A BIT MORE ON SMT: There are lots of people doing DIY audio projects. Some are engineers, some are detail freaks, and some just get lucky. Hand soldering (or using DIY "reflow" equipment) for SMT parts is tricky business.
On top of obvious damage like cracking parts by heating them too fast or too high of moisture content, you can get them too hot and/or harm SMT parts in more subtle ways. Capacitors and linear semiconductors are especially susceptible to this sort of non-obvious damage.
The datasheets are full of disclaimers that read something like "heating the die to temperatures above xxx C may cause parametric changes"--i.e. the part may still work but no longer as intended. Guides on reflow warn of several kinds of damage including this one mentioning "high temperatures (beyond 260 °C) may cause damage to the internal dies of SMT components as well as foster intermetallic growth":
Wikipedia Reflow Soldering
I'm not trying to say don't ever try surface mount at home, but I think it's more tricky than many seem to realize. The parts are designed to be soldered with a liquid paste with very different properties than hand solder. And, more important, you're supposed to use computer controlled reflow lines that perform very specific timed and zoned temperature ramps.
With purely digital parts it's less of a concern as generally they either work or they don't. You might shorten their life hand soldering them but at least it's fairly obvious when they don't work right. Linear analog parts, however, are not so black and white.
That high-end SMT op amp you paid good money for might degrade to more of a $0.25 cheap part, or worse, if you get it too hot, subject it to excessive thermal shock, or highly uneven temperatures. I've even seen hand soldered SMT parts where there's a crack in the package on the bottom (invisible) side that lets the atmosphere in and slowly degrades the part over time.
So, in this case, I think it's better to be safe than sorry. Because most are not set up to fully measure their finished project meets all of the specs--especially one like this that's capable of extremely low distortion--it's best to have as little that can go wrong as possible in the assembly. That greatly helps the odds others will get the same performance as the reference design.
Ultimately, it's all about balancing trade offs. But I try make informed decisions myself, and where I can, help others do the same. DIY SMT, especially with analog, is risky.
On top of obvious damage like cracking parts by heating them too fast or too high of moisture content, you can get them too hot and/or harm SMT parts in more subtle ways. Capacitors and linear semiconductors are especially susceptible to this sort of non-obvious damage.
The datasheets are full of disclaimers that read something like "heating the die to temperatures above xxx C may cause parametric changes"--i.e. the part may still work but no longer as intended. Guides on reflow warn of several kinds of damage including this one mentioning "high temperatures (beyond 260 °C) may cause damage to the internal dies of SMT components as well as foster intermetallic growth":
Wikipedia Reflow Soldering
I'm not trying to say don't ever try surface mount at home, but I think it's more tricky than many seem to realize. The parts are designed to be soldered with a liquid paste with very different properties than hand solder. And, more important, you're supposed to use computer controlled reflow lines that perform very specific timed and zoned temperature ramps.
With purely digital parts it's less of a concern as generally they either work or they don't. You might shorten their life hand soldering them but at least it's fairly obvious when they don't work right. Linear analog parts, however, are not so black and white.
That high-end SMT op amp you paid good money for might degrade to more of a $0.25 cheap part, or worse, if you get it too hot, subject it to excessive thermal shock, or highly uneven temperatures. I've even seen hand soldered SMT parts where there's a crack in the package on the bottom (invisible) side that lets the atmosphere in and slowly degrades the part over time.
So, in this case, I think it's better to be safe than sorry. Because most are not set up to fully measure their finished project meets all of the specs--especially one like this that's capable of extremely low distortion--it's best to have as little that can go wrong as possible in the assembly. That greatly helps the odds others will get the same performance as the reference design.
Ultimately, it's all about balancing trade offs. But I try make informed decisions myself, and where I can, help others do the same. DIY SMT, especially with analog, is risky.
So much for Ford (National Semi) vs Chevy (TI):
TI to acquire National Semiconductor
{EDITORIAL COMMENTARY} I don't want to get too far off track here, but I believe competition is everything when it comes to avoiding mediocrity. And we seem to have less and less of it almost by the day. If TI buys Analog Devices and Linear Tech the party will be largely over.
TI has already turned some amazing entrepreneurial semiconductor companies into slow moving, budget constrained, mindless divisions lost in the vast TI bureaucracy. All the best employees typically leave (or are forced out) and those who are left hate their new job compared to their old one. Of course National wasn't exactly a start up, but still. { /EDITORIAL COMMENTARY }
TI to acquire National Semiconductor
{EDITORIAL COMMENTARY} I don't want to get too far off track here, but I believe competition is everything when it comes to avoiding mediocrity. And we seem to have less and less of it almost by the day. If TI buys Analog Devices and Linear Tech the party will be largely over.
TI has already turned some amazing entrepreneurial semiconductor companies into slow moving, budget constrained, mindless divisions lost in the vast TI bureaucracy. All the best employees typically leave (or are forced out) and those who are left hate their new job compared to their old one. Of course National wasn't exactly a start up, but still. { /EDITORIAL COMMENTARY }
in an early through hole prototype I used LM6172 dual with both op amps in the chip paralleled with 10 Ohm current sharing R (but not Class A biased) for the fast output op amp - adequate I,V for my 300 Ohm HD-600 headphones - the 100 MHz GBW LM6172 are fast enough to run with local gain equal to the overall amp closed loop gain of +5 to +10 so that the input op amp is ~ working "unity gain" (and the LM6172 is "well behaved" with higher local loop gain - I wouldn't try to use them in unity gain - so the protoype used the same full closed loop local feedback gain in each of the dual with separate feedback networks)
it turns out that pretty good distortion performance is possible even with TL071 input/global loop feedback op amp (not the -160 dB though) as long as its output stage doesn't have to drive any load or swing to the ps rails
you may need a local feedback "lead" C around the input op amp if you use a much faster input/global feedback op amp - the output op amp local feedback corner frequency at ~GBW/local output gain should be at least 2x the input op amp unity gain frequency
most iem have insane high sensitivity: 100-200 mVrms will drive them to 120 dB SPL - so amplification isn't really what you need with desktop CD/DVD/DAC giving 2 Vrms, many portable players 1 Vrms
I've repeatedly proposed custom step down Xmfr for iem drive - but I find them uncomfortable so I've never spec'd or built anything for them
it turns out that pretty good distortion performance is possible even with TL071 input/global loop feedback op amp (not the -160 dB though) as long as its output stage doesn't have to drive any load or swing to the ps rails
you may need a local feedback "lead" C around the input op amp if you use a much faster input/global feedback op amp - the output op amp local feedback corner frequency at ~GBW/local output gain should be at least 2x the input op amp unity gain frequency
most iem have insane high sensitivity: 100-200 mVrms will drive them to 120 dB SPL - so amplification isn't really what you need with desktop CD/DVD/DAC giving 2 Vrms, many portable players 1 Vrms
I've repeatedly proposed custom step down Xmfr for iem drive - but I find them uncomfortable so I've never spec'd or built anything for them
Last edited:
I think you make it sound a little worse than it really is.
I've been soldering SMT by hand since I started this hobby about 6 years ago. I've yet to run into any serious problems. Granted, I stick to leaded parts and not stuff like BGA.
I've even built the design this thread is about and soldered the LME49600 to the PCB by hand on my first design. The second design I made converted it to a through hole part to save space and allow a real heatsink to be attached.
Hey theAnonymous1! I was hoping you would spot this thread eventually. Your LME49600 / LME49713 designs are what got me interested enough in the forum to join last year. 
I just noticed opc's comments he posted in the "wire" thread about the National DC servo circuit in the datasheet headphone amp not working well. That explains something. I SPICE'd just that servo circuit a couple of days ago and have been struggling with it ever since. I figured that I must be missing something. As I recall you had a few mV of output DC offset too.
I just noticed opc's comments he posted in the "wire" thread about the National DC servo circuit in the datasheet headphone amp not working well. That explains something. I SPICE'd just that servo circuit a couple of days ago and have been struggling with it ever since. I figured that I must be missing something. As I recall you had a few mV of output DC offset too.
Last edited:
Heh, I don't know if I should feel flattered or feel like I cheated you somehow; considering I barely know what I'm doing.
I've built three LME49600 boards and one LME49713 board so far and they have all had varying offsets, the highest being 20mV. I still can't definitively figure out what's causing it. It seems to me like it's the op-amps natural offset amplified by the gain. The servo doesn't null it out for whatever reason. Resistor tolerance doesn't seem to change anything.
@theAnonymous1 thanks for the comments. I was mostly just posting factual info about SMT that some may not be aware of. Someone like yourself with six years of SMT experience is not my main concern. But, for anyone, the 5 pin exposed tab LME49600 is comparatively easy to solder, and much harder to harm, than the 28 pin hidden "tab" TPA6120. And, obviously, the through hole BUF634 is more foolproof still.
@jcx, I've never tried paralleling op amps. I've always picked an op amp with enough drive capability to begin with. But, at some point, you might exceed the peak current and/or dissipation capability of even the best op amps that meet the other specs you require. So, for those applications, paralleled devices could be an interesting solution if you don't want to use some sort of output buffer IC or a discrete output stage. There are many ways to slice the pie.
I think providing 2 or more gain options is a reasonable solution for highly efficient IEMs. The Benchmark DAC1 series uses internal jumpers and they work well. Do you have suggestions on where/how to best implement a gain option?
If I understood correctly, you were proposing running the input stage as a unity gain inverting buffer and having all the gain take place in the TPA6120 output stage with local feedback for both? Is that correct? Is that something you've prototyped at some point? So for a gain option, would you propose jumpers altering the gain of the TPA6120 output stage? Or set it to the lower gain required and add optional gain to the first stage? Or none of the above?
@jcx, I've never tried paralleling op amps. I've always picked an op amp with enough drive capability to begin with. But, at some point, you might exceed the peak current and/or dissipation capability of even the best op amps that meet the other specs you require. So, for those applications, paralleled devices could be an interesting solution if you don't want to use some sort of output buffer IC or a discrete output stage. There are many ways to slice the pie.
I think providing 2 or more gain options is a reasonable solution for highly efficient IEMs. The Benchmark DAC1 series uses internal jumpers and they work well. Do you have suggestions on where/how to best implement a gain option?
If I understood correctly, you were proposing running the input stage as a unity gain inverting buffer and having all the gain take place in the TPA6120 output stage with local feedback for both? Is that correct? Is that something you've prototyped at some point? So for a gain option, would you propose jumpers altering the gain of the TPA6120 output stage? Or set it to the lower gain required and add optional gain to the first stage? Or none of the above?
Hello,
Presbyopia starts coming on when you are a teenager.
I can do tweezers, liquid flux, a dab of solder on a pad, solder wick and hot air. The tiny surface mount parts require a different set of tools and tricks to do by hand. They are not so fun, they are doable if you are so inclined. They are not for the beginner.
A friend of mine crafts handmade knives, tiny ones as objects of art and real neck slicers.
Tiny work does not thrill me.
DT
All just for fun!
Presbyopia starts coming on when you are a teenager.
I can do tweezers, liquid flux, a dab of solder on a pad, solder wick and hot air. The tiny surface mount parts require a different set of tools and tricks to do by hand. They are not so fun, they are doable if you are so inclined. They are not for the beginner.
A friend of mine crafts handmade knives, tiny ones as objects of art and real neck slicers.
Tiny work does not thrill me.
DT
All just for fun!
for the types of multiloop composite op amp circuits I like the input op amp global gain setting network feedback is taken from the output of the whole amp, the output op amp stage is fully inside the input op amp global feedback loop, along with any "current sharing" resistors when a parallel output stage is used (unlike the HA1 example)
when you look "dynamically" at loop gain, if the output op amp local feedback is configured to provide the same Av as the Global loop, then the input op amp feedback through the output stage gain, global feedback network is ~ 1x
just multiply through the output op amp local gain and the global feedback network attenuation
with the output stage local and overall global gain setting feedback matching then the global feedback loop gain intercept is about the unity gain frequency of the input op amp and the phase margin is the unity gain phase margin of the input op amp minus the phase shift from the output stage working with its local feedback - leading to the requirement for a very fast output op amp
the required coordination of the output stage local feedback and the Global feedback makes gain switching more difficult
gain switching the feedback around 100MHz op amps isn't always easy either
a big diy project drawback is that op amp "rolling" in general requires understanding both loops gain setting stability consequences and making feedback component changes with varying op amp GBW, phase margin if you're really trying to optimise performance
you don't have to match the output stage local feedback gain setting to the global feedback determined gain - but that is the most added loop gain you can easily use with "unity gain" compensated input op amps - you can always use less gain in the output - like the unity gain buffer you're using with the LME49600
unity gain output buffers "in the loop" are probably a popular diy choice because you don't have to understand as much about op amp feedback loop stability and input op amp "rolling" is easier
these multiloop circuits can be ran as inverting gain stages like I show in the 1st post in my composite op amp thread (with more complicated loop gain shaping),
http://www.diyaudio.com/forums/solid-state/45794-high-loop-gain-composite-op-amp-circuits.html
they can also be used in noninverting mode as in post #3 (schematic may be harder to follow because it includes bridged output)
both examples use less gain in the output op amp local feedback (at the Global loop gain intercept frequency) which gives up a little performance for a little more global loop stability margin
in noninverting mode your performance will be limited by the input op amp common mode nonlinearity and nonlinear input parasitic impedance interaction with source, feedback resistance
Last edited:
i reread my post and want to make it clear that my role and any input in any of these projects i mentioned is as a simple pick and place builder, nothing more, i just happen to have built quite a number of the projects you mention plus more.
i also didnt mean to harp on 'the wire' so much, i only offered it as an example of the national buffer driving about as difficult headphone load as can be imagined (apart from stats and some rather high impedance older cans of course), to offer confirmation that the lme49600/10 does not produce any offset in this design, plus provides the 250ma with the right layout as far as i can tell, negating the need to stack them Imo and to serve as a warning not to underestimate the power supply requirements when using these buffers.
i have build 2 of PA's qrv08 and to me they lack in comparison to 'the wire', particularly in bass performance, perhaps its the wimpy onboard transformers, the sound is very very clean, but does not provide the same authority
step down transformer to drive multidriver iems jcx? really? i wont right that off before i actually try it as i have been meaning to, especially coming from you, but do you own any iems like this? like little low impedance difficult to drive speakers, not linear in the slightest
i also didnt mean to harp on 'the wire' so much, i only offered it as an example of the national buffer driving about as difficult headphone load as can be imagined (apart from stats and some rather high impedance older cans of course), to offer confirmation that the lme49600/10 does not produce any offset in this design, plus provides the 250ma with the right layout as far as i can tell, negating the need to stack them Imo and to serve as a warning not to underestimate the power supply requirements when using these buffers.
i have build 2 of PA's qrv08 and to me they lack in comparison to 'the wire', particularly in bass performance, perhaps its the wimpy onboard transformers, the sound is very very clean, but does not provide the same authority
step down transformer to drive multidriver iems jcx? really? i wont right that off before i actually try it as i have been meaning to, especially coming from you, but do you own any iems like this? like little low impedance difficult to drive speakers, not linear in the slightest
Last edited:
"DIY SMT, especially with analog, is risky. "
I think if you are prepared to invest in a decent soldering iron and some optics (figure $150 to $200 for both), some tweezers, and you are careful, it really is not that bad. I started using SMD about a year ago and have been surprised at how quickly I got the hand of it. Also, saves a lot of drilling on big projects! to solder a 1206 resistor or cap, I don't think the iron is in contact with each pad longer than 1-2 secs. The IR reflow oven is I think equally as stressful, if not more so. Take a look at SuzyJ's amp - she used all SMD to create a very compact neat looking result.
I think if you are prepared to invest in a decent soldering iron and some optics (figure $150 to $200 for both), some tweezers, and you are careful, it really is not that bad. I started using SMD about a year ago and have been surprised at how quickly I got the hand of it. Also, saves a lot of drilling on big projects! to solder a 1206 resistor or cap, I don't think the iron is in contact with each pad longer than 1-2 secs. The IR reflow oven is I think equally as stressful, if not more so. Take a look at SuzyJ's amp - she used all SMD to create a very compact neat looking result.
Ok, Peranders has revised the Wiki and apparently his website (or I was blind yesterday). You can find the QRV09 web page here:
Sjostrom QRV09 Webpage
And a ZIP file with a PDF of the schematics and board layout here:
qrv09r0.zip
For an extra $8 or so, he'll pre-solder the TPA6120 to the board hopefully solving that tricky problem. That adds to the DIY-friendly factor.
The QRV09 schematic shows local feedback around the TPA6120 and around the input stage with no global feedback. Unfortunately, it also shows the TI recommended 10 ohm resistor in the output (along with an optional 110 ohm resistor). TI says:
The series output resistor should be kept at a minimum of 10 Ω. It is small enough so that the effect on the load is minimal, but large enough to provide the protection necessary such that the output of the amplifier sees little capacitance.
IMHO, they're dead wrong about the "effect on the load is minimal" if said load includes most balanced armature IEMs and even many 16 ohm conventional dynamic headphones. They will show significant frequency response deviations from that much series resistance. It will also degrade the damping factor and raise the system Q hanging the bass performance (generally for the worse if the headphone were designed, as many are, for a zero ohm source).
I have a board on order with Peranders and one of my first goals will be to hopefully replace that 10 ohm resistor with a parallel resistor inductor combination to drop the output impedance in the audio band while hopefully still keeping the design stable. But, I have to confess, this is another thing I don't much like about the TPA6120. Being an extremely "fast" current feedback device it's much more prone to stability problems. I have to agree with JCX it would be a bad idea to try and put gain options in the TPA6120 feedback loop. The data sheet spends quite a bit of time on the topic and even suggests using a network analyzer to determine the stability of the final design.
Some don't like series inductors in the output as they can cause phase shift at high audible frequencies. The National LME49600 doesn't require a series resistor or inductor between it and the headphones.
And, while I'm voicing TPA6120 concerns, I'm looking forward to measuring the channel separation under a variety of conditions. TI claims 90 dB, which if true under real world conditions, is plenty good. But because both channels are in a single part, I'm a bit concerned it's not going to be that good in reality. The LME49600, being a single channel device, also has an advantage here.
Of course you can always use two TPA6120's and not use half of each in a "dual mono" configuration. Or you can try to parallel them but that might create other stability problems (something JCX could probably help out with?). But then you can't use the Peranders board.
So, in summary, I have both the Chinese HA TPA6120 DAC and the Peranders QRV09 board on the way. At this point it seems to make the most sense to see how these perform before going much further with a LME49600 design.
Anyone who's on the fence about the QRV09 might want to jump on that bandwagon now before it's too late. If that board measures really well, and I can overcome the TI mandated 10 ohm output impedance, I likely won't do a LME49600/BUF634 design.
Sjostrom QRV09 Webpage
And a ZIP file with a PDF of the schematics and board layout here:
qrv09r0.zip
For an extra $8 or so, he'll pre-solder the TPA6120 to the board hopefully solving that tricky problem. That adds to the DIY-friendly factor.
The QRV09 schematic shows local feedback around the TPA6120 and around the input stage with no global feedback. Unfortunately, it also shows the TI recommended 10 ohm resistor in the output (along with an optional 110 ohm resistor). TI says:
The series output resistor should be kept at a minimum of 10 Ω. It is small enough so that the effect on the load is minimal, but large enough to provide the protection necessary such that the output of the amplifier sees little capacitance.
IMHO, they're dead wrong about the "effect on the load is minimal" if said load includes most balanced armature IEMs and even many 16 ohm conventional dynamic headphones. They will show significant frequency response deviations from that much series resistance. It will also degrade the damping factor and raise the system Q hanging the bass performance (generally for the worse if the headphone were designed, as many are, for a zero ohm source).
I have a board on order with Peranders and one of my first goals will be to hopefully replace that 10 ohm resistor with a parallel resistor inductor combination to drop the output impedance in the audio band while hopefully still keeping the design stable. But, I have to confess, this is another thing I don't much like about the TPA6120. Being an extremely "fast" current feedback device it's much more prone to stability problems. I have to agree with JCX it would be a bad idea to try and put gain options in the TPA6120 feedback loop. The data sheet spends quite a bit of time on the topic and even suggests using a network analyzer to determine the stability of the final design.
Some don't like series inductors in the output as they can cause phase shift at high audible frequencies. The National LME49600 doesn't require a series resistor or inductor between it and the headphones.
And, while I'm voicing TPA6120 concerns, I'm looking forward to measuring the channel separation under a variety of conditions. TI claims 90 dB, which if true under real world conditions, is plenty good. But because both channels are in a single part, I'm a bit concerned it's not going to be that good in reality. The LME49600, being a single channel device, also has an advantage here.
Of course you can always use two TPA6120's and not use half of each in a "dual mono" configuration. Or you can try to parallel them but that might create other stability problems (something JCX could probably help out with?). But then you can't use the Peranders board.
So, in summary, I have both the Chinese HA TPA6120 DAC and the Peranders QRV09 board on the way. At this point it seems to make the most sense to see how these perform before going much further with a LME49600 design.
Anyone who's on the fence about the QRV09 might want to jump on that bandwagon now before it's too late. If that board measures really well, and I can overcome the TI mandated 10 ohm output impedance, I likely won't do a LME49600/BUF634 design.
Last edited:
in audio power amps the load isolation series impedance uses an inductor to keep the audio frequency efficiency/damping ratio high while "disconnecting" the amp output from the load (mostly cable C causes stability problems) at high frequency near the loop gain intercept/corner frequency
the TPA6120 is so fast that the load isolation impedance can be lower than you normally see in audio power amps which may have only ~ 1 MHz loop gain intercept and at best 30-50 MHz output Q - requiring a few uH inductance to get load C isolation near 1 MHz and beyond
the TPA can have >50 MHz corner frequency and uses GHz Q so the ~10 Ohm series load isolation impedance doesn't have to fully develop until beyond ~10 MHz so < ~1uH inductance can be OK
lossy ferrite bead can be a good technical fit to the load isolation requirement - you may even find some that physically fit the QRV09 10 Ohm output series R smt pad
to avoid the often claimed audio distortion of output load isolating ferrites I spec beads with Isat ratings ~ 10x of the op amp short circuit current http://www.steward.com/pdfs/brochures/Broch008.pdf
those really concerned by audiophile tweaker's unverified claims about ferrite "ruining the sound" can always use air core inductor and parallel R - old carbon comp with a winding over the body is one option
even John Curl seems to agree that < ~ 1uH series load isolation L doesn't harm "the sound" with dynamic speaker loads
except for the multi-armature high end iem all other dynamic headphones have much higher Z than loudspeakers, and usually only a single bass impedance peak followed by a slow high frequency Z rise due to coil inductance - this makes their response very insensitive to small series load isolation L
I can't see worrying about the special impedance requirements of multi-armature iem when drive level mismatch, amplified input noise caused hiss from a general purpose amp will be audible - these special iem require a specialized amp design
the TPA6120 is so fast that the load isolation impedance can be lower than you normally see in audio power amps which may have only ~ 1 MHz loop gain intercept and at best 30-50 MHz output Q - requiring a few uH inductance to get load C isolation near 1 MHz and beyond
the TPA can have >50 MHz corner frequency and uses GHz Q so the ~10 Ohm series load isolation impedance doesn't have to fully develop until beyond ~10 MHz so < ~1uH inductance can be OK
lossy ferrite bead can be a good technical fit to the load isolation requirement - you may even find some that physically fit the QRV09 10 Ohm output series R smt pad
to avoid the often claimed audio distortion of output load isolating ferrites I spec beads with Isat ratings ~ 10x of the op amp short circuit current http://www.steward.com/pdfs/brochures/Broch008.pdf
those really concerned by audiophile tweaker's unverified claims about ferrite "ruining the sound" can always use air core inductor and parallel R - old carbon comp with a winding over the body is one option
even John Curl seems to agree that < ~ 1uH series load isolation L doesn't harm "the sound" with dynamic speaker loads
except for the multi-armature high end iem all other dynamic headphones have much higher Z than loudspeakers, and usually only a single bass impedance peak followed by a slow high frequency Z rise due to coil inductance - this makes their response very insensitive to small series load isolation L
I can't see worrying about the special impedance requirements of multi-armature iem when drive level mismatch, amplified input noise caused hiss from a general purpose amp will be audible - these special iem require a specialized amp design
Last edited:
I've made a CF bastard child of the LME49600 reference design using LME49713 (and minus the LME49600). The person I built it for liked it better than the LME49600 I had previously built for them.
Plan to build myself another one soon. I also made a layout that allows switching the LME49600 in/out of the op-amp feedback loop with a jumper. Curious to hear what kind of difference it makes.
Plan to build myself another one soon. I also made a layout that allows switching the LME49600 in/out of the op-amp feedback loop with a jumper. Curious to hear what kind of difference it makes.
Thanks JCX. I'll look for a 1 uH (if you're confident in that value?) SMT inductor that might work and perhaps could even be "SMT stacked" on top of a 10 ohm SMT resistor to damp any resonances (ultrasonic, RF, or otherwise). I can also test for inductor saturation issues with the dScope and my Agilent scope.
I agree kludging in an air wound inductor on an old-school carbon resistor would be the audiophile approved solution. But given how fussy the TPA6120 is, that might create other problems with parasitics, longer lead lengths, etc.
We have some time, so if you're bored, please feel free to toss some values in whatever sims you already have of the TPA6120 and see what you come up with for the minimum L? I can also order multiple values and test stability, phase shift, saturation, etc. starting with the smallest and moving up if needed.
I agree kludging in an air wound inductor on an old-school carbon resistor would be the audiophile approved solution. But given how fussy the TPA6120 is, that might create other problems with parasitics, longer lead lengths, etc.
We have some time, so if you're bored, please feel free to toss some values in whatever sims you already have of the TPA6120 and see what you come up with for the minimum L? I can also order multiple values and test stability, phase shift, saturation, etc. starting with the smallest and moving up if needed.
FYI, I calculate the parts cost from US suppliers for the QRV09 at about $40 - $55 depending on how many different suppliers you want to use, shipping costs, etc. And the PCB is about $20 with shipping from Peranders. So that leaves a volume pot ($2 - ??) head phone jack ($1 - ??), input connectors ($1 - ??) and a line cord/AC input socket ($2 - ??). So that's about $80+ or so... add another $20 - $50 for a decent enclosure and you're at $100 - $130+ for the total project depending on how fancy you want to get with the off board stuff and enclosure. You could save about $5 with the LM4562 vs the AD8610.
That's within $10 of the National reference design I was considering with the same power supply.
And it looks possible, at least on paper, to solve the QRV09's 10 ohm output impedance issue without a very clean modification for those concerned about that.
JCX, a quick follow up on the above... It looks like 500mA is about the most saturation current you can get in a 1 uH 0805 SMT inductor (rated currents go up to about 1.2 amps but that's well into saturation). A typical part is the Taiyo Yuden CB2012T1R0M. That's not exactly a huge margin but should work OK. Again, the right testing will show if there are any issues. I'm hopeful stacking an 0805 inductor with a 10 resistor will do the trick.
That's within $10 of the National reference design I was considering with the same power supply.
And it looks possible, at least on paper, to solve the QRV09's 10 ohm output impedance issue without a very clean modification for those concerned about that.
JCX, a quick follow up on the above... It looks like 500mA is about the most saturation current you can get in a 1 uH 0805 SMT inductor (rated currents go up to about 1.2 amps but that's well into saturation). A typical part is the Taiyo Yuden CB2012T1R0M. That's not exactly a huge margin but should work OK. Again, the right testing will show if there are any issues. I'm hopeful stacking an 0805 inductor with a 10 resistor will do the trick.
I would tend to agree. I think I've lost interest in National's design at this point. I'll be really curious to read your reviews of the TPA6120 HA and P-A's board.
0805 is ludicrous size for something that may see short circuit conditions
from what I can see of the QRV09 schematic thumbnail it looks like you could use the thru hole jumper positions to put a leaded bead core (hair pin bend the lead to fit) in series by shorting one 0805 pad?
I would like to see the Idc impedance curves but the 5A Steward 28L0138-10R may work
I much prefer lossy bead core in this app - wound "inductor" may have too low self resonant frequency and unnecessary Rdc
from what I can see of the QRV09 schematic thumbnail it looks like you could use the thru hole jumper positions to put a leaded bead core (hair pin bend the lead to fit) in series by shorting one 0805 pad?
I would like to see the Idc impedance curves but the 5A Steward 28L0138-10R may work
I much prefer lossy bead core in this app - wound "inductor" may have too low self resonant frequency and unnecessary Rdc
Last edited:
JCX, I think the output impedance jumper block is a useful feature that more headphone amps should have. If someone, for example, has two or more different headphones (I know lots of people with closer to a dozen) and one is designed for zero ohms and the other the IEC 120 ohm standard, they can use the jumper to have their cake and eat it too. Obviously you could also install a switch off board but it's an elegant solution to have the jumper.
So if you meant using those holes, that's a trade off but could work. You mentioned "thumbnails"? Oddly, Peranders has removed the documentation link from the wiki. I'll email it to you so you have the full size images in the PDF.
The datasheet for the CB2012T1R0M 0805 part is here:
http://www.yuden.co.jp/ut/product/pdf/wound02_e.pdf
The self resonant frequency is 100 Mhz and the DCR is only 0.15 ohms which is plenty low. The saturation current is 500 mA and the rated current is 900 mA.
I agree 500 mA is a bit marginal but it's likely close to the real world current output of the TPA6120. The TI datasheet is rather vague about what conditions it will put out 700 mA (they don't specify the load--is that short circuit current?). And, in real world use, even if it can manage more than 500 mA it's unlikely anyone will need that much current.
On 15 volt rails TI specifies a 12 volt output swing into 32 ohms. That's a 3 volt drop from the rails. So with the QRV09's 12 volt rails, you'll get about 9 volts peak into 32 ohms. That's 280 mA peak of peak current.
Into 16 ohms, It would have to manage better than 8 volts. And on 12 volt rails, I'm betting that's about the limit of the TPA6120. And, regardless, that's 2000 mW!!!
So, real world, with any load over 18 ohms you cannot saturate the 0805 inductor no matter how loud someone wants to listen. And even with lower impedance (or highly reactive) loads, I suspect the TPA6120 will clip at very close to the same point the 0805 inductor saturates. And considering 50 mW into 16 ohms is well into hearing damage territory with any low impedance headphones I know of, even if you multiply that by 10X, you still get 500 mW which the inductor can do with ease.
But, obviously, someone could certainly use a bigger inductor as long as the physical wiring/mounting didn't create a TPA6120 stability problem. Using the jumper block through hole points should work but you either give up on the selectable impedance feature or have to move it off board.
I'm starting to feelthis should be in the Peranders QRV09 thread... but here is fine for now. I don't want to step on any toes in the other thread. Not everyone agrees with my views about output impedance but the trend seems to be towards "zero ohm" for a lot of good reasons. See:
http://www.diyaudio.com/forums/headphones/162962-maximum-allowable-headphone-amp-output-impedance-2.html#post2522343
So if you meant using those holes, that's a trade off but could work. You mentioned "thumbnails"? Oddly, Peranders has removed the documentation link from the wiki. I'll email it to you so you have the full size images in the PDF.
The datasheet for the CB2012T1R0M 0805 part is here:
http://www.yuden.co.jp/ut/product/pdf/wound02_e.pdf
The self resonant frequency is 100 Mhz and the DCR is only 0.15 ohms which is plenty low. The saturation current is 500 mA and the rated current is 900 mA.
I agree 500 mA is a bit marginal but it's likely close to the real world current output of the TPA6120. The TI datasheet is rather vague about what conditions it will put out 700 mA (they don't specify the load--is that short circuit current?). And, in real world use, even if it can manage more than 500 mA it's unlikely anyone will need that much current.
On 15 volt rails TI specifies a 12 volt output swing into 32 ohms. That's a 3 volt drop from the rails. So with the QRV09's 12 volt rails, you'll get about 9 volts peak into 32 ohms. That's 280 mA peak of peak current.
Into 16 ohms, It would have to manage better than 8 volts. And on 12 volt rails, I'm betting that's about the limit of the TPA6120. And, regardless, that's 2000 mW!!!
So, real world, with any load over 18 ohms you cannot saturate the 0805 inductor no matter how loud someone wants to listen. And even with lower impedance (or highly reactive) loads, I suspect the TPA6120 will clip at very close to the same point the 0805 inductor saturates. And considering 50 mW into 16 ohms is well into hearing damage territory with any low impedance headphones I know of, even if you multiply that by 10X, you still get 500 mW which the inductor can do with ease.
But, obviously, someone could certainly use a bigger inductor as long as the physical wiring/mounting didn't create a TPA6120 stability problem. Using the jumper block through hole points should work but you either give up on the selectable impedance feature or have to move it off board.
I'm starting to feelthis should be in the Peranders QRV09 thread... but here is fine for now. I don't want to step on any toes in the other thread. Not everyone agrees with my views about output impedance but the trend seems to be towards "zero ohm" for a lot of good reasons. See:
http://www.diyaudio.com/forums/headphones/162962-maximum-allowable-headphone-amp-output-impedance-2.html#post2522343
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.