I am considering adding some dithering to a delta sigma PWM signal. Has anyone done this here, and if so, could they reccomend the best place to add the dither in the loop? I am currently still researching this topic, so if anyone has links or good references about this, please send them my way. _____________________________ Posted through www.DSPRelated.com
adding dither to delta sigma PWM control
Started by ●July 15, 2013
Reply by ●July 15, 20132013-07-15
On 7/15/2013 1:12 PM, jacobfenton wrote:> I am considering adding some dithering to a delta sigma PWM signal.Delta sigma and PWM are two big differences. Delta sigma is linear modulation while PWM isn't.> Has > anyone done this here, and if so, could they reccomend the best place to > add the dither in the loop? I am currently still researching this topic, so > if anyone has links or good references about this, please send them my way.You can dither PWM. However it gets intricate if you do noise shaping. Noise spectrum will be spreaded in some way by PWM nonlinearity. Vladimir Vassilevsky DSP and Mixed Signal Designs www.abvolt.com
Reply by ●July 15, 20132013-07-15
"jacobfenton" <51114@dsprelated> writes:> I am considering adding some dithering to a delta sigma PWM signal. Has > anyone done this here, and if so, could they reccomend the best place to > add the dither in the loop? I am currently still researching this topic, so > if anyone has links or good references about this, please send them my > way.Delta sigma and PWM are two different things. As in all dithering, the right place to add it is just before the quantizer. That goes for a delta sigma DAC as well, so it would be in the modulator loop before the quantizer. See p.20, Figure 11, here: http://www.digitalsignallabs.com/presentation.pdf -- Randy Yates Digital Signal Labs http://www.digitalsignallabs.com
Reply by ●July 15, 20132013-07-15
On Mon, 15 Jul 2013 13:12:40 -0500, jacobfenton wrote:> I am considering adding some dithering to a delta sigma PWM signal. Has > anyone done this here, and if so, could they reccomend the best place to > add the dither in the loop? I am currently still researching this topic, > so if anyone has links or good references about this, please send them > my way.Why? Delta-sigma converters kinda add their own dithering anyway. Are you trying to break up tones, or what? -- Tim Wescott Wescott Design Services http://www.wescottdesign.com
Reply by ●July 15, 20132013-07-15
>On Mon, 15 Jul 2013 13:12:40 -0500, jacobfenton wrote: > >> I am considering adding some dithering to a delta sigma PWM signal. Has >> anyone done this here, and if so, could they reccomend the best placeto>> add the dither in the loop? I am currently still researching thistopic,>> so if anyone has links or good references about this, please send them >> my way. > >Why? Delta-sigma converters kinda add their own dithering anyway. Are >you trying to break up tones, or what? > >-- > >Tim Wescott >Wescott Design Services >http://www.wescottdesign.com > >Sorry for the confusion, I should remove PWM from the post. I am just using a delta sigma. I was considering adding dithering, to reduce the low frequency energy when the loop is running near 100% on. _____________________________ Posted through www.DSPRelated.com
Reply by ●July 15, 20132013-07-15
On Mon, 15 Jul 2013 14:58:50 -0500, jacobfenton wrote:>>On Mon, 15 Jul 2013 13:12:40 -0500, jacobfenton wrote: >> >>> I am considering adding some dithering to a delta sigma PWM signal. >>> Has anyone done this here, and if so, could they reccomend the best >>> place > to >>> add the dither in the loop? I am currently still researching this > topic, >>> so if anyone has links or good references about this, please send them >>> my way. >> >>Why? Delta-sigma converters kinda add their own dithering anyway. Are >>you trying to break up tones, or what? >> >>-- >> >>Tim Wescott Wescott Design Services http://www.wescottdesign.com >> >> > Sorry for the confusion, I should remove PWM from the post. I am just > using a delta sigma. > > I was considering adding dithering, to reduce the low frequency energy > when the loop is running near 100% on.Ah. That may make sense. I've heard (here, in fact) of doing such. I think the dodges are, in order: 1: A sigma-delta with a 1st-order loop filter. Bone-head easy but surprisingly useful, it's the most I've ever done in a real application. 2: A sigma-delta with a 2nd-order loop filter. Makes a much better converter, but you actually have to exercise the brain cells to make it work -- and it's heavier lifting for a sigma-delta in an ISR. 3: A 2nd-order sigma delta with random dither. 4: Higher than 2nd-order sigma delta, with the gains arranged so that the loop is chaotic. Think of it as a sigma-delta with auto-dither. BUT -- I ain't no sigma-delta expert!! -- Tim Wescott Wescott Design Services http://www.wescottdesign.com
Reply by ●July 15, 20132013-07-15
"jacobfenton" <51114@dsprelated> writes:> [...] > I was considering adding dithering, to reduce the low frequency energy when > the loop is running near 100% on.That makes zero sense to me. 100% on implies your target is DC - why would you want to reduce low frequency energy? Like a drag racer reducing fuel flow at max acceleration... -- Randy Yates Digital Signal Labs http://www.digitalsignallabs.com
Reply by ●July 15, 20132013-07-15
>"jacobfenton" <51114@dsprelated> writes: >> [...] >> I was considering adding dithering, to reduce the low frequency energywhen>> the loop is running near 100% on. > >That makes zero sense to me. 100% on implies your target is DC - why >would you want to reduce low frequency energy? Like a drag racer >reducing fuel flow at max acceleration... >-- >Randy Yates >Digital Signal Labs >http://www.digitalsignallabs.com >Not exactly at 100% on, so not totally DC. Above 50% on, the frequecy components of the delta sigma starts to drop, so I was hoping that adding some dither would allow me to spread out the energy a bit more. I may be totally wrong, as I am still investigating all this. I will let you know if I have any positive results. _____________________________ Posted through www.DSPRelated.com
Reply by ●July 15, 20132013-07-15
>On Mon, 15 Jul 2013 14:58:50 -0500, jacobfenton wrote: > >>>On Mon, 15 Jul 2013 13:12:40 -0500, jacobfenton wrote: >>> >>>> I am considering adding some dithering to a delta sigma PWM signal. >>>> Has anyone done this here, and if so, could they reccomend the best >>>> place >> to >>>> add the dither in the loop? I am currently still researching this >> topic, >>>> so if anyone has links or good references about this, please sendthem>>>> my way. >>> >>>Why? Delta-sigma converters kinda add their own dithering anyway. Are >>>you trying to break up tones, or what? >>> >>>-- >>> >>>Tim Wescott Wescott Design Services http://www.wescottdesign.com >>> >>> >> Sorry for the confusion, I should remove PWM from the post. I am just >> using a delta sigma. >> >> I was considering adding dithering, to reduce the low frequency energy >> when the loop is running near 100% on. > >Ah. That may make sense. I've heard (here, in fact) of doing such. I >think the dodges are, in order: > >1: A sigma-delta with a 1st-order loop filter. Bone-head easy but > surprisingly useful, it's the most I've ever done in a real > application. > >2: A sigma-delta with a 2nd-order loop filter. Makes a much better > converter, but you actually have to exercise the brain cells > to make it work -- and it's heavier lifting for a sigma-delta > in an ISR. > >3: A 2nd-order sigma delta with random dither. > >4: Higher than 2nd-order sigma delta, with the gains arranged so > that the loop is chaotic. Think of it as a sigma-delta with > auto-dither. > >BUT -- I ain't no sigma-delta expert!! > >-- > >Tim Wescott >Wescott Design Services >http://www.wescottdesign.com > >Yes, so far the 1st order modulator is so simple and works that its hard to justify spending time analyzing the second order one. _____________________________ Posted through www.DSPRelated.com
Reply by ●July 15, 20132013-07-15
The in-band snr of any sigma-delta converter is reduced as you approach the clipping limits. This is because you end up with long strings of +1's or -1's without many transitions, and therefore there is less opportunity for feedback noise-shaping to work its magic by switching back and forth. Unfortunately this is the "law of the land" and there's nothing to be done about it, other than trying to repeal information theory. If you need more dynamic range you can always add more quantizer bits, or increase the over-sampling ratio, or use a higher-order loop filter. Bob
