National LME49600 Reference Design Project

[RocketScientist]

Just an update. I expressed a bit of skepticism in another thread over the popular belief the metal TO99 LME49710 sounds better than the SOIC plastic version. It turns out there may be at least a tiny bit of fact behind that--at least in terms of measurable performance. But only when it's driving a relatively low impedance load hard enough to cause significant extra power dissipation. Because that series has relatively high bias, we're talking a serious load.

Long story short, there is a bit more thermal distortion in the plastic package when the op amp is working hard into a tough load. Mind you, we're still talking about vanishingly low levels of distortion even in the plastic package, but I have it on good authority you can measure the difference with the right equipment. The metal package keeps the die at a more constant temp under stressful dynamic conditions.

[IanAS]

All I can say to that is that on a system so revealing that I and friends can hear the differences between, eg, different types of speakers protection relays, (they're all too bad so we so don't use them) we could very easily hear how much better those metal canned 49710HAs were compared to the metal caned dual 49720HA and how much better they were than the plastic dual DIL 49710.

We ran them in for weeks, comparing at weekly intervals.

Driving what I presume to be a low current through a 200 ohm chip resistor of the DC coupled sound card out put stage, through a meter of wire, into a 39k load and pair of transistor bases at a DC coupled power amplifier.

We didn't need to run the new 49710HAs in actually, they were good straight away in a variety of ways.

Another pal has a less revealing system and his DAC had the DIL 49710 as the DAC chip follower, followed by an AD797. I simply swapped the DIL 49710 for the TO99 49710HA and he was immediately pleased. Even his wife commented later that day the system sounded better, and she's not 'into Hi-Fi' and was not aware the chips had been swapped.

I presume there is no thermal stress going on but I guess there could still be thermal gradients. I wondered if it was about RFI screening or microphonics screening, or the leads; the TO99 appear to have copper round leads which are bendy rather than the rather brittle and flat chip pins. Maybe the dual sharing the same PSU pins is why the singles sound better, even though with the sound card they are sharing the same socket. In the sound card, there is just the one op-amp after the DAC chip, with capacitors and resistors around it so that it's filtering. Maybe the effect would be different in a simple unity buffer. I don't know. To all listeners they sounded better by far though.

If a person hasn't heard it for them self then they're not in a position to say they don't sound better merely because said person can't understand how they can sound better.

[theAnonymous1]

Originally Posted by IanAS

.....or the leads; the TO99 appear to have copper round leads which are bendy rather than the rather brittle and flat chip pins.

Good ol' fashion magnetic steel actually; nickel plated.

How does it sound now?

[IanAS]

Originally Posted by theAnonymous1

Good ol' fashion magnetic steel

Thanks for the correction My assumption without checking them before posting. Normally the steel leads 'stick' to my cutters and I don't recall that happening with these. So I got the last one out that I cut the leads down on, which was last July or sometime ago and indeed there is no copper colour showing from the cut ends. So I got out a magnet and the leads did attract to it. So did the can.

However it still sounds just as good. Maybe I need to get some more of that steel in there

Would their magnetism be filtering out some RFI?

Whatever the cause, we're happy with the effect.

I'd like to compare them with the AD797's in that friends DAC. A few weeks ago I bought some of those dual surface mount to DIL brown dog adapters so I can, at some point, get some surface mount versions and see what they sound like and compare those with the LME49990.

[IanAS]

Someone mentioned small 4 pole connectors: http://www.farnell.com/datasheets/8946.pdf

Silver plated copper alloy contacts.

Looks like these will fit between the PCI card slots at the back of a PC allowing a line socket to be wired to the card then poked through the slots and dangled to meet a cable plug. Card can still be removed for further work.

[IanAS]

Browse for Products | Farnell United Kingdom

Or some cheaper Multicomp branded ones from CPC which are maybe not silver plated and maybe lower copper content in the brass / bronze: Your Search Results | CPC

[RocketScientist]

I finally have all the parts for the QRV09 build and have also been testing another TPA6120 design. I have to say I'm impressed so far but I haven't yet tried using an inductor on the output to bridge the TI specified 10+ ohms of series resistance. I'm still working on a full set of "baseline" measurements before I start making modifications.

The chip runs somewhat warm even on +/- 9 volt rails so it's clear the pad underneath needs to be properly soldered. And, witht the QRV09 board, I'm not sure how to do that without drilling a bigger hole in the board to fit the soldering iron tip through.

I mainly wanted everyone to know I'm still making progress and still hope to arrive at a practical ultra low distortion design--even if it ends up being someone else's.

In the meantime, I've tested and reviewed the AMB Labs Mini3 which is based on the AD8397 high current op amp. The AD8397 itself seems to do a respectable job--all things considered. But the single 9 volt battery and virtual ground (rail splitter) really limit the performance in several ways. It was educational if nothing else.

I was rather disappointed a well proven DIY design that's been popular for years didn't come even close to several of the published measurements (partly made with RMAA). And I learned a lot about virtual grounds and "3 channel" designs. The short answer: Avoid them if possible--just as JCX has advised earlier in this thread. Here's the review:

AMB Mini3 Measurements & Review

[RocketScientist]

I have completed the bulk of my testing on my Sjostrom QRV09 and it does most everything very well but there are a few issues.

The 10 ohm series output resistor previously discussed can be shunted with an inductor to greatly lower the output impedance without sustained oscillation even with abusively capacitive loads. I've also explored the clipping behavior and there are no issues even with reactive loads.

But a 10 Khz square wave does ring more than I would like with 0.01 uF or more capacitance on the output with the output inductor mod. I don't know what the "worst case" is for real world headphones? With the headphones I've tried the output looks relatively good (in some cases better than with just the 10 ohm resistor).

Using an 0805 chip inductor also modestly increases distortion in the audio band. And using a 1 ohm output resistor in place of 10 ohms (no inductor) is a bad idea as I found signs of serious instability and quickly removed the load so as not to risk damage to the TPA6120 from severe oscillation.

I have more experience with power amplifier stability for speakers where you have to design for long runs of unknown speaker wire, complex crossover networks, etc. And you typically have options for various global and local feedback tweaks to increase stability, base stopper resistors, etc. But that's not the case with the QRV09.

In the Sjostrom Per Anders design the TPA6120 is operated at unity gain with local feedback removing the usual options for using the feedback loop to compensate for reactive loads. And I can't, of course, change anything inside the TPA6120 (unlike with a discrete amp). So I'm in somewhat uncharted waters trying to reduce the ringing.

I have a couple more things I want to explore and I'll be publishing a blog article on the QRV09 soon. In the meantime, if anyone has any ideas on worst case headphones loads, addressing the ringing, etc. please let me know?

[jcx]

ringing in the RLC output network/load C is outside of the amp's feedback loop - as long as it puts out a well controlled square wave at the op amp output/feedback node there is little you can do with compensation unless you want to roll off the amp at the ringing frequency - which will change with Cload, cable length

CFA op amps do have unobvious frenquency compensation in the choice of feedback R - you can use bigger feedback R even in unity gain to over-comensate the TPA - I would do this if you're seeing 50-100 MHz ringing at the TPA output pin - and probe C can be an issue in measuring this

[RocketScientist]

Thanks JCX. I tried another 0805 chip ferrite and it yields the best results yet with minimal ringing even with a lot of added capacitance. The ferrite also nicely controls the overshoot of the stock design with typical headphones without excessively rounding the 10 Khz square wave. It's better than the inductors in that regard. And this ferrite causes the least audio performance degradation of anything I've tried. So I'm going to re-run the applicable dScope tests and publish the article.

I also checked the feedback as you suggested and, as expected, it is free of ringing. So the ringing appears to be a resonance confined to the output and doesn't seem to indicate marginal overall stability. The ringing is at about 500 Khz with a 0.047 uF capacitor in parallel with a pair of Sennheiser CX300 headphones but it's very minimal with the current ferrites.

Please correct me if I'm wrong, but because the TPA6120 is configured with only local feedback and no gain it should have maximum stability? Emitter followers, and other output stages, can have inherent "local" instability issues. And I assume that's the main risk here and partly why TI recommends the 10 ohm output resistor? So as long as the output stage within the TPA6120 is happy, and there are not power supply and/or PCB layout issues, it should be solid, correct?

I've done quite a bit of testing, including some 20 feet of cable with various loads, added capacitance, and headphones at the other end. But my concern is there could be some weird pair of headphones out there that might trigger instability in the TPA6120 without the 10 ohms to isolate the reactive load. The ferrite is only effective at very high frequencies. I don't want someone to fry their amp or their headphones based on my modification.