1
\$\begingroup\$

I have a 5V power supply that is converted to 3.3V using a buck converter. Now I want a 1.8V supply. What is the industry standard way to feed a 1.8V buck converter? Do we feed it from the 5V supply or from the 3.3V supply?

From an efficiency point of view, I understand that the most efficient way may depend on the loads. Still, on buck converters, the efficiency gets worse, generally, as Vout is further away from Vin. But if I feed the 1.8V converter from the 3.3V one, I already have the power loss of the first one, albeit maybe getting a small boon from the increased current draw if I am on the side of the efficiency curve that favors an increase in current.

Nevertheless, what is generally done in this case? Two buck converters converting from the same supply, or one converting from the other?

\$\endgroup\$
  • 5
    \$\begingroup\$ I don't think there is a standard. We just do what is right for our special case. Maybe your 3.3V regulator is better off not handling the additional load so you can use a smaller one, maybe the case you are making is true and the better choice. I think both are a valid choice. \$\endgroup\$ – Arsenal Sep 12 '17 at 16:08
3
\$\begingroup\$

As you said yourself, there might be a fine balance in terms of overall efficiency. And cost. You gain a bit here, but lose a bit there. If you have curves of efficiency versus load and input voltage, you probably can calculate theoretical optimum. However if the load is variable independently, the optimization space will explode, and some use models would need to be involved. It could be a good semester project for a third-year EE student.

However, there might be one advantage of having 5->3.3>1.8V arrangement. Usually a microprocessor system requires certain power-on sequencing, and usually higher voltage needs to come up first, middle-one next, and lowest one last. In sequential arrangement this sequencing will come naturally, while in parallel arrangement you would need enable inputs and RC delays. But again, the sequencing might have certain special timing, which will be easier to implement with proper RC on enables than in sequential arrangement.

Actually, a better way is to implement complex power rails using a programmable PMIC with multiple outputs. This is the "industry standard" today.

\$\endgroup\$
1
\$\begingroup\$

Usually it's a good idea to keep DC/DC converters separate. Chaining them up may cause various interferences between them, as voltage ripple from one might mess up the one that comes afterwards. Of course, it is easy to avoid this problem by appropriate filtering in between, but this requires testing of the design under all operation conditions to ensure proper function. As 5V is quite low, you are not forced to chain the DC/DC converters. Just build a 5V->3.3V and a 5V->1.8V. Most of the modern chips allow to do proper power sequencing by using the power good outputs and then enable inputs. If you have high output loads (several amps) it is a good idea to use a single, multi-phase DC/DC converter, in order to keep the current ripple on the input low.

\$\endgroup\$
  • \$\begingroup\$ That's what I was thinking, thank you for the additional details. \$\endgroup\$ – rmarques Sep 13 '17 at 9:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.