It isn't wide-band noise though, its specific image tones. For example in measuring THD at 15kHz you're normally assuming that once you notch out the fundamental what's left is only 30k/45k/60k etc. But in the case of a NOS DAC with only minimal (6dB/8ve) filtering there's a massive component at...

What commercial NOS DAC has a sufficiently steep analog filter? I don't know of one that could keep the THD+N reading below 0.1% at all frequencies to 20kHz.

<pedantry> That's a pretty standard usage of 'v' to indicate ü as the Chinese name's written in pinyin. See here : http://pinyin.info/news/2009/v-for-u/ </pedantry>

No, I'm using a multibit DAC of my own design.

I tried 30Hz with my DAC, couldn't hear the faint sawtooth so I guess most likely it is your DAC. I can hear distortion at higher frequencies though with my DAC - not sure if that's the DAC, the amp, my headphones or my ears....

How are you generating the single frequency tones and what kind of distortion is it you hear ? Could you describe its character?

I'd be a little surprised to see AK4499 show up much under $500, unless the chip price reduces significantly. At present over on Octopart its listed at $80 in small quantities, falling to around $65 in high volume. To make using it worthwhile it needs a fair spend on power supply regulators and...

Yes, those transformers were what I was talking about. I agree 70euro is fairly cheap for a commercial signal trafo.

I'd guess deleting the output transformers shaved a very significant percentage off the BOM cost.

Strictly speaking its not current output, its an array of equal-valued resistors fed from a reference voltage. The output stage though should present the lowest possible impedance to the DAC (in order to achieve the lowest THD), in this sense its interfaced as if it were a current output DAC...

Post #21 - 'its possible to infer the saturation limit, <130mA'. The amp takes quite a lot more than 130mA. Curiously though Elvee says it doesn't really matter - yet @tomchr's measurements show ultrasonic noise as an issue.

The designer's power input filter doesn't work very well - its into saturation, as pointed out on the DIYA thread and potentially confirmed by @tomchr's measurements. Fortunately its DIY so you can improve it yourself, just reduce the number of turns on the core. 3 would be about right. There...

Common-mode noise is normally the culprit for this and yep, its generally at a much lower level with a 'linear' supply (as opposed to an SMPSU).

Haven't had a serious listen to TDA1541A but TDA1387 is awesome for RBCD. When you factor in the price (next to nothing) they're unbeatable.

As a (somewhat pedantic) point of fact, neither TDA1541A nor TDA1387 are R2R DACs as they don't use resistors to create the bit weightings. Both of them use proprietary Philips technology.