Forums

conditioning a fast current-out DAC

Started by John Larkin December 30, 2016
Fast DACs usually have differential current outputs. Properly
converting that to a clean single-ended signal is a little tricky.

This looks good:

https://dl.dropboxusercontent.com/u/53724080/Circuits/Fast_Stuff/V377_DAC.JPG

The AD8130 is a cool amp. It has two differential inputs and the
output is the sum of the two diff inputs. (One gotcha is that the
inputs are only linear up to 2 volts or so.)

I need maybe 2.5 MHz overall bandwidth, so I started with a 3-pole
2.75 MHz Bessel filter. Then I added C1 to gobble DAC glitches and add
another lowpass pole, which necessitated tweaking the LC filter a bit.
This case is simple enough that a little guesswork and Spicing got the
step response back to looking nice.

This is the situation where we'll grossly over-clock the DAC, 120 MHz
maybe, and dither the data to improve resolution and linearity.


-- 

John Larkin         Highland Technology, Inc
picosecond timing   precision measurement 

jlarkin att highlandtechnology dott com
http://www.highlandtechnology.com

On 12/30/2016 5:34 PM, John Larkin wrote:
> > Fast DACs usually have differential current outputs. Properly > converting that to a clean single-ended signal is a little tricky. > > This looks good: > > https://dl.dropboxusercontent.com/u/53724080/Circuits/Fast_Stuff/V377_DAC.JPG > > The AD8130 is a cool amp. It has two differential inputs and the > output is the sum of the two diff inputs. (One gotcha is that the > inputs are only linear up to 2 volts or so.) > > I need maybe 2.5 MHz overall bandwidth, so I started with a 3-pole > 2.75 MHz Bessel filter. Then I added C1 to gobble DAC glitches and add > another lowpass pole, which necessitated tweaking the LC filter a bit. > This case is simple enough that a little guesswork and Spicing got the > step response back to looking nice. > > This is the situation where we'll grossly over-clock the DAC, 120 MHz > maybe, and dither the data to improve resolution and linearity. > >
It looks like a PI filter to me. Are PI and Bessel filters the same?
On Sat, 31 Dec 2016 09:56:10 -0600, John S <Sophi.2@invalid.org>
wrote:

>On 12/30/2016 5:34 PM, John Larkin wrote: >> >> Fast DACs usually have differential current outputs. Properly >> converting that to a clean single-ended signal is a little tricky. >> >> This looks good: >> >> https://dl.dropboxusercontent.com/u/53724080/Circuits/Fast_Stuff/V377_DAC.JPG >> >> The AD8130 is a cool amp. It has two differential inputs and the >> output is the sum of the two diff inputs. (One gotcha is that the >> inputs are only linear up to 2 volts or so.) >> >> I need maybe 2.5 MHz overall bandwidth, so I started with a 3-pole >> 2.75 MHz Bessel filter. Then I added C1 to gobble DAC glitches and add >> another lowpass pole, which necessitated tweaking the LC filter a bit. >> This case is simple enough that a little guesswork and Spicing got the >> step response back to looking nice. >> >> This is the situation where we'll grossly over-clock the DAC, 120 MHz >> maybe, and dither the data to improve resolution and linearity. >> >> > >It looks like a PI filter to me. Are PI and Bessel filters the same?
Pi is the topology. Bessel is the frequency/step response. Single-ended LC lowpass filters are usually of the form llll---llll---llll C C C C and can have all sorts of different responses, depending on the values. Bessel is good for waveform generators. It has a clean step response, no overshoot or ringing. We'll be doing a bunch of digital filtering too, before the data gets to the DACs. -- John Larkin Highland Technology, Inc lunatic fringe electronics
John Larkin wrote...
> > Fast DACs usually have differential current outputs. Properly > converting that to a clean single-ended signal is a little tricky. > This looks good: >https://dl.dropboxusercontent.com/u/53724080/Circuits/Fast_Stuff/V377_DAC.JPG > > The AD8130 is a cool amp. It has two differential inputs > and the output is the sum of the two diff inputs.
Yes, it's an interesting design concept, with two cancelling gm inputs, one can do nice things. We feature it doing cool stuff in AoE III, e.g. see Figure 5.71, page 353. And it has excellent CMRR, Table 5.7 directs you to trace 12 in Figure 5.73. But gm stages suffer from an intrinsically high distortion, made acceptable only by the cancelling action. Not a problem for many applications, like the CAT5 cable equalizer in our figure. There's an amplifier type that's better: the fully- differential amplifiers, Chapter 5, section 5.17, and Table 5.10 Many of these have distortion below -80dB and to even -105dB at 10MHz, see Figure 5.106. Compare this to -74dB for the AD8130, at 1V in. Many will happily take your DAC's current outputs directly at their differential summing junctions, arguably a better way to deal with current DACs. -- Thanks, - Win
On 31 Dec 2016 09:42:54 -0800, Winfield Hill
<hill@rowland.harvard.edu> wrote:

>John Larkin wrote... >> >> Fast DACs usually have differential current outputs. Properly >> converting that to a clean single-ended signal is a little tricky. >> This looks good: >>https://dl.dropboxusercontent.com/u/53724080/Circuits/Fast_Stuff/V377_DAC.JPG >> >> The AD8130 is a cool amp. It has two differential inputs >> and the output is the sum of the two diff inputs. > > Yes, it's an interesting design concept, with two > cancelling gm inputs, one can do nice things. We > feature it doing cool stuff in AoE III, e.g. see > Figure 5.71, page 353. And it has excellent CMRR, > Table 5.7 directs you to trace 12 in Figure 5.73. > > But gm stages suffer from an intrinsically high > distortion, made acceptable only by the cancelling > action. Not a problem for many applications, like > the CAT5 cable equalizer in our figure. > > There's an amplifier type that's better: the fully- > differential amplifiers, Chapter 5, section 5.17, > and Table 5.10 Many of these have distortion below > -80dB and to even -105dB at 10MHz, see Figure 5.106. > Compare this to -74dB for the AD8130, at 1V in.
-74 is fine for my arb. It will be used mostly for simulating signals from rotating machines. The output amp will have more THD than the 8130.
> > Many will happily take your DAC's current outputs > directly at their differential summing junctions, > arguably a better way to deal with current DACs.
I don't want differential gain, I want clean diff-to-single-ended conversion. We've done that with fast opamps, but the AD8130 circuit has low parts count and it's symmetric and elegant. Lots of 100 ohm resistors and 150 pF caps. It's sort of a sport here to keep the BOM line count down, to minimize pick-and-place feeders. I may as well make the output resistors 100r 0805s like the rest of the 100's. An 0805 with good copper pours will dissipate as much heat as a 1206. -- John Larkin Highland Technology, Inc lunatic fringe electronics