Comments on digitally-controlled LDO

The digitally-controlled LDO has been receiving significant attentions in recent years; in 2017 ISSCC, there is a special session focusing on the digitally-controlled LDO.

The presented papers in 2017 ISSCC reveal the advantage of digitally-controlled LDO in low-voltage low-power applications. It is interesting to note that most of the said papers are from academia.

My questions are,
1. Has digitally-controlled LDO been accepted/recognized by industrial? if not, what is the real practice in state-of-the-art industrial products?
2. For low-voltage low-power applications, particularly for the noise-insensitive digital circuit therein, is there a real need for the LDO design to migrate from the conventional analog domain to the emerging digital domain?

To my understanding, though conventional analog LDO does not lend itself to the low-voltage applications, the issue can be resolved. For an example, by locally boosting the supply of the control circuit.

Hi,

I´m not sure what you are talking about.

It seems you more talk about switch mode regulators vs linear regulators.

Besides this there are papers, discussing:
* LDOs with digital interface (I2C fro example) to set the output voltage
* something like "voltage regulator offering digitally controllable dynamic voltage scaling (DVS) for near/sub-threshold applications"

Could you please clarify?

Klaus

Sorry for any confusion caused.

I'm not talking about switch mode regulators vs linear regulators. Instead, I'm talking two categories of linear regulators.
"A 100nA-to-2mA successive-approximation digital LDO with PD compensation and sub-LSB duty control achieving a 15.1ns response time at 0.5V" (ISSCC 2017) is a good example of the digitally-controlled LDO I've mentioned.

My question is,
Has the digitally-controlled LDO been adopted in any mature product? If not, why?

When I used to work for a large power supply company over 10 years ago, Intel had already produced a bus specification for controlling the power supply which supplied the microcontroller's core voltage.

Core voltages vary between 0.8 and 2.7 volts if I remember correctly, and are adjusted continuously depending on the processor's workload.
The most demanding specification, besides an extremely tight regulation for a unit supplying 50 to 100 amps, was the loop bandwidth required to switch between voltages extremely quickly.

So yes, digitally controlled supplies have been around for a while.

If you can create a digitally controlled power supply IC, I believe there will be so much customer.Especially amateurs even small companies..
For instance, 0-30 V and 0-3A digitally controlled IC regulator...Sounds good huh ?

JZJIANG said:

My question is,
Has the digitally-controlled LDO been adopted in any mature product? If not, why?

Click to expand...

DC-LDOs are very immature itselfs (several papers through ca 4 years) so cannot be a part of any mature design.

The digital scheme with the SAR would have to be clocked
and then the LDO becomes noisy (or noisier). If the application
is a digital load that makes plenty of its own noise, then fine.
But many LDO applications are for stuff that needs to be
very quiet (like RF, and low jitter clock source PLLs, etc.).
You might be able to filter it inside the package with due
care in design.

PMOS LDOs aren't good for very low input voltages, but
ULDOs (with an auxiliary housekeeping supply, say 5V)
and a low voltage high current supply (say, 1.1V to make
1.00 clean) are now somewhat common.

Digital control loops can (not to say, always will) be a
way to get faster load-step response. But so is using a
faster technology to do analog loops. The big reason
for popularity of the field is the immaturity of the field,
nobody but the Red Rag is going to publish your paper
on Yet Another Linear LDO and even they have their
standards. If "publish or perish" is your job description,
what else are you going to do but work on something
that can get peer reviewed traction?

Thank you very much for sharing your experience, which is really helpful.

I guess you are talking about the application incorporating Dynamic Voltage Scaling (DVS) technique.

However, I'm a little bit confused about the term "digitally controlled supplies" you mentioned. How does the digital control perform, and is it similar to what the papers presented (i.e., performing digital loop regulation and controlling the number of turned-on power transistors)? If so, what is the advantage of the digital control approach compared to the analog control approach?

Though wide loop bandwidth is required, analog loop exploiting advanced technology nodes still can somewhat meet the requirement (see below the response from "dick_freebird"). Further, as digitally-controlled LDO is inferior in terms of noise and PSRR, it seems that analog control approach is more approperiate, even for low-voltage applications incorporating DVS technique.

Looking forward to your advice. Thank you.

Thank you for your insightful vision, particularly your comment "Digital control loops can (not to say, always will) be a way to get faster load-step response. But so is using a faster technology to do analog loops."

I totally agree with your opinion, and that is the reason I post this thread. As a student, I believe ISSCC papers demonstrate the leading technologies in both academia and industrial area. The digitally-controlled LDOs are novel and interesting (from my point of view), however, I'm not sufficiently convinced about their meaning, and what applications they really fit.

It seems conventional analog control approach and its variants are able to accommodate various applications (including the low-voltage applications) without noise penalty.