice, audiophile friendly to those with irrational fear of smd, are 5x the cost of lme49990, for worse performance
it wasnt copied from the wire, the wire came later, but does not take any cues from the ebay board
ahh ok cool, but yeah still best to go with the wire if you can pick up a pcb, or ask opc if you can make one up using the files. youve taken too long realizing it while posting in the O2 thread right next door. pretty sure there'll be someone willing to let go of one of the SE-SE rework boards, which just need a tiny single patch. these were additional to the GB when something got switched in the pcb files, so were let go on the cheap (snapped up in a matter of hours). the SE-SE was the best performer in some aspects
I spent a lot of time in the Wire thread when I joined a few months ago, but it never resonanted with me like the O2. I don't think I even posted in the Wire thread. I knew about the National design at that point and I think I dismissed the Wire because cost and performance didn't line up as well as with the National design. I'm serious. National has readings from 20-20k and they have output swings shown at typical loads for 100-1k-10k. From what I can infer, the Wire will get me within spec, but National's design still performs better for only a few more dollars.
So in planning for the future of ultra-low impedance IEMs, I'm using numbers from AKG's ridiculously expensive K3003. I believe they represent where we'll be in a few years for normal cost products, so I'm working with their specs. That's 8Ω and 0.0025mW (0.000025W) for a 78dB listening level. I messed up in my previous projections based on forgetting a couple zeros (mW to W conversion). That translates to 0.45mV and 0.56mA. Anyway, at those specs, the National design is projected to deliver ~0.005% THD+N @ 1kHz. That's a bit higher than I hoped because it's going to be nearly 10x the distortion at 20kHz based on their curves. For these criteria, I can't find a better design out there.
I wonder what AKG is thinking with a design that's guaranteed to sound distorted on pretty much everything... and cost over $1k. I'm interested to see how the O2 or ODA measures with an 8Ω load at reasonable listening levels. My projections put it >0.1% THD+N in the higher frequencies (above 10kHz). I'm looking for better than that with the most difficult products and I hope I'm wrong with those numbers anyway.
The O2 is the best thing since sliced bread for my preferred headphones at any reasonable cost. Everything, including IEMs, benefit from the noise supression, regulated power delivery, and low output impedance. One of the things working against me, though, is the efficiency of IEMs. I need something that has unity gain and fractional gain options. Building that into the National design may negatively impact its performance.
So in planning for the future of ultra-low impedance IEMs, I'm using numbers from AKG's ridiculously expensive K3003. I believe they represent where we'll be in a few years for normal cost products, so I'm working with their specs. That's 8Ω and 0.0025mW (0.000025W) for a 78dB listening level. I messed up in my previous projections based on forgetting a couple zeros (mW to W conversion). That translates to 0.45mV and 0.56mA. Anyway, at those specs, the National design is projected to deliver ~0.005% THD+N @ 1kHz. That's a bit higher than I hoped because it's going to be nearly 10x the distortion at 20kHz based on their curves. For these criteria, I can't find a better design out there.
I wonder what AKG is thinking with a design that's guaranteed to sound distorted on pretty much everything... and cost over $1k. I'm interested to see how the O2 or ODA measures with an 8Ω load at reasonable listening levels. My projections put it >0.1% THD+N in the higher frequencies (above 10kHz). I'm looking for better than that with the most difficult products and I hope I'm wrong with those numbers anyway.
The O2 is the best thing since sliced bread for my preferred headphones at any reasonable cost. Everything, including IEMs, benefit from the noise supression, regulated power delivery, and low output impedance. One of the things working against me, though, is the efficiency of IEMs. I need something that has unity gain and fractional gain options. Building that into the National design may negatively impact its performance.
i think the only difference you will find between them is the national design has better measuring equipment because opc was bouncing off the audio precision noise floor at -140db. i wouldnt plan for the stupid AKG, they will be driven TERRIBLY by the daps people will use them with and you can bet people wont support that 8 ohms driven by the average dap output z of 2-10 ohms will sound like sh1t. also, is THD the only spec you read? that servo in the national design will mean its a whole lot less linear than the wire.
but hey do what you like and the point is kinda moot, you cant buy the wire even if you wanted to, all several hundred pcbs are sold out
the national design in the real world has problems with output offset and this fact actually impacted the sales of the national buffer for a couple of years. but i'll take an actual real world measurement by a hobbiest donating his time for his own pleasure in building and sharing ie. very little if anything to gain; over a company trying to sell product
but hey do what you like and the point is kinda moot, you cant buy the wire even if you wanted to, all several hundred pcbs are sold out
the national design in the real world has problems with output offset and this fact actually impacted the sales of the national buffer for a couple of years. but i'll take an actual real world measurement by a hobbiest donating his time for his own pleasure in building and sharing ie. very little if anything to gain; over a company trying to sell product
Last edited:
Whats about to use TO99 LME49713HA (two or more in parallel)?
One could drive about +- 100mA
http://www.national.com/ds/LM/LME49713.pdf
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.