From: Utrecht, NL
Registration: Feb 2005
| posted March 19, 2005 06:15 PM||profile edit reply w/quote IP|
Gordon, Thanks for reaction, lots to think about!
-Okay, I will pay more attention to the datasheets, man. Be in for some surprises. See below, where I will do some interpretation of specs.
- Don't feel attacked. You write as if you have to defend yourself against all kinds of words that I have put in your mouth. Lighten up, I did not.
- I never called you a liar. Not explicitly, not implicitly. Do not throw that at me. It's not fair.
- I never said you have to believe me. Just try and see for yourself. If you want to have an open mind, then you are condemned to investigate yourself. And interpret your findings. And be stimulated by others. Especially when they say things that differ from your own conclusions.
- I never said that you called the THS4062 top notch. I just repeated a word YOU started to use somewhere, as if I had used it before, which I had not.
-Printed the Albini article. Thanks, interesting. will read it.
-Most of the Modders and DIYers have VERY limited experience and knowledge for getting the best out of a part. Hobbyists. Most of them run after firmly stated opinions and get stuck with bad designs and sub-optimal sound. Typical behaviour. Humans need signals, directions, guidance.
But as long as they wish to listen to fixed voltage regulators for example I certainly cannot take those opinions seriously. But I do not say that they are lying. Just that they ain't right.
Only a very limited number of those people really know enough to extract all of the qualities out of the parts.
Mostly THOSE are the people with an open mind. And the guts to investigate a claim, or a spec, or a phenomenon.
And the guts to draw the right conclusion, whether they like the outcome, or not.
- You too, you run after OPA627 community. This opamp is not as good as everyone wants to believe. Its sound is colored. Why? I don't know, never deeply investigated. But I had to dismiss it because of this. Ever heard it side by side with the AD797? A tricky opamp, hard to get stable (you need 600MHz or 1Gs/s scope), but very, very good when used properly. Which only VERY few people do!
- Especially do not believe companies. They need to sell.
- All batteries sound different. Are you listening to your opamp or to your battery? Or your bypass caps? Multilayer ceramics? Slow electrolytics? Are you hearing the dielectric properties of a chemical substance?
Run on batteries, listen carefully, change your electrolytics, listen again, you'll hear differences. Shouldn't be possible, if batteries were perfect.
Batteries are helpful, not holy.
Pick up the power supply line with a decent series capacitor (ground reference at input of that same device (lynx?)), feed it to a microphone amplifier and listen to it when your gear is working. You hear music. On the power supply lines.
_ Not all that is not the best is bad. I never said that. Compared to a OPA2604 indeed a 2134 is not bad, even slightly better maybe.
But compared to a REALLY good opamp the 2134 is bad. Not "fair". The difference is big.
It is all about perspective. The moment you know how REALLY good something can sound, the differences become VERY apparent, and big gaps between opamps become clear.
Look back 8 years. Remember what you thought was top quality back then. Is it still that good? Or has your knowledge and experience advanced? This is also an illustration in how far you yourself have advanced.
-Award winning audio gear does not tell me much, except for that they did a good job in implementation of mediocre stuff, and achieved a (slightly?) better balance than others. That's an art, and I have high respect for those people that make a successfull product on sound quality merits. I work with this award-winning stuff all day, design award-winning stuff myself and know how gross those claims can be.
Award-winning audio / hifi is about things that sound "pleasant".
I won't degrade Lexicon. They need to earn a living too. Nothing wrong with that, but budget is budget. If the OPA2134 is good enough for the price, then they CERTAINLY will not put a more expensive opamp in!
-You're right about the distortion figures in the THS datasheets. I did not read carefully enough, I take that back.
The THS 4062 and THS4032 seem to have quite similar distortion figures. Above 100kHz. Below that, there is no data, unfortunately. 3rd harmonic for 4032 is better below 1MHZ and worse above 1M but that's due to the difference in speed (is it? see below), and not so relevant for audio, I suppose.
-But the input voltage noise of the two tells another story. In I/V converters not very important, I agree, but in all normal audio-applications this is VERY important. Here all noise counts and colors your signal 9 times!
-About speed: The 180MHz of the THS4062 is somewhat flattered: only in non-inverting mode, gain +1. Being a unity gain stable voltage feedback opamp, that will be half (I.E. 90MHz) at double gain (+2). THS4032 is 100MHz here.
In an I/V converter the opamp is inverting. So, also here the tables have turned. Because the THS4032 is somewhat decompensated (and hence not unity-gain stable) it is 100MHz (@G=-1) where the 4062 gets stuck at 50MHz.
So essentially in all areas the THS4032 is faster, except at G=+1, which it can't do...
-About settling time: The single most important feature for I/V converters. What happens to your audio signal when the square wave is ringing? There is a lot of energy generated in ringing, it is fed back by the feedback network right into your dac's output!...which is all translated in audible disturbances, noise modulation, coloration, power supply modulation, ground modulation, reference voltage modulation, and what else.
When an opamp stops ringing halfway compared to the other type, that means half (okay, 65% in this case, 90vs140ns) the time duration of unwanted energy.
-No we do not evaluate sound by the datasheet. But the datasheet is a very interesting (although complicated) source of information from which we can make the assumption if a part has potential or not. And if settling time, noise and the like sport good figures, then we can assume that the opamp has potential.
The other way round: If I have tried an opamp, and I found it to be sounding very good, then I like to blame some good specs for that. However I admit that that's quite unscientific. But if you have a lot of experience, then you can (I can at least) see from a datasheet if an opamp will be worth trying or worth to be left alone.
-Burr Brown do anything to optimize their specs. Nothing wrong with that. But that does not say ANYTHING about sound quality. At Ti they do not evaluate opamps nor dacs for sound quality. They want a part to be technically correct. Specs correct. Reproducability optimized. Ofcourse.
Re. Lynx: Okay if you say so. I can't believe that. Is it used in an electrically clean environment? Or in a computer?
Did you do broadband measuring (0Hz-100/200/300/1000whateverMHz) of your supply lines and ground? What will be the PSRR of your opamps and other chips on this card at 10MHz? and at 100MHz? How much of this pollution will get into your audio path, and ground path, and reference voltages? What will that do? Nothing? because it is 100MHz? It will be half rectified by all kinds of semiconductors in your chips, stored in parasitic capacitances, released a little later, generate offsets, modulate or be modulated by your audio signal, generating all kinds of intermodulation products, noise modulations etc etc. There you go: polluting your sound.
-You make an assumption that is very peculiar:
"The reason why I couldn't hear the difference was in the fact, that the spectrum of distortions hasn't changed with op amps swap."
How do you know? Where is the proof of the correlation? It is only a convenient explanation which unfortunately has the power of making you stop thinking.
This surely can't be the end, can it?
There are two things you know: 1. the distortion spectrum did not change, and 2. the sound did not change.
This is totally insufficient to draw ANY conclusion from, otherwise than that apparently the things you actually DID measure, appeared to have not changed.
Try listen to decent loudspeakers. That is a lot more revealing. You need to develop a sound field in front of you where subtle phase- and time differences an low-level information take care of proper stage display, front-to-back stage/hall info, placement of instruments in the available space, etc. Headphones do not give you (all of) those clues. It all happens inside your head, which is a very unnatural thing.
Re.: my Juli@: you'll be surprised how good those AKM's sound. I was surprised too. And I already had a hint, because of the Yamaha. Again you're fixed to specs 'n data. I'm not. I listen to my own recordings of symphony orchestras in real concert halls and relate the sound quality to actual music from which I know how it actually sounded in the hall!
Evidence is unstable. Correlation fails. RMAA is measuring static data. Good for providing static specs.
But useless for obtaining relevant info on sound quality. That relies on dynamic behaviour of the active parts like A/D's, DACs, opamps. RMAA does not measure any dynamic property. For example: why not make a 20Hz-20kHz sweep at -1dBFS and do an uncorrelated amplitude sweep from -50...-130dB at various frequencies, and look at the intermodulation products and amplitude linearity results? Or the other way round; an amplitude sweep between 0dBFS and -130dBFS of a range of frequencies, and run a frequency sweep at the same time at -40, -50, -60, -70, -80, -90, -100 etc dB... what would we see? I don't know but would for sure be interesting, don't you think?
If you hear differences in sound, you may or may not see differences in test- and measurement results.
But who is telling you that those are correlated? What evidence would you provide to show beyond any scientific doubt that the things you have measured are indeed exactly the same things that cause the differences in sound you hear?
This is the heart of the problem: it is impossible.
For now, the evidence is absent, and we have to rely on our own judgement.
That is what makes sound quality so interesting...and so controversial.
Gordon: Why did Crystal design the CS43122 if they already had the 4396? The only difference is 2dB in dynamic range (as far as I see).
To me it is the same chip in another package, maybe with 1 slightly thicker ground wire from die to pin18, or 1 bit more in the modulator, like 5bits for CS43122 and only 4bits for the CD4396? Maybe you know?
Temporary: Stereo crosstalk is greatly improved by all actions described in previous postings.
It is not attribuable to one single source. Depends mainly on quality of board layout, thickness of copper tracks and ground impedances, power supply decoupling, and clock, these factors determine if it is digital crosstalk or analog.
Especially digital crosstalk is not to be overlooked: digital transitions all over the place create noise on power lines and "ground bounce"; and that noise generates jitter everywhere, which is modulated by your audio signal. This "data-jitter" modulates the clock frequency around your dac and your crosstalk is a fact.
Take care of a very good clock, and decouple all power supply pins and reference pins of all chips -analog as well as digital- with 100uF minimum high quality electrolytics (Panasonic FC or Sanyo's Oscon or equivalent stuff) AND high-speed surface-mount capacitors with low-dielectric absorption properties. NO ceramics! Better use 0805 or 1206 PPS film, widely available nowadays from Panasonic and Epcos, to name a few.
Block incoming RF, by using good common-mode inductors in power-supply feed lines as well as ground that is coming from your computer ground plane.
With these measures you will greatly decrease dynamic crosstalk, and you'll end up with a much wider and much more transparent sound field.