Low power SEPIC SMPS

[cupoftea]

Hi,
We need a SEPIC for 5V 5mA out.
Vin is 35-45V
Non isolated
Fsw....we can be flexible here.

Operating quiescent current should be <400uA.

Digikey is down at the mo' do you know how i can find a simple current mode controller for this?...UCC28C43 has 2.3mA bias current so i need to be able to search on IQ (Quiescent current)...and that means digikey.

 

[BradtheRad]

Since your power needs are small, the classic 2-transistor current regulator should be suitable. With load absent or present it draws less than 100uA. By using PNP type the load can be at 0V ground.

--- Updated Sunday at 2:00 AM ---

Link below runs schematic in Falstad's animated interactive simulator:

tinyurl.com/29cq9la8

 

[FvM]

Why SEPIC?
How about TPS54062 or LMR36500?

 

[cupoftea]

Thanks,
We like the TPS54062 and LMR36500.
But their datasheets do not tell of their "Switching bias current"....so we dare not go with them.

TPS54062

TPS54062 data sheet, product information and support | TI.com

TI’s TPS54062 is a 4.7V to 60V Input, 50mA Synchronous Step-Down Converter with Low IQ. Find parameters, ordering and quality information

www.ti.com

LMR36500

LMR36500 data sheet, product information and support | TI.com

TI’s LMR36500 is a 3-V to 65-V, 50-mA ultra-small synchronous buck converter with 6 μA IQ. Find parameters, ordering and quality information

www.ti.com

So However, in their datasheets it gives only the quiescent current draw for the "non-switching" state.
Often these little Buck controllers comprise switching-BJTs which result in a high operating current when they are switching,
but this is not stated in their datasheet.
This effect can actually be seen with some of the analog.com Buck controllers where
they consume a lot more IQ when they are switching...eg LT1912 etc...this can be seen on the LTspice simulation attached.
The datasheet does not allure to this extra IQ.
For example, the attached LT1912 LTspice simulation shows the LT1912 drawing 9.7mA (average) from the 35V input.
(the output load at 5V is only 5mA!!)
The LT1912 datasheet lists only 600uA for "non-switching" IQ....but that is no good at all when you realise that when switching you end up with a couple of mA's more than that. This is the great deficiency of all datasheets in this matter, as they all do this....from any semico.
A couple of mA's of "switching" bias current would be a killer for our application....a total killer!

LT1912

https://www.analog.com/media/en/technical-documentation/data-sheets/1912fa.pdf

Who knows what the actual "real" switching bias current will be with TPS54062....who knows , it could be several mA's!!

This is why the SEPIC situation, with external FET is so good in our low power application, because when you have an external FET, you can calculate exactly what is your "switching" bias current.
The number of companies i have seen where they have had to scrap 10000's of PCBs due to this is amazing! Its a simple and very easy mistake to make.....because the datasheet is a "bit economical with the truth"(!!!)

So this is why we need a low "switching bias current" current mode controller with an external FET....eg like a UCC28C43 but with only say <400uA or less of bias current...the "switching bias current" doesnt matter here because we will control the switching and the external FET....but trying to find a low switching bias current version of like a UCC28C43 on the web is murder...especially with digikey being down.

The AP63200 gives actual "efficiency curves", and down to 2mA load...so with this case its good.....you can be sure the bias current is not high with the AP63200. There is no way they could get 70% efficiency at 2mA load if there was high "swirtching bias current".
In fact, i've just noticed that TPS54062 does give an efficiency curve for a 5V 1.2mA output...and at 400kHz, it shows that the "switching bias current" must be "less than 400uA". (because its 30% efficient at 5v, 1.2mA load from a 10V input.)

AP63200

https://www.diodes.com/datasheet/download/AP63200-AP63201-AP63203-AP63205.pdf

..The problem is that there are no "<50mA Buck controllers with internal FETs/BJTs" that have low "switching bias current"...as stated, most of them dont even state it.

Also, the main reason to want an alternative to the Buck controllers is because with a SEPIC, we can have an external FET and
actively and significantly damp the switching transitions of the FET so as to reduce EMC problems.

Also, regarding cost, the TPS54062 and LMR36500 are about a $1 each.....and the AP63200 is about $0.25....so even though the TPS54062/LMR36500
are better, we would be forced to go with the AP63200.
This is a nuisance, because even with heavy soft start, the AP63200 will be drawing way more startup current than we would like.
-But we must put up with that.

So yes, we are deciding on a Low cost Buck or an EMC friendly SEPIC at the moment.

But the SEPIC with external FET looks best....though please come forward if you may, with low bias current current mode controllers?...eg like a UCC28C43 but with <400uA of bias current...no semico search engine allows you to search on this!

Since your power needs are small, the classic 2-transistor current regulator should be suitable. With load absent or present it draws less than 100uA.

..Thanks Brad, your words are extremely wise...with a 35V input, and an output of 5V and 5mA, it is very plausible that a linear regulator solution would end up being just as efficient as a low cost SMPS solution. A high cost, customized SMPS solution would win but nobody wants high cost.

 

[betwixt]

Even passing <100uA down a resistor and Zener to bias a series pass transistor should work and be a simple solution.

Brian.

 

[cupoftea]

Since your power needs are small, the classic 2-transistor current regulator should be suitable. With load absent or present it draws less than 100uA.

Thanks, yes, for most of our products we do just that....in fact its a Darlington BJT so its even less than 100uA in the zener.

But for one of the products, we need greater efficiency than this.
As discussed the spec is 35Vin and 5V, 5mA out.

 

[FvM]

AP63200 is 32V Vin,max and doesn't meet the original requirements, only AP66200 offers >= 45V Vin,max. It's also a 2 A buck, but at least with soft start feature.

So However, in their datasheets it gives only the quiescent current draw for the "non-switching" state.
Often these little Buck controllers comprise switching-BJTs which result in a high operating current when they are switching,
but this is not stated in their datasheet.

Efficiency figures give additional info about current consumption in operation.

Component costs can be a reason not to use a dedicated low-power switcher, however knowing your typical designs with dozens of additional parts, I seriously doubt that overall costs of your design will undercut a slim LMR36500 circuit.

I didn't yet experience SEPIC as EMI friendly. I won't consider it unless I need buck-boost function.

 

[cupoftea]

I didn't yet experience SEPIC as EMI friendly.

Thanks, yes, i agree..but with SEPIC it will do step down with only low side drive and sense..and the external fet can be transition damped which massively alleviates EMC...the switching edge is the prime cause of all EMC ills in any SMPS. Damp it well and you kill EMC problems.

We cannot transition damp a buck when its fet is inside the controller.

 

[D.A.(Tony)Stewart]

Why would you want a SEPIC converter when LDO will consume less power?
Your output is only 25 mW.

 

[Easy peasy]

for such low power there are discrete circuits that will give you a SEPIC ( if you absolutely need this topology ) and will automatically drop in freq at light load to give your low quiescent current, 120V power xtor, very simple inductors/caps/diodes - even a cmos 555 control would meet this spec ( we have done these ) and drive the xtor just fine.

 

[D.A.(Tony)Stewart]

What are your reasons for such an illogical requirement?

 

[cupoftea]

There's lots of them and the inefficiency and lost power will add up.

--- Updated Monday at 8:58 PM ---

even a cmos 555 control would meet this spec ( we have done these ) and drive the xtor just fine.

Thanks, i remember you previously kindly gave a schem of a 555 with some regulation circuity around it.

 

[mtwieg]

So to summarize, first you insist datasheets don't show the "real" quiescent loss:

Thanks,
We like the TPS54062 and LMR36500.
But their datasheets do not tell of their "Switching bias current"....so we dare not go with them.
...
Who knows what the actual "real" switching bias current will be with TPS54062....who knows , it could be several mA's!!

And then you admit that they do (via their efficiency curves):

The AP63200 gives actual "efficiency curves", and down to 2mA load...so with this case its good.....you can be sure the bias current is not high with the AP63200. There is no way they could get 70% efficiency at 2mA load if there was high "swirtching bias current".

In fact, i've just noticed that TPS54062 does give an efficiency curve for a 5V 1.2mA output...and at 400kHz, it shows that the "switching bias current" must be "less than 400uA". (because its 30% efficient at 5v, 1.2mA load from a 10V input.)

And then you say this (despite the LMR36500 meeting this description exactly):

..The problem is that there are no "<50mA Buck controllers with internal FETs/BJTs" that have low "switching bias current"...as stated, most of them dont even state it.

Oh but now cost might matter more than efficiency:

Also, regarding cost, the TPS54062 and LMR36500 are about a $1 each.....and the AP63200 is about $0.25....so even though the TPS54062/LMR36500
are better, we would be forced to go with the AP63200.
This is a nuisance, because even with heavy soft start, the AP63200 will be drawing way more startup current than we would like.
-But we must put up with that.

And then you restate your request for the thing someone just pointed you to, as if it never happened:

But the SEPIC with external FET looks best....though please come forward if you may, with low bias current current mode controllers?...eg like a UCC28C43 but with <400uA of bias current...no semico search engine allows you to search on this!

Did I get that right?

--- Updated Today at 2:36 PM ---

Thanks, yes, i agree..but with SEPIC it will do step down with only low side drive and sense..and the external fet can be transition damped which massively alleviates EMC...the switching edge is the prime cause of all EMC ills in any SMPS. Damp it well and you kill EMC problems.

We cannot transition damp a buck when its fet is inside the controller.

I highly doubt a 25mW DCDC is going to be an EMC issue problem.

 

[crutschow]

There's lots of them and the inefficiency and lost power will add up.

How many is "lots"?

 

[betwixt]

Going back to the single transistor solution, worst case idle dissipation if you run 100uA through the Zener is is 100uA x 45V = 4.5mW increasing to ~ 230mW total under full load.

Brian.