High resolution VCO-based ADC
View Patent ↗An analog to digital conversion (ADC) circuit includes a voltage-controlled oscillator (VCO)-based quantizer that receives a voltage input signal to be quantized and provides a digital output. A predictor samples the digital output, evaluates correlation between successive samples, and predicts a predicted input sample from the correlation to minimize voltage-to-frequency transfer of the VCO. A feedback loop L1 with a digital to analog converter (DAC) receives the predicted input sample, converts it and subtracts it from the voltage input signal. A feedback loop L2 adds the predicted sample to the digital output.
1. An analog to digital conversion (ADC) circuit comprising:
a voltage-controlled oscillator (VCO)-based quantizer that receives a voltage input signal to be quantized and provides a digital output;
a predictor that samples the digital output, evaluates correlation between successive samples, and predicts a predicted quantization error correction from the correlation to minimize voltage-to-frequency transfer of the VCO;
a feedback loop L1 with a digital to analog converter (DAC) receiving the predicted quantization error correction, converting it and adding it to the voltage input signal; and
a feedback loop L2 to add the predicted quantization error to the digital output.
2. The ADC circuit of claim 1 , wherein a z-transform function of the predictor is 2−z −1 .
3. The ADC of claim 1 , comprising gain error correction logic (GEC) in feedback loop L2 to equalize gain between the feedback loops L1 and L2.
4. The ADC of claim 3 , comprising injection of a two-level (±1) zero mean random dither d into the DAC and the GEC.
5. The ADC of claim 4 , wherein the GEC logic multiplies the dither d with a GEC input u to obtain a product x, wherein the input u consists of:
u=G ADC G DAC d+C
where c comprises components uncorrelated with d.
6. The ADC of claim 5 , wherein the GEC averages the product x to obtain average gain Ĝ, which is the estimate of loop gain G ADC G DAC , and then corrects loop gain by normalizing its value with the estimate via first inverting the estimate and then multiplying input u with the inverse.
7. The ADC of claim 6 , wherein the estimate is provided by least mean square (LMS) engine that updates the loop gain value and a Newton-Raphson's Reciprocal solver to compute the inverse.
8. The ADC of claim 1 , wherein the predicted quantization error correction is truncated by truncation logic before input to the DAC.
9. The ADC of claim 8 , wherein the truncation logic comprises a right-shift followed by a left-shift in the feedback loop L2 to reduce input bit-width to the DAC.
10. The ADC of claim 9 , wherein each of the pair of VCOs are implemented as two stages with a first stage having a current-reuse structure and a second stage that is a source degenerated common source.
11. The ADC of claim 8 , wherein the truncation logic comprises a noise shaping ΔΣ truncator.
12. The ADC of claim 11 , wherein the noise shaping ΔΣ truncator high-pass shapes truncation error.
13. The ADC of claim 1 , wherein the VCO-based quantizer comprises a pair of VCOs arranged differentially.
14. The ADC of claim 1 , wherein the VCO-based quantizer comprises a three-stage current-starved ring oscillator with an output of one of three stages being tapped and level shifted by a level shifter to full swing, wherein a left shifter output clocks an edge counter whose output is sampled by a main sampling clock, and the sampled output is differenced to finish quantization.
15. The ADC of claim 14 , comprising gray-encoding to smooth crossing of clock domains between the edge counter output and the main sampling clock.
16. The ADC of claim 1 , wherein the VCO-based quantizer comprises multi-tap phase decoding.
17. The ADC of claim 1 , wherein the DAC comprises a segmented-tree dynamic matching element DAC.
18. The ADC of claim 1 , comprising a split-ADC architecture with two identical sub-ADCs converting the same input signal, x, forming two channels A and B and a dither generator producing a two-level 1st-order highpass-shaped sequence, d, with zero mean that is injected into both of the two channels.
19. The ADC of claim 18 , wherein the resulting channel A and B output, y A and y B , can be approximated by:
y A =G A ·x+P A1 ·d,
y B =G B ·x−P B1 ·d,
where G A is the sub-ADC gain of channel A and P A1 is the P 1 path gain of channel A, similarly for channel B, and the overall ADC output, D OUT , is obtained by summing y A and y B .
20. The ADC of claim 19 , wherein P 1 gain estimation is performed by taking the difference between y A and y B , which is 2P 1 ·d, which multiplication is comprises sign flipping, a P1 gain error, e, is obtained by subtracting 2, a digital accumulator integrates e and generates a gain control word (GCW) to adjust variable gain of the VCO-based quantizer.