Go Back   Home > Forums > Amplifiers > Headphone Systems

Headphone Systems Everything to do with Headphones

Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Reply
 
Thread Tools Search this Thread
Old 1st April 2011, 12:36 PM   #11
IanAS is offline IanAS  England
diyAudio Member
 
Join Date: Dec 2004
Location: London
Gradually developed it from a good sound card into an outstanding one.

Every change carefully done with extensive listening tests at every position of capacitor of chip change. I didn't just throw a bunch of caps onto it. I tried standard and NX Black gates in most positions and preferred the Nichicon Polymers, better more informative bass. Very good transient response, very clean. The BGs bass was larger but muddled.

http://www.avsforum.com/avs-vb/showt...5#post19865855
I posted further info lower down and on the next page.

The sound card is a work in progress. I have some Vishays to throw on in place of the surface mount chips and also, just in case it helps, replace the chip caps with polystyrene, much larger but I think can be fitted.

Then I thought I'd try and fit a couple or four LME49600, or eight if in push pull, or BUF634 and what ever is needed, to make a headphone output onto a daughter board. Or, a multiple transistor output stage with a Vishay for each emitter resistor.

Last edited by IanAS; 1st April 2011 at 12:52 PM.
  Reply With Quote
Old 1st April 2011, 04:21 PM   #12
diyAudio Member
 
RocketScientist's Avatar
 
Join Date: Mar 2008
Location: Northwest
Quote:
Originally Posted by Atilla View Post
Oh lord, what have you done to the poor Auzentech ;(
To each their own. As I said earlier... if it puts a bigger smile on your face...

I'm thinking of spinning a quick and dirty board to use as a test mule to explore things like the parallel buffers. Besides the sighted listening tests, the only engineering reason I can see for paralleled buffers might be thermal. National implies at high temps the LME49600's current limiting could be a problem. Adding the heatsinks Agdr mentioned might be another solution if that excessive dissipation proves to be a real world problem (increasing the foil area only helps so much--especially with 1 oz copper). And as he suggested, it might just be a matter of altering the footprint/pads to accommodate optional heatsinks.

I plan to make measurements both ways (single device and paralleled devices). If nothing else that should at least yield a bit more objective information for others to base their decisions on.

So... dare I bring this up? I'm curious what others think is best for the power supply? National uses batteries in the reference design and I'm guessing most don't think 9 volt batteries are ideal. The main options I'm considering:
  • Just do an amp PCB with pads for the DC rails and leave it up to the user how they want to power it.
  • Include a relatively simple power supply on the board that accepts AC from a transformer. Users who don't want to use it can simply not populate the components.
  • Design it around a $35 external "brick" multi-output DC power supply such as the Meanwell P25A Series with a mating connector on the PCB.
  • Provide for an on-board split rail DC-DC converter module so it will run from a variety of inexpensive DC wall transformers or other DC sources and use a standard 5mm/2mm DC barrel power connector on the PCB.
Any of the above allow for using an off-board elaborate power supply if desired. All but the first choice provide a convenient single board solution. IMHO, with the right PCB layout there's little advantage to putting the power supply on its own board. It just adds expense and hassle to the construction.

The Meanwell power supply comes with a DIN connector and the PCB could accommodate a mating connector making for a very simple plug-and-play solution that's also fully "agency approved"--no wiring, nothing to mount to the chassis, little to go wrong, no exposed dangerous AC, etc. The same is true of the last solution with a DC wall transformer.

It also depends on what voltage rails a person wants (how much output swing you need from the amp). The Meanwell and DC-DC solutions would already be regulated and could just be LC filtered to further lower the noise and ripple. Or, the rails could be regulated again (say from +/- 15 down to +/- 12 (LDO) or +/- 9 volts (7809/7909) with on-board regulators at the expense of some output capability. The National board works very well with 9 volt rails.

The Meanwell triple output supply also has the benefit of a 5 volt supply to operate things like DC protection relays and associated circuitry keeping the split rail supply "pure" for just audio use. The same could be done with the on board DC-DC solution. The raw DC input could power the protection circuit.

The last two solutions also keep AC power completely out of the amp which should mean less hum and noise.

For those skeptical of switching supplies and/or DC-DC converters, it's worth pointing out a lot of high-end audio gear is now "going green" and using switching supplies. And some well reviewed DACs (like those from HRT) use DC-DC converters to generate split rails from USB power. Properly used, switching sources can provide less noise in the audible range than conventional linear supplies. But I know some audiophiles have a psychological bias against them.

I plan to size the board to slide into a reasonably priced enclosure such as the Hammond 1455Q and also have mounting holes so it can be easily mounted in other enclosures.

For a solution with external DC power, all the connections could likely be kept along one edge to require only one panel to be machined and minimize point-to-point wiring. Anyone could skip installing one or more of the PCB mount connectors and use any sort of chassis and wiring arrangement they wanted.

Overall I like the idea of a self contained board that's fairly usable (like the National eval board) "as is" without a bunch of Rube Goldberg wiring required. The last 2 power options would provide an elegant plug-and-play solution without needing an enclosure, off-board connectors, etc.

Thoughts?
__________________
http://nwavguy.com - Personal non-commercial audio blog

Last edited by RocketScientist; 1st April 2011 at 04:23 PM.
  Reply With Quote
Old 1st April 2011, 05:09 PM   #13
IanAS is offline IanAS  England
diyAudio Member
 
Join Date: Dec 2004
Location: London
Quote:
Originally Posted by RocketScientist View Post
Besides the sighted listening tests, the only engineering reason I can see for paralleled buffers might be thermal
And to lower the output impedance.

Thermal is not an issue is it? 'The Wire' had no heat sinks. If it becomes an issue, then use one, or a bigger one. The solder down tab is about saving money in large scale production.

My Senn HD650 are 300 ohms only in some places. Nearer to 500 ohms in others. I don't really know, but I saw some graphs which suggested that the higher the output resistance, the quieter the mid bass and treble as those are the 650's higher impedance areas. Output Ohms coil Z = Less current at the higher Z areas? That's probably not what's going on.

Maybe at say 10 Ohms output resistance, there is no further benefit. I don't know. With power amps, the lower the better.

I've read a few times that paralleled output devices produced more and better bass on headphones. Just the same as power amps.

Quote:
Originally Posted by RocketScientist View Post
I plan to make measurements both ways (single device and paralleled devices). If nothing else that should at least yield a bit more objective information for others to base their decisions on.
That will be interesting Maybe it's a back EMF thing. Lower output impedance reducing the effect on the feedback.

Last edited by IanAS; 1st April 2011 at 05:35 PM.
  Reply With Quote
Old 2nd April 2011, 02:04 AM   #14
agdr is offline agdr  United States
diyAudio Member
 
agdr's Avatar
 
Join Date: Sep 2010
RocketScientist: Good work! You are being thorough. That is always a good thing.

I finally remembered where I had seen that 10R output resistance figure for the BUF634. From this Burr Brown app note on using an op amp with the BUF634

http://focus.ti.com/general/docs/lit...5&fileType=pdf

Some good stuff in there although they discuss a 50 ohm output video amp. They go through some math with one op amp powering 3 BUF634s in parallel. From page 2, "The output resistance of the BUF634 is about 10 ohms."

Quote:
Originally Posted by RocketScientist View Post
I'm also a bit concerned about the loading to the gain stage with paralleled buffers. You get roughly twice the capacitance and half the impedance. It's a direct connection to the gain stage with no series resistance..
Good thought! From that app note, which also directly coupled the op amp to all 3 paralleled BUF634s:

"No series resistors were used between the output of op amp A1 and the buffer inputs since they would form a low pass filter in combination with the input capacitance of the buffers. Any phase shift resulting from this low-pass could cause the entire circuit to oscillate..."

So you are right, no series resistors between stages at least for the BUF634, and double (or in this case triple) the load capacitance presented to the output of the op amp. Something to look into.

Yep, definitely do some tests on single buffer vs. paralleled if you have the time. The results would be really interesting. My expectation would be that 10 ohms in the chip in series with the 44 ohm resistive part of my Shure's impedance over much of the frequency range = not so good in terms of power delivery. But the same with a 300 ohm set of cans, probably no significant difference.

Quote:
Originally Posted by RocketScientist View Post
How well they share current might be subject to random matching, temperature, etc.
Now this is really interesting. In that app note they go through the math to calculate the compensation current passed between paralleled BUF634s for the minimum and maximum input offset voltages, given the internal resistances that help balance the group up. Good grief, the BUF634 has a +/- 100mV max input offset voltage. I knew the LME49600 had a typical around 16mV, but its max is +/- 60mV. At those offsets they come up with several mA (3mA - 10mA).

I would agree with your comments about sharing the gain stage and servo in a single package. The servo half should be operating at a very low frequency vs. the gain stage. Shouldn't be much interaction there on the die or the shared power supply leads. And I also agree that is probably why National did it that way - they certainly did have access to chips for that eval board as you noted.

On that DC offset, I agree that 10mV or so is probably not going to affect the headphones much, but why is it there? Seems like that circuit should be able to do better. Probably time for some circuit measurements to find out what isn't working the way it should. Or it is working exactly the way it should and my expectations are off.

One thought is that huge +/-16mV -> +/- 100mV input offset voltage for the LME49600 and BUF634. Maybe that is pressing up against the edge of the design to compensate. Might also be interesting to throw the circuit into SPICE and do some sensitivity analysis by varying each part's value a few percent for effect on the output offset.

On that heatsink, I was mainly thinking in terms of reducing the amount of heatsink board foil needed while keeping the chip temp the same. However, you bring up an excellent point. There is going to be some temperature distribution on the foil from the chip out to the pad edge. If it is like most things I would guess gaussian. Adding that heatsink would likely flatten the curve, reducing chip temperature, given the recommended amount of foil (from National). That is probably better. I would take a cooler chip temp over saving some board space any day, and it preserves National's full recommended foil area.

IanAS: I love that skywire picture with the two metal can LME49710s! That works. I'm going to have to skywire a couple of parts myself on a project this weekend.

Last edited by agdr; 2nd April 2011 at 02:23 AM.
  Reply With Quote
Old 2nd April 2011, 02:06 AM   #15
diyAudio Member
 
Join Date: Aug 2009
Location: Sacramento
Hello,
$0.02 worth for grins. For reference I did a perf-board headphone amplifier inside a computer AB switch box. The amplifier was a BUF 634 installed inside the feedback loop of a LM 4562 Op-Amp. This little amplifier built in a couple of hours is the reality standard and for a few dollars in parts beats most in a heads up contest. This design was cobbled from TI example circuits on their page. No ownership here.
I recommend some simple basic amplifier as a reality standard to compare to the product being created here. It is not difficult to build something from a bag of chips.
DT
All just for fun!
  Reply With Quote
Old 2nd April 2011, 02:35 AM   #16
diyAudio Member
 
RocketScientist's Avatar
 
Join Date: Mar 2008
Location: Northwest
Quote:
Originally Posted by DualTriode View Post
Hello,
$0.02 worth for grins. For reference I did a perf-board headphone amplifier inside a computer AB switch box. The amplifier was a BUF 634 installed inside the feedback loop of a LM 4562 Op-Amp. This little amplifier built in a couple of hours is the reality standard and for a few dollars in parts beats most in a heads up contest. This design was cobbled from TI example circuits on their page. No ownership here.
I recommend some simple basic amplifier as a reality standard to compare to the product being created here. It is not difficult to build something from a bag of chips.
If I understand you correctly, that's in effect what I'm trying to do. My intent is a relatively low cost design that will at least measure roughly as good or better than most any headphone amp at any price. I don't want to turn this into a silver soldered, $200 power corded, unobtanium wonder.

Also, you may know this already, but the LM4562 is essentially identical to the LME49720 and LME49860. It tends to be cheaper but it's the same, or nearly the same, chunk of silicon as the other parts. This was discussed a few years ago by Audioman54.
__________________
http://nwavguy.com - Personal non-commercial audio blog
  Reply With Quote
Old 2nd April 2011, 02:56 AM   #17
diyAudio Member
 
RocketScientist's Avatar
 
Join Date: Mar 2008
Location: Northwest
Quote:
Originally Posted by IanAS View Post
Thermal is not an issue is it? 'The Wire' had no heat sinks. If it becomes an issue, then use one, or a bigger one. The solder down tab is about saving money in large scale production.
I don't know yet how hot the parts are going to run. The datasheet calls out 3 - 6 square inches of copper per device. I suspect 6 square inches is pushing the point of diminishing returns (i.e. 100 square inches of PCB foil probably wouldn't keep the part much cooler). I suspect, if they get too hot at all, it would be with < 32 ohm headphones that are seriously inefficient playing highly compressed music played at hearing damage levels.

Someone also might set their amp on top of other gear that runs hot, seal it up inside an airtight enclosure, use it outdoors on their deck in the sun, etc. I don't know what sort of ambient temps, ventilation, etc. National was assuming.

Quote:
And to lower the output impedance.

Maybe at say 10 Ohms output resistance, there is no further benefit. I don't know. With power amps, the lower the better.
Because of the overall feedback, the output impedance of even a single device should be well under 1 ohm as long as it doesn't clip or current limit. Paralleling buffers won't change it much at all. But paralleled devices will give you roughly twice the ultimate current capability. And that could be a valid reason for doing it.

National rates the operating current at 250 mA and the short circuit current at 490 mA both on 15 volt rails. If it can really deliver 250 mA into typical headphone loads, that's plenty for nearly any sane application. But someone with low impedance really inefficient cans who likes it loud and is playing something with really wide dynamic range may hit the current limit on extreme peaks. The data sheet adds:

Thermal dissipation may be the factor that limits the continuous output
current. The maximum output voltage swing magnitude varies with
junction temperature and output current.


The good news is most of the above are easy enough to measure. I can, for example, simulate a worst case load while playing really wide dynamic range music at ear splitting levels and check for current limiting with a single buffer. I can also measure the output impedance of both versions. Etc.
__________________
http://nwavguy.com - Personal non-commercial audio blog
  Reply With Quote
Old 2nd April 2011, 03:13 AM   #18
diyAudio Member
 
RocketScientist's Avatar
 
Join Date: Mar 2008
Location: Northwest
Quote:
Originally Posted by agdr View Post
I finally remembered where I had seen that 10R output resistance figure for the BUF634. From this Burr Brown app note on using an op amp with the BUF634

http://focus.ti.com/general/docs/lit...5&fileType=pdf

Some good stuff in there although they discuss a 50 ohm output video amp. They go through some math with one op amp powering 3 BUF634s in parallel. From page 2, "The output resistance of the BUF634 is about 10 ohms."
Thanks for the link. That 10 ohms is, I think, the effective emitter resistance of the output devices in the BUF634.

Quote:
Yep, definitely do some tests on single buffer vs. paralleled if you have the time. The results would be really interesting. My expectation would be that 10 ohms in the chip in series with the 44 ohm resistive part of my Shure's impedance over much of the frequency range = not so good in terms of power delivery. But the same with a 300 ohm set of cans, probably no significant difference.
The 10 ohms, in effect, disappears due to the feedback. Any drop is compensated for by the feedback loop. So the actual output impedance should be well under 1 ohm as long as the amp doesn't clip or current limit.

Quote:
Good grief, the BUF634 has a +/- 100mV max input offset voltage. I knew the LME49600 had a typical around 16mV, but its max is +/- 60mV. At those offsets they come up with several mA (3mA - 10mA).
I need to study it more closely, but on first glance I think what they're talking about is the paralleled BUF634's "fighting" each other with respect to different offsets. Any DC difference in their output will have to be compensated for by the other(s) as the net current must sum to zero if the output voltage is (essentially) zero.

Quote:
On that DC offset, I agree that 10mV or so is probably not going to affect the headphones much, but why is it there? Seems like that circuit should be able to do better. Probably time for some circuit measurements to find out what isn't working the way it should. Or it is working exactly the way it should and my expectations are off.

One thought is that huge +/-16mV -> +/- 100mV input offset voltage for the LME49600 and BUF634. Maybe that is pressing up against the edge of the design to compensate. Might also be interesting to throw the circuit into SPICE and do some sensitivity analysis by varying each part's value a few percent for effect on the output offset.
You're welcome to run a sim if you want. I'm happy to build it and measure what's going on. Generally servos work consistently well until they hit their limit and then things get suddenly ugly. That's why I suspected RFI, grounding, etc.

For those who might not know, RF picked up by audio circuits can easily be rectified into DC by any of the many transistor junctions. The induced DC can look like a valid signal to the circuit. So unexpected DC offset is a rather common symptom of poor RF immunity.
__________________
http://nwavguy.com - Personal non-commercial audio blog
  Reply With Quote
Old 2nd April 2011, 06:47 AM   #19
diyAudio Member
 
Join Date: Aug 2009
Location: Sacramento
Quote:
Originally Posted by RocketScientist View Post
If I understand you correctly, that's in effect what I'm trying to do. My intent is a relatively low cost design that will at least measure roughly as good or better than most any headphone amp at any price. I don't want to turn this into a silver soldered, $200 power corded, unobtanium wonder.

Also, you may know this already, but the LM4562 is essentially identical to the LME49720 and LME49860. It tends to be cheaper but it's the same, or nearly the same, chunk of silicon as the other parts. This was discussed a few years ago by Audioman54.
Hello,
OK, hats off, I am with you. That exact same TI pdf is what inspired me.
With a pair of 12 volt gel cells +- 12 volts with the BW pin shunted (15 ma class A) heat was not an issue at the buffer. There was no heat to the touch, with the IR thermometer there were only a few degrees delta t. I was ready to attach heat sinks but they were not needed.
With the BUF in the Op-Amp feedback loop the DC offset is limited to the offset of the Op-Amp only.
A thought about the DC offset, perhaps buy 10 or 20 Op-Amps put them in a jig and select the Op-Amps at the low end of the DC offset range. Save the others for use where there is a blocking capacitor being used.
A servo will add alot of parts and complexity.
DT
All just for fun!
  Reply With Quote
Old 2nd April 2011, 09:25 AM   #20
Ferrari is offline Ferrari  Netherlands
diyAudio Member
 
Join Date: Jul 2006
RocketScientist, how about adding a third channel (a copy of the L/R channel, with unity gain) to create a 3-channel headphone amp?
The 3-channel topology keeps headphone return currents from disturbing the signal ground.

Btw, count me in as interested.
  Reply With Quote

Reply


Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
National LME49810TB/LME49830TB UltraHighFidelity HighPower Amplifier Reference Design KLe Chip Amps 75 21st October 2013 12:47 PM
Help with Dayton Reference 2-way Project coldcut Multi-Way 10 23rd May 2010 01:07 PM
How about a 3 inch reference project? type Full Range 52 13th October 2007 04:17 AM


New To Site? Need Help?

All times are GMT. The time now is 02:04 PM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio

Content Relevant URLs by vBSEO 3.3.2