A discussion of gliding and a comparison of methods: pedal control vs. neutral

The CAN-View data capture and analysis fun continues, especially on a lazy Independence Day with nothing to do after eating too many burgers ....

As pulse and glide regulars know, the conventional method of triggering a glide is through go-pedal control -- completely lifting off and feathering back down until all arrows disappear on the MFD Energy monitor. It is fairly easy but takes a bit of practice to hit and hold.

My CAN-View observations of pedal position values have suggested that a pedal controlled (PC) glide can be executed with pedal position anywhere from 1-10%. Despite the MFD showing no arrows, there actually is a small amount of current flow out of the battery. But interestingly, there appears to be no relationship between pedal position and current flow until after the 10% threshold is exceeded:



This and my seat-of-the-pants sensations of drag suggest that the car no longer is gliding above 10% pedal position. Presumably the no-arrows condition disappears at that mark, but since my CV uses the MFD for its display, I can't monitor it and the Energy screen simultaneously to verify.

There is a relationship between speed and current flow with PC glides:




Some have suggested using neutral for glides. It's easy and requires no practice. After recording these PC glide data, and then observing on the display a fairly low amount and narrow range of current flow (seemingly less than 1.5 amps) with N glides, I have increasingly used neutral -- and, of course, I have recorded data:



This confirms that current flow moves within a more narrow range than with PC glides, at speeds below 30 MPH. Interestingly as speed approaches 30 it appears to creep up a bit, and then above 30 it's all over the map, even occasionally into the battery. Finally, what's most interesting to me is this blip:



This is a neutral glide that I let transition to warp neutral (i.e., >40 MPH) coming down a fairly steep bridge approach. It carries me across the bridge and then up a short incline to a toll booth, where the last data point is.

On this and other steep hills previously I've noticed a brief surge of positive current at speeds around the mid-30s, often spiking to well over 10 amps, along with a simultaneous sensation of increased drag as though I had let it slip into regenerative coasting. (It might have been over 10 amps here too but CAN-View captures data only every two seconds or so, and that's about the duration of the positive surge.) Then something I hadn't noticed before are the fluctuations of current flow in warp neutral being more erratic than most of those at slower speeds. I'd be curious to hear from any of the experts an explanation of these phenomena.

So the question became: "Which method produces less drag?" With the answer to that might come the answer to this one: "Which method is more efficient?" So I did a head-to-head test of the two methods.

The test consisted of four runs with each method on a four-lane road through a lightly traveled industrial park. The route is shown here. There was a net elevation decrease of approximately 36' from start to finish.

I aimed to begin each run at approximately 28 MPH. Each glide was allowed to continue to a defined ending point 0.6 mile from the starting point.

Outside air temperatures were in the low 90s, and winds were out of the WSW at around 15 MPH -- both as reported by Weather Underground's almanac information for my locale on test day.

Climate control was off to avoid the significant current draw from air conditioning. Both front windows and the right rear window were opened for driver comfort and ventilation for the hybrid battery.

Results, with charts for speeds and hybrid battery current flow:





As suspected, there clearly is more drag with neutral glides than with pedal-controlled glides.

The CAN-View data capture records battery SOC in half-percent increments. All glides began with SOC between 55.5% and 59%. SOC predictably dropped more with PC glides than with N glides: 1%-1.5% with each of the former and 0%-0.5% with each of the latter.

So it seems that with lower current flow, N-glides are more efficient, especially between 30 and 40 MPH. I'll hedge a bit by saying this conclusion is subject to correction, given that I don't understand everything going on under the hood as well as some. But it seems generally accepted that use of battery power and subsequent replenishment introduce inefficiencies from conversion losses, and better from an efficiency standpoint is to avoid the battery as much as possible.

I have incorporated these findings into my everyday driving, using N glides preferentially when safe, especially between 30 and 40 MPH. I use PC glides when preservation of momentum (e.g., heavier traffic) is more important than preservation of SOC, or when I want to quickly alternate between a glide or outright coasting (e.g., timing my approach to a red light).

It should be stated that using neutral on downgrades is illegal in many states. But states with such a law usually don't seem to have a similar prohibition on using neutral on level terrain or uphill, as suggested by a recent CleanMPG thread on the subject.

 

It might be better said that N glides use less battery power rather than conserve more. Hobbit has reported that current flow seen with PC glides in fact is indicating a small push. I suppose you could say a PC glide actually is a low-demand EV mode of sorts.

There can be a drop in momentum shifting back into D if the shift occurs with your foot off the pedal -- you have regenerative coasting until you begin the next pulse. But what's great for this purpose is that, while in neutral, the pedal is totally disassociated from the drive system. With practice you can learn approximately where to position the pedal for the upcoming glide, and you can have it in position well in advance of the shift. Then the ICE spins up almost immediately with the shift.

As for MPG results, I can't draw any firm conclusions yet. It seems intuitive they would improve. It is generally agreed that deliberate and sustained higher-demand EV mode operation hurts fuel economy. There is no reason in my mind that this "low-demand EV" would help it. The battery charge still has to be replenished from the ICE.

My fuel economy has improved recently, but that's to be expected anyway because of summer temperatures. I certainly haven't seen any dropoff. But I am reluctant to say, "Hey, look, my fuel economy is better because of [fill in the blank]" without a controlled test. There are so many variables that affect FE from trip to trip and from tank to tank -- weather chief among them. I'd like to test this with a series of runs on the same day in the same weather conditions, unimpeded and unhurried by traffic.

Instead of saying there is more drag with N glides (which some may interpret as a negative), my conclusion might be stated better that there is less battery assist.

Oh, and Tony ... I concede the point. About the burgers, that is.

 

Me too! I’m a relatively uneducated (in technology, anyway) data geek. I’m on the lookout for those more knowledgeable than I to tell me exactly what some of these data mean, and to correct any inaccurate conclusions.

You are correct in concept, and probably in the ballpark with 1/3, though I’d be reluctant to pin it down with a precise number. Incremental SOC changes obviously are pretty small. It would be nice to see SOC readings for this purpose at least down to a tenth of a percent. But the CAN-View data capture reports it only in half-percent increments, so there is considerable opportunity for rounding errors.

After reading your post, I did some additional math on my data. The average SOC loss for N glides was 22% (relative percent, not absolute) of that for PC glides: 0.25% per run vs. 1.125%. Then I averaged current flow and approximate time for each run and figured the average amp-hours used. N glides used 39% of the current of PC glides: 0.091 Ah per run vs. 0.035. But with that calculation comes another opportunity for error. Current flow of course is continuous, whereas CV’s data capture essentially does data sampling. And the sampling interval actually is slightly less than two seconds and seems to vary somewhat. I use two seconds (and state that they are approximate time intervals) for convenience and to keep the math simple.

I don’t consider battery energy to be providing any push during a neutral glide. The car’s “basal metabolic rate,” so to speak, is close to 1A. I generally run with the radio on, and I did so during these tests, so that adds a few mA. There’s not much further to go to reach the average 1.3A of the N glide runs. EDIT: I am also powering my laptop via a plug-in inverter during my tests, so there's a few more mA.

On the other hand, as I mentioned earlier, Hobbit states that there is some push with PC glides. The CV data stream also includes battery voltage, so I calculated average battery power for each glide to be 286W for N glides and 820W for PC glides. After converting the difference to horsepower we could say that PC glides provide approximately 0.7 hp extra thrust.

Thanks, by the way, for giving me the excuse to run these numbers!

Try this shift mid-glide a few times on a steep hill, and I think your butt check results will be revised.

OK, I’m a little puzzled. I’m not sure what you’re saying about not going “into coast” after shifting into D. And what is the “35 mph speed requirement for gliding” you’re referring to? Since you mention S3b, maybe it’s referring to the speed threshold between S3a and S3b. Finally, what is the “delatching” you’re referring to?

Anyway, my tests were all done with car in S4. For S3, as I’m sure you know, you can’t glide below 34 MPH without either an EV switch or beginning the glide above that speed.

 

OK, it's starting to sink in. I read on a "See Spot run" level, so it takes awhile sometimes.

I do routinely experience a short episode of the car refusing to glide after entering S4, but it has nothing to do with my glide method -- nor, I think, whether I'm gliding at all, but I can't say that for sure because it's so infrequent that I'm not doing some semblance of P&G at those speeds. Anyway, according to Hobbit, it's from the secondary cooling loop opening to completely heat what remains in the thermos. (I hope I've paraphrased accurately.) But it's not the behavior you describe, which I've never noticed. Instead it refuses to glide (actually more typical of S2) but it resolves spontaneously after a minute or so.

Having said that, I'll keep an eye out for what you describe, especially as weather starts to cool in a couple of months. It's been only recently and after gathering these data that I've been using N glides with any regularity.

 

OK, it's done, at least with one long-ish run with each method:






The test was performed on this route. The route is mostly four lanes, 45 MPH, lightly traveled, in the vicinity of some industry stuck in the middle of an otherwise rural area. It is relatively flat with some very gentle hills. Net elevation change from end to end is less than 15'.

Two continuous round trips, a total of 12.4 miles each, were made for each method, beginning and ending at the northern end. That end has a stoplight, so time spent awaiting the light after the first round trip is highlighted on the charts. The neutral glide test was performed first.

Weather was mostly sunny, temperature around 90F, light winds out of the SW. I used mobile broadband to stay continuously connected to Weather Underground, monitoring real-time temperature and wind readings at the closest weather station less than a mile away. Temperature increased by one degree and wind speed by one MPH between starting and ending the tests; wind direction remained constant. Modeling performed with the Prius MPG Simulator suggests that these changes would mostly offset each other insofar as their effect on fuel economy is concerned, so they were considered insignificant.

As before, I kept climate control off to eliminate its effect on battery discharge. All windows were opened for comfort and hybrid battery ventilation.

Hybrid battery temperature, incidentally, held fairly steady. For the N glide test all modules remained at 100F throughout. For the PC glide test, the coolest module stayed at 100F while the hottest ranged from 100F to 102F.

The pulse range was 15-40 MPH, other than (obviously) leaving or approaching the ends of the test route. A bit of on-the-fly judgment was required, along with an occasional rapid pedal adjustment, in executing pulses approaching the ends, trying to time the speed and distance of those pulses to minimize the need to brake. Braking (predictably) still was required at both ends, but the charts suggest I succeeded in keeping it to a minimum.

Because of the previously-discussed issue of state of charge (SOC), I attempted to begin each test with an identical SOC. Obviously, however, the tests were not likely to conclude with the same SOC, and the charts demonstrate that to be the case. In fact, one of the more significant data findings (to me) is the steady rise in SOC during the N glide test as compared to the regular undulations seen during the PC glide test. So obviously the charge replenishment seen with the PC test pulses was enough only to offset the discharge during glides -- this despite those pulses obviously producing more current flow into the battery than the N test pulses.

To adjust for the effect of greater battery discharge during PC glides, I included a period of force-charging at the end of the PC glide test to bring the charge to the same level as where it ended with the N glide test. (Force-charging is done, with the car in “D,” by depressing the brake [to keep the car stationary] and accelerator simultaneously, the latter to force the ICE to light and rev enough to generate enough current to charge the battery.) Here are the various SOC readings via the CV display and data capture:

	Column 1	Column 2	Column 3	Column 4	Column 5
0	SOC reading	N glide[/B]	PC glide[/B]
1	Beginning	display	56%	56%
2	Beginning	recorded data	56.5%	56.0%
3	Ending	display	before force-charge	60%	55%
4	Ending	recorded data	before force-charge	60.5%	55.5%
5	Ending	display	after force-charge	NA	60%
6	Ending	recorded data	after force-charge	NA	60%

The charts show raw ending MPG figures for both tests and the force charge-adjusted MPG for the PC test. With the force-charge adjustment, the N glide test shows modestly better fuel economy.

Some limitations to the test: Even though the distance was identical and weather conditions nearly so, minor variations in technique and weather could have been confounders. This includes the variations in number of P&G cycles – not surprising, given the greater speed decay and shorter duration of N glides – and the aforementioned adjustments of pulse speed and length approaching the ends of the course. Also, there could have been transient wind speed and direction changes not reflected in the local weather station's recordings.

So the two methods appear at least comparable for fuel economy, with a possible slight edge in favor of using neutral to glide. With the test limitations and the relatively minor adjusted difference, I still am reluctant to state that unequivocally. A longer test on a closed course with no stops would be ideal, though probably impractical and maybe even impossible in my area.