While this is theoretically possible, it's highly impractical. Burr-Brown's co-linear architecture comes quickly to mind, as an example of utilizing two D/A units of X-bits resolution, to created a single composite D/A unit of X + 1 bits resolution. Creating a composite 24-bit D/A out of multiple 16-bit D/A units would require 256 such 16-bits units. Setting aside the technical problems of accurate gain scaling across those 256 units, and all the requisite control logic.
I saw an MSB DAC tested to darn near 24 bit performance.
To hit 24-bit performance means a signal-to-noise ratio circa 144 dB. At that point, the D/A conversion is just one piece of the puzzle. Keeping the analog that clean I should think requires genius-level circuit layout and shielding, and multiple very cleverly done power supplies, and ultra-clean analog circuits.
I applaud the DIY spirit though. Maybe utilize multiple paralleled 24 bit devices. Also, start with the very highest spec'd devices-the performance can't be better than that, eh?
To hit 24-bit performance means a signal-to-noise ratio circa 144 dB. At that point, the D/A conversion is just one piece of the puzzle. Keeping the analog that clean I should think requires genius-level circuit layout and shielding, and multiple very cleverly done power supplies, and ultra-clean analog circuits.
I applaud the DIY spirit though. Maybe utilize multiple paralleled 24 bit devices. Also, start with the very highest spec'd devices-the performance can't be better than that, eh?
regarding the MSB dac they are using vishay VAR resistors costing about 10usd each which has very decent precision 0.005%. I agree that we need is much much higher precision but thats what the case how did MSB achieve with just 80 pairs of resistors?
did they parallel / cascade? how did they get to it?
The psu voltage regulator with ultra low noise should be possible with this chip from TI
Linear Regulator (LDO) - Single Channel LDO - TPS7A4701 - TI.com
There are even some zetex versions which gives dead noise floors for psu.
do you know how MSB has achieved 24bit? I dont think its impossible to make it
did they parallel / cascade? how did they get to it?
The psu voltage regulator with ultra low noise should be possible with this chip from TI
Linear Regulator (LDO) - Single Channel LDO - TPS7A4701 - TI.com
There are even some zetex versions which gives dead noise floors for psu.
do you know how MSB has achieved 24bit? I dont think its impossible to make it
Why do you want 24bits?
TDA1545 is low enough power and small enough in size to contemplate paralleling enough. I reckon it needs 256 of them to get to 24bit resolution. The noise won't be 24bits though because doubling the number of DACs only gives 3dB noise improvement. So while a single TDA1545 is around 100dB SNR the array of 256 will only improve the noise by 24dB - that's around -124dB so sllghtly better than 20bits.
TDA1545 is low enough power and small enough in size to contemplate paralleling enough. I reckon it needs 256 of them to get to 24bit resolution. The noise won't be 24bits though because doubling the number of DACs only gives 3dB noise improvement. So while a single TDA1545 is around 100dB SNR the array of 256 will only improve the noise by 24dB - that's around -124dB so sllghtly better than 20bits.
dont tell me I am mad but im using a Analog Digital Receiver chip which has PCM outputs in I2S format and we want to use ladder dac for 5 channel. I know you may call me nuts but believe me or not we are doing something substantial as even drivers are being custom designed to a very extremely low distortion. The tweeter diaphragms are even from the one who give it to scanspeak and seas and dynaudio.
we are trying to make something earthly natural sound quality even in 5.1 format so we will be using ladder dac for 5 channel but not for the .1 subwoofer channel.
So my biggest doubt is that when using DTS HD Master Audio format does it really takes 24 bit?
For CDs im still happy with 16 or 18 bit ladder dac which is somewhat easily achievable but the prob comes with 24 channel...
Regarding the parallel DAC operation is it like just plain paralleling of the 16 bits DACs?
Im thinking of achieving 18 bit and then go upto 21bit that should be generally sufficient I believe what do you say?
we are trying to make something earthly natural sound quality even in 5.1 format so we will be using ladder dac for 5 channel but not for the .1 subwoofer channel.
So my biggest doubt is that when using DTS HD Master Audio format does it really takes 24 bit?
For CDs im still happy with 16 or 18 bit ladder dac which is somewhat easily achievable but the prob comes with 24 channel...
Regarding the parallel DAC operation is it like just plain paralleling of the 16 bits DACs?
Im thinking of achieving 18 bit and then go upto 21bit that should be generally sufficient I believe what do you say?
My understanding is, they used technology from Fantastic Voyage to hire and miniaturize ex-parochial school nuns. They sit inside the DAC, and if one of the resistors makes noise, they hit it with a calibrated audio ruler. This keeps the resistors very quiet.
I think the actual S/N was around -136 or something. Probably far beyond any recording in existence.
Hey, now I'm wondering…if a DAC makes noise in the forest, and nobody is there to hear it, is it really making noise?
I'll tell you you're mad for sure but then if you're not mad why are you on DIYA? Its 'Projects by fanatics, for fanatics' isn't it ?
I'm with you 100%, just as mad as you are. I'd even use a multibit DAC for the bass.
I think its marketing - no ADC exists even remotely close to 24bits in SNR.
No, it needs some DSP to work out how to allocate all the bits below the 16th. Paralleling 256 16bit DACs just gives one 16bit DAC with 256X the output current. Really the number of bits you need is determined by what noise level you're targeting at the listening position vs the peak SPL you want. Work backwards from that to determine the required SNR of your DAC. My guess is you won't need more than 20bits tops with the quietest possible listening room.
In most cases the answer is no. With a nun based DAC this condition is only possible if a sufficient number of deaf nuns can be located, which is difficult. A more general answer to the OP's question is only one 16 bit DAC is required to implement a 24 bit DAC since all that's needed is to multiply the output of an 8 bit DAC with the 16 bit DAC. The TDA1545's full scale output current is 1mA typ, so the multiplication results in a full scale 24 bit output of 256mA. The DAC can't make noise without some kind of assistance, so let us assume an 8 ohm compression driver attached to a 103 dB SPL W/m horn is used for I/V. Full scale output power is 500mW, so one must be approximately 64 km away from the DAC before its sound falls below 0dB SPL 2uPa and becomes inaudible. Sufficient sparsely populated forests exist to satisfy this condition. So it follows by definition from the nature of the question it's not possible to observe the DAC and thereby collapse its wave function into a definitive state of noise making.if a DAC makes noise in the forest, and nobody is there to hear it, is it really making noise?
Unfortunately this arrangement is highly susceptible to user error as it's easy to mislocate the required 128 km diameter circle such that someone could end up within it and potentially hear the DAC, thereby spoiling the carefully planned arrangement. The best solution I'm aware of for this is to operate the least efficient driver you can find in a nude dipole configuration to maximize its self cancellation. This is especially effective in mitigating those undesirable bass frequencies which propagate well over landscape. I would suggest looking carefully at 8 inch subs as these often roll off quickly in the highs and can be found with passband efficiencies around 80 dB SPL W/m, meaning 20Hz efficiencies below 40 dB SPL W/m are possible. It's also preferable to choose dense forest as this maximizes sound attenuation with distance from the DAC. With proper design I believe circle diameters well under 2 km should be achievable.
I'll tell you you're mad for sure but then if you're not mad why are you on DIYA? Its 'Projects by fanatics, for fanatics' isn't it ?
I'm with you 100%, just as mad as you are. I'd even use a multibit DAC for the bass.
I think its marketing - no ADC exists even remotely close to 24bits in SNR.
No, it needs some DSP to work out how to allocate all the bits below the 16th. Paralleling 256 16bit DACs just gives one 16bit DAC with 256X the output current. Really the number of bits you need is determined by what noise level you're targeting at the listening position vs the peak SPL you want. Work backwards from that to determine the required SNR of your DAC. My guess is you won't need more than 20bits tops with the quietest possible listening room.
ok then lets consider we are happy to use 16 bit ladder dac and what happens if we feed 24bit DTS HD master Audio data into the 16bit dac does it work? Or it gives any pops while playing so that the extra bits above 16 end up not processed?
As i have read that if you look even the best Burr Brown pcm1792 might not come close to a 8 bit ladder dac quality so I am very happy using a 16 bit all I have to know is that in order to pump in the 24 bit data do we need a 24bit to 16 bit conversion to be done? if so how do we do it? or is it not really required?
24bits down to 16 is a truncation. Truncations are non-linear operations so the way to deal with it is to dither at the LSB with TPDF (the addition of two uncorrelated random numbers linearly distributed between -0.5 and 0.5) dither.
Certainly the S-D parts have fairly crummy (in terms of element matching) low-bit DACs inside them. They get around this shortcoming by doing DSP which turns the element mismatches into (shaped) noise.
Certainly the S-D parts have fairly crummy (in terms of element matching) low-bit DACs inside them. They get around this shortcoming by doing DSP which turns the element mismatches into (shaped) noise.
Is this to be a digital filter-less 16-bit DAC system? If there is to be a digital interpolation filter located in the system, then it very likely provides for 24-bit to 16-bit format conversion, including generating and applying the appropriate dither to remove truncation distortion.
Also, do you recall where you read that the PCM1792 does not come close to the quality of an 8-bit ladder DAC? While the PCM1792 contains a 5-bit native converter unit, sigma-delta operation lowers it's in-band noise and distortion far, far beneath that produced by an 8-bit ladder DAC. However, as Abraxilito will readily tell you, out-of-band quantization noise will be higher at some ultrasonic frequency band. The impact of such noise is a hot topic of debate, particularly with 1-bit, where such noise is most severe.
Last edited:
well Im thinking of using this way that use a Xonar HDAV 1.3 Slim card which has HDMI output with PCM but we can force the card to output only 16bit 96khz output from the card so that we can use ladder dac for the output? Asus' Xonar HDAV (And Xonar HDAV Slim) - The HTPC / Windows 7 Chronicles: You Asked, We Answer!
what do you say?
what do you say?
Is this DAC ok to use are the Resistor values of 1ohm and 2ohm are correct?
http://www.sonicillusions.co.uk/images/dac18.png
http://www.sonicillusions.co.uk/images/dac18.png
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.