I'm convinced that if we perform the right rituals with the correct offerings, we will be able to summon Nwavguy to DIYAudio 😀
The ODA has twice the output capacity of O2, so if O2 is able to pump out 350mw, then the ODA can do 700mw?
The ODA has 3x the current output capability of the O2 and can run on +/-12.5V rails for any load, or +/-15Vdc for loads 300R and up vs. the +/-12.0V rails in the O2. Here is how that translates into output power at the various typical headphone loads:
ODA Headphone Amp
ODA Headphone Amp
Those two posts are in the forum where the fellow who is doing the dScope tests hangs out. Someone over there had wanted to know output power.
The O2's maximum power at the various loads are in the graph "THD+N output & max power on AC" in NwAvGuy's blog here:
NwAvGuy: O2 Headphone Amp
So comparing the two, here are the results with the ODA not using the attenuation resistors and with +/-12.5V power rails in the ODA:
15R load: O2 = 337mW, 16R load ODA = 2250mW (=2.25W). ODA has 6.7x the O2 max power.
33R load: O2 = 613mW, 32R load ODA = 1640mW (=1.64W). ODA has 2.7x the O2 max power.
150R load O2 = 355mW, 120R load ODA = 438mW. ODA has 1.2x the O2 max power.
And then going up to +/15V rails for the ODA in that second A.C. post gives:
600R load: O2 = 88mW, ODA = 158mW. ODA has 1.8x the O2 max power.
Last edited:
Cool, cool, thanks. I'll look at it more closely when I'm awake.
Another question: So we know what balanced cables do. What about balanced amp/dacs? What's the point of that? Obviously, one function is to accept balanced cables, but anything else?
Thanks
Another question: So we know what balanced cables do. What about balanced amp/dacs? What's the point of that? Obviously, one function is to accept balanced cables, but anything else?
Thanks
agdr - I had a thought, if there was a stepped attenuator instead of a pot, could that have higher specs (apart from channel balance) than the current pot solution, depending on which resistors were used?
@Darkwizzie - The only benefit to balanced I can see would be if you were trying to drive a really low sensitivity headphone, or had very long cables. Or, maybe, if the noise vs voltage graph of your amp was exponential.
@Darkwizzie - The only benefit to balanced I can see would be if you were trying to drive a really low sensitivity headphone, or had very long cables. Or, maybe, if the noise vs voltage graph of your amp was exponential.
NwAvGuy said it best here
Scroll down to the "What they're not" section, then the second item down on "no balanced outputs".
So there you go, balanced being officially branded as "snake oil" for home use by NwAvGuy. 🙂
There is one clever thing that can be done with headphones wired for balanced outputs with separate ground wires going to each driver. They can be driven with a differential set of outputs, essentially doubling the output swing and 4x the power. Here is a good way to think of it. Imagine you have a standard headphone amp with one left channel and one right channel. But take that and run the "hot" of the right channel out to one lead of the left headphone driver, then run the "hot" lead of the left channel out to the other lead of the left headphone driver. You don't use the ground connections on the amplifier outputs at all. Then feed the input of this amp with a balanced signal (which is differential by nature).
What happens is that when one channel of the amp is heading up to the positive rail with a signal the other channel is heading down to the negative rail. As far as the left headphone cup is concerned it is seeing 2x the voltage swing. The output transistors or chips in the amp also have to dissipate double the power. Two headphone amps would also have to be used since both channels in each amp work to power just one side of the headphone.
I wouldn't build an amp like this though because someone could accidentally plug in a "standard" headphone that has the grounds of the two drivers connected. That would short out one channel of the amp. 🙂
Last edited:
It definitely would. A stepped attenuator would have much closer channel-to-channel tracking. It would also have less crosstalk between the channels and could potentially have less current noise, if low current noise resistors were used.
Sounds like a win - why not implement that into ODA? 😀
Oh yeah, about the power, I've finally learned how to use the power rating of an amp and calculate if an amp is capable of powering a given headphone. Yipeeee!
Finished building #1 of the 2 ODAs I have today !
Agdr's build instructions are very good. A great job mate.. only thing I did wrong was put a LM337 where I wanted a LM317 🙂 All sorted.
I'm using a 240V 18V AC toroid to drive my amp. I'm getting regulated +12.472V and -12.437V rails, from the DC supply of 26.65V and -26.49V. I measured and adjusted my DC offsets to about 0.8uV.. pretty neat that it's that low.
Once I get my Keithley 2015 running properly I'll throw it at some of the THD tests 🙂
Sounds pretty good on my Q701s, and HD600s. Plenty loud and sounds good enough.
The LM317/337 get a bit warm.. will get bigger heatsinks 🙂
Awesome build and result. Thank you Martin !!
Agdr's build instructions are very good. A great job mate.. only thing I did wrong was put a LM337 where I wanted a LM317 🙂 All sorted.
I'm using a 240V 18V AC toroid to drive my amp. I'm getting regulated +12.472V and -12.437V rails, from the DC supply of 26.65V and -26.49V. I measured and adjusted my DC offsets to about 0.8uV.. pretty neat that it's that low.
Once I get my Keithley 2015 running properly I'll throw it at some of the THD tests 🙂
Sounds pretty good on my Q701s, and HD600s. Plenty loud and sounds good enough.
The LM317/337 get a bit warm.. will get bigger heatsinks 🙂
Awesome build and result. Thank you Martin !!
Hey congratulations on your build! 🙂
Your power rail voltage are right on the money. I purposely designed them to be just slightly below the 12.50Vdc. I figured that since there will always be a small amount of variation with the regulator's internal reference that I would rather have the result be slightly below rather than slightly above the target.
That heat sink issue didn't hit me until I read your post, but sure enough I always used the temporary heat sinks with a 16Vac transformer. The extra 2Vac for the 18Vac could produce a fair amount of extra heat, depending upon current. Maybe you could just bolt them onto a single large plate to take the place of the back panel, with the insulating washers installed of course.
I plugged the 97dB/mW and 300R for the HD600s into a spreadsheet and with the 7.25Vrms swing (with the +/-12.5Vdc rails and the attenuation resistors shorted) you get a whopping 119.5dB SPL!. 😀 Even with the attenuation resistors you would still get 113dB SPL.
With the Q701's at 105dB/V and 62 ohms you would hit 120dB SPL at 5.62Vrms, before the full 7.25Vrms on the volume control. With the attenuation resistors you would get around 116dB SPL.
So you may want to give it a try with the attenuation resistors un-jumpered, if you haven't already. Looks like that would still give plenty of volume with both headphones, but would add the 1/2x "gain" (attenuation) in the lowest gain position to cut the level of any "hot" sources (like a couple of mine) in half. 🙂
The big problem there is size. 🙂 The stepped attenuators tend to be huge. But it is certainly possible to just panel mount one, using a larger case, and run the wires down to the volume pot holes in the PC board.
Oh yeah, about the power, I've finally learned how to use the power rating of an amp and calculate if an amp is capable of powering a given headphone. Yipeeee!
I have a spreadsheets posted for that too, here: https://drive.google.com/folderview?id=0B67cJELZW-i8SXZrcktvTGl2Y28&usp=sharing
The spreadsheet actually came from a post over on Head-Fi. I have the credits for the post in those files with the spreadsheets. Just enter the headphone impedance and either dB/mW or dB/V sensitivity (delete anything in whichever field isn't used) and your target SPL dB level. It will calculate rms voltage and current needed from the amplifier. Remember that for the ODA the total current available is around 420mA since it is divided up over 6 op amps. The number in the spreadsheet is the total for all 6. The maximum voltage swing for the +/-12.5Vdc rail ODA version is 7.25Vrms, while the swing for the +/-15Vdc version is around 9.5Vdc.
Last edited:
Yep, going to get a large heatsink.. now to build the hardwood headphone stand and integrate the amp into it 🙂
Interesting comparison.. Both headphones plugged in at the same time, although the Q701s are theoretically louder by sensitivity alone, the HD600s seem louder.
I also threw a square wave into the amp and got a 500ns rise time, only 1% overshoot 😎
Will also look carefully at the RMS output per channel tonight to see how equal the gain is per stage.
Pixola...
Congrats on your ODA build!!
I have been enjoying mine for a few months now and its been a great amp for me and have really enjoyed it. My O2's just don't get used much anymore!
Good luck on number #2.
Alex
Congrats on your ODA build!!
I have been enjoying mine for a few months now and its been a great amp for me and have really enjoyed it. My O2's just don't get used much anymore!
Good luck on number #2.
Alex
Good stuff! 😀
The gain circuit should be very accurate, with those 0.1% resistors, but the pot won't be. The Alps and Bourns 9mm pots are only good for 20%. One suggestion there, since you are making a case anyway, is to replace the pot with a stepped attenuator wired into the PCB pot holes, like Moragg's suggestion a few posts back. That would up the channel-channel tracking accuracy and should decrease crosstalk.
Hey please post a picture of your finished case/stand! That is going to look sharp. 🙂
Using a sound card based Spectrum Analyzer, I'm noticing a couple of artefacts. Whilst I know this doesn't give me dScope type results, it can still show up more obvious issues and help me understand the design.
If I set the gain to position 1 (left most) on ODA 2.1, and input a 1kHz sine wave, I see a fairly clean signal at the headphone output (0.005% THD, SNR 75dB).
If I set the gain to any other position, I see fairly significant 2nd 3rd and 4th order harmonics and understandably the THD rises by one order of magnitude to 0.05%.
I'm going to flip the board tonight and trace through the signal and see where the harmonics are being introduced.
Untitled-1 by irovingi, on Flickr
Untitled-2 by irovingi, on Flickr
If I set the gain to position 1 (left most) on ODA 2.1, and input a 1kHz sine wave, I see a fairly clean signal at the headphone output (0.005% THD, SNR 75dB).
If I set the gain to any other position, I see fairly significant 2nd 3rd and 4th order harmonics and understandably the THD rises by one order of magnitude to 0.05%.
I'm going to flip the board tonight and trace through the signal and see where the harmonics are being introduced.
Untitled-1 by irovingi, on FlickrUntitled-2 by irovingi, on Flickr
Last edited:
Good testing work! The O2 has the same thing, but the 2nd harmonic is down at -95dBr and the 3rd at -105dBr. Take a look at NwAvGuy's dScope plot here - scroll down to the heading "THD+N 15 ohms specturm":
NwAvGuy: O2 Headphone Amp
Also note his test conditions at the bottom of the graph - volume knob 100%, 400mVrms input (for dBr), and bandwidth 22KHz. The ODA dScope test over on Audio Circle showed the same 2nd and 3rd harmonics, but at about the same low levels as the O2. The soundcard tester probably has a higher noise floor. My QA400 tester, which is just a soundcard in a box, does the same.
I don't remember which gains he tested it at though. I'll have to go back and look at the video. At any rate, you are definitely right that even if the absolute levels are off a bit from a dScope due to noise floor, the relative results are just fine. In other words there is definitely a 2nd and 3rd harmonic at some level.
On your signal-tracing testing you may want to compare the output of the gain stage at JP14 and JP15 with the amp output to compare harmonic generation in the gain stage vs. the output. I've been planning on asking the dScope fellow to do the same with his next round of testing. With the harmonics going up with gain it could very well be the gain stage. Given the star ground segments I would suggest grounding the distortion tester off the metal shell of the gain switch, which is on the input segment ground star, for the JP14 and JP15 tests.
I've actually been thinking about substituting a 5K pot for the 1K at having that dScoped. The Johnson noise would be higher with the 5K, but it would be 1/2 the load on the gain chips which could reduce THD. So it is possible the net result with THD+N could actually go down. One of those things a person would never know for sure without a dScope test.
Last edited:
AGDR...
Very interesting... haven't heard anything from Mike yet?
The 5K pot would be a very interesting experiment!
You need to build some ODA's , sell them and buy a d'Scope!!
Smiling..
Alex
Very interesting... haven't heard anything from Mike yet?
The 5K pot would be a very interesting experiment!
You need to build some ODA's , sell them and buy a d'Scope!!
Smiling..
Alex
Not yet! 🙂 I'll be quite eager to see that next round of dScope results. Just recap from the above posts, Mike sent the ODA back to me at my request and I changed a few things to put it back in a "standard" configuration. The one I sent him had 0.5R balancing resistors in one channel and 2R in the other, just to get a dScope reading on which led to the lowest THD. It was the 0.5R as expected. I also had the pre-amp wired hooked up and wired into the input circuit past the input select switch. For the base set of dScope readings I wanted that disconnected. So nothing should be worse with the new set of dScope readings and few things may be slightly better.
In the LME49990 datasheet graph #7 shows the THD+N for +/-15Vdc rails (the closest that have to +/-12.5), a 600 ohm load and 8Vrms swing at 0.000005%. They don't state the closed loop voltage gain, but back up in the tables the THD number is given with an Av=1 condition, so safe to assume the graph is done at unity gain.
Mike's ODA dScope results are back in post 524. Into a 15 ohm load at the O2's 293mW maximum power level (to get apples to apples vs. the NwAvGuy O2 plots) here measured THD+N at 0.00207% for one channel and 0.00203% for the other. The ODA power can go way up from there of course. At 1.17W, 4.4 times the O2's maximum level, Mike measured 0.00201% THD+N. I seem to recall he did those measurements on the second gain setting step, but I'll have to review the video to be sure.
But yes, I can definitely see the benefit in having a dScope or AP at hand while designing and testing. That allowed NwAvGuy to try many different circuit ideas for the lowest distortion and noise levels.
Last edited:
Hmm.. Isolated the source of the 2nd harmonic noise to IC2..
Input
Output
Also AGDR, reading the notes about the comparator inputs, my LEDs are reading:
LED1 2.621V, LED2 2.599V.
Wondering if by chance the LEDs I've used aren't OK..
Input
An externally hosted image should be here but it was not working when we last tested it.
Debug_IC2-3 by irovingi, on Flickr
Output
An externally hosted image should be here but it was not working when we last tested it.
Debug_IC2-6 by irovingi, on Flickr
Also AGDR, reading the notes about the comparator inputs, my LEDs are reading:
LED1 2.621V, LED2 2.599V.
Wondering if by chance the LEDs I've used aren't OK..
I've been meaning to ask you about that. At the time you did your SMD part order from Mouser I think they were out of stock of the 40mcd 130 degree lens SMD LEDs, the Mouser 638-152BHC-AN1P22T. You probably wound up getting the alternate BOM choice, the 100mcd 60 degree lens, #638-1121BHCAP2R12T . There is a small forward voltage difference between the two, even though they are both blue. The 130 degree are the ones I used in the build instructions. Mouser has the 130 degree back in stock and I have a bunch now. Please let me know if you wound up with the 60 degree parts - if so I'll put some 130 degree in the mail to you. 🙂
Interesting about the harmonic spectrum! Are you getting the same with IC1's output, or is it just IC2 and that one channel? I noticed a pertinent post from Richard Marsh in his headamp there here (post #822):
http://www.diyaudio.com/forums/headphone-systems/211394-marsh-headphone-amp-linear-audio-17.html#post4071186
that without a notch filter anything below -100dB on the soundcard testers is useless and buried in the noise (for absolute value), based on comparison with his Audio Precision tester. I've read similar from opc before, with his AP. So it all comes down to the level of those harmonics. I haven't gone back an reviewed the dScope results yet but as I recall with the 15R load the 2nd harmonic was slightly more than the O2, like -92dB vs. -90dB, but the other harmonics were less than the O2. But keep in mind his testing so far was all with the 15R load and 2V, 4V and 6Vrms of swing, pretty much the absolute worst case for generating THD.
But good observation, the harmonics do seem to be coming from the gain stage. That is most likely due to the non-inverting op-amp configuration. The alternative with inverting stages would probably look like this post I was helping a fellow with:
http://www.diyaudio.com/forums/headphone-systems/262222-opa2134-input-stage.html
a inverting buffer feeding an inverting attenuator. The noise would be higher, with those 10K resistors in series with the input and feedback, but the THD should be less. I have to say that inverting attenuator is interesting. A combined attenuation/gain stage. I have that one on my "things to look into" list! 🙂 At some point I way whip a small circuit board with just those parts and see if I can get Mike at A.C. to dScope test it. See what the tradeoff between increased noise and (assumed) decreased THD looks like.