Skip to main content

Yay or Nay? A Closer Look at AnDapt’s PMIC On-Demand Technology


Innovations on making product features customizable are recently gaining popularity. Take Andapt for example, a fabless start-up that unveiled its Multi-Rail Power Platform technology for On-Demand PMIC applications a few months back. (read all about it here: Will PMIC On-Demand Replace Catalog Power Devices?) Their online platform, WebAmp, enables the consumer to configure the PMIC based on desired specifications. Fortunately, I got a hands-on experience during the trial period (without the physical board (AmP8DB1) or adaptor (AmpLink)). In my opinion, their GUI is friendly but it lacks a verification method for tuning (i.e. the entered combination of specs). How would we know if it will perform as expected or if there are contradicting indications that yield queer behavior? Also, there is not just one IP available, but many that cater to a differing number of channels and voltage requirements (each with their own price tag).

Every new emerging technology has the potential to overshadow its predecessor. In the case of Andapt’s AmP, it can replace conventional catalog power devices. You may wonder what the advantages and disadvantages are between catalog and on-demand PMICs. Are on-demand PMICs a better choice in terms of economy? How about performance? This will be the focus of the article, aimed at providing a more in-depth appraisal and comparison between the two. However, before sound personal judgment can be passed, it is necessary to elaborate first on the PMIC development process, then AnDapt’s AmP technology.



The IC Development Process


Many may have a bird’s eye view of this. A slab of pure silicon sliced to ultra-thin wafers, then diced for fitting into small IC packages after patterns and structures have been etched. Sounds simple, but when the entire process is laid out, certain pivotal and unelidable procedures will surface. It will be evident that these factors could limit an on-demand solution unless work-arounds are implemented.



Figure 1. Flowchart of a typical IC development process (right).

Every product starts with a request or a need. A customer lists the specifications and the circuit designers meticulously review them. The review gives insights on possible problems, enabling the designers to make preemptive decisions before they happen. After reaching a consensus on all stages, the IC design begins with a draft either from existing schematics or from scratch (which is riskier, more difficult and more demanding of stringent tests/requirements). If the simulation results are okay (breakdown, ESD, etc.), the schematic’s layout follows. The final GDS file undergoes many checking runs too, both visually and through verification utilities. Some common examples of such tests include a D.R.C. (design rule check), an M.S.C. (metal stress check), and an L.V.S. (layout vs. schematic) to name a few. What follows composes most of the price and time in IC development. Manufacture of samples for design verification and evaluation proves costly, but it assures quality and compliance with all the customer’s requirements. Actual testing can take time based on differences between simulation and evaluation results. At worst, the process can reset back to the circuit design phase. A lot of effort is poured into this stage because slip-ups can trigger a domino effect resulting in an irreversible catastrophe. The final transfer to mass production can also be delayed if anything is amiss.

An on-demand PMIC skips the entire testing stage. This means no data can be used to guarantee the PMICs performance out in the field. Will it oscillate given this load? Are there any peculiar runts or spikes at this temperature? Will this voltage trigger a protection feature? In addition, if any of these events happen, the troubleshooter has nothing at hand to start with. In contrast, a conventional PMIC performing strangely can be easily debugged by comparing stark symptoms with evaluation data. Adjustments can be made to the circuit design and re-entered into the IC development process. Ever wondered why there are many series of certain IC products that seem to do the exact same thing?

(Note: AnDapt addresses this by earning industry-standard certifications for its product. However, it only answers the problem to a certain extent because security and assurance are still limited to the degree of testing done.)



AnDapt’s Adaptive Multi-Rail Power Platform ICs


Nitpicking aside, the technology is an impressive feat of engineering. Can you imagine a PMIC that adapts – or should I say – andapts to a required set of electrical characteristics (no pun intended)? There are 9 members for the 12V platform family with a 6A maximum output current cap (not bad). It communicates by SPI or I2C control, interfaced with a USB Type-B to USB Type-A cable and adapter (same as with most PMICs) to the computer. Figure 2 shows a lucid explanation of how the interface works.



Figure 2. Interface diagram extracted from the AmPDB1 demo board datasheet.

From my understanding of what multi-rail is, I’m expecting each input supply to the channels to have OCP enabled. This means if any of the channels are drawing too much current, the chip will automatically turn off.

I don’t have access to a copy of AnDapt’s patent but I can try to predict how the adaptive nature of their IC works. Hopefully, my experience can guide me towards a close assumption.

The channels of commercial PMICs can be subdivided into 3 general classes: operational amplifiers (that usually operate as voltage followers), low-dropout regulators (LDOs), and DC-DC switching converters (buck, boost, buck-boost, inverting, etc.). So, I am going to assume they made 3 of these per channel. For example, one power block (channel) will comprise of an op-amp architecture, another an LDO, and another a DCDC. Each of them are enabled/disabled through the “digital fabric” (possibly with a multiplexer that chooses the output and/or an enable pin).

After all, DCDC switching converters differ only by the arrangement of the capacitor, inductor, and diode (which are external in this case). LDOs simply use a pass transistor (I doubt they changed this, even the latest LDO architecture still uses a pass transistor). Op-amps are simply op-amps (though the feedback loop can have a compensation network). Other particular applications (such as battery chargers) are variants with external components (mayhap a sense resistor for the Coulomb Counter).

Now we have a guess on how AnDapt manages to shift between different kinds of outputs. 

How about the specifications? My hypothesis – lots and lots of trimming networks (and maybe scalable integrated MOSFETs just for the LDOs).

The output voltage level can be adjusted by trimming the reference/bandgap voltage. Poles and zeroes can be adjusted by trimming the compensation networks (given a specific output capacitor – now you wouldn’t wonder why they’re asking for the value of output capacitance in their GUI). UVLO and OCP levels are adjusted by trimming the sampling resistance (actually, with over years of experience with ASICs and PMICs, I can firmly attest that sampling anything almost ALWAYS involves a resistor/another circuit element modified to have resistance of a particular value that will affect the trigger level). Finally, the soft start feature is just a counter interfaced to a DAC, so the DAC is the one most probably trimmed by the digital fabric. 

Hmmm… reverse engineering a technology is always fun even though I could’ve completely missed the target. How about you, what do you think is the structure behind AnDapt’s on-demand PMIC?


AnDapt’s AmP’s Niche in the Market

It isn’t possible to design an on-demand PMIC that can adjust to all kinds of electrical specifications. In fact, if you search family members of a PMIC product, you may find that it requires a lot of pinouts and indications that are significantly different from the original.

Why?

Because most products that PMICs cater to are diverse and unique as well (forgive me if my reiterations of conspicuous facts vex the reader, I just can’t risk losing chain of thought). Take for example PMICs for mobile phones. The power footprint of a smart phone running iOS (Apple) will certainly be different from one running Android (say Samsung). Thus, one can expect the specifications between these two PMICs to be completely alien from each other in spite of them being designed for the same product.

Which is why AnDapt has AmP4D and AmP12D (currently under development in their online platform). It adds more room for functionality but I still have doubts it is enough to address all requirements of today’s applications.

Perhaps on-demand PMICs are best used when low volume and quick-design turnaround are important (just like FPGAs and standard cell libraries in digital applications). Note that FPGAs did not replace its custom IC counterparts because it was too expensive for mass production. The same case may apply to on-demand PMICs.

In your opinion, do on-demand PMICs have the potential to outshine its conventional counterparts?




Perspectives and Comments


The following are consolidated online comments on Andapt’s AmP technology.



Elizabethsimon commented:

Nothing on the roadmap to cover 15V input??? 

Where I work, we have an internally developed supply that delivers 15V to our boards... 

Also, those of us who power things off car batteries would be more interested if you had something that would work from say 10 to 16V



Antedeluvian commented:

Looks intriguing. I poked around on the web site and of course it is all wine and roses, but there are quite a few products that are undescribed (look at battery chargers), so I suppose I will have to wait. 

An irritant is that you need to register before you can even glimpse a data sheet. I really don't want to be a recipient of more data when I don't know if I even want to see that data.


Something to think about. Normally the IP is buried in the cost of a non-configurable chip. I am not sure I understand this model. 

It's a bit "Applish" since you are not allowed to write about the devices without approval. That puts pay to just playing around and blogging about it. 

It also seems that some, if not all of the devices carry a monthly licensing fee (if I understood it correctly). 

I need to wrap my head around this concept- I know it works like that with IP in FPGAs, but LDOs?





Crusty commented:

As Rcurl says this looks interesting. 
As you know I have moved on from Atmel to Cypress PSoC devices like the PSoC 5 this is a great chip as I can have 4 different voltages for the 4 banks of I/O pins. Along with the ability to work on voltages up to 5 volt. This power block had me interested, but thats where it stays after I looked at the pinout of the chip. They are all underneath and will not fit well with my experimentation mode which tends to the use of visible pins when soldering. I can understand the need to have a hearty heat sink area on the base if using lots of amps, but this will not go down too well in the Arduino / Maker market. That is unless someone comes up with a simple block with pins brought out to connectors. 
Will they offer a $10 dev board as Cypress does, this really sells the kit.

Rcurl commented:

This looks really interesting, but it's one of those things that will have to rattle around in my head for a while until I figure out what to do with it. 
How much does the Amplink adapter cost? I'm curious as to why there are so many pins on the interface connector. 
It's hard to believe that you can handle 72 amps continuous in that tiny chip 



utahimagineering_llc commented:

I'd say no. I've looked at this part, talked to them about it and it has its places but power supply design is really about whats the problem your trying to solve. In this case, it has a bunch of logic internally, current sensors, etc so it might save you some space IF you need that. They still have very little silicon actually out as well. For a MP design, such as a systems board, where you need several rails sprinkled around the board, its a possible solution if you are not concerned about board space and efficiency. For my case, I have very little room and need much better efficiency so it simply doesn't play in my space.

What is going to spur a revolution? Switched cap and GaN is going to have a much bigger role going forward."

*I’ve also tried asking Dave Jones from the EEVBlog and he comments he has never heard of Andapt and is not familiar with it (it isn’t that popular yet after all). Maybe it needs to get a bit more traction in the PMIC market.











Comments

Popular posts from this blog

Calculator Techniques for the Casio FX-991ES and FX-991EX Unraveled

In solving engineering problems, one may not have the luxury of time. Most situations demand immediate results. The price of falling behind schedule is costly and demeaning to one's reputation. Therefore, every bit of precaution must be taken to expedite calculations. The following introduces methods to tackle these problems speedily using a Casio calculator FX-991ES and FX-991EX.


►For algebraic problems where you need to find the exact value of a dependent or independent variable, just use the CALC or [ES] Mode 5 functions or [EX] MENU A functions.


►For definite differentiation and integration problems, simply use the d/dx and integral operators in the COMP mode.


►For models that follow the differential equation: dP/dx=kt and models that follow a geometric function(i.e. A*B^x).

[ES]
-Simply go to Mode 3 (STAT) (5)      e^x
-For geometric functions Mode 3 (STAT) 6 A*B^x
-(Why? Because the solution to the D.E. dP/dx=kt is an exponential function e^x.
When we know the boundary con…

What Ifs? Sigmoid Function vs. Error Function in Machine Learning through Logistic Regression

While brushing up on some study materials in Mathematics, a familiar function piqued my interest. It was the error function, the solution to the non-elementary integral exp(-x^2) and whose complement is used in determining the conditional probability of bit error due to noise:






Or quite simply, the probability of error due to noise.




But the real point of interest was the nature of the curve of the function shown below.



Now, why be so interested in such a function? When I compared erf(x) with the sigmoid function commonly used in defining the decision boundary in machine learning algorithms, it returned a steeper slope. Then the thought came to me. What would be the differences of using the error function instead of the sigmoid function? Would the cost improve? Would the training accuracy improve?

And so my curiosity got the better of me and I played around with both of the functions to see what would happen.

Sigmoid vs. Error
First of all, replacing the sigmoid function with the error…

What are all these Nanogenerator stuff, Anyway?

Was there a time when you were introduced to the piezoelectric effect in one of your Physics classes and wondered, “If piezoelectric crystals generate voltage when subjected to vibration, can’t we harness this voltage to power our electronics?” It was a pretty interesting afterthought. What about the voltage developed from the Seebeck effect? There are a lot of naturally occurring temperature gradients in our environment such as the thermal gradient between our body, the engines we use, even our gadgets and ambient temperature. It would feel wasteful to watch all the energy from these potential sources dissipate to the empty void. Apparently, such sources would only yield power just enough for the mobile phone of an ant. But recent developments in materials science as well as improvements in power consumption of modern electronics have aroused interest anew. Thus, in 2006 the first nanogenerator emerged drawing energy through the piezoelectric and semiconductor characteristics of a …