As someone who works in embedded/digital design, I don't quite get this post. Can someone explain if I'm missing something obvious here?
- There's a huge range of great lab supplies between random $50 bulk regulator and a 10k supply from R&S/Keithley/Keysight which have OVP, OCP, SCPI, and other monitoring.
- The preface would probably be better posed if it were comparing a low/medium cost lab supply with load, against a higher end supply.
- There was no discussion on performance or capability of this approach, i.e. the load isn't going to be responsive enough to help reject any CM/DM noise, and would likely add some of it's own.
- If you really need to monitor the DUT, then the integrated monitoring in these kinds of supplies/loads are not going to be up to the task, and again, low-cost 'correct' tools will massively outperform this 'hack'.
- One feature that low-cost loads do hold over similarly priced lab PSU's is the ability to program relatively high speed transient tests, but this wasn't discussed. The obvious follow-on would then be trying to justify the lack of waveform control over using a cheap sig-gen.
- What class of security/reverse engineering tasks call for 4-quadrant SMU?
Also, if you're doing this for 'precision at low cost', shouldn't you be putting the sense leads on the DUT rather than the supply?
I'm the author of the post, and one of my colleagues alerted me to your questions. I'll try and address them the best I can:
Re: Range of Lab Supplies: Absolutely, there's a wide range of lab supplies available. The post emphasizes a cost-effective strategy for scenarios where budgets are tight or where flexibility is needed beyond what mid-range supplies offer. Even in a professional environment, access to "That" power supply might impossible due to other people using it, or maybe you need a procurement that works with your company's budget and timelines. In our line of work, we're generally powering equipment up that may not have come with a power supply, so we generally expect well-behaved hardware on the down-stream side. We've been pretty successfull with mail-order $30 high-power supplies and the goal here was to offer a little more protection without applying the same protection requirements to each supply we procure. In the end, it saves our customers money and allows us to be incredibly competive in the marketplace. Hopefully that clears up the preface context.
Re: Performance and Capability: The focus here was more on cost-effectiveness and versatility. While acknowledging the load's limitations in noise rejection, it's a trade-off for a broader range of testing capabilities at a lower cost. Also in this trade is the general expectation for a pretty well-behaved load/DUT. For specific noise-sensitive applications, additional filtering or more specialized equipment would indeed be necessary. In our industry we often use downstream filtering to gain more insight on what the hardware is doing, which allows us to extract a notion of state. This aids the reversal and VR process substantially.
Re: DUT Monitoring, Transient Tests, and Waveform Control--again, we're not interested in testing the performance of the supply, though it is a nice feature to use to do a quick burnin/smoke-test when we do receive them. Luckily the loads we typically investigate are well-behaved, production tested electronics--and if we're really lucky, we can break as many as we want (and sometimes we do!).
Re: SMU's: SMUs in RE/VR can be a huge asset to have when trying to gain introspection or validating/testing side-channel analysis. We could easily write an entire blog post on how SMUs enable our mission. One of my personal favorites is using them to I/V characterize unknown component pins to undestand construction and possible funciton. On a lot of components/processors, the physical output stage has a glaring difference on input vs output pins, and both differ from bidir's as well. We also have a bit of IP related to this work that helps us with the competitive edge.
I definitely could have elaborated more on why we choose to do 4-wire sense at the supply. With an electronic load, it doesn't actually do any supply or load regulation. It only uses the sense wires as a trip-point monitor for OVP/UVLO. Using it as a pass-through, we leave it wide open to be lowest impedance as possible. Since we are way more worried about the OVP condition with these cheap alibaba supplies, it's slightly safer to monitor as close to the source as possible. UVLO isn't a huge problem becuase at that point, we'll start seeing signs of it in the RE/VR process.
Your points are well-taken, and thanks for asking! Please lmk if there are follow-up questions.
- There's a huge range of great lab supplies between random $50 bulk regulator and a 10k supply from R&S/Keithley/Keysight which have OVP, OCP, SCPI, and other monitoring.
- The preface would probably be better posed if it were comparing a low/medium cost lab supply with load, against a higher end supply.
- There was no discussion on performance or capability of this approach, i.e. the load isn't going to be responsive enough to help reject any CM/DM noise, and would likely add some of it's own.
- If you really need to monitor the DUT, then the integrated monitoring in these kinds of supplies/loads are not going to be up to the task, and again, low-cost 'correct' tools will massively outperform this 'hack'.
- One feature that low-cost loads do hold over similarly priced lab PSU's is the ability to program relatively high speed transient tests, but this wasn't discussed. The obvious follow-on would then be trying to justify the lack of waveform control over using a cheap sig-gen.
- What class of security/reverse engineering tasks call for 4-quadrant SMU?
Also, if you're doing this for 'precision at low cost', shouldn't you be putting the sense leads on the DUT rather than the supply?