High Vin to Low Vout Buck converter with high frequency

[dmg]

Hello,

I need to power a mobile communications module (2G/3.5G GSM/HSPA) from an automotive 24V power rail. That means nominal 10V to 32V Vin range, and the converter should go higher to allow for operation during transient at the clamping voltage of the input protection. The module is powered at 3.8V DC nominal and in GSM mode it will consume current in up to 2A pulses when transmitting with 1/8 duty cycle.

The manufacturer also gives specifications for allowable ripple and at each frequency band on the DC suplly rail and voltage drop, and it tells you that if you use a switching regulator it needs to switch at above 600 kHz or else they can't guarantee compliance.

I'm constrained in PCB area and the design should be low-ish cost. The problem is that the requirements are contradictory. Most regulators capable of the required Vin range won't work because the application requires too low Ton for the switch, and out of those claiming that they can do the job because they use topologies optimized for this, most come from manufacurers like allegro that don't provide data on ripple/transient reponse nor any simulation model nor any way to evaluate how they work apart from buying the eval board and tinkering with it.

So, due to the lack of options availabe for the job I thought about cascading switching regulators to reduce Vin to Vout in two stages and allow for more relaxed duty cycle and Ton requirements but the area constrainment makes me not like this option due to extra space required for two converters, and I've also never cascaded two switchmode regulators operating at similar frequencies and for what I've read it's not that immediate to assess the performance of both of them working in this configuration.

Could you recommend me a topology for this application?

[Benta]

No suggestion, but note that the regulator should survive an 80 V load dump, as well as reverse polarity.

[danadak]

Micrel / Microchip, TI make a number of fast parts -


http://www.microchip.com/design-centers/analog/power-management/overview

http://www.ti.com/lsds/ti/power-management/buck-converter-integrated-switch-products.page



Regards, Dana.

[MagicSmoker]

...
The manufacturer also gives specifications for allowable ripple and at each frequency band on the DC suplly rail and voltage drop, and it tells you that if you use a switching regulator it needs to switch at above 600 kHz or else they can't guarantee compliance.
...
Could you recommend me a topology for this application?

If you need really low ripple and can trade off some efficiency to get it then a buck converter followed by a linear regulator will get you done.

Otherwise, you can reduce ripple in a buck fairly well by simply increasing the inductance above the usual recommendation of targeting 40% peak to peak ripple current at full load/max Vin (lower ripple current into the output capacitor means lower ripple voltage), and, of course, increasing the output capacitance, and/or using a small LC post-filter will help as well.

What is the specific ripple tolerance (Vpp) allowed, and what kind of transient response is required (ie - what is the maximum step change in load current and over what period of time does it occur)?

[David Hess]

If you used a tapped buck inductor, then the duty cycle requirements for a high step down ratio are relaxed.  Check out the LT1074 datasheet for an example.

[dmg]

...
The manufacturer also gives specifications for allowable ripple and at each frequency band on the DC suplly rail and voltage drop, and it tells you that if you use a switching regulator it needs to switch at above 600 kHz or else they can't guarantee compliance.
...
Could you recommend me a topology for this application?

If you need really low ripple and can trade off some efficiency to get it then a buck converter followed by a linear regulator will get you done.

Otherwise, you can reduce ripple in a buck fairly well by simply increasing the inductance above the usual recommendation of targeting 40% peak to peak ripple current at full load/max Vin (lower ripple current into the output capacitor means lower ripple voltage), and, of course, increasing the output capacitance, and/or using a small LC post-filter will help as well.

What is the specific ripple tolerance (Vpp) allowed, and what kind of transient response is required (ie - what is the maximum step change in load current and over what period of time does it occur)?

"
Yes, after evaluating some more my options that's the way I think I'm going, and it might even "improve" my efficiency in this situation. When calculating losses for marginally capable-for-the-job buck converters of normal topologies the switching losses go up to the atmosphere, the power dissipation on the swich rises and the efficiency goes down the drain to a 40-ish to 50-ish % at average load. Yes, it's that bad.

So, what I'm going to do is use a buck converter to go from VBat to say 5V switching at low-ish frequency (100-200 kHz) that relaxes Ton requirements for the converter and then eat the resulting ripple with a high PSRR linear regulator. The problem is now to find a suitable linear regulator. I've found some that meet the peak current requirement, have 40+ dB PSRR at 100kHz and keep it high up to 1 MHz and beyond and are reasonably priced, but only available on large packages. I know several models of high PSRR regulators that I normally use as post-regulators for powering RF stuff but those are either too expensive or don't go past 1A. Do you happen to know any good candidate? It has to supply 2A peaks with 1/8 dutycyles in some situations, but the rest of the time it won't be asked for more than 200-300 mA.

[MagicSmoker]

Not to state the obvious, but I would look for LDO regulators that are specifically intended for powering GSM HSPAs because these do, indeed, have a peculiar and demanding load profile so there will certainly be regulators aimed at the task. One possible candidate is Exar XRP6272, but check with various GSM suppliers to see if they have something more specific.

Failing all else, you could always make a wide bandwidth LDO linear regulator with a MOSFET and high speed op-amp, as you have a higher supply voltage available to drive the MOSFET gate.

[dmg]

Not to state the obvious, but I would look for LDO regulators that are specifically intended for powering GSM HSPAs because these do, indeed, have a peculiar and demanding load profile so there will certainly be regulators aimed at the task. One possible candidate is Exar XRP6272, but check with various GSM suppliers to see if they have something more specific.

Failing all else, you could always make a wide bandwidth LDO linear regulator with a MOSFET and high speed op-amp, as you have a higher supply voltage available to drive the MOSFET gate.

Thanks! That Exar part slipped through my search process and it is a perfect match for my application.