> Microchip have just published a data sheet for a boost converter
> working from a single cell for processors. Not sure if it goes to 5V
> output though. May just be 3.3V.
If it's the MCP1623/24 I found the datasheet easily enough, but
can't find the actual device to either buy or sample, which is a bit
of a tease. And no mention that I can see anywhere of if/when it
will be available
which looks, at least superficially, the same as the MCP1623/24.
Unavailable for sampling but can be purchased June, $0.46/100
(compare with $1.80/100 for ST L6920)
BCCs: Nicish synchronous boost converter for eg uC supply
> > Microchip have just published a data sheet for a boost converter
> > working from a single cell for processors. Not sure if it goes to 5V
> > output though. May just be 3.3V.
Vout is 2 - 5.5 V programmable.
Vin start is 0.65V - 5.5 V
Vin run is >= 0.35V.
Start at 0.65V is only with 1 mA load current so not applicable if you
cannot deload it during startup.
For most purposes Vin >= 1.0 V will suffice.
Claimed "up to typical" efficiency figure is a meaningful but
potentially misleading term.
At 2.5V Vin (and probably 3.3V out) it will deliver > 90% efficiency
at 2ma - 150 mA load.
At 1.2 Vin and ?3.3 Vout? it delivers > 80% at 4-185 mA load.
There are a wide range of Asian sourced parts* which perform a
similar function but which have non synchronous output but which sell
for around $US0.10 in 1000's. Internal and external FET versions are
available - the latter adding a FET cost BUT also allowing essentially
unlimited output current as desired. Efficiency is down on the
synchronous version but not vastly so.
(At 3V3 an output Schottky diode adds notionally say 0V4V/3v3 = 12%
losses so you'd expect it to be well down on ideal. )
(At 5V it's 0V4/5V = 8%.) The synchronous rectifier should be far
closer to ideal if done properly.
Peak FET switch current is about double for the 1640.
The 1623 is PWM only while the 1624 can switch between PWM and PFM
allowing greater efficiency at low loads due to pulse skipping.
Both versions offer "true low disconnect" by turning off the
synchronous output PFET during shutdown. This is often very useful as
it removes the battery voltage which is otherwise present on the load
in a non-synchronous design during converter shutdown (less a
Schottky diode drop). Inability to remove voltage from the output can
be extremely undesirable.
The 1640 has variants which are not available in the 1623/24 which
instead of turning off the PFEt in disable mode, turn it hard on. This
allows the battery to be connected to the output in shutdown. If Vin
is high enough this is most useful.
Eg a system may require > = 3.3 V out and be operated by 3 x NimH
cells. When V p er cell > ~= 1.15 V the battery can be connected to
the output directly and the converter disabled. When Vbat falls too
low the converter is enabled. This mode is not available with the
1623/1624 meaning that it cannot be used with 3 x NimH and a 3.3V
output;. At most 2 x NimH can be used.
Similarly 1 x LiIon cannot be used at 3V3 out but is OK at 5V out.
I haven't compared the 1640 / 1623 specs in detail but it seems like
the 1640 is the superior part unless there are some extra features
waiting to be found (quite possible).
> If it's the MCP1623/24 I found the datasheet easily enough, but
> can't find the actual device to either buy or sample, which is a bit
> of a tease. And no mention that I can see anywhere of if/when it
> will be available
>
> Press release leads to MCP1640
>
> http://www.microchip.com/stellent/idcplg?IdcService=SS_GET_PAGE&nodeId=2018&mcparam=en547321
>
> which looks, at least superficially, the same as the MCP1623/24.
> Unavailable for sampling but can be purchased June, $0.46/100
> (compare with $1.80/100 for ST L6920)
Russell McMahon wrote:
>
> Peak FET switch current is about double for the 1640.
>
> ...
>
> The 1640 has variants which are not available in the 1623/24 which
> instead of turning off the PFEt in disable mode, turn it hard on. This
> allows the battery to be connected to the output in shutdown. If Vin
> is high enough this is most useful.
>
I've ordered 3 free samples of the 1640B as I was looking for something
similar. There is a Linear Technology part I was going to try.
the 1640 seems to come in plain, B, C, D etc, and my understanding is
you either order an turn off OR hard on version of FET?
I think I have ordered the version that turns off.
> [changed subject]
>
>> Microchip have just published a data sheet for a boost converter
>> working from a single cell for processors. Not sure if it goes to 5V
>> output though. May just be 3.3V.
>
> If it's the MCP1623/24 I found the datasheet easily enough, but
> can't find the actual device to either buy or sample, which is a bit
> of a tease. And no mention that I can see anywhere of if/when it
> will be available
>
> Press release leads to MCP1640
yes it was the MCP1623/4 that I was thinking of. I had seen only the
datasheet and not the product page.
> Having my first proper play this evening with a battery booster,
> the ST L6920
>
> I ordered some 10uH inductors from Farnell. The 1077049 to be
> exact, the Murata 22R103C. It's a 6mm x 8mm cylinder, through-
> hole
>
> With 3.8V in / 5V out, the circuit is drawing around 15uA without
> a load from Vout to V
>
> 4k7 (1.07mA load) draws 1.51mA from the battery
> LED + 680R (4.87mA load) draws 6.69mA
> 1k (5mA load) draws 6.96mA
> LED + 1k (9.87mA load) draws 13.65mA
>
> Efficiency is a pretty consistent 72%. I expected > 90%
Take the 5mA case. Input power is 3.8V * 6.96mA = 26.4mw
Output power is 5V * 5mA = 25mW
ivp wrote:
> With 3.8V in / 5V out, the circuit is drawing around 15uA without
> a load from Vout to V
>
> 4k7 (1.07mA load) draws 1.51mA from the battery
> LED + 680R (4.87mA load) draws 6.69mA
> 1k (5mA load) draws 6.96mA
> LED + 1k (9.87mA load) draws 13.65mA
>
> Efficiency is a pretty consistent 72%. I expected > 90%
At these low voltages, the diode drop becomes significant. Is the diode a
schottky? If not, make it so. If so, is this chip capable of syncronous
rectification?
10uH sounds rather low unless you have a fast switching frequency, probably
around 1MHz or at least over 500KHz. Otherwise, the inductor is unlikely
where the lost power is being dissipated.
>> 10uH sounds rather low unless you have a fast switching frequency,
>> probably around 1MHz or at least over 500KHz
No clock per se.
Terminates input ramp at 5 uS or 1A - seamless swap between PFM and
PWM as load varies.
> As pointed out, I missed the very obvious wattage comparison
Energy out per energy in, of which Wattage ratios are equivalent, is
the only meaningful efficiency measure. Current ratios and voltage
ratios depend on the boost voltage ratios and losses and more and so
by themselves give no indication of efficiency. IF you hold other
terms constant then the ratio of say 'output currents at constant Vout
for fixed power in' does give an indication of what's happening, but
this leads to such restricted measurement and setup situations that Vo
x Io / (Vin x Iin) is by far the best measure.
Be sure to measure Vin on the circuit side of any Iin ammeter.
Be aware that high frequency power supply noise on power wiring (in
and out) can play havoc with some meters). Largish cap (100 uF?)
across meter terminals can help.
> Energy out per energy in, of which Wattage ratios are equivalent, is
> the only meaningful efficiency measure.
...
> ... Vo x Io / (Vin x Iin) is by far the best measure.
Pedant's hat on to respond to my own post.
I mean "energy out/Energy" or "Wattage out/Wattage in" (not the
identical thing but equivalent results when interpreted sensibly.
ie by all means measure Vo^2/Rload or Iload^2 xRload if more convenient.
V^2/R is usually easier for a fixed resistive load, and V x I for an
eg LED load.
Note that, when optimising efficiencies or comparing converters you
have to decide whether to include or exclude any losses in sense
resistors and output rectifiers in your calculations. The merit in
including or excluding will depend on what you are trying to measure.
If end to end efficiency is what counts then higher losses in sense
components in one system (eg may use a 1,.2 V reference drop across a
resistor in series with a LED (a very unsensible circuit arrangement
but not unknown)) is fair game for exclusion from the delivered power
calculation. Similarly, when comparing synchronous vs Schottky
rectified systems for overall goodness you would account for the
Schottky losses in efficiency calculations. However, when eg
optimising inductor performance or switching frequency etc to
determine energy conversion efficiency of the converter proper you may
wish to consider the sense resistor as part of the load - it's not the
converter's fault that the designer has chosen to allow stupidly high
losses in the output circuitry, but may skew the results when trying
to optimise some other part of the system.
>
> Be sure to measure Vin on the circuit side of any Iin ammeter.
> Be aware that high frequency power supply noise on power wiring (in
> and out) can play havoc with some meters). Largish cap (100 uF?)
> across meter terminals can help.
>
>
>
> Russell
>
> --
Since the noise most likely is in the 100s of MHz range, a smallish cap perhaps
would be better. A ferrite clamp can also help.
/Ruben
==============================
Ruben Jönsson
AB Liros Electronic
Box 9124, 200 39 Malmö, Sweden
TEL INT +46 40142078
FAX INT +46 40947388 .....rubenKILLspam@spam@pp.sbbs.se
==============================
>
>> Be sure to measure Vin on the circuit side of any Iin ammeter.
>> Be aware that high frequency power supply noise on power wiring (in
>>and out) can play havoc with some meters). Largish cap (100 uF?)
>> across meter terminals can help.
>
>.
>
>> Russell
>
> --
>Since the noise most likely is in the 100s of MHz range, a smallish cap
>perhaps
>would be better. A ferrite clamp can also help.
Yes, Anything low ESR would seem to me be the best option. I think most low
voltage, large caps are (<0.5ohm), but you get an odd one now and then that
can be very high ESR (especially old ones, as they deteriorate with age),
but keeps it's rated capacitance. Maybe a large cap and a small ceramic
would be in order just to be sure. ESR meters are always a handy tool..
> The 1640 has variants which are not available in the 1623/24
> which instead of turning off the PFEt in disable mode, turn it hard
> on. This allows the battery to be connected to the output in
> shutdown. If Vin is high enough this is most useful
Let me see if I understand this. The datasheet says
4.1.4 INPUT BYPASS OPTION
The MCP1640C/D devices incorporate the input bypass shutdown
option. With the EN input pulled low, the output is connected to the
input using the internal P-Channel MOSFET. In this mode, the
current draw from the input (battery) is less than 1 uA with no load
The Input Bypass mode is used when the input voltage range is high
enough for the load to operate in Sleep or low Iq mode. When a
higher regulated output voltage is necessary to operate the application,
the EN input is pulled high enabling the boost converter
Now, I've got a sensor which works from 3.2V to 5.5V. If I
use 3 x alkalines (4.95V when fresh) and set the converter to
eg 3.3V output, the converter can be shut down until battery V
is below 3.3V. Apart from the 1uA quiescent during bypass, the
converter efficiency will not figure in the battery drain until then.
The PIC (12F675) associated with the sensor can periodically
monitor the battery voltage and enable the converter when needed
The figure I haven't seen in the datasheet is the "low Iq" value,
but I suspect it might be 1uA, which isn't enough for the sensor
to run, although is the 1uA just to turn the FET on, with the load
able to take its operating current through the FET ? The block
diagram isn't full enough, and the description flips unclearly
between current and voltage
"A "true" load disconnect mode provides input to output isolation
while disabled by removing the normal boost regulator diode
path from input to output. A bypass mode option connects the
input to the output using the integrated low resistance P-Channel
MOSFET, which provides a low bias keep alive voltage for circuits
operating in Deep Sleep mode. Both options consume less than
1 uA of input current"
"low bias keep alive voltage" ?
The converter consumes 1uA or that's all the converter will
pass for the load ?
2010/5/10 Ruben Jönsson <EraseMErubenspam_OUTTakeThisOuTpp.sbbs.se>:
>>
>> Be sure to measure Vin on the circuit side of any Iin ammeter.
>> Be aware that high frequency power supply noise on power wiring (in
>> and out) can play havoc with some meters). Largish cap (100 uF?)
>> across meter terminals can help.
>>
>> --
>
> Since the noise most likely is in the 100s of MHz range, a smallish cap perhaps
> would be better. A ferrite clamp can also help.
>
Somehow I think you and Russell must be speaking differently
with regard to the meaning of "noise". Maybe you and
Russell can clarify.
> > The 1640 has variants which are not available in the 1623/24
> > which instead of turning off the PFEt in disable mode, turn it hard
> > on. This allows the battery to be connected to the output in
> > shutdown. If Vin is high enough this is most useful
Quick answer to what I think the question is from what I recall the
datasheet said. The following or the correct version thereof is in the
datasheet.
"Pass through when off" version will allow straight through operation
as you describe. It will turn the converter on when the enable line
goes high.
When the enable line is low the device looks like an on PFET in series
with battery. At this stage the device drwas 1 uA. Current able to be
drawn is probably limited by PFET rating. 500 mA or 800 mA or
whatever. See datasheet. PFET needs no drive current, being a FET.
When converter is running but unloaded the current drain is typically
14 uA (from memory). This is perhaps at 3V3 or whatever but will
probably be similar regardless. Any load adds to this.
>> Be sure to measure Vin on the circuit side of any Iin ammeter.
>> Be aware that high frequency power supply noise on power wiring (in
>> and out) can play havoc with some meters). Largish cap (100 uF?)
>> across meter terminals can help.
> Since the noise most likely is in the 100s of MHz range, a smallish cap perhaps
> would be better. A ferrite clamp can also help.
As the smps oscillator typically runs in the 100's of kHz region,
that's where the noise is.
The capacitor value is based on my practical experience of such
situations. In extremis I have been known to parallel eg 100 NF with
an ecap and even use a series inductor or small resistor between meter
and measured circuit, BUT usually an ecap alone works well. Such
filtering relies on forming an RC filter and as the series resistance
involved is in the milliohms range (track to board edge, short leads)
a 10's of uF value tends not to go astray. My practical experience is
that some meters can be badly affected by such noise and that a cap
fixes it to the extent that using any of a wide range of meters then
gives similar results.
Without such a cap, efficiencies of in excess of 100% can readily be
obtained :-).
> When the enable line is low the device looks like an on PFET in
> series with battery. At this stage the device drwas 1 uA. Current
> able to be drawn is probably limited by PFET rating. 500 mA or
> 800 mA or whatever
Yes, that's what I'd like to believe what they mean by "bypass",
and it's common sense to expect that. I just found the explanation
a little fuzzy and incomplete
> > When the enable line is low the device looks like an on PFET in
> > series with battery. At this stage the device drwas 1 uA. Current
> > able to be drawn is probably limited by PFET rating. 500 mA or
> > 800 mA or whatever
> Yes, that's what I'd like to believe what they mean by "bypass",
> and it's common sense to expect that. I just found the explanation
> a little fuzzy and incomplete
The design makes it easy for them.
The output rectifier diode is a PFET.
When the chip is disabled:
- In one version they disable the FET and you get output isolation.
- In the other version they enable it and you get a through connection.
There is a possible "gotcha" here - BUT they may have very
specifically avoided it.
With a "normal" P MOSFET as the synchronous rectifier you would have a
reverse body diode which caused interesting results whichever way it's
connected under certain conditions.
eg the norm is to have the body diode in the same direction as it
would be if a conventional eg Schottky diode was used for output
(cathode to output). This means that you then do not get battery
cutoff by turning off the PFET.
>From their block diagram it appears that the body diode is absent - on
reflection this is probably the case with all synchronous rectifier
ICs with this topology and I have never looked into it. The FET is a
truly isolated one, unlike discrete units which have the substrate
implicitly connected to Drain by virtue of the means of construction.
Something new every day ... :-). Someone may care to correct me on
this.
Russell McMahon wrote:
>
> >From their block diagram it appears that the body diode is absent - on
> reflection this is probably the case with all synchronous rectifier
> ICs with this topology and I have never looked into it. The FET is a
> truly isolated one, unlike discrete units which have the substrate
> implicitly connected to Drain by virtue of the means of construction.
> Something new every day ... :-). Someone may care to correct me on
> this.
>
They mention that the output capacitor will not discharge back into
input supply. Since it's a P FET, isn't the substrate at Vin ? So as you
say how do they isolate when Vin > Vout? Millions of years ago I
remember signal MOSFETs that had no source/drain diode. The 4066 quad
switch obviously has "isolating" FETs. But it's unusual to see one that
can do 800mA (mentioned someplace as max SW terminal current).
My samples of this are shipping from Microchip on 15th. I'll let you
know how I get on with a 3.3V from 1 AA alkaline, 3.3V from 2x NiMH
maybe, and 5V from 2x AA (Alkaline or NiMH). Also what a single cell ->
3.3V followed by second 1640 to get 5V so as to have 3.3V and 5V.
They show a single Li cell. I don't fancy using this with a 2.5V to 4.2V
LiPoly for 5V. Seems like wrong starting point in terms of charger
circuit and discharge cut off. Maybe they mean a disposable primary
Lithium battery?
> >> Be sure to measure Vin on the circuit side of any Iin ammeter.
> >> Be aware that high frequency power supply noise on power wiring (in
> >> and out) can play havoc with some meters). Largish cap (100 uF?)
> >> across meter terminals can help.
>
> > Since the noise most likely is in the 100s of MHz range, a smallish cap
> perhaps
> > would be better. A ferrite clamp can also help.
>
> As the smps oscillator typically runs in the 100's of kHz region,
> that's where the noise is.
I am talking about conducted and radiated EMI which, with the waveforms shown
in figures 2-16 - 2-18 in the MCP1640 datasheet, will be much higher than 100's
of kHz.
Check it out with a spectrum analyzer.
The higher frequencies will most likely interfere as radiated noise with the
testleads working as an antenna.
The levels will of course be highly depending on the layout and decoupling of
the circuit.
/Ruben
==============================
Ruben Jönsson
AB Liros Electronic
Box 9124, 200 39 Malmö, Sweden
TEL INT +46 40142078
FAX INT +46 40947388 @spam@rubenKILLspampp.sbbs.se
==============================
> > Since the noise most likely is in the 100s of MHz range, a smallish cap
> > perhaps
> > would be better. A ferrite clamp can also help.
> As the smps oscillator typically runs in the 100's of kHz region,
> that's where the noise is.
>
I am talking about conducted and radiated EMI which, with the
waveforms shown
in figures 2-16 - 2-18 in the MCP1640 datasheet, will be much higher than
100's
of kHz. Check it out with a spectrum analyzer.
The higher frequencies will most likely interfere as radiated noise with the
testleads working as an antenna. The levels will of course be highly
depending on the layout and decoupling of the circuit.
/>
Whereas, I am talking about the noise which mucketh up some of my meters on
occasion when I use them in exactly the manner described in this particular
application. ie a current meter feeding the smps board and a volt meter
measuring the voltage on the PCB side of the current meter. In this location
a meter is exposed to any less than perfect smps power supply filtering at
smps frequencies as the smps draws cyclical current from the psu. When
running on batteries in final form (which is the fate of most units that I
deal with) the cyclical currents are drawn from the battery with no meter
present and no great harm results.
However, it is my direct observation that some meters are "noised up" by
being employed with low level 100 kHz or so noise superimposed on the 100's
of mA of DC that is typically imvolved in my applications. And that an
electrolytic capacitor across the metetr leads usually makes things work OK.
As previously noted, parallel small C's and a series L or R can also help
where neceessary. usually they are not necessary. (I have used ferrite
clamps and psu lead toroids and beads in other psu noise applications but
not for meter filtering.)
Power supply current pulses at smps cycle rate are unavoidable. In an ideal
design these will be all supplied by the onboard capacitor with the external
psu or battery supplying the cap at the DC mean. This ideal is unachievable
as it assumes zero cap ESR and an ideal current source feed from the
external supply. To the extent that this ideality is violated you will see
current pulses at smps cycle rate on the input supply. [[An analysis of
"what current flows where when" in a boost converter can be eye opening the
first time it is seen, but regardless of where it may be expected to come
from, in practice cyclical inductor feed comes from the psu leads.]].
Undoubtedly radiated and conducted signals that interfere in more
concentional manners also have to be dealt with, but the above is a real
world test and measurement issue.
I have a spectrum analyser of an excessively clunky nature that it sees
vanishingy little use day to day when I am prototyping smps circuits. An
AM/FM radio and an el-cheapo black and white manually tunable TV make
excellent 'wsniffers' for most purposes - with the radio alone sufficing in
most cases for rough checking.
> The 1640 has variants which are not available in the 1623/24
> which instead of turning off the PFEt in disable mode, turn it hard
> on. This allows the battery to be connected to the output in
> shutdown. If Vin is high enough this is most useful
Meant to mention I saw an application circuit to do what I was
thinking it could. Figure 6.1 in the datasheet
I received some early samples of the MCP1640, price looked great, package, etc.
So I put one in a design...works great, just based on the app notes in the data sheet.
However, as some have mentioned....can't yet buy the parts. So I did another similar design for the same client...but I used a TI TPS61220 for the boost. Costs more, but I can actually buy the part.
I haven't spun the other board yet..still waiting on the clients evaluation to determine when it might go into production and if they aren't ready for it till July then I will keep the part on there.
> From: Alan B Pearce <RemoveMEAlan.B.PearceTakeThisOuTstfc.ac.uk>
> Subject: Re: [EE] Microchip boost converters
> To: "Microcontroller discussion list - Public." <spamBeGonepiclistspamBeGonemit.edu>
> Date: Saturday, May 8, 2010, 9:53 AM
> > [changed subject]
> >
> >> Microchip have just published a data sheet for a
> boost converter
> >> working from a single cell for processors. Not
> sure if it goes to 5V
> >> output though. May just be 3.3V.
> >
> > If it's the MCP1623/24 I found the datasheet easily
> enough, but
> > can't find the actual device to either buy or sample,
> which is a bit
> > of a tease. And no mention that I can see anywhere of
> if/when it
> > will be available
> >
> > Press release leads to MCP1640
>
> yes it was the MCP1623/4 that I was thinking of. I had seen
> only the
> datasheet and not the product page.
>
> haven't looked at the MCP1640.
>
> I received some early samples of the MCP1640, price looked great,
> package, etc.
>
> So I put one in a design...works great, just based on the app notes
> in the data sheet
Cool. Mine are in transit, looking forward to trying them out
> TI TPS61220 for the boost
Thanks for the tip. Just in case the MCP doesn't work out
clear.net.nz>:
