The LiPo Project

[Dougigs]

Jazper -- I wonder if we can deal with potential reference noise in a way that doesn't involve sticking an LMx17 in the line. The voltage drop produced by a regulator is worrisome, since we're only starting with 11.1v at best -- I'd rather find a way to make the full voltage available. Right now we have a MOSFET in the way, but we can find ones with resistances in the handful of miliohms. A regulator, aside from PCB space, seems like a pretty big blockage...

You mentioned current-sink approaches to this problem. Care to elaborate? I know, in theory, that a very small resistor followed by a current sink can get rid of zener/reference/regulator noise without much loading.

One other thing: There are some extremely low-noise voltage references out there.... Unfortunately, they all consume current in the ma range; all the microamp references don't have the best noise specifications. Has anyone produced a low-noise, ultralow-current reference?

 

[Jazper]

Quote:

Originally Posted by guzzler Eh? Your own USB DAC has a smaller chip than that, SSOP. It's 0.65mm pitch as well, and I agree that soldering it is fine; just take the time to line it up properly.

I didn't solder it up

but I do have a weller station which can do smd quite easily. I'm more worried about those who really don't like smd.

Re: current sink

It's good in theory, but it'll knock down the voltage, which is not what you want.

 

[Jazper]

The TL431 might be an acceptable voltage reference for this...

http://www.sjostromaudio.com/phpBB2/viewtopic.php?t=16

http://coset.se/tmp/Audio/PerAnders/TL431B/TL431-D.pdf

 

[DaKi][er]

Really, how much of an issue is this noise in the reference voltage? It is only for the comparator to switch off the load from the batteries, we are only concerned with switching it off when it gets into the 10's of mV range and that noise in the uV range wont concern us and I think that Jazper is worrying about nothing here. And if you really want to, a ceramic or something bypassing it would improve it somewhat

 

[guzzler]

I've sold 100 of the DACs not to mention 150 of the S/PDIF boards, so I don't think it's really about "those who don't really like SMD", it's meant to be enjoyable and about trying something new. I've not heard of anyone who's messed up the PCM2702, although that could just be that hear from those that suceed

I agree with dakiller here, all you need is a reference; it doesn't have to be low noise as long as it's within the tolerance of your comparater.

R1/R2 is a simple resistor divider, so from the article:

Quote:

The circuit is set up for a single-cell Li-Ion battery, where the lockout voltage (the voltage at which the protection circuit disconnects the load from the battery) is 3.0 V. This voltage, set by the ratio of R1 and R2, is sensed at node A. When the battery voltage drops below 3.0 V, node A falls below the threshold at node B, which is defined as: VB = 1.25 V + I × R4 = 1.37 V where I = (VT - 1.25 V)/(R3 + R4) = 800 nA and VT = lockout voltage.

What you're looking for is V[size=xx-small]R2[/size] [ie, node A] < VB

V[size=xx-small]R2[/size] = (V[size=xx-small]T[/size] * R2)/(R1 + R2)

 

[Dougigs]

Good stuff. Thanks for the thinking, Guzzler -- we can work out resistances for various typical voltage values (are 0.1 per cent resistors commonly available in SMD?)

I did some reading up on voltage references, and I think we're safe. As DAKiller notes, we're just driving a binary switch so there's almost no way it could affect the supply voltage. And even with non-low-noise references, we're talking in the low hundreds of ppm -- so for our tolerances, this shouldn't affect the comparator.

Once I get a couple voltage references, I'll hook the circuit up to the scope this weekend and see what the noise looks like. If worst comes to worst I'll add a small ceramic cap.

Jazper: The problem with the voltage reference you cite is that it requires a minimum of 0.4ma (and likely 1ma) to operate. We need something that consumes current in the low tens of microamps, because this circuit needs to operate all the time, even after the battery's been cut off. Now, with voltage references, there seems to be an inverse realtionship between current consumption and noise -- the higher-current ones are lower noise. But, again, I don't see any way the noise could appear on the supply. Also, the noise they tend to be concerned about is below 10hz, i.e. wobble, which isn't going to create audio problems and which doesn't affect our comparator (because it has a hysteresis mechanism).

Micropower voltage references are quite widely available. The ones we want use bandgap technology, which is very stable as well as requiring only microamps of current. Given the elimination of noise as a problem, we can search for the ones that use the very lowest current.

 

[skyskraper]

RS has 0.1% smd resistors i think

 

[Dougigs]

Quote:

RS has 0.1% smd resistors i think

Yeah, I checked, they do -- though only in the dreaded 0603 "speck of pepper" package. Other places (Digikey, Farnell, Rapid) sell them in the 0805 package, which sounds a bit more workable... there don't seem to be precision resistors in the more appealing 1206 package.

Farnell actually sells them as "0% tolerance," which sounds like good marketing if not terribly good science.

 

[DaKi][er]

Quote:

Originally Posted by Dougigs Yeah, I checked, they do -- though only in the dreaded 0603 "speck of pepper" package. Other places (Digikey, Farnell, Rapid) sell them in the 0805 package, which sounds a bit more workable... there don't seem to be precision resistors in the more appealing 1206 package. Farnell actually sells them as "0% tolerance," which sounds like good marketing if not terribly good science.

Mouser have 0.1% in 1206 and in single quantities

(Edit: on looking in the catalog here, they all stop at 1M for their bigest value

on checking digikey, so do they) also digikey will get you with their multiple of 10 on most resistors, adds up when you pay 80c to $1 each

If you plan on building a few of them and making the boards for others it is very nice to be able to get everything from one 1 of the big distributors with 2 at most, or think about combining harder to get parts with boards (not that the project is anywhere near this stage, just keeping the idea in the back of the mind while you go along and consider it when choosing parts

 

[Dougigs]

Well, there are a couple ways to handle the precision resistor issue:

1) Have slots for 2 1206-package SMD resistors in series, so people can either combine 2 0.1 per cent values to get the precise measure, or use more commonly available 1 per cent resistors and measure them to hand-pick tighter tolerances (which is something you can easily do with 1206 size); or

2) use non-smd resistors in this position -- 1/8 watt 0.1 per cent are very widely availabe and don't increase board height much.

Actually, we can kinda do both by having 2 series 1206 slots with holes for a leaded resistor.

 

[Dougigs]

Actually, the reference specified in that LT schematic is almost perfect: it claims a current consumption of 800 nanoamps.

But it costs between $6.50 and $11.00 at DigiKey, and isn't available from the major UK/European suppliers.

However, the REF1004 and LT1004 units work just the same, are very widely available and cost an awful lot less. Their current consumption is in the microamps rather than the nanoamps, but is still low enough not to be a worry. And they're 1.2V rather than 1.25V -- not important at all; in fact you could put just about any reference voltage here (you just need to calibrate r1 - r4 to account for your battery voltage and your cutoff point vs. the reference voltage).

Anyway, the two chip families are pin-compatible (both are in SOIC-8 packages with the cathode on pin 4 and the anode on pin 6).

I've ordered a TO92 of the REF1004 for breadboarding and SMD version for board use.

 

[Dougigs]

Well, this is an easier question. Obviously this should be a single op amp, so we don't burn any battery current operating unused op amp segments.

Basically, you can pick any single SO8-package op amp, as long as

a) its supply voltage is more than your maximum battery power, and

b) it uses very, very little quiescent supply current (Iq) -- definitely well under 100uA and ideally in the tens of microamps. In other words, a micropower op amp.

There are a good number of op amps that fit these specs, and some of them are very affordable (under $3/1.50 pounds). It does not need to be a rail-to-rail chip for this circuit, nor does it need to be a single-supply op amp -- but a lot of op amps with these specs are rail-to-rail and single-supply, and that doesn't hurt.



EDIT: BUT -- There is nothing quite like the TLV2401 from TI. It's 1.45 pounds at RS, $1.68 at DigiKey, claims to consume 800nano amps -- like, less than 1uA -- and takes single-supply voltages up to 16V (and just for fun has reverse-voltage protection up to 18V).

http://focus.ti.com/lit/ds/symlink/tlv2401.pdf

Well, that's my pick. It'd take about 250 years to drain your battery.

 

[Dougigs]

Ok, I've played with this circuit a bit and the resistor values are close to being settled.

What I'm trying to determine, again, are the right values for a 9V cutoff point (i.e. using an 11.1V three-cell stack). A number of value-combinations are possible in a voltage divider like this, but here are three factors to consider:

1) We don't really want R1, R2 or R3 to drop below the megaohm values, since we want this circuit to consume current in the low microamps.

2) It's worth keeping in mind that values above 1M are scanty, and since we want a fair degree of precision, I'd like to pick values that can be made from no more than, say, three commonly-available resistors in series. I think we have to use non-SMD resistors for R1-R3 because SMDs are hard to source in megaohm values.

3) On that point, it seems like a bad idea to go above, say, 5M since that could entail long strings of resistors (and we're working on a tiny board here).

So this combination seems to fit best (keeping in mind that I'm specifying a 1.2V reference chip):

- R4 stays the same (150K)
- R3 drops to 2M (from 2.05M) for convenience

-- that gives us a point B voltage of 1.744V when the battery drops to 9V. So we want to calculate a voltage divider for point A (R1&R2) that gives us 1.743V or less.

The best combination I can come up with seems to be R1 = 4.2M and R2 = 1M; that gives us 1.7308V. If we want to get more accurate, R2 can be 1.0082 (a 1M + a 8.2K resistor) = 1.7422V, which seems pretty good for our purposes.

If someone would like to check my math, I'd be grateful (you shouldn't trust a political science dropout with Ohm's law). In any case, I've got it up on a breadboard now and it seems to work nicely at these values.

This way we can build R1 out of 3 resistors (MRS25 resistors, which are very small and widely available, come in a good range of megaohm values including 2.2M... I'm assuming that other lines in this size have a 2.2M). R2 is maximum of 2 resistors, R3 is 2 and R4 is a single resistor.

Once I've tested it thoroughly for cutoff, hysteresis and noise, I'll post a PCB design, probably tomorrow.

 

[DaKi][er]

Quote:

Originally Posted by Dougigs Ok, I've played with this circuit a bit and the resistor values are close to being settled. What I'm trying to determine, again, are the right values for a 9V cutoff point (i.e. using an 11.1V three-cell stack). A number of value-combinations are possible in a voltage divider like this, but here are three factors to consider: 1) We don't really want R1, R2 or R3 to drop below the megaohm values, since we want this circuit to consume current in the low microamps. 2) It's worth keeping in mind that values above 1M are scanty, and since we want a fair degree of precision, I'd like to pick values that can be made from no more than, say, three commonly-available resistors in series. I think we have to use non-SMD resistors for R1-R3 because SMDs are hard to source in megaohm values. 3) On that point, it seems like a bad idea to go above, say, 5M since that could entail long strings of resistors (and we're working on a tiny board here). So this combination seems to fit best (keeping in mind that I'm specifying a 1.2V reference chip): - R4 stays the same (150K) - R3 drops to 2M (from 2.05M) for convenience -- that gives us a point B voltage of 1.744V when the battery drops to 9V. So we want to calculate a voltage divider for point A (R1&R2) that gives us 1.743V or less. The best combination I can come up with seems to be R1 = 4.2M and R2 = 1M; that gives us 1.7308V. If we want to get more accurate, R2 can be 1.0082 (a 1M + a 8.2K resistor) = 1.7422V, which seems pretty good for our purposes. If someone would like to check my math, I'd be grateful (you shouldn't trust a political science dropout with Ohm's law). In any case, I've got it up on a breadboard now and it seems to work nicely at these values. This way we can build R1 out of 3 resistors (MRS25 resistors, which are very small and widely available, come in a good range of megaohm values including 2.2M... I'm assuming that other lines in this size have a 2.2M). R2 is maximum of 2 resistors, R3 is 2 and R4 is a single resistor. Once I've tested it thoroughly for cutoff, hysteresis and noise, I'll post a PCB design, probably tomorrow.

Just one thing to watch for is your resistor tolerances, as a 1M resistor can be anything upto +/-10K ohm and still be in 1%, matching can be done quite easilly if you got equipment upto it (a 6.5 digit 4wire ohm meter helps, but your average diy'er dont have that gear)

 

[Dougigs]

Yeah, that's a good point, and it can actually work in our favour -- if we pick somewhat wide-tolerance resistors it's actually possible to hand-pick one that's much closer to the actual desired value.

Though from my experience it's like trying to find two perfectly matching socks from your sock drawer -- the values end up all over the place except exactly where you want them.

The precise target value for R2, by the way, is 1,008,764 ohms. That is assuming that R1 is exactly 4,200,000 ohms. A tolerance of 0.1 per cent would make this a lot easier (fortunately there's good availability at reasonable prices for 0.1s these days).