About 30 uV (unweighted). A bit over 20 uV (A-weighted).
I suppose the audibility depends on the shape of the noise spectrum. It may be worth comparing the A-weighted numbers rather than the unweighted numbers. The noise of the MOD86 is basically flat with a very, very slight 1/f component.
You can get down to 10x (20 dB). I don't recall there being a noise advantage of doing so, though, it probably makes an improvement at the system level as the SNR from your source can be higher.
With the MOD86 you can lower the gain further by choosing a THAT1203 or THAT1206 for the front end. The MOD286 does not have this option.
As Andrew points to, a headphone amp might be the ticket. I'm working on such an amp and expect the noise to be really, really hard to measure. Single-digit uV (unweighted) for sure. THD well below that of the MOD86 (which is already world class). Drawback: 500 mA max output current per channel.
Expect more info on this to emerge at the end of this month. I plan to have it ready for the Head-Fi meet in Calgary on the 25th.
Tom
Last edited:
I call that topology "opamp with big shoes". It can certainly be done.
Except the LME49990 is obsolete. It was discontinued by TI last year.
BTW: With these extremely wide-band amps (LME49600: 180 MHz!) you really need to pay attention to the circuit so you don't inadvertently make a power oscillator. My 400 MHz oscilloscope and 2.9 GHz spectrum analyzer were very handy during the debugging of my headphone amp prototype. If the controlling opamp in the opamp + LME49600 circuit has any oddities in the open loop response it is also easy to have a circuit that'll work well at Cload = 0 pF and Cload = 10 nF but will oscillate at 100 pF < Cload < 10 nF. Given that headphone cables add some 30 pF/foot, the consequences of such an amp design are obvious. It is possible to compensate for this. Just be careful and keep your eyes open.
Tom
Awesome. Thank you for sharing your experience.
The noise of the MOD86 is dominated by the THAT1200. The THAT1206 is 1 dB quieter so if you switch to the THAT1206, the output noise should be ~1 dB lower as well (back of envelope math, usual disclaimers apply).
Tom
500 mA, peak. Not 500 mA RMS. So 500 mW into 4 Ω, 1 W into 8 Ω. More relevantly: 3 W into 20 Ω (current limited), 3 W into 32 Ω (voltage limited). Just shy of 400 mW into 300 Ω (voltage limited).
Plenty for a high-efficiency driver and more than enough to launch your headphones into space. Yet, some will want 5-10 W to drive their headphones.
Tom
Plenty for a high-efficiency driver and more than enough to launch your headphones into space. Yet, some will want 5-10 W to drive their headphones.
Tom
I'm not sure how sensitive the JBL 2435HPL is, but my B&C DE25-8 is a touch under 110dB/W/M. One big difference in my system (WRT noise) is the passive crossover between the amp and the compression driver. This reduces mid band noise by ~10dB. Have you thought of implementing your CD compensation passively? This will improve your noise floor by around 10dB where your ears are most sensitive. The rest of the crossover function could be done actively.
For the MOD286 is changing the sensitivity to 4V (Balanced) just a matter of changing the amp's gain, or is there an input stage that dictates the high sensitivity?
The output stage has 20 dB of gain. While I have found a couple of ways to reduce the gain in the circuit simulator, none of the ways translated into good performance across the various oddities and corner cases on the LM3886. I can put an attenuator in front, but that defeats the purpose of the whole gain structure optimization.
Most people actually request higher gain (which is possible by swapping out a resistor).
Tom
Most people actually request higher gain (which is possible by swapping out a resistor).
Tom
I would like to avoid it if possible.
I still use a protection cap though.
In my JBL M2 clones I use the stock passive network that applies a ~9dB lpad and 1kHz 1st order filter to complement the active setting.
In this configuration, and with a 32 ohm compression driver, I cannot hear any noise coming from the nc400.
But still, that is a big waste of power burned into resistors: I'd rather use a dead quiet 10W amp than a "noisy" (all things relative...) 100W amp with a 10dB lpad.
A pair of 110dB/W @ 1m speakers will need ~ 4W per channel to achieve 110dB at a 2.5m listening position.
Can the other speakers achieve that 110dB at the listening seat?
4W into 8ohms is equivalent to 8Vpk and 1Apk
Can the other speakers achieve that 110dB at the listening seat?
4W into 8ohms is equivalent to 8Vpk and 1Apk
Tom,
Will the MOD286 have differential inputs as well? You noted the different THAT input module options for MOD86 but that it wouldn't be an option for MOD286. Don't know if that means no differential inputs or just a different differential (would that be Diff-squared?) for MOD286.
Will the MOD286 have differential inputs as well? You noted the different THAT input module options for MOD86 but that it wouldn't be an option for MOD286. Don't know if that means no differential inputs or just a different differential (would that be Diff-squared?) for MOD286.
That is basically what the LPUHP would give me, but a tiny bit more power would not hurt, especially if I want something versatile that can also drive a high load and/or a less sensitive driver at times.
Not that I really need it, but yes they can 🙂
I currently have nc400's and LPUHP, so I am pretty much covered, but I would love to have a more versatile amplifier that I can use in various scenario without searching for the perfect match for a given load and sensitivity every time.
Something as "silent" as the LPUHP, while being almost as powerful as the nc400.
Something like a THX AAA maybe ?
Quick question... why bother with a dc servo when the LME49710 has excellent dc offset specs?
Drift over months/years. From board flux or other residue. Certainly possible for the chips input stage to drift as well.
Last edited:
So doesn't the servo op-amp suffer from drift/flux etc?
(Must admit I remove the flux, and never noticed much drift even with discrete ciruits even over many years, but hey ho).
(Must admit I remove the flux, and never noticed much drift even with discrete ciruits even over many years, but hey ho).
If that's the case it's very risky!
More likely it's to cope with the +/-10mV max o/p offset of the the THATS1200 (should have looked at the DS before I asked my question I suppose).
Last edited: