Some people enrich the sound of their systems using carbon resistors. Nothing bad though. Just a preference.
“I personally believe that THD should be as low as possible and dominated by the 2nd harmonic. Harmonics above the 4th should be below the noise floor. Noise should be a low number on a spec sheet and inaudible. Slew rate spec should be >10x what the signal requires. If you like a system that scores poorly on these metrics, then I am not designing for you.”
Getting harmonics above the 4th below the noise floor is no mean feat given measurement systems that are available today. The spot noise floor on a QA401 with the inputs shorted is -155 dBV, and with a typical low noise DUT connected you’re at -135 dBV. (these measurements were done with 2.94 Hz bin width and 50 averages).
Getting harmonics above the 4th below the noise floor is no mean feat given measurement systems that are available today. The spot noise floor on a QA401 with the inputs shorted is -155 dBV, and with a typical low noise DUT connected you’re at -135 dBV. (these measurements were done with 2.94 Hz bin width and 50 averages).
Last edited:
Some are, I would agree. Would you care to offer an example or three of what you think is a 'good enough' dac to achieve excellent depth reproduction?
EDIT: On the subject of a measurement for depth, I think you gave some good advice (although I might nitpick a little here and there), but it didn't seem a quantitative metric of depth reproduction quality. That is, not exactly a measurement per se.
Last edited:
Yes, sorta. Good tube amps can produce an enjoyable stereo illusion, but is it accurate? My guess would be that some of the perceived imaging is synthetic, caused by small dynamic nonlinearities in tube circuits (that are pretty well understood by some musical instrument tube amp designers, but often out of mind for other tube hi-fi amp designers). That is to say, there may be a bit of exaggeration of the audible cues of imaging produced along with other distortion products. Since we tend to measure distortion using stationary test signals, and do it one channel at a time, we don't normally look for little dynamic circuit behaviors that might affect imaging of music signals. Again, a guess only at this point.
Last edited:
https://www.firstwatt.com/pdf/art_h2_v1.pdf
"So why is the phase important? Well, it's a subtle thing. I don't suppose everyone can hear
it, and fewer particularly care, but from listening tests we learn that there is a tendency to
interpret negative phase 2nd as giving a deeper soundstage and improved localization than
otherwise. Positive phase seems to put the instruments and vocals closer and a little more
in-your-face with enhanced detail."
"So why is the phase important? Well, it's a subtle thing. I don't suppose everyone can hear
it, and fewer particularly care, but from listening tests we learn that there is a tendency to
interpret negative phase 2nd as giving a deeper soundstage and improved localization than
otherwise. Positive phase seems to put the instruments and vocals closer and a little more
in-your-face with enhanced detail."
Does anyone have subjective listening impressions of the same circuit with OPA1656 vs OPA1642? Considering use of the latter as 1656 has too much interest and getting hard to find reliably for production.
Interesting thread. I was probably the first to retrofit an EMU 1820m recording interface with LM4562 years ago and the SQ improvement was quite substantial. I still use that unit today.
I recently bought another 1820m since they are now obsolete, though the ADC/DAC chips are still quite good even by todays standards, and the units can be had for next to nothing. The one I just picked up is stock and is still packed with NJM2068M. So I was able to get a fresh reminder of just how much improved the one with 4562 in it actually sounds, lol.
That said, the biggest issue with the 4562 is quiescent current, which is typically about twice that (if not more) compared to most alternatives. Not too much a problem except for retrofit scenarios having 20-30 opamps...both the power supply and thermal budgets can get blown out in a hurry.
In my first retrofit of the 1820m I failed to recognize these facts, but fortunately it had enough margin for both counts to allow it to run hours a day all these years. But it does get quite warm to the touch.
All to say, the stock one I just picked up is getting filled with opa1656. Since I'm a musician first as opposed to audiophile, I have zero concern with 1656's higher noise floor at low frequencies, as its noise floor in the "money region" is actually lower than most if not all competitors. And the quiescent current per device is half that of the 4562 I used in the first one.
I'll post in a few weeks my subjective take on the results along with RMAA objective results... Wish me luck! And thanks for everyone sharing thoughts on 1656. It makes it much easier to commit to doing a 30 opamp retrofit with zero audition beforehand. I did that last one with zero audition ad it worked out splendidly hopefully this does too!
I recently bought another 1820m since they are now obsolete, though the ADC/DAC chips are still quite good even by todays standards, and the units can be had for next to nothing. The one I just picked up is stock and is still packed with NJM2068M. So I was able to get a fresh reminder of just how much improved the one with 4562 in it actually sounds, lol.
That said, the biggest issue with the 4562 is quiescent current, which is typically about twice that (if not more) compared to most alternatives. Not too much a problem except for retrofit scenarios having 20-30 opamps...both the power supply and thermal budgets can get blown out in a hurry.
In my first retrofit of the 1820m I failed to recognize these facts, but fortunately it had enough margin for both counts to allow it to run hours a day all these years. But it does get quite warm to the touch.
All to say, the stock one I just picked up is getting filled with opa1656. Since I'm a musician first as opposed to audiophile, I have zero concern with 1656's higher noise floor at low frequencies, as its noise floor in the "money region" is actually lower than most if not all competitors. And the quiescent current per device is half that of the 4562 I used in the first one.
I'll post in a few weeks my subjective take on the results along with RMAA objective results... Wish me luck! And thanks for everyone sharing thoughts on 1656. It makes it much easier to commit to doing a 30 opamp retrofit with zero audition beforehand. I did that last one with zero audition ad it worked out splendidly hopefully this does too!
The first problem is trying to find a perfect op amp with RRIO. This always requires an engineering tradeoff somewhere.
It would be much better to boost supply voltages and look at the greater number of op amp options. If running off 5V USB supplies, then create one's own very quiet switcher to boost voltage tailored for your specific application.
For R-R inputs on an op amp, one can use complementary inputs, but then there is always a Vos &/or Gm glitch in the common mode input range as current shifts across the complementary input pairs.
One can use a depletion input differential pair, but then the input noise depends on common mode input voltage. The process technology options are limited.
One can use a voltage boost to allow the input stage to operate at a voltage outside the supply rails, but then there is added charge pump switching noise.
RRIO op amps are mainly for ease of use/design-in.
As production shifts across the Fabs, the scarcity of op amps may shift. Finding reliable availability may shift across op amp families. Keep in mind, there is currently an industry wide semiconductor shortage and customer volumes shift in unexpected ways.
Worldwide chip shortage expected to last into next year, and that’s good news for semiconductor stocks - MarketWatch
Global chip shortage to last till 2022, causing bump in semiconductor stock value
Porsche CEO warns of '''very serious''' global chip shortage
Siglent SDS2000X Plus - Page 97
Last edited:
Good post!
Once I saw on TV a shootout Jaguar vs Porsche with many car experts and journalists as judges. They said on paper the Porsche was better in all compartments, even the stereo of the Porsche sounded better but all preferred the Jaguar because of its handling and driving feeling. Can't remember exactly but it was like 12 vs 0 in favor of Jaguar!
There's a ton of measurements that aren't generally taken.
I mean, I can develop a system with wonderful THD at 1kHz, that will sound like absolute crap. This isn't out of the ordinary either, marketing comes in with a price point and specific specifications, and the engineers figure it out. The whole game is a tradeoff. Look at datasheets in-depth and you'll see it.
I mean, I can develop a system with wonderful THD at 1kHz, that will sound like absolute crap. This isn't out of the ordinary either, marketing comes in with a price point and specific specifications, and the engineers figure it out. The whole game is a tradeoff. Look at datasheets in-depth and you'll see it.
In an IC op amp, raw cost of the silicon chip is not really a factor. Yield factors in if the specs are too aggressive for the process, but with our capability today to simulate the silicon, that is less of a problem. Now, if there's low yielding wafer, that is because something was wrong in the processing.
Another big factor that enters in is testing and trim during testing. Basically, if the tests are hard to run or time-consuming that adds to the cost. Finally, the costs of design, fabrication, packaging, and testing have to be amortized over the expected sales numbers for the product. A surprisingly high number of ICs sell hardly any numbers, and the ones that sell well have to cover the costs of the duds, as well.
If part's volume is expected to be low, it's going to be expensive. Someone needed a special part, was willing to pay well for it, and now it's still in the catalog.
Then there is what people are willing to pay for the part. If it's cheap to make, but it performs better than the competition, then you still charge what the competition charges, or only slightly less. People will pay for the price, just like in high end audio.
Another big factor that enters in is testing and trim during testing. Basically, if the tests are hard to run or time-consuming that adds to the cost. Finally, the costs of design, fabrication, packaging, and testing have to be amortized over the expected sales numbers for the product. A surprisingly high number of ICs sell hardly any numbers, and the ones that sell well have to cover the costs of the duds, as well.
If part's volume is expected to be low, it's going to be expensive. Someone needed a special part, was willing to pay well for it, and now it's still in the catalog.
Then there is what people are willing to pay for the part. If it's cheap to make, but it performs better than the competition, then you still charge what the competition charges, or only slightly less. People will pay for the price, just like in high end audio.
How about using Rod Elliott's S.I.M. (Sound Impairment Monitor) to compare opamps?
Sound Impairment Monitor (SIM) - Is This The Answer?
Sound Impairment Monitor (SIM) - Is This The Answer?
S.I.M probably can't find every problem. Being able to substitute a spectrum analyzer and an FFT analyzer for the scope might add some useful capability. The other problem is that S.I.M is only intended for checking mono, not stereo. Dynamic differences between channels can be missed. Then there is the matter that no system is perfect, so there is always going to be the issue of detectable small deviations from perfect and trying to decide if they are truly inaudible to every one of the 7.8 billion humans on earth. In other words, how does one interpret the possible importance of measurement results that are not gross in nature? Also, how can one include a DAC in a system measurement without also including an ADC at the same time? What about ground loops between the measurement system and the DUT (including when testing for effects of environmental RFI/EMI ingress)?
Last edited:
I have never really understood all the hype around different sounding op amps.
If I put in a music signal then if its not coming out amplifier amplified exactly the same its not a good op amp.
Given the circuit has negative feedback there is little reason why it wont come out the same.
I understand why the old 741 is crap due to poor bandwidth, but most modern stuff is pretty good.
Claims of op amps sounding amazingly different are either lies or just pure imagination in my opinion.
If I put in a music signal then if its not coming out amplifier amplified exactly the same its not a good op amp.
Given the circuit has negative feedback there is little reason why it wont come out the same.
I understand why the old 741 is crap due to poor bandwidth, but most modern stuff is pretty good.
Claims of op amps sounding amazingly different are either lies or just pure imagination in my opinion.