A number of years in the past, I printed a design thought (See Determine 1: “Cancel PWM DAC ripple with analog subtraction,” November 28, 2017) displaying a easy trick for the cancellation of PWM output ripple. It employs passive summation of the same old PWM sign with its AC-coupled inverse, thus dramatically attenuating the undesired AC (ripple) element with out affecting the needed DC element.
Determine 1 Cancel PWM DAC ripple with analog subtraction.
Fortunately, “Cancel…” was properly acquired typically. It even earned just a few pages masking it and a few terrific refinements by authors Winfield Hill and Paul Horowitz within the well-liked “The Artwork of Electronics: The x-Chapters, 4x.25 Ripple Discount in PWM,” pp383-385.
One in every of these TAoE enhancements addressed the truth that PWM DAC accuracy tends to be restricted by dependence on digital logic energy provides which, as H&H aptly level out, “are notoriously inaccurate.” Their intelligent repair (see Determine 2) is so as to add precision analog switches and an correct voltage reference to the same old PWM topology, thus isolating the DAC output from logic provide inaccuracy. One swap generates a precision model of the same old PWM waveform whereas the opposite generates its inverse. Including the AC-coupled element of the latter to the DC of the previous performs the ripple-canceling analog subtraction as described in my 2015 design thought (DI).
Determine 2 Sketch of TAoE PWM accuracy enchancment idea
That is in fact a chic, completely workable, and efficient resolution to the PWM accuracy downside.
Nonetheless, since two analog swap packages are required for every DAC, if a number of channels of PWM DA conversion are wanted the bundle depend could turn out to be inconveniently massive. Additionally, because the variety of channels will increase, so does the overall loading positioned upon Vref, which, relying on the reference gadget used, could ultimately degrade the accuracy of the reference voltage.
A easy variation on Determine 2 is subsequently proven in Determine 3, by which the PWM logic management sign is used to derive the ripple-cancellation sign, thus saving an analog swap and chopping by half Vref loading. This requires scaling the R2C2 ripple-subtraction parts by an element X = Vlogic/Vref, which in fact will solely work in addition to X is secure. Notice additionally that the PWM setpoint should be 1’s complemented (V = -V – 1), which is well finished in software program.
Determine 3 PWM logic sign used for Ripple Cancellation, thus saving a swap.
In case Vlogic isn’t secure sufficient for Determine 3 to work, Determine 4 reveals one more completely different implementation of the TAoE accuracy-promoting idea that makes use of a technique of exactly servoing pulse space (amplitude x period x frequency) to an correct reference voltage, as proven in one other latest DI within the context of VFC analog to digital conversion.
Determine 4 Q2 pulse space servoed to voltage reference.
Determine 4 depends on the flexibility of adjustable voltage references just like the TL431 to behave as analog comparators/integrators with correct built-in thresholds, as proven in Determine 5.
Determine 5 LT431 as precision comparator/distinction amplifier with built-in Vref (from LT431 datasheet).
The LT431 (U1) integrates the typical distinction of U2 pin2 to its inner 2.50V reference and forces equality by growing Q2 pulse period when Fclk x pulse period x 5V = pulse space is lower than 2.5V, and reducing it when larger than 2.5. R1C4 units the time fixed of the combination, R4 gives compensation for U1’s enter bias present (~1.8µA), and R5 gives pullup for U1’s output.
U2’s inherent inner gadget matching (propagation and transition instances, and many others.) trigger the opposite three bits to precisely observe the flip-flop inside U2’s suggestions loop, thus offering correct calibration for all three PWM channels. The related RC networks implement ripple cancelation, and if all three are populated, present three correct DACs from solely two gadget packages.
After all, if fewer than three channels are wanted, not all of the ripple cancellation networks must be populated.
Stephen Woodward’s relationship with EDN’s DI column goes again fairly a methods. In all, a complete of 64 submissions have been accepted since his first contribution was printed in 1974.
Associated Content material
