Cold-start engine knock mystery @198,000 mi (and engine condition generally)

Since around 198,000 miles I've had an ominous knock/tap from the engine. Strangely I've just found this thread which seems to be describing something similar, but it is about much newer Prii, and mine never had this sound until last month.

The weather had just gotten chilly. On a cold start, first thing in the morning, there is an irregular, sharply metallic, shaking-the-junk-drawer knock that starts the moment the engine actually fires (a few seconds after MG1 starts spinning it) and subsides after a couple seconds. This can be loud enough to startle somebody standing outside the car on the far side of the driveway.

What's left after the irregular knock subsides is a perfectly-regular, RPM-linked, knock/tap, not loud but noticeable at idle, louder when some moderate load is put on the engine. Quieter once the engine fully warms up, but still noticeable.

To me it sounds like RPM/2. Some things I've read suggest it has to be valve train then and not, say, a rod, because the cams run at RPM/2. (Is that a safe call? I've never heard rod knock in person - what does that really sound like? I'm guessing the rod's under downward force for 3/4ths of the cycle and the force would only change direction above and below the intake stroke - does that make a gallopy double-knock sound at RPM/2, or is it not that simple?)

Listening

So far, I've:

1. Taken it to an independent mechanic to listen. Weather had a warm snap that day and he didn't hear it. (I don't get this part: in cold weather I can hear it even after the engine warms up. But in warm weather it's hard to hear. Huh? The thermostat works: engine warms up to the same temp either way. Weird.)
2. Made a recording on the next cold morning (right after the guy couldn't hear it). Made another appointment. Weather snapped warm again but he could hear the recording. He thought valve clearance or chain slap. (I'm skeptical about the chain idea because the sound is so regular.)
3. Taken it to the dealer. Another warm snap day of the appointment. Played the recording. They suggested I leave it overnight during the next cold snap so they could start it first thing in the morning.
4. Left it overnight. They said they heard it on startup but it didn't stay loud enough long enough to trace the sound. Offered to sell me a valve adjustment and/or a water pump.

Ok, at least that's two sources that didn't think it was a rod, so I'm relaxed enough to take a break from getting engine prices and spend a little more time on diagnosis.

Valve clearances at 198,000 miles

Last weekend I measured the current valve clearances. I now know why the dealer quotes ~ $800 to adjust these. Good grief! And I wasn't even removing the cams, as I'd have to for adjusting instead of just checking.

Measured (spec is intake .17mm to .23mm exhaust .27mm to .33mm cold):

	Cylinder	Valve	Inch	mm
1	1	ex1	.010"	.25mm	tight
2	1	ex2	.011"	.28mm
3	1	in1	.007"	.18mm
4	1	in2	.007"	.18mm
5	2	ex1	.012"	.30mm
6	2	ex2	.011"	.28mm
7	2	in1	.006"	.15mm	tight
8	2	in2	.007"	.18mm
9	3	ex1	.010"	.25mm	tight
10	3	ex2	.010"	.25mm	tight
11	3	in1	.007"	.18mm
12	3	in2	.007"	.18mm
13	4	ex1	.0095"	.24mm	tight
14	4	ex2	.010"	.25mm	tight
15	4	in1	.007"	.18mm
16	4	in2	.007"	.18mm

Note: these measurements weren't on a stone cold engine. It had been shut down for about 2 1/2 hours on a windy, 60-ish-degree day by the time I made the first measurement. It was all cool to the touch. Also, instead of feelers, I measured by setting a dial indicator to the top of each lifter, then working the lifter up and down with a magnet-on-a-stick. (Seemed more repeatable to me - I never liked having to judge what "slight drag" on a feeler should feel like.)

Looks like nothing is very badly off there (not bad for 198,000 miles?), and what is off is tight, not loose. I guess eventually I should tear in again and fix the tight clearances to avoid burning valves (unless I decide it's less hassle to just swap an engine). But to explain the noise, I would have expected loose clearances, so I don't think I've found any smoking gun in the clearances, have I?

One observation

When I first got the valve cover off, I noticed the #4 exhaust cams did not seem to show the same film of oil I saw on the others. A clue? Earlier with engine running, my stethoscope found the valve train noise extra prominent in the area of the valve cover over #4 exhaust.

I don't understand the lubrication system well enough to know how oil reaches the cam lobes. The valve cover itself is surprisingly complicated: there's a steel stamped baffle inside with some holes that look as if they might pour oil over the lobes, if it were somehow fed with oil up from the head, but I never spotted anything that looked like a passage to feed it. And right over where the #4 exhaust lobes would be, the baffle didn't seem to have any holes anyway, so if that was why the lobes looked dry then for some reason it was built that way 11 years ago.

Another explanation might be just that the last thing I had done after the engine last ran, and before taking the valve cover off, was a leakdown test. I had turned the crank a few times around by hand, which might have left the lobes a little dry with no oil pressure to speak of. I don't know.

Leakdown test results

I don't have any compression baseline from when I bought the car, because I don't have a Prius-aware scanner that can tell it to crank for a compression test. But last week I picked up an OTC 5609 Cylinder Leakage Tester, basically a regulator and two pressure gauges with a calibrated orifice between, using an air compressor to test without having to crank the engine. It's cheaper than any Prius-aware scanner I know of. (Heck, even if you don't have an air compressor you can buy a small one and it's still cheaper.) It also seems more repeatable and useful than the old way of checking compression.

The reading is made by setting the regulator to a chosen pressure on the inlet gauge (in my case 75 or 90 psi) and reading the outlet gauge (downstream of the orifice), which can be converted to "%" with a chart in the tool case.

	Cyl	@75psi	@90psi	%
1	1	70	82	7% 9%
2	3	69	missed	8%
3	4	72	85	4% 6%
4	2	73	87	3% 3%

(I missed one reading when the note-taking app I was using didn't record me.) Strictly speaking, the "%" figures can't be compared to others using different testers because apparently different testers can use different orifice dimensions (unless they're sold to meet the FAA standard for aircraft). Mine doesn't say anything about its orifice size.

Still, what I read online makes these readings sound ok for an 11 year old engine. The spark plugs have the light tan deposits considered normal.

There doesn't seem to be much wrong with this engine after 11 years, except for this one noise. Trouble is, if it doesn't turn out to be valve train, I don't think the double-decker oil pan would even let me inspect the bearings without pulling the engine, and at that point it's just as easy to put a different one in.

I still like the car too much to want to punt.

Any thoughts?

-Chap

Edit: update tables in ancient post because XenForo keeps changing how tables work

 

Oh, sorry. Since I bought the car (at 125,228) I've been using Carquest's house-brand full synthetic (it's an Ashland label) in the user-manual spec'd 5w-30 weight. I think I did one change of Castrol synthetic before I stumbled on the house-brand stuff.

The morning temperatures the last month have been all over the map - ~ 30 to 60 F depending on the day. It always warms up on the day of a mechanic appointment. Warm morning => they don't hear the noise.

That kind of thing occurred to me. At the moment I just lack the experience with audio/signal processing software to easily do stuff like that; I'm sure there are packages that don't make it very hard. I just made that recording as a phone video and dragged it into GarageBand to export the audio. GarageBand shows a cute waveform representation and I toyed briefly with trying to pick out the knock period and confirm its relationship to RPM, but I probably need some software a little bit geekier than GarageBand for that.

I wonder if there's a way to hook a circuit to the cam position sensor and make an audible tick that could be recorded. The cam encoder looks as if it just has small bumps 90 degrees apart, with one I think missing as an absolute index.

I guess the drawback to using either spark or cam sensor for indexing is that both timings vary. The crank sensor might be better for true position (maybe combined with spark or cam to show which half of the cycle - ok, it's getting complicated.)

I should mention the recording starts just before ready-on, with the big rattle a few seconds later. Then it runs for a bit, and you can hear me shift into D (those shifter clunks are really loud) and do a force-charge to put gradually more load on. Around 1:35 where the ScanGauge "LOD" hits about 57% the regular knocking is especially audible.

Yer kidding - the workaround for xenforo adding ridiculous vertical space above tables is to strip out all the newlines within the tables??

Thanks for the tip....

-Chap

 

Is it really that the cam sensor drives the spark, or more that the ECM consumes the crank and cam sensor signals, magical computations happen, and then the ECM drives the spark?

I don't think the cam sensor has much resolution; when I looked at the rotor it seemed to have only 3 raised blips on it 90 degrees apart (with no blip where the fourth one would go), which is probably about the bare minimum signal the ECM could use to determine the VVT cam angle, and only by combining it with the crank sensor signal.


I should have checked: hobbit's got all the info, natch.

If you're accustomed to Audacity, according to this it ought to be possible to import M4A in versions later than 1.2.x.

I'm sorry, I had the choice of exporting with the MP3 or AAC codec and I thought the latter was common enough these days.

-Chap

 

As far as I know in a cold start MG1 always spins the engine at around 1200 rpm for a few seconds with no injection or spark, gets the oil circulating, and then the ECM starts driving the injectors and sparks and the engine is truly started. Ordinarily that isn't noticeable because spinning up to 1200 sounds like starting to most people, and the actual start a few seconds later doesn't sound very much different (unless the engine is mine, recently). That's why you can often see postings in here like "my engine starts
fine, runs for a few seconds, then stalls" when somebody has an engine that's really not starting at all, but they hear it being spun up by MG1 for as long as it takes the ECM to realize it hasn't started and give up.

So I'm pretty sure the answer to my weird delayed knock puzzle is simply that the knock begins when the engine actually starts. Even so it's a little bit puzzling because it almost sounds as if oil is reaching some bearing late, when I thought the whole point of the prolonged spin-up before starting was to have the oil pressurized in advance.

I don't have an oil pressure gauge, but I have thought about getting one. I'm not looking forward to trying to get to the pressure switch to install a gauge. Looking in the manual makes me wonder whether I could even get to it with the intake manifold in place (but I haven't gone out to look carefully and see if there's a practical way).

If I have to do hours of contortionism just to diagnose what's going on with this engine, I might just want to spend the hours pulling it out and putting a newer one in.

-Chap

 

Has anybody reading this ever been into the valve train or happened to learn how the oil is meant to be supplied to the cam lobes and lifters? I don't see them mentioned in the oil flow diagram in the manual.

I think the camshafts are hollow and carry oil to their own bearings ... do the bearing caps then have jets that oil the lobes, or what? Is oil just escaping the sides of the bearings enough? In my short time in there I didn't happen to clearly see how the lobes are supposed to be lubed. (I mentioned the elaborate construction of the valve cover itself, which I don't understand at all, but I never saw anything that looked like a passage to carry oil up there. Unless it catches oil slung off the chain, or something ... ?)

Thanks,
-Chap

 

This is from under the valve cover back when I took the valve lash and leakdown measurements ... I've had the video in my phone since then, but needed to grab a good frame to show what I was talking about with the #4 exhaust lobes. (Oddly, the latest Android camera makes it very easy to pick out a still while recording video, but if there's a way to pick one out later I never found it, and wound up doing that part on a Mac in way too many steps.)

The #4 exhaust lobes are the ones at the upper right. You can see how all the other lobes visible in the frame are visibly wet with yellowish oil (e.g. the #4 intakes) or even darker varnishy oil (we see the low side of the #3 exhaust lobes, where they never touch to get wiped clean). By contrast the #4 exhaust lobes look like squeaky-clean silvery metal - they were the only ones to look like that. Heck, even the boss the exhaust lifter rides in seems to be clean of the dark varnish that's managed to cover everything else.

I don't know if there's a clue here, or if there is, a clue to what.

Everything is pretty dark with this varnish, which, from what I read, is not the same as sludge. I didn't see what looked like any sludgy accumulation. The varnish is mostly a thin tenacious coating that doesn't rub off easily - though in a couple of places on the valve cover and the #1 cam bearing cap, it can flake off into thin flat flakes.

Any thoughts?

Thanks,
-Chap

 

Update: I've run one oil change interval with a cleaning additive. It doesn't seem to have affected the sounds much, but the weather has warmed, so comparison is difficult. For what it's worth, it does seem to have some cleaning effect, just looking at my oil drain plug. Also my rate of oil consumption over the oil change interval seems to be back to what I was used to (last summer/fall it seemed to be getting worse). But I never actually made quantitative notes on consumption and this was an unusually short OCI so a grain of salt may be needed here.

I have read some other posts possibly linking sounds like mine to cylinder power imbalances with causes in the fuel or ignition systems. There's a relatively inexpensive injector testing and reconditioning service I might try next. Again at 204,000 miles it won't hurt to get those parts cleaned, it's not a whole lot of money, and the test reports will be interesting even if there isn't any effect on this knocking issue.

Somewhere I read of an old-timer technique to check for grossly excessive rod clearances by checking piston motion with a probe through the spark plug bore while rocking the crank. That sounds like the sort of thing I would think of and I'm kind of surprised I didn't. I've got something like 6,000 more miles on the car now since I first noticed the noise and it's not growing rapidly worse, and a couple mechanics have said it doesn't sound like rod knock, so I'm chiefly looking elsewhere, but this test seems simple enough I think I'll try it when I next have time to have the plugs out.

-Chap

 

Anybody happen to know the size/thread of the oil pressure switch?

Also, is it relatively easy to reach from below with the engine under covers removed? I'm hoping it is, 'cause it sure seems buried from the top.

-Chap