Hi everyone, I'm not sure it is totally related to DSP, however I feel that you are the best group to answer y questions me. I have the following noise figure for a 12-bit DAC (AD9762) Noise density = 50 pA/√Hz for FS = 125 MHz and IOUT = 20 mA For the 14-bit DAC of the same family (AD9764), I have Noise density = 50 pA/√Hz for FS = 125 MHz and IOUT = 20 mA First, it seems a bit suspicious that 2 DACs with different resolutions would have the same noise figure. Second, I think that 50 pA/√Hz is a bit low for a 12-bit DAC. If BW = Fs/2 or 62.5 MHz then Noise = 50 pA/√Hz * sqrt(62.5e6) = 395 nA. the "SNR" would be SNR = 20log10(0.7070*20mA/395nA) = 91 dB while a 12-bit DAC should not have a resolution higher than about 74 dB (6.02*12 + 1.76 dB). I'm not 100% sure about my calculations, but there is something weird here I think. What do you guys think? If someone has a good reference\application note on DAC quantization noise, it would be welcome. Regards
DAC output noise interpretation
Started by ●September 30, 2014
Reply by ●September 30, 20142014-09-30
On Tuesday, September 30, 2014 4:12:18 PM UTC-4, benjamin....@gmail.com wrote:> Hi everyone, > > > > I'm not sure it is totally related to DSP, however I feel that you are the best group to answer y questions me. > > > > I have the following noise figure for a 12-bit DAC (AD9762) > > > > Noise density = 50 pA/√Hz for FS = 125 MHz and IOUT = 20 mA > > > > > > For the 14-bit DAC of the same family (AD9764), I have > > > > Noise density = 50 pA/√Hz for FS = 125 MHz and IOUT = 20 mA > > > > > > First, it seems a bit suspicious that 2 DACs with different resolutions would have the same noise figure. > > > > Second, I think that 50 pA/√Hz is a bit low for a 12-bit DAC. > > > > If BW = Fs/2 or 62.5 MHz then > > > > Noise = 50 pA/√Hz * sqrt(62.5e6) = 395 nA. > > > > the "SNR" would be SNR = 20log10(0.7070*20mA/395nA) = 91 dB while a 12-bit DAC should not have a resolution higher than about 74 dB (6.02*12 + 1.76 dB). > > > > I'm not 100% sure about my calculations, but there is something weird here I think. What do you guys think? If someone has a good reference\application note on DAC quantization noise, it would be welcome. > > > > Regardsmy first ___guess___ would be that these noise specs are for the inherent noise in the DAC hardware and not for the quantization noise associated with digitization of the signal. In other words, if you set the DAC to a fixed output code and the output was DC and never changing, you would still have a noise level of 50 pA per root Hz. When you start moving bits around, the quantization noise is additional. Mark
Reply by ●September 30, 20142014-09-30
The noise you are reading about at is analog noise in the current-steering output stage. This is the noise you would measure with the input set to a static digital code. Once you apply a busy digital signal you will see the quantization noise at the level predicted by theory. The analog noise will vary with DC input code as there are both common- mode as well as differential sources of noise. The common- mode components cancel out with a mid- scale input. Typically the designers will attempt to make the analog noise much lower than the quantization noise to give the best performance. Bob
Reply by ●September 30, 20142014-09-30
On Tue, 30 Sep 2014 13:12:18 -0700, benjamin.couillard wrote:> Hi everyone, > > I'm not sure it is totally related to DSP, however I feel that you are > the best group to answer y questions me. > > I have the following noise figure for a 12-bit DAC (AD9762) > > Noise density = 50 pA/√Hz for FS = 125 MHz and IOUT = 20 mA > > > For the 14-bit DAC of the same family (AD9764), I have > > Noise density = 50 pA/√Hz for FS = 125 MHz and IOUT = 20 mA > > > First, it seems a bit suspicious that 2 DACs with different resolutions > would have the same noise figure.You are misusing the term "noise figure". It has a specific meaning, and that meaning is not the level of noise on an output. See http:// en.wikipedia.org/wiki/Noise_figure.> Second, I think that 50 pA/√Hz is a bit low for a 12-bit DAC. > > If BW = Fs/2 or 62.5 MHz then > > Noise = 50 pA/√Hz * sqrt(62.5e6) = 395 nA. > > the "SNR" would be SNR = 20log10(0.7070*20mA/395nA) = 91 dB while a > 12-bit DAC should not have a resolution higher than about 74 dB (6.02*12 > + 1.76 dB). > > I'm not 100% sure about my calculations, but there is something weird > here I think. What do you guys think? If someone has a good > reference\application note on DAC quantization noise, it would be > welcome.The number you are quoting is output noise. That is the noise contributed by the output stage of the DAC, regardless of what the rest of the chip is doing. What it should mean is that if you hold the DAC steady at one output value, the actual output current of the DAC will vary with a noise that has a density of 50 pA per root Hz. The output noise density will be a _part_ of the overall output noise, and will contribute to the overall noise figure. But it won't be the whole of the output noise. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com
Reply by ●September 30, 20142014-09-30
Thanks for the explanations everyone. @Tim : Okay, thanks for the clarification about noise level and noise figure.
Reply by ●September 30, 20142014-09-30
radams2000@gmail.com wrote: (snip)> Typically the designers will attempt to make the analog noise much > lower than the quantization noise to give the best performance.Or make it equal to the quantization noise, and not worry about dithering. -- glen
Reply by ●September 30, 20142014-09-30
Glen In this case it's a DAC so the noise is added at the output and doesn't contribute to dither. Bob
Reply by ●October 1, 20142014-10-01
radams2000@gmail.com wrote:> In this case it's a DAC so the noise is added at the > output and doesn't contribute to dither.Oh, yes. When I first read it I was thinking it was asking about an ADC, then, even though I notice later that it was a DAC, I forgot when I wrote that one. Though it could be that the output goes to an ADC ... -- glen
Reply by ●October 1, 20142014-10-01
On Tue, 30 Sep 2014 13:12:18 -0700, benjamin.couillard wrote: [snip]> First, it seems a bit suspicious that 2 DACs with different resolutions > would have the same noise figure.It would be fair to guess that that entire range of DACs (8 bit through 14 bit) are actually the same silicon, graded for performance. I haven't read all the datasheets, but I anticipate that they all have identical specs except for the linearity (expressed in a number of different ways, e.g. SFDR, DNL) and of course the number of bits. BTW, they're not recommended for new design. Regards, Allan
Reply by ●October 3, 20142014-10-03
>On Tuesday, September 30, 2014 4:12:18 PM UTC-4, benjamin....@gmail.comwro=>te: >> Hi everyone, >>=20 >>=20 >>=20 >> I'm not sure it is totally related to DSP, however I feel that you areth=>e best group to answer y questions me. >>=20 >>=20 >>=20 >> I have the following noise figure for a 12-bit DAC (AD9762) >>=20 >>=20 >>=20 >> Noise density =3D 50 pA/=E2=88=9AHz for FS =3D 125 MHz and IOUT =3D 20mA>>=20 >>=20 >>=20 >>=20 >>=20 >> For the 14-bit DAC of the same family (AD9764), I have >>=20 >>=20 >>=20 >> Noise density =3D 50 pA/=E2=88=9AHz for FS =3D 125 MHz and IOUT =3D 20mA>>=20 >>=20 >>=20 >>=20 >>=20 >> First, it seems a bit suspicious that 2 DACs with different resolutionsw=>ould have the same noise figure. >>=20 >>=20 >>=20 >> Second, I think that 50 pA/=E2=88=9AHz is a bit low for a 12-bitDAC.=20>>=20 >>=20 >>=20 >> If BW =3D Fs/2 or 62.5 MHz then >>=20 >>=20 >>=20 >> Noise =3D 50 pA/=E2=88=9AHz * sqrt(62.5e6) =3D 395 nA. >>=20 >>=20 >>=20 >> the "SNR" would be SNR =3D 20log10(0.7070*20mA/395nA) =3D 91 dB while a1=>2-bit DAC should not have a resolution higher than about 74 dB (6.02*12 +1=>.76 dB). >>=20 >>=20 >>=20 >> I'm not 100% sure about my calculations, but there is something weirdher=>e I think. What do you guys think? If someone has a goodreference\applicat=>ion note on DAC quantization noise, it would be welcome. >>=20 >>=20 >>=20 >> Regards > >my first ___guess___ would be that these noise specs are for the inherentn=>oise in the DAC hardware and not for the quantization noise associatedwith=> digitization of the signal. > >In other words, if you set the DAC to a fixed output code and the outputwa=>s DC and never changing, you would still have a noise level of 50 pA perr=>oot Hz. > >When you start moving bits around, the quantization noise is additional. > >Mark >DAC's don't really produce quantization noise since the discrete-time signal applied to the DAC is already quantized. Only the process that synthesized the discrete-time samples can cause quantization error. The only noise generated by the discrete-time to continuous-time conversion done by a DAC is the noise caused by the devices as you've mentioned ... and DAC non-linearities, and other distortions such as aperture distortion, etc. -Doug _____________________________ Posted through www.DSPRelated.com
