It is a significant advantage when the input section is a switched-capacitor circuit, which has the disadvantages that the ADC then has no suppression of aliases around multiples of the sigma-delta modulator sample rate and that the circuit that drives the ADC has to handle a spiky input current.
There are two peculiarities I forgot to mention:
3. It is more difficult in an ADC to reduce the high-frequency content with some passive filtering or with a FIRDAC, because any extra filter you put in the loop affects the loop stability. In a DAC, such filters are outside the sigma-delta loop.
4. For multibit or quasi-multibit modulators ($), dithering the (digital) requantizer of a sigma-delta DAC according to theory is easier than dithering the (analogue) quantizer of an ADC according to theory. Adding an analogue signal with precisely the right amplitude and probability distribution is not straightforward, so one usually settles for just using a few levels with equal probabilities.
With such improper dithering, it is difficult to ensure that there will never be any tones around fs/2 that could intermodulate to audible tones in the op-amp.
Regarding op-amp swing, internal op-amps usually have far less supply voltage than the +/- 15 V that many external op-amps can handle, so the internal op-amps have to swing closer to the rails.
($): Single-bit modulators can never be dithered according to theory.
Last edited:
Internal opamps are presumably CMOS? Output stage opamps are often bipolar and or contain input protection diodes. Seems like forward biased junctions are where its easiest for demodulation and intermodulation to occur?
Still trying to enumerate any possibly significant differences... Why? I know of some dac designers who have experimented with RF suppression before opamp stages and found readily measurable and or audible improvements when filtering was carefully designed. Mostly I hear about these issues when designing with ESS dac chips. Some of what is going on might be time-variant distortion and or signal-correlated noise, since it does not always appear to show up well on typical audio FFTs. IMHO highly doubtful its all imaginary. More likely we are sometimes 'looking for car keys under the streetlight.'
Still trying to enumerate any possibly significant differences... Why? I know of some dac designers who have experimented with RF suppression before opamp stages and found readily measurable and or audible improvements when filtering was carefully designed. Mostly I hear about these issues when designing with ESS dac chips. Some of what is going on might be time-variant distortion and or signal-correlated noise, since it does not always appear to show up well on typical audio FFTs. IMHO highly doubtful its all imaginary. More likely we are sometimes 'looking for car keys under the streetlight.'
Last edited:
Most sigma-delta ADC and DAC chips are made in CMOS processes, so the internal op-amps are then also CMOS. They may or may not be class A.
The excursions should not be big enough for protection diodes to turn on, but a forward-biased base-emitter junction/differential pair can very well start intermodulating for out-of-band signals. Mind you, the voltage-to-current transfer of a weakly inverted MOS differential pair is only a bit less curved than that of a bipolar pair.
Do you generally prefer op-amps with a large slew-rate-limit-to-gain-bandwidth-product ratio? That ratio is a measure for the "linear" input voltage range of the input stage at frequencies where the feedback has no loop gain anymore.
The excursions should not be big enough for protection diodes to turn on, but a forward-biased base-emitter junction/differential pair can very well start intermodulating for out-of-band signals. Mind you, the voltage-to-current transfer of a weakly inverted MOS differential pair is only a bit less curved than that of a bipolar pair.
Do you generally prefer op-amps with a large slew-rate-limit-to-gain-bandwidth-product ratio? That ratio is a measure for the "linear" input voltage range of the input stage at frequencies where the feedback has no loop gain anymore.
IIRC according to TI engineers posting in the forum, input protection diodes are one known source of demodulation/intermodulation with resulting bias shifts and DC output offsets in opamps. Again IIRC they say more modern opamps are designed for improved RF immunity, but specific details were not given. It may be they are only required to pass one steady-state RF signal test, don't know.
Regarding my preference for opamps in dac output stages, it entirely dependent on SQ. Different dac chips and layouts may favor one opamp or another. OPA1612 has been a pretty safe bet, but not always the best.
Regarding my preference for opamps in dac output stages, it entirely dependent on SQ. Different dac chips and layouts may favor one opamp or another. OPA1612 has been a pretty safe bet, but not always the best.
You can get all sorts of funny low-level artefacts out of sigma-delta ADCs and DACs, especially when they are not dithered in accordance with nonsubtractive dither theory. Tones that get frequency modulated by the signal are the most well-known. When there is a bit of noise in the modulator, those tones can become noise bumps.
The chaotic mode of my valve DAC has nonstationary wideband noise that sounds like woosh-woosh-woosh, with the frequency of the envelope proportional to the input signal. The slightest amount of signal makes it inaudible, but on versions 1 and 2.1, you can hear it when playing silence with the volume turned way up. (I added a small offset to also make it inaudible during silence in version 3.)
There is no way it would show up on a DFT plot. Maybe you could see it if you would make several DFT plots and compare their noise floors. It should be visible on a sensitive oscilloscope with a timebase of a few seconds when the out-of-band noise is filtered off sufficiently.
Last edited:
I think you are pretty close to the mark here, no pun, WRT driving ADC IP's. Generally ADC IP's have quite decent size smd film caps right at the IP's to shunt all this HF noise and it's very easy to configure the OPA drive circuit to handle the capacitive load
and provide the required low drive Z for low distortion.
TCD
Speaking of R2R, some tape machine porn.
A few interesting / quirky design features for those that know tape.
\___/
TCD
\___/TCD
I think you have strange hearing. Our brains do a LOT of processing so this wouldn't be an abnormal thing for you to be very different from most people.
If it's for the better the only evidence we have at this time is the convenience factor. These are THD chaser words. There's more going on than you're probably aware of... It isn't the distortion. In fact a lot of very old gear had less linear distortion, by the truck load. Also there are other technical, objective, things that a lot of very old equipment had advantages in. You think they're fools but they were not. The best recordings there are, that I've ever heard, are very old. There just isn't any real comparison.
1992? Odd time to pick. Depends on whether you're talking digital or analog. Digital? Very much so. Just look at the top 100 from then.
Whoa, hold on there. Firstly tape is a lot like our hearing, the frequency limits exist but the amount of resolution is another thing. Much like our ears that transmit terabytes worth of data a second to our brain, tape is more dense for resolution. I've got a friend working on new ways to digitally transmit a better copy of music without the typical DAC/DSD conversion process. He brings this up constantly... I'm not sure what he's working on will work in the US without better internet though.
Objectively better is an odd reference in some ways. You may be able to measure numerous things it appears better with, but the value of each measurement is not the same. Much like the guy with the DDDAC TDA1543, clearly the measurement they laughed at wasn't the most important. I have a TDA1543 DAC as well and I'd take it over friends' DAC's that costs an absurd amount; purely based on the sound I hear. Although I typically don't have the DAC plugged in because... I own records and don't have any reason to listen to the inferior sound.
I think digital can be greatly improved, as I've heard it sound decent (red book, btw, never heard anything else sound better). Sadly there's a lot of knowledge that is mixed up in the design of the equipment that is a big enough rats nest that I think some day I'll just have to work on designing new stuff.
I love it, I'm going to use this some day. This is a person with ears.
I'm not against digital... I just think I need to work on solutions to make it appreciable compared to analog. Sadly no one is going to re-record...
