Diamond Buffers - Page 3

Quote:

Originally Posted by Chris J 

I don't think the moderator's mind naming him, but I wouldn't post a link either! LOL!

Chicken! Bok bok bok!

se

.

.

.

when a amp circuit using the TPA6120 chip doesn't come within 20 dB of the manufacturer's plots I have to assume implementation problems - common impedance grounding errors, poor power supply, inadequate pwr pad heatsinking could all contribute

the series L doesn't have to be a tiny 0805 chip ferrite  - the required inductance is low enough to use "perfectly linear" air core inductor

even lossy ferrites can give "unmeasurable" audio frequency distortion if upsized large enough to avoid saturation - I used a 5 A rated part

I still maintain the TPA6120 beats the performance of previous unity gain buffer chips

if you look for measurements of Harris, BUF634 ect their "open loop" distortion is much worse at low Ohm loads - the LME buffer datasheet only shows the distortion with the buffer chip inside a multiloop's added feedback loop

I don't expect even by itself the TPA6120 has audible errors, but I prefer the multiloop feedback topology with a "good", "audio" op amp input, global feedback around the TPA, in which case you could in principle move your plot's distortion traces down 40-60 dB more with good implementation

I believe that with advanced multiloop composite circuit, "good", "audio" op amp input, global feedback around the TPA, that active device distortions are not visible without long averaging to see below the noise floor - layout, passive device performance become bigger limits to the amp's distortion performance 

.

What is this? A TPA6120 love/hatefest?

Even @ 20dB above the TI plots we're still talking angels/pinheads territory. That's 0.0014% @ 80mW into 600 ohms. That's one of the truly amazing things about it. It can fail to meet it's spec by a factor of 10 and its noise + distortion will still be below a level almost universally considered to be inaudible, by a factor of 10.

This kind of argument only encourages the view that distortion at these levels is significant. It's not. You guys are supposed to know this, and probably do, but you let the fact get subsumed in arguments which are really intended to demonstrate your superiority, your greater discrimination, your greater knowledge of the subject, but only end up being used as ammunition by people who are groping for justifications for their prejudices.

How 'bout your greater self-control?

w

I have to agree with jcx.

As someone pointed out a few posts ago, spec. sheets provided by the manufacturer's legally cannot mislead the consumer of those products.

Obviously the manufacturer has optimized the test layout to show best performance.

However, a good designer should be able to replicate that design.

I'll guess that whoever implemented those designs posted with the poor performance did not spend enough time and effort to optimize the PC layout, including power supply implementation, power supply decoupling, etc.

Perhaps they created a design under constraint of time, budget, measurement equipment and released something "good enough".

The spec sheet states the IC must be heatsunk, I've never seen an STX, is the TPA properly bonded to a heatsink? Did the STX designers do the headphone amp as an afterthought?

.

obviously some will still use discrete buffers, "diamond" or not - others insist on all discrete - despite massive advances in monolithic technology

imaginations may be captured by "classic" hybrid or monolithic buffer chips from the popular diy projects

I just claim the TPA6120 is technically worthy of consideration - betters many other implementation's specs within its I,V, power limits, allows an extra degree of freedom in design by being able to deliver V gain with local feedback, inside a multiloop

the datasheet specs are qualified for gain, load conditions - all of the data should have been taken from/reproducible on the demo board - which is fully documented with pcb artwork, BOM - TI wants customers to be able to achieve the datasheet numbers

http://www.ti.com/lit/ug/slou169/slou169.pdf

Quote:

Originally Posted by stv014 

 

The manufacturers might not lie, but that does not mean that they cannot test their products under conditions that make them look as good as possible. Other than the already mentioned resistors in the feedback loop, details like whether the amplifier is used in an inverting (no common mode voltage swing) or non-inverting configuration could make a difference. The fact that three independent implementations of the TPA6120 have highly similar distortion driving low impedance loads (and I do not even know if there are any that were confirmed to reproduce the datasheet performance) suggests that either the problem is inherent to the TPA6120, or the chip is difficult to implement properly.

Sure.

If the manufacturers can make them look as good as possible, then the consumers should be able to duplicate those results, TI isn't out to delude anyone.

But who said it was supposed to be easy?

It's easy to screw up with poor PCB layout, poor power supply implementation, ignoring half the spec sheet..

TI gives you lots of suggestions, makes me wonder if the designers bothered to apply them.

Anyone notice that TI typically uses +/-12 or +/- 15 Volt rails for quoted specs and graphs? I suspect this is to make the TPA looks it's best.

Does the STX have +/- 12 V rails?

.

.

arguments of buffers aside, that looks like a very interesting amplifier design to my very untrained eyes. Especially intriguing is the idea of sinking multiple transistors to the same small heat-sink to minimize temperature variation, although I wonder if any sacrifices were made in layout efficiency to facilitate that. Regardless, it might be an amp I have to try someday!

Quote:

Originally Posted by stv014 

 

Here is a high resolution picture of the PCB of the STX, although it probably does not tell much, as it is a multi-layer board and many of the traces are not visible. One thing that can be seen is that it apparently does not have as much cooling as the reference implementation, where the TPA6120 is soldered onto a large ground-filled area. You can also see a simplified schematic diagram (it does not show any power supply capacitors, some relays, and RF filtering at the output connectors) here. For completeness, here and here are some distortion graphs I created with only one channel driven (front panel/22 Ω load/~1.5 Vrms, and the headphone jack on the card itself/56 Ω load/~2 Vrms, respectively). These also show cross-channel THD; note that at high frequencies those traces are dominated by noise and interference.

Yep, the picture of the PCB does not tell much. Neither does the schematic.

Labels, what do the lines on the graphs represent? Is it calibrated?

Are you measuring the whole signal chain or just the TPA6120?

I really don't know why TI publishes one spec but STX gets different results.

Just a guess here, the STX was designed by a team of digital designers and one of them was saddled with the task of designing the headphone amp, and since he is not an analog guy, he doesn't know how to finesse the circuit. They use cheap test equipment so can't see what they are doing, it sounds good enough so they don't care, who knows? The circuit does a lousy job of rejecting noise from the rest of the circuit.........um, um.....I'm tapped out. Others can invent other theories. Unless someone has a Deepthroat at STX?

Wakibaki will be happy to see this discussion seems to have ran it's course...................

We will now go back to our regular programming.