Better to charger higher or run battery lower?

[darth_vad3r]

I imagine the onboard charger efficiency depends, to a small extent, on how clean the incoming AC wave form is. This could be part of the difference you’re seeing.

Yes, so many factors.

I decided to look into this one a bit more:

I may also try to incorporate I-squared R loss into the model in addition to the fixed overhead and fixed % losses to see if I can get a better fit.

... and the TL;DR conclusion is it seems insignificant

Using Google to tell me that AWG gauge 1 wire has 0.124 ohm resistance per 1000 ft at 20C, and estimating 10 feet of wire in the system, I get I-squared R losses of:

• 10 A: 0.124 W
• 20 A: 0.496 W
• 30 A: 1.116 W

i.e. not a heck of a lot.

I'm guessing there's less than 10 feet of travel and they probably use thicker gauge wires or "equivalent to wire buses" that have lower resistance too.

Even if we use AWG gauge 6, 0.3951 ohm/1000ft, for 10ft we get:

• 10 A: 0.3951 W
• 20 A: 1.580 W
• 30 A: 3.556 W

i.e. *still* not a heck of a lot. Less than 0.1%.

 

[Allistah]

With my 14-50 outlet install, I ran about 45’ of 6ga wire. Still probably not a lot of loss.

I’m curious about 30-40 amps in regard to efficiency. I wonder if it stops improving or if it still continues to be just slightly more efficient at 40A vs 30A.

 

[darth_vad3r]

With my 14-50 outlet install, I ran about 45’ of 6ga wire. Still probably not a lot of loss.

I’m curious about 30-40 amps in regard to efficiency. I wonder if it stops improving or if it still continues to be just slightly more efficient at 40A vs 30A.

Unfortunately I won't be able to test beyond 32 A since I have the SR+ ... I suspect it will continue to improve slightly as the constant(ish) overhead continues to be a diminishing percentage of the total power. If you charge quicker, less power is wasted keeping the car on.

The higher current adds a tiny amount of extra losses, but nothing compared to the overhead losses.

 

[Allistah]

Ah ok.. so maybe I’ll start charging at 40A then.. I’m guessing the rate itself is fine for the batteries since super chargers charge at such a high rate and they’re fine..

 

[AlanSubie4Life]

You can see the difference is not insignificant. It was over 2% at 30A, and about 1.6-1.7% at 20A and 10A.

There are 6 new datapoints (from 18 underlying samples) of the charge efficiency at 10, 20, 30A at ~196V

If you take your raw data from the two Chargepoints (if you have the exact voltage at each current level for both Chargepoints), and then plot them all on the same plot, using POWER as the x-axis variable (rather than current), it might be interesting to see whether the data align a bit better. 196V is ~5% lower than 208V (maybe the other one was not 208V of course), so you'd expect it to not align exactly correctly, if you plot vs. current.

Obviously the presence of multiple 16A charging units in the car (which switch based on current, perhaps - we don't know how they behave exactly) will confuse a picture plotted vs. power, rather than current. But just a suggestion as another way to look at it.

 

[darth_vad3r]

If you take your raw data from the two Chargepoints (if you have the exact voltage at each current level for both Chargepoints), and then plot them all on the same plot, using POWER as the x-axis variable (rather than current), it might be interesting to see whether the data align a bit better. 196V is ~5% lower than 208V (maybe the other one was not 208V of course), so you'd expect it to not align exactly correctly, if you plot vs. current.

Obviously the presence of multiple 16A charging units in the car (which switch based on current, perhaps - we don't know how they behave exactly) will confuse a picture plotted vs. power, rather than current. But just a suggestion as another way to look at it.

Ya, the other wasn’t at a full 208 either so the difference was around 4%. The second one was notably lower though which is why I noted it. The data ‘points’ added should be shifted a little to the left but I was just being lazy and slapping them onto the same chart for comparison.

I will try to re-plot maybe tonight with efficiency vs power.

If I repeat the test at home with 120V I’ll make two sets of charts ... will be interesting to see if the small part of the graphs that will overlap will line up well or not (i.e. L2 at 5-6A vs L1 at 11-12A).

 

[darth_vad3r]

Updated info from a new test performed in the same manner as previously (see my prior quoted posts above for info), visualized with 2 theoretical models of the charge efficiency using 250 or 350 watts of fixed charging overhead and 95% underlying charger efficiency.

This was for a 208-volt L2 charger, with a Model 3 SR+, multiple data points taken (2-3) at each amp setting in the car, from 5->30A.

Note, I did not notice any drop in efficiency crossing from 16A to 17A as one might theorize if that triggers a 2nd 16A charge unit to kick in and lower the efficiency.

View attachment 442358

The 350 W overhead line (blue-dashed) is a fairly good fit.

Yet another chart. This was at a different, slightly lower powered "6 kW" chargepoint ... for whatever reason the numbers look lower than the previous data, but I slapped them onto the same chart nonetheless to compare with the prior chart.

There are 6 new datapoints (from 18 underlying samples) of the charge efficiency at 10, 20, 30A at ~196V taken with the same data sampling from the API technique, but from both inside the car, and outside the car with the doors locked (and the display off).

You can see the difference is not insignificant. It was over 2% at 30A, and about 1.6-1.7% at 20A and 10A. I would have expected it to be less at 30A but for some reason it wasn't. I probably need a lot more data samples to weed out random rounding sampling errors.

View attachment 442840

So, in the end, the conclusion is that butt in seat, screen on (daytime mode), no HVAC, costs about 34-141 watts over the car being locked and screen off and driver not present.

Corrected 250 to 350 W in last line of the first post above, and new version of the 2nd chart in 2nd post above with x-axis now being Charge Power (kW) instead of Current (amps) so that the new data points from test #2 with the measured differences between LCD screen on and off (at a different slightly lower-powered ChargePoint) slot into a more 'proper' spot on the chart:

(I also added a grey line for what is my projection of what would have been measured with the LCD off in the first test)

 

[darth_vad3r]

I think eventually I'd like to see 3 lines on one chart, with x-axis being power, y-axis efficiency (with the 3 lines being those relatively easily available to me: charging at 120V from 5-12A, ~196V from 5-30A, and ~208V).

Typically only see the highest of these voltages at lower amps, and at 30A the voltage drops a bit, but nominally I think both these chargers are fed 208V supply, it's just that one is further away from the panel than the other ... or some other issue.

 

[AlanSubie4Life]

Corrected 250 to 350 W in last line of the first post above, and new version of the 2nd chart in 2nd post above with x-axis now being Charge Power (kW) instead of Current (amps) so that the new data points from test #2 with the measured differences between LCD screen on and off (at a different slightly lower-powered ChargePoint) slot into a more 'proper' spot on the chart:

(I also added a grey line for what is my projection of what would have been measured with the LCD off in the first test)

View attachment 445018

Cool. I wonder whether the rolloff on the low end with LCD off (which is way below what the 250W projection predicts) would actually be quite that low for a more extended charge where you left the vehicle alone...perhaps there are some overhead loads at the beginning of charging, soon after the door is closed/screen is off, which take a while to go away - so for a more extended charge in this test maybe you would get closer to the red prediction?

In other words maybe the true “minimum overhead” is lower than what your gathered data suggests...maybe the overhead was slightly inflated by the time limitations of the way this data was gathered...?

 

[darth_vad3r]

Cool. I wonder whether the rolloff on the low end with LCD off (which is way below what the 250W projection predicts) would actually be quite that low for a more extended charge where you left the vehicle alone...perhaps there are some overhead loads at the beginning of charging, soon after the door is closed/screen is off, which take a while to go away - so for a more extended charge in this test maybe you would get closer to the red prediction?

In other words maybe the true “minimum overhead” is lower than what your gathered data suggests...maybe the overhead was slightly inflated by the time limitations of the way this data was gathered...?

When you say “beginning” of charge are you assuming time is increasing to the right? Or something else?

The solid black line was gathered from one single continuous ChargePoint session. I started the test before the end of a “regularly scheduled” 30 A charge session, maybe about 90 minutes into it? So the test started “warm” and at the high end, starting at 30 A I slowly lowered the current setting in the car, 1 amp at a time, let it settle, then took 3 data points from the API for each level. For good measure, I swept back up from 5A to 30A before finishing and gathered a few more samples along the way every 5A to see if anything looked dramatically different.

I suspect the larger ”error” at low power is just that the model is imperfect. e,g. for a better fit the -5% constant charger loss should perhaps be modeled as slope or curve itself. Not sure. Could the charger efficiency be inversely correlated with voltage, irrespective of power? The voltage level tends to rise as you lower the load. Not sure that 1 or 2% difference is enough to make it ‘fit’.

So ... Googling AC to DC efficiency curves tells me this was it (curve).

Most of the curves I saw generally look the same as mine. Some have a steep climb and then almost flat but a slight long taper. Others more close to my curve’s shape.

Using this instead of a flat 95% probably makes the model fit a lot better at the low end and a little bit better at the top end as well:

Image source:
RAC15-K/RAC20-K Series Converters - RECOM Power | DigiKey

 

[ZOMGVTEK]

One thing to note. Although this method may show the energy put into the pack, it does not take into account intercalation efficiency or how the charge rate impacts heating and the duration the car stays awake past completion. And you pay for the energy consumed from the meter, so ideally if you want to get super fancy the cumulative resistance of the line past the meter to the car is the relevant bit. And you'd need to double the length of the wire to get round trip resistance. But yeah, if you run #1 and charge at 32A on 200A service, that's gonna be relatively small. For people with long runs, 100A service, and less ideal installs, this can be quite a bit. Unlikely it would skew the results downward too much, but the response is non linear.

 

[darth_vad3r]

One thing to note. Although this method may show the energy put into the pack, it does not take into account intercalation efficiency or how the charge rate impacts heating and the duration the car stays awake past completion.

It’s uncertain if the car’s reporting of “energy added” takes into account the car’s best guess at actual net energy added taking into account all DC-side and battery losses or not.

And you pay for the energy consumed from the meter, so ideally if you want to get super fancy the cumulative resistance of the line past the meter to the car is the relevant bit. And you'd need to double the length of the wire to get round trip resistance. But yeah, if you run #1 and charge at 32A on 200A service, that's gonna be relatively small. For people with long runs, 100A service, and less ideal installs, this can be quite a bit. Unlikely it would skew the results downward too much, but the response is non linear.

Somewhere here (another thread?) I ran the I-squared R loss numbers for 10-20 feet only (for the car) and determined it was so tiny it wasn’t worth modelling. If you have a 100’ run that might make it less tiny, but it’s still dwarfed by the car overhead and also a lot smaller than the ~5% loss on the charger.

E.g. Using AWG 6 I get ~1% (~80 W) for a 30m run (60m total ~200’) at 32A.
It’s 20 W at 16A, and 180 W at 48 A for AWG 6.
I get 12.5, 50, 113 W for AWG 4.
8, 31.5, 71 W for AWG 2
6, 25, 56 W for AWG 1.
These are for 30 metre runs (60m total ~= 200’ (196-ish))

Those scale linearly with length of run, so cut them in half for a 50’ run.

Anyways, presumably all our cars are the same(ish), at least by trim, so having the car-only efficiency ignoring the losses from the wall to the car is still useful.

 

[AlanSubie4Life]

Although this method may show the energy put into the pack, it does not take into account intercalation efficiency

I assume to some extent it must account for the input intercalation efficiency, since these values from the API do seem to be a measure of energy put into the pack in some fashion. Note that the EPA indicates that 78kWh can be pulled from the battery per the Tesla-provided test results. And API pulls similar to those used above to generate the plots (which were done on an SR+) extrapolate to 76kWh added for a full charge (0->310 miles) on the AWD vehicle.

That being said, there is definitely a discrepancy between energy added to the pack vs. what shows up on the trip meter (for AWD it's 245Wh/rmi for the energy you put in, and 230Wh/rmi for the energy you take out). It's somewhat of a semantic question, since all that matters is how far you can go for a given amount of energy from the meter. I could argue that the Trip meter in the car reads low by the ratio 230/245, but it might actually be reading accurately, but unfortunately you can't get the same amount of energy out of the pack as you put into it (that hypothesis is not supported by the EPA numbers though). But it doesn't really matter; one way or another the energy is either undercounted by the trip meter, or lost due to intercalation or whatever.

If you're trying to assign an exact "true" efficiency (however you decide to define that) to the charging process, yes, it's unclear what exactly is being counted by the API pulls used here (as mentioned immediately above). But it also doesn't matter much if the objective is to compare efficiency at different charge rates.

I suspect the larger ”error” at low power is just that the model is imperfect. e,g. for a better fit the -5% constant charger loss should perhaps be modeled as slope or curve itself.

For sure that's a factor. Actually, I was just getting at the fact that your datapoints for LCD off deviate considerably from your 250W model fit (not even looking at the low end). That suggests that there is quite a bit of static loss - more than 250W...which is just more than I expected...though it could well be correct. In any case, this additional loss really adds up and hurts efficiency at the low end - and that's not even counting the additional inefficiency due to running the AC->DC converter at low input power (which yes, was not in your model).

So: I was just wondering whether, for a more extended charge interval, with the car unbothered, while charging at say 2kW input power, whether the resulting data would line up better with your 250W overhead model than your existing data. That's all. Just wondering whether the somewhat shorter measurement intervals introduced some error. It might not have.

It's all fairly unimportant of course. Just trying to understand the minor non-idealities that aren't being captured by your model.

 

[darth_vad3r]

I was just getting at the fact that your datapoints for LCD off deviate considerably from your 250W model fit (not even looking at the low end). That suggests that there is quite a bit of static loss - more than 250W...which is just more than I expected...though it could well be correct. In any case, this additional loss really adds up and hurts efficiency at the low end - and that's not even counting the additional inefficiency due to running the AC->DC converter at low input power (which yes, was not in your model).

So: I was just wondering whether, for a more extended charge interval, with the car unbothered, while charging at say 2kW input power, whether the resulting data would line up better with your 250W overhead model than your existing data. That's all. Just wondering whether the somewhat shorter measurement intervals introduced some error. It might not have.

It's all fairly unimportant of course. Just trying to understand the minor non-idealities that aren't being captured by your model.

Gotcha. It could even be that measuring over a short interval mid-charge caught it while the battery coolant pump was running (or running more than average). Similar to trying to measure AC consumption at a random point in time while it typically cycles on or off or at least varies its output to maintain cabin temp—or like you said at the start, when for AC for example, there’s a massive power spike to get the initial cabin temp down).

If we charted thousands of people’s sessions like ABRP does we could paint a better picture and fit a curve to the median efficiency.

If the app or the API let me modify the setting for current (amps) remotely I could also gather more interesting data easily for longer periods while the car is locked with no accessories or screen on. Alas, it doesn’t.

 

[darth_vad3r]

Image source:
RAC15-K/RAC20-K Series Converters - RECOM Power | DigiKey

That chart got me thinking about the 2 or 3x 16 A sub-component chargers that seem to make up the “32 A” or “48 A” chargers on the SR and LR.

I would expect Tesla to either (a) optimize for component wear/redundancy and only turn on multiple chargers when required, or (b) optimize for charging efficiency (which might also help with wear?).

If the efficiency profile is similar to above, you’d turn on a 2nd charger maybe as early as 40-50% of load to get 2 @ 20-25% which looks slightly more efficient.

Oh wait. Hmm. They don’t let you charge below 5A, so maybe that cuts out the bottom 10-16% of the chart already for 48A and 32A chargers. Or I guess it should be the bottom 31% relative to 16? Hmm.

Actually, if that’s the way to do it, then using a model of flat efficiency from 5 to 16A on an individual charger might be pretty close.

It might go from 5 to 32A something like this for 2 chargers on the SRs...
5+0
...
9+0
5+5
5.5+5.5
6+6
...
16+16

On an LR it might split to 5+5+5 at 15A.

Hmm... I feel a new chart is in order here from the same data. Assuming constant overhead, chart the underlying charger efficiency vs amps that is implied by the measured efficiency to try and glean any pattern there.

It does seem like would be a bit of a jaggy slope down from about 9A to 30A, with a ramp up from 5A to ~8-9A

I’ll have to remember if one of my 16A chargers fails to do a test and make a chart first before getting it fixed

 

[AlanSubie4Life]

t might go from 5 to 32A something like this for 2 chargers on the SRs...
5+0
...
9+0
5+5
5.5+5.5
6+6
...
16+16

On an LR it might split to 5+5+5 at 15A.

Yeah, lots of possibilities, that's what I was pondering in the past. I have assumed they optimize for efficiency, but who knows?

Just on a related note, here's a video showing what I believe are the three modules on the PCS, if I understand correctly. I thought they were separate (which would reduce repair costs I guess), but it looks likely they'd have to replace all 2/3 if you have a failure of one, as they are all on the same board.

 

[darth_vad3r]

Yeah, lots of possibilities, that's what I was pondering in the past. I have assumed they optimize for efficiency, but who knows?

I asssume the same and have the same “but who knows” condition

Just on a related note, here's a video showing what I believe are the three modules on the PCS, if I understand correctly. I thought they were separate (which would reduce repair costs I guess), but it looks likely they'd have to replace all 2/3 if you have a failure of one, as they are all on the same board.

That jives with what someone with the failure said here recently about being told they have to replace the ‘master charger’ or something to that effect. I assumed that hinted at some relationship between chargers, but the part number posted by someone referring to the Tesla parts site listed only one part. Not sure why you’d call it the “master” charger then? Unless there is yet another redundant tiny(?) charger built-in somewhere on the pack / in the penthouse for when the “master” charger fails completely?

It seems to me they’ve split it into pieces, at least in part, for redundancy. If one fails you can still charge. Not necessarily for replacement costs.

They have redundant power and data lines, redundant power steering, “everything” is redundant.

I caught a sentence in the autonomy day that I forgot about actually ... Elon mentioned you could lose the “main pack” and still drive on the “auxiliary pack”.

I haven’t seen this mentioned in any tear downs, so I wonder if it’s built into the “main pack” but a portion is wired separately. Or what. He can’t mean the 12V. Did they hide some uncooled 2170’s somewhere nobody has looked yet? Doubtful. Elon using hyperbole or inaccurate explanation? Possible if not probable? LOL.

Found it. It was right after the 3rd presenter who I’d gone back to skim through after he was confirmed to not be at Tesla any more... ya, so here cued up at 3:00:19 mark, for about 30s. “Steering and braking” from the “auxiliary power system” ... I guess that’s the 12V then ... no Easter egg packs

 

[AlanSubie4Life]

“Steering and braking” from the “auxiliary power system” ... I guess that’s the 12V then .

I guess that implies the power steering and brake boosters run off the 12V system, which makes sense, though I hadn't thought it through before. Unless there's a boost converter from 12V->400V (I assume not). Anyway there's 0.5kWh in the 12V or so, so that would last a while since it doesn't need to power the motors.

 

[darth_vad3r]

I guess that implies the power steering and brake boosters run off the 12V system, which makes sense, though I hadn't thought it through before. Unless there's a boost converter from 12V->400V (I assume not). Anyway there's 0.5kWh in the 12V or so, so that would last a while since it doesn't need to power the motors.

Should last long enough to slow down to a stop and pull over (hopefully) safely.