On Tue, 02 Oct 2007 17:19:06 -0700, meow2222@care2.com
wrote:


>> To arrive at ~ 5V alone it could be done, but to remain at ~
>> 5V under high current may not be so easy with sub-C cells.
>
>Neither of us really knows what size cells, but of course theyll be
>sized for the job. However given that D cells can knock out 60A
>there doesnt seem much doubt that it'll all fit in a standard 5.25"
>case.

With Sub-C yes, but I am not convinced you could get enough
D cells plus supportive circuitry including the connectors,
wiring harness, etc, in that space (not just volume but also
the form factor dimensions.

Maybe it simply uses two drive bays instead, if the case has
the bays that isn't a problem. If it were important to have
this internal I would just as soon select a case with two
bays available so as to be able to reach higher capacity.



>I'd look at NiFe rather than NiCd though. With their very high
>current ability they can be made smaller and more V stable.

It'll drive the price up even more when it seems cost
prohibitive already vs a standard external UPS.

>> The motors don't care if the voltage drops while a PC does.
>> Running a pack with a tap in it to derive 5V and 3V power
>> might be too much current to keep voltage up. Even so, D is
>> definitely an improvement and yet getting ever closer to
>> exceeding available space in a drive bay. Remember it's not
>> just the batteries but the circuit, insulating material or
>> standoffs, the incoming and outgoing connectors or wiring
>> harness.
>
>those other parts are fairly small volume-wise.

Each alone, yes, but they'll all have to be on a circuit
board isolated and/or mounted to the casing in the rear, and
all these low volume parts add up, as well as requiring
fairly beefy wiring or traces to keep voltage drop to a
minimum since it is already a large concern by extending the
primary PSU harness to this circuit, through it and away
from it. Maybe it's not impossible but to know for sure
with a given battery cell size you will have to make a
prototype or at least a detailed model.



>> Some might only need to avoid losing data, but how many
>> would take a device with lesser versatility and capacity
>> just to save a bit of space outside of the PC while taking
>> up space and having to route a second thicker wiring harness
>> inside the PC?
>
>I'm not planning any extra wiring required inside the PC. Why do you
>think there would need to be? V_drop too high?

Yes, as well as degradation of the wiring and connectors.
You can't just push multiple times as much current through
these connectors and wiring for more than a few seconds
without it causing a problem and even for a few seconds
there is definitely going to be a voltage drop. Given brief
enough runtime that these parts stay within thermal
thresholds, you could just factor for the lower delivered
voltage and start out with a higher supply voltage, but this
current supplied will vary per system so it becomes more
complex to calculate or not a universal solution because you
have multiple variable factors effecting whether your
battery supply remains within the upper and lower limits the
system needs, particularly for 3.3 and 5V. You could make
it all easier by just using a regulated (switching) supply
after the battery, but this also raises the cost and size.


>
>Why would a business user choose an inbuilt miniups?
>1. Because they dont want to fork out on a large whole site UPS, nor
>have loads of mains plugin UPSes cluttering up the place.

As I'd stated previously, there is no new technology
suddenly making this possible where it wasn't before. I
have not heard of any demand for such a product except this
one thread on usenet.

>2. Lower cost than an external plugin UPS

.... but it's not, I guarantee this won't sell for less than
a $30 UPS. I'm thinking due to low volume the retail price
would be closer to $200. Remember, it's not just bare parts
costs, there's construction, development, advertising,
distribution, etc, and then there's the profit that has to
be made off the product.


>3. Ease: just tick the box on the form and you get a data-safe PC.
>4. Any company PCs used offsite will always be UPS protected
>this way

Maybe I'm wrong and your idea puts you one step away from
millions in profit... but I doubt it. Build a prototype and
try it with multiple varying platforms. Keep a BOM and then
take a poll of who will pay $X for Y performing internal
UPS.


>
>
>> There's also another problem in that PC PSU
>> with rail sense leads use these to sense at the ATX
>> connector under the presumption it is plugged into the load,
>> while the resultant voltage at the load will be lower than
>> it senses due to the intermediate stage of having to go
>> through the battery pack subcircuit.
>
>main psu doesnt go through the batt pack, the wiring in the pc is
>not affected, the miniups just plugs into a molex connector, end of
>job.

It has to have the main psu running through it.

Remember, the computer has to get power from both supplies.
They both have to be wired to the (motherboard for example)
part powered to deliver the power to it, BUT you can't just
use a splitter, because the output from the battery pack
being commonly connected to the main PSU would result in
high current into the battery pack or vice versa, depending
on which had a higher actual voltage level. With more
sophisticated electronics this could be overcome but not as
you'd described it. The cheapest solution would be to just
use isolation diodes, but once again the volume of the parts
goes up, towards being too big to fit in the bay space.




>If running the Pc's original wiring loom proves too high a
>V_drop
>to set the batt pack out a little higher and run it open loop, then
>there would also need to be an exrta sense wire tagged on. Very
>easy.

As I mentioned above, you can't just connect it to the PSU
wiring harness, because you have two supplies with
inevitably, slightly different voltage levels. With an
isolation method you could use the PSU wiring harness but
you still have to break that wiring harness by splicing into
it before it gets to the components from the PSU, or have an
inbuilt circuit on the internal UPS to do this.



>
>
>> >> Also we haven't even considered the little odds and ends to
>> >> implement this, like the wiring harnesses, connectors, etc,
>> >> requiring the primary AC-DC PSU to input to this battery
>> >> pack
>> >
>> >I think we may have been talking at cross purposes then. I was
>> >proposing doing away with all that, just trickle charging the battery
>> >pack from the PC's molex connector. Would need nothing more
>> >than a basic bottom price linear regulator.
>>
>> Pack input isn't the problem, that'll work for charging but
>> you still need to isolate the two supplies from each other.
>> You have two power sources and you have input to the pack
>> through this regulator, but it still has to get to the
>> motherboard without being a common rail to the PSU, and both
>> have to be plugged into everything so the original PSU plus
>> then has to be unplugged (else a lot of ugly grafting of
>> leads onto the existing PSU wiring harness) . Thus, you
>> have to unplug the PSU from the mobo and put it's connector
>> onto another circuit board (or dongle style split) along
>> with the output from the battery pack to the motheboard and
>> other components needing power.
>
>You're way overcomplicating this. The battery pack output stays
>connected via a relay whenever the machine is on.

You mean disconnected, right? If so, ok BUT this is yet
another part taking up space that had not been mentioned
yet, and multi-pole or multiple relays capable of this much
DC current aren't particularly small, and further PCB space
and traces continue to add to the volume required.



>At power down
>it senses that and continues supplying power for 15s (ie it doesnt
>distinguish mains failure from normal power down) then cuts out.

15 seconds is not long enough to make this product worth
buying/owning/etc, IMO.

It may also need to distinguish mains failure from normal
power down.



>
>A possible way to control charging is to connect the half way point
>on the batt pack to each rail (0v or 12v) via a basic reg, thus
>charging each half of the pack alternately. Doide & R ensures the
>pack and its halves can discharge very fast but only charge at a
>trickle.

?? What do you mean connect the half way point?

Yes you could get a usable result with resistors and diodes,
providing all parts are suitably spec'd. It still may not
be a univeral solution since different systems have
different current requirements and the voltage will vary
based on current. IMO, it really needs to stay within ATX
specs which is 5%. I suspect a difference in current from
one system to the next, alone will exceed this 5% meaning it
would have to be custom designed for each system...
something an OEM could do but would eliminate it from being
an aftermarket product unless the system integrator knows
exactly how to spec the current requirement and there were
multiple models of this internal UPS to meet each different
system requirement.

It would be much easier to just use a DC-DC switching
regulator supply after the battery pack. That makes it
nearly universal (within the current capability it can
handle as a max. amount).



>
>
>> >Same molex serves the
>> >PC with power during the mini-ups's brief run. I dont know what the
>> >continuous current rating of those molexes is, but if the UPS only
>> >powers the system for 20 seconds its not going to be a problem.
>>
>> 6A, IIRC, but there's the voltage drop problem across the
>> connector and single supply leads making it too lossy.
>
>Do you know what the R or range of R is? I've not measured it. The
>ATX Vspec allows a fair amount of swing IIRC


If the system has enough power interruptions to make this
UPS viable, we're talking about a continual overcurrent
which will degrade the connectors (first, and then wiring).
The ATX spec allows 5% but with battery voltage dropping
while discharging, battery voltage dropping with changes
(increases) in current, and the wiring plus connectors
causing further drop with increases in current, very quickly
the UPS needs matched to a system within a certain current
usage, and this also greatly depends on what that system is
doing at the moment the power goes out since it is very easy
for modern processors/GPU/etc to have changes in current
over 30% if not a lot more.



>
>> You'll also have to switch the lead from input to output
>> around the charging circuit if I understand what you're
>> suggesting,
>
>trying to avoid any such switching if poss. If we switched the batt
>pack between charge and out, a power failure would mess with the
>pc before the relay clicked in.

Well... for the purposes of what I wrote, a relay is a
switch.


>> then have the power going through these single
>> supply leads up to the PSU wiring harness distribution point
>> and down into the main PSU harness, which is quite a
>> distance to travel for the currents in a PC.
>
>yeah. If too much R and load variation we could resort to an extra
>sense wire to fix it. That would require a high current reg though,
>best avoided if poss.

The sense wire would need to be from the load to the
feedback in the main PSU, meaning the system as it started
can't remain intact, it has to have the PSU wiring modified
regardless of whether that went to the UPS before the
components.

As for a sense wire from the UPS to the load, that too would
be a good approach to take to keep the voltage right, as a
feedback to the UPS which has a regulation stage on the
output. This is really the only universal solution to
ensure compatibility, the way it needs to be done to work
well. As for high current regulator, you want a switching
regulator circuit anyway not a linear because linear is too
lossy. A typical voltage drop across one might be roughly
1.5 to 2.5V, which times at least a dozen amps total as a
sum of all power rails to power the system, is 18W+ loss,
reducing the battery pack's runtime. An LDO linear could be
used instead with lower loss but still it's a significant
loss when running from battery power and with the battery
capacity constrained by available space in a drive bay.

The current isn't necessarily much of a problem for the
regulator though, as I'd mentioned in a past rely all one
would have to do is put a pass transistor into the regulator
subcircuit. Examples of this can be found in some linear
regulator datasheets, for example I think National's LM317
datasheet has one.

http://www.national.com/ds.cgi/LM/LM117.pdf
pg 16, but using pass transistor(s) capable of the current
instead of the default 3 * LM195 shown.

I went off on a tangent there, a switching supply output
from the pack is really the best alternative.

>> An OEM could come closer to the goal because they can limit
>> the unit's compatibility with certain systems' unique
>> current requirements. On the other hand we'd be cursing all
>> that much more about the proprietary way they ended up
>> getting the job done, OEMs don't need much of an excuse to
>> do screwy things that make it either impossible or very
>> costly to replace anything... keeping in mind that even if
>> the mini-UPS were to work there will always be a certain
>> failure rate and replacement requirement.
>
>They do things proprietary ways mainly because they can save
>money. 20p off 10,000 units is £200, os if they can do it non
>standardly for a few p less they will. But I dont see proprietary vs
>not having a big effect on viability.

Proprietary vs univeral means the OEM can determine the
current consumption per rail of the target system, and match
the DC ups to it. Without having the target system's
current requirement there is another variable in whether the
UPS' DC output stays in spec, unless it has the regulation
stage following it before the load and ideally the feedback
remote sense lead.



>One PC mfr would only capture
>part of the market, but people buying new systems are an attractive
>target for addons, especially if they promise to safeguard against
>business data corruption or loss, and do it at lower cost than
>anyone else's UPS, and with less hassle..

yes, and that is exactly why it needs the regulation stage
to make it universal instead of just running straight from
the pack to the load, which is basically the same reason why
a switching main PSU also needs (already has) a feedback
loop so it can adjust for differing loads.