Another problem with pots below 10K that I ran into with the ODA's 1K pot in the middle: the 9mm audio taper pots are only rated at 25mW, which works out to be 5Vrms maximum continuous output from the VAS, or with "music power" assuming 2x peaks, say 10Vrms. I solved that by using a linear 1K pot, the linear 9mm are rated at 50mW (guitar amps often use linear from what I can tell), and/or allowing the use of attenuation resistors if using the audio taper. But still yet another restriction on using pots below 10K. 5K 9mm are available but less common to find in stock and only an incremental Johnson noise improvement over 10K. I had used 1K for lowest Johnson noise and to take advantage of the LME49990's ability (gain stage) to drive 600R, pot in parallel with feedback resistors.
BTW- hearing that tomchr's HP-1 uses a pot in the middle I've officially "retired" my ODA design out on my website (in another thread here) and have put in a link and suggestion to look at his HP-1 instead.
tomchr's HP-1 sounds like a more modern design and has the objective testing with a "real" tester that was the whole point of NwAvGuy's O2 and his mentioned but never-released ODA (O2 Desktop Amplifer). The AC-AC wall transformer my version of the ODA (and NwAvGuy's O2!) is designed for is nearly un-obtainable now, as is the discontinued LME49990 chip. I put a clipping detector in the ODA to solve that VAS stage potential for clipping-without-notification in the O2, but Tom's solution is even better, use higher voltage power rails. I couldn't do that with the ODA given the 3 parallel SIP NJM4556A chips on each channel. They have high idle current resulting in large idle package dissipation at even +/-15Vdc rails with no way to heat sink, hence the standard +/-12.5Vdc rails. tomchr: excellent work on your HP-1 design, IMHO!
BTW- hearing that tomchr's HP-1 uses a pot in the middle I've officially "retired" my ODA design out on my website (in another thread here) and have put in a link and suggestion to look at his HP-1 instead.
tomchr's HP-1 sounds like a more modern design and has the objective testing with a "real" tester that was the whole point of NwAvGuy's O2 and his mentioned but never-released ODA (O2 Desktop Amplifer). The AC-AC wall transformer my version of the ODA (and NwAvGuy's O2!) is designed for is nearly un-obtainable now, as is the discontinued LME49990 chip. I put a clipping detector in the ODA to solve that VAS stage potential for clipping-without-notification in the O2, but Tom's solution is even better, use higher voltage power rails. I couldn't do that with the ODA given the 3 parallel SIP NJM4556A chips on each channel. They have high idle current resulting in large idle package dissipation at even +/-15Vdc rails with no way to heat sink, hence the standard +/-12.5Vdc rails. tomchr: excellent work on your HP-1 design, IMHO!
Last edited:
Would be really interesting to repeat the measurements with agdr's booster board in place of the 4556s!
Yes it would!
I've been thinking about that. Good timing too, I just received a bunch of boards back from fabrication, a new V3.6 with a few minor changes. I want to pay for the testing though, as per Tom's website. Tom - I'll send you a PM tomorrow to get a quote.
Last edited:
He did a lot of thd+n measurements at 400mv, 15 ohms. He had thd+n x V graphs so people could see what the amp was capable of. In comparison, one of the more popular alternatives states thd without noise with no load.
OK then, given rail specs and 2x gain of the output stage you were talking about 12VRMS at VAS output, and not at amp input.
Sure, but as I wrote multiple times, the effects of a misconfigured gain switch with a standard pot-before-vas design, are pretty obvious to the normal user.
If you are at 1600 position with your volume and still low, time for higher gain.
If you are at 0800 position and its too loud (or too noisy), go to lower gain.
With a pot-after-vas design, an higher than necessary gain setting might lead to clipping even at low SPL listening sessions.
Which is a bit of a counter intuitive user experience.
Sure, though, honestly, the vast majority of an HP amp users these says, are going to feed it with a DAC output (whose output Z is usually around the 100 Ohms).
I would usually tend to design based on what a typical human ear can hear.
As I am writing this reply, I am listening to music through a FiiO E12, and a 96dB/mW efficient HP (closed back).
The E12 is pretty much a clone of The Wire HP amp (which is nothing more than the initial example schematics which appeared in the TI tech sheets for the LMEs long time ago - opamp vas + buffer). So, standard design (no rocket science here) with pot before VAS. Also pretty cheap.
At high gain, and my typical listening SPL (guess around 80..85 dB), the amp is virtually silent to my ears.
So, would I get extra 10dB of SNR with an alternate design?
Sure, if it came with no drawbacks.
Similarly, I’d get a reduction from 0.001% THD to 0.0001% THD, if it came for free.
Human ear can't tell the difference, but sure.
So, at the end, is all about weighing Pros and Cons of given design choices.
If it was absolute, given that pot-after-vas is sure neither new nor rocket science, we’d see a lot of HP amps using it … which is not the case.
Last edited:
Well, if you'd rather have 12 µV of noise on the output, which is audible in IEMs, go right ahead. Put the VAS after the volume pot.
I do this for a living and aimed squarely at the high-end market with the HP-1. I can't sell an amp that has 12 µV of noise on the output. I trust that my customers are savvy enough to be able to use a gain switch, or to educate themselves by reading the manual. As you say, many use a DAC, in which case you can set-and-forget the gain switch, thus, rendering this discussion rather moot.
Tom
I do this for a living and aimed squarely at the high-end market with the HP-1. I can't sell an amp that has 12 µV of noise on the output. I trust that my customers are savvy enough to be able to use a gain switch, or to educate themselves by reading the manual. As you say, many use a DAC, in which case you can set-and-forget the gain switch, thus, rendering this discussion rather moot.
Tom
Thanks. I appreciate it. Just toss me an email (through my website) when you're ready.
Tom
Tom, would you mind performing one additional measurement that yours truly would have liked to see for ages?
I would really like to see THD(f) / THD(level) for the gain stage alone. So volume set for maybe 0.5-1 Vrms of output at the point of input clipping, mains operation preferred, both gain settings. (Bonus points for other opamps or feedback network resistor values just in case you happen to have too much time on your hands.)
Anyway, it looks like the good ol' NJM4556 would have a fair bit of common-mode distortion, not unexpected for such an oldschool part (first used in 1984!) with a buffered VAS (in sim I always had to give the emitter follower part of the VAS a current source to keep that from happening). You can also see that the 3rd harmonic becomes more prominent at high frequencies into the 32 ohm load, no doubt crossover. The slight increase in the bass region points towards thermal feedback towards the input stage, a problem addressed in modern designs but presumably not yet back then. That's one area where a MUSES02 with its copper lead frame might do better than its plain ol' NJM4580 counterpart.
On the whole, measured performance may not be state of the art but looks pretty darn decent anyway. IMHO this would be hard to fault audibly (except maybe with an HE-6) unless you happened to clip the input. Lack of user-friendliness in regard to the latter is the only other noteworthy minus of this design IMHO.
Oh, and worst-case source impedance for a 10k pot after a low-impedance source is 2k5, not 5k. (It's 5k || 5k.)
I would really like to see THD(f) / THD(level) for the gain stage alone. So volume set for maybe 0.5-1 Vrms of output at the point of input clipping, mains operation preferred, both gain settings. (Bonus points for other opamps or feedback network resistor values just in case you happen to have too much time on your hands.)
Anyway, it looks like the good ol' NJM4556 would have a fair bit of common-mode distortion, not unexpected for such an oldschool part (first used in 1984!) with a buffered VAS (in sim I always had to give the emitter follower part of the VAS a current source to keep that from happening). You can also see that the 3rd harmonic becomes more prominent at high frequencies into the 32 ohm load, no doubt crossover. The slight increase in the bass region points towards thermal feedback towards the input stage, a problem addressed in modern designs but presumably not yet back then. That's one area where a MUSES02 with its copper lead frame might do better than its plain ol' NJM4580 counterpart.
On the whole, measured performance may not be state of the art but looks pretty darn decent anyway. IMHO this would be hard to fault audibly (except maybe with an HE-6) unless you happened to clip the input. Lack of user-friendliness in regard to the latter is the only other noteworthy minus of this design IMHO.
Oh, and worst-case source impedance for a 10k pot after a low-impedance source is 2k5, not 5k. (It's 5k || 5k.)
Last edited:
You're right. My bad. I thought those noise numbers looked a bit high. My point of putting the gain first for better noise performance still stands, though.
Tom
The thing is, in real life you are not going to have pot in the mid position with iem or any other sensitive headphones plugged in.
Also, if you take a typical source with nominal 2Vrms output and S/N, lets say, 120dB, you'll have around 14nV/sqrtHz of noise on the input of the amp (assuming that 1/f noise impact is insignificant). Take the worst case mid position, where 10k pot gives ~6.5nV/sqrtHz, and since it is 1:2 for the signal noise - 7nV/sqrtHz is left from the source. So the worst case scenario reduces S/N to 117dB - it is barely a loss. And if you take opamp noise in consideration the difference will be even lower.
For a lower lever source situation will be a bit worse, but it will not be an easy task to find one with 120dB S/N.
Also, if you take a typical source with nominal 2Vrms output and S/N, lets say, 120dB, you'll have around 14nV/sqrtHz of noise on the input of the amp (assuming that 1/f noise impact is insignificant). Take the worst case mid position, where 10k pot gives ~6.5nV/sqrtHz, and since it is 1:2 for the signal noise - 7nV/sqrtHz is left from the source. So the worst case scenario reduces S/N to 117dB - it is barely a loss. And if you take opamp noise in consideration the difference will be even lower.
For a lower lever source situation will be a bit worse, but it will not be an easy task to find one with 120dB S/N.
Last edited:
O2 AP testing
I just remembered something. Tyll Hertsens bought an AP in 2012 and NwAvGuy helped him set up a suite of headphone amplifier tests for it:
InnerFidelity Headphone Amp Measurement Program Update July 2012 | InnerFidelity
NwAvGuy's Heaphone Amp Measurement Recommendations | InnerFidelity
I've been plodding through all the dScope tests NwAvGuy did on the O2 in his blog post here
NwAvGuy: O2 Headphone Amp
trying to understand the various test conditions for each. At the bottom, below the last graph but before comments he has a paragraph titled "verification of measurements":
"I’m hoping to have the O2’s more critical measurements verified by someone independent with an Audio Precision, dScope, or similar equipment. If there are any measurements anyone is especially suspicious of, please let me know so I can make sure they are verified."
I laughed at that a bit given that it only has taken 5 1/2 years here. But I had forgotten about the stuff Tyll did. I'm not sure if he ever did publish a full set of verification tests for the O2. By now his AP is probably the "old" model too.
I just remembered something. Tyll Hertsens bought an AP in 2012 and NwAvGuy helped him set up a suite of headphone amplifier tests for it:
InnerFidelity Headphone Amp Measurement Program Update July 2012 | InnerFidelity
NwAvGuy's Heaphone Amp Measurement Recommendations | InnerFidelity
I've been plodding through all the dScope tests NwAvGuy did on the O2 in his blog post here
NwAvGuy: O2 Headphone Amp
trying to understand the various test conditions for each. At the bottom, below the last graph but before comments he has a paragraph titled "verification of measurements":
"I’m hoping to have the O2’s more critical measurements verified by someone independent with an Audio Precision, dScope, or similar equipment. If there are any measurements anyone is especially suspicious of, please let me know so I can make sure they are verified."
I laughed at that a bit given that it only has taken 5 1/2 years here.
But I had forgotten about the stuff Tyll did. I'm not sure if he ever did publish a full set of verification tests for the O2. By now his AP is probably the "old" model too.
Last edited:
Nobody stopped you from buying your very own AP during those 5.5 years...
Yeah. It's the SYS-2700 they're using most likely. Could be an older SYS-2xxx. The SYS-2700 is very good. To get meaningfully better you need an APx555. They start at $32k for the base model. The main difference between the 2700 and the newer APx series is the software. There were hardware changes too, but the software is really where it's at. The software for the SYS-2700 looked like it was written by an intern. It was the archetypical quirky engineering software. It was awful. The APx software is a stark contrast to this. It is **AWESOME** AP has won awards for it and I can see why. It is a joy to use.
Tom
Yeah. It's the SYS-2700 they're using most likely. Could be an older SYS-2xxx. The SYS-2700 is very good. To get meaningfully better you need an APx555. They start at $32k for the base model. The main difference between the 2700 and the newer APx series is the software. There were hardware changes too, but the software is really where it's at. The software for the SYS-2700 looked like it was written by an intern. It was the archetypical quirky engineering software. It was awful. The APx software is a stark contrast to this. It is **AWESOME** AP has won awards for it and I can see why. It is a joy to use.
Tom
Interesting! I have read some posts to that effect over the years about the sys-2700 software. Apparently pretty user-unfriendly, although functional.
Wow, $32K! That is about twice what I thought they cost. Goes to show that folks are getting their money's worth in an amp (headphone or speaker) that has been properly measured with an AP. I've toyed around with the idea of getting one over the years (lol, still thinking they were $16k or so), but decided that was just more than I wanted to sink into the hobby. I would say it is vital tool though for a serious audio business! I'm glad you are offering the testing consulting service.
Audio Analyzer - SR1
http://www.used-line.com/list-analyzers/miscellaneous/stanford-research-systems-sr1
However, you can do a lot with an audio card, if you really want to.
http://www.used-line.com/list-analyzers/miscellaneous/stanford-research-systems-sr1
However, you can do a lot with an audio card, if you really want to.
Last edited:
Yep. Cheap test equipment is an oxymoron. The APx525 is more "affordable" but still five digits.
Also, don't forget that the calibration of one of those will set you back a pretty penny too. Last I had a SYS-2700 in for calibration (4-5 years ago I think), the calibration alone was $1500. Yep. Not cheap.
That is certainly true. Many sound cards perform quite well at their sweet spot.
Tom
It is not exactly what I meant. To get a good distortion/IMD measurement, you need to suppres main tone(s) anyway, to see their products well (that what some audio analyzers do). Once you've done that, ADC distortions become less of a problem.
Sergey888 - I do remember that you use a SR1! Those look like nice analyzers.
Good point about the notch filter. I'm still working on one. Over in the equipment forum I've now had fabricated a fliege notch filter, Akerberg Mossberg fully differential, and recently a Hall. The Fliege oscillates, the Hall has high THD that I haven't had a chance to track down yet (at least the active hall, on my list to wire it up as passive and try that), and haven't even had a chance to mess with the A-M notch other than powering it up. I know that the "gold standard" every one uses is Twin-T, but I'm avoiding that given the 3 simultaneous adjustments. I keep searching for a single-adjustment notch.
tomchr - I agree about calibration! Yeah that is the thing everyone forgets. High precision test equipment is not much better than a paperweight if not calibrated on a regular basis. I'm going to be sending in 3 DMMs myself to Tek and Fluke in the next month or two for their 2 year cals.
I think I figured out where I got that $16K figure. I think I had read somewhere that was a the price of a new dScope III.
Good point about the notch filter. I'm still working on one.
Over in the equipment forum I've now had fabricated a fliege notch filter, Akerberg Mossberg fully differential, and recently a Hall. The Fliege oscillates, the Hall has high THD that I haven't had a chance to track down yet (at least the active hall, on my list to wire it up as passive and try that), and haven't even had a chance to mess with the A-M notch other than powering it up. I know that the "gold standard" every one uses is Twin-T, but I'm avoiding that given the 3 simultaneous adjustments. I keep searching for a single-adjustment notch.tomchr - I agree about calibration! Yeah that is the thing everyone forgets. High precision test equipment is not much better than a paperweight if not calibrated on a regular basis. I'm going to be sending in 3 DMMs myself to Tek and Fluke in the next month or two for their 2 year cals.
I think I figured out where I got that $16K figure. I think I had read somewhere that was a the price of a new dScope III.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.