Mistakes were made, my R34 Story

[GTSBoy]

The trouble with phone apps is that they really do not know the sensitivity of the mic on each specific phone, whether there is some sort of cover or skin that can change that, etc etc. So the readings off them are not calibrated. You certainly couldn't be sure that you were even within 3 dB.

I have the same app on two (actually more than that) different phones/tablets and the readings in the same room at the same time are never the same.

I don't mind the apps - they are as useful as a dyno - if you're testing for delta from thing1 to thing2. But not to try to generate some sort of absolute value.

[Dose Pipe Sutututu]

Done, I'll do mine soon too - whenver I take out the car next.

I can cheat and make it quieter by turning off VCT at idle too - but I lose the braps

[Dose Pipe Sutututu]

2 hours ago, GTSBoy said:

The trouble with phone apps is that they really do not know the sensitivity of the mic on each specific phone, whether there is some sort of cover or skin that can change that, etc etc. So the readings off them are not calibrated. You certainly couldn't be sure that you were even within 3 dB.

this, hence the comment about shipping the dB meter lol.

 

@Kinkstaah do you have a Google Pixel 9 Pro XL by any chance? lol

[Kinkstaah]

I definitely don't.... I have a Samsung S9.

Though for giggles I did this as I was messing with idle calcs anyway and thought, well.
This is with the Civic parked right behind it. Basically almost touching. I have no idea if that affects anything but this is hardly a fully scientific test lol.

Varex Closed:

Varex Open:

[Duncan]

I'll bet an app is close enough to be usable for dB; since it is a logarithmic scale; even a 10% error in readings wouldn't make 1 dB difference

[GTSBoy]

2 hours ago, Duncan said:

even a 10% error in readings wouldn't make 1 dB difference

'sif the error will be that low!

[MBS206]

My embedded systems thoughts have me sitting with GTS on this.

Variation between same phone hardware, should be small.

However, the internal "intensity" or "volume" amount that say Google passes to the app, will be quite different, as the underlying hardware will be passing different levels for the same volume to the Google OS.

Until the app creator has had each individual phone, and set benchmarks and calibrations for each, the amount of error can be quite huge.

 

It can even be observed by using different phones, recording the same noise, and then playing it back, they end up soon ding different.

A big reason for it, is even the different types of mics used in phones have different responses, and different frequency ranges. Then you need to get into the DSP, and the variations in those, their sample rates which then effect their frequency range, and then the quality of the DSP, and what type of hardware conversion they do to for the ADC within the DSP.

Oh, and let's not forget at the low level phones are designed to cutout loud sounds. It's one of the reasons they suck in really loud environments (eg concerts). The louder you yell, the more you'll get cutout too

Note DSP is Digital Signal Processor

ADC is the analogue to data converter.

I don't have any real data on what the variation would truly be, however, chat GPT says in general, their output is typically between +/-2dB to +/-5dB of what you're really measuring. So realistically, anything from 4 to 10dB variation is possible even with the same devices.

[Kinkstaah]

Well, I'm back from the dyno today.

Some things do partially make sense. The pod filter/airbox delete picked up between 6-10rwkw on 98 - because heat soak does kind of affect things and there was playing with tune/timing/AFR.

Oddly enough, the car was running much leaner than before. So lean it was audibly pinging on the dyno which I got video of:
 

 
We're talking going from ~12.7 in the past to the first run being at like ~14.0. It is now tuned to ~12.5 on the Dyno, which correlates to about ~12.1 on my wideband in the car. These matched last time, which is very odd. The dyno plots only show the dyno's reported AFR - should be last time, yet now it no longer agrees and was way leaner.

Nobody has an explanation for how a pod can make the car run notably leaner, yet not really give any more power when you add fuel in.

A few different types of intake design were tested:

 

 
There was no difference other than IAT with the pod sticking out of the bay. The pod sticking out of the bay (but connected) is actually still warmer than what I usually see on the road.

Removing the pod entirely lost about ~2kw. But to be fair, all of the runs could be argued to vary by that amount when temperatures climb etc etc. It's safe to say that the filter isn't causing any restrictions of any note that can be reasonably altered in any way. This is in line with what I'd expect given the Engine Masters testing.

323KW on 98 and ~335KW on E85 is actually a pretty solid result, up about ~45kw from 99% of LS1 cammed combos, with generally much larger cams/exhaust etc as well.

It is after all up 42KW (98) and 54KW (E85) from before.
+10KW from a pod and removing the box is cheap as chips compared to what the head work cost per kw

No, I did not get to drop the exhaust and test. When it comes to exhaust... it all just seems to change frequencies and cost or gain 2hp here or there. I don't realistically think I'll drop this to test it - because there's not much else I can really do about it/route it any other way/make it bigger/just bought mufflers. Engine masters beat the hell out of headers with a hammer to deliberately kink them and didn't lose power at all, I sincerely doubt that going larger primaries would help. If it were even possible for clearance/conversion reasons... which it's not...

I may throw the E85 in there at some point and do a drag run to see what MPH it traps for science.

It isn't lost on me that ~320kw Skylines do trap about the same MPH that ~370kw F-Body/Corvettes do in the USA for the same  or similar weight. (122-125mph).

Of course, if I go there and trap 104mph or something then I'll just 'accidentally' have an accident on the way home from the drag strip and buy a M4.

[MBS206]

HPA do an interesting video on tuning, and Adrian shows on the dyno, moving from 10:1 to 12:1 (And I think even 13:1) and the power moves SFA.

However, a couple of degrees of timing has a HUGE impact.
Exhaust side of things will more be the length of the exhaust and how many bends and turns you have. The straighter you can keep it, the better. By a lot. Oh, and getting the diameter right. Bigger isn't always better. Going to big can actually lose you power.

[admS15]

Decent result Greg, sounds tuff AF👍

 

When is the Whipple or procharger getting fitted?😉

[Kinkstaah]

I'm back from the dyno - again!

I went looking for someone who knew LS's and had a roller dyno, to see how it shaped up compared to everything else and confirm the powerband really is peaking where Mr Mamo says it should.

TLDR: The dyno result I got this time definitely had the shape of how it feels on the road and finally 'makes sense'.

Also we had a bit more time to play with timing on the dyno, it turns out the common practice in LS is to lower the timing around peak torque and restore it to max after. So given a car was on the dyno and mostly dialled in already, it was time for tweaking.

Luis at APS is definitely knowledgable when it came to this and had overlays ready to go and was happy to share. If you map out your cylinder airmass you start seeing graphs that look a LOT like the engine's torque curve.

The good thing also is if you map out your timing curve when you're avoiding knock... this curve very much looks like the inverse of the airmass curve.

The result? Well it's another 10.7kw/14hp kw from where I drove it in at. Pretty much everywhere, too.

As to how much this car actually makes in Hub Dyno numbers, American Dyno numbers, or Mainline dyno numbers, I say I don't know and it's gone up ~25kw since I started tinkering lol. It IS interesting how the shorter ratio gears I have aren't scaled right on this dyno - 6840RPM is 199KMH, not 175KMH.

I have also seen other printouts here with cars with less mods at much higher "kmh" for their RPM due Commodores having 3.45's or longer (!) rear diff ratios maxing out 4th gear which is the 1:1 gear on the T56.

Does this matter? No, not really.

The real answer is go to the strip and see what it traps, but:

I guess I should have gone last Sunday...

[GTSBoy]

You Gregged a whole racetrack!?

[MBS206]

All I hear is ROADTRIP!

[Kinkstaah]

On 01/05/2025 at 4:11 PM, Kinkstaah said:

I believe the resistance chart is this

How the f**k do I convert this to a voltage?
Once I have a voltage... I then have to perform this transform of said voltage to show it in the scanner:

So I found this:

https://www.efihardware.com/temperature-sensor-voltage-calculator

I didn't know what the pullup resistor is. So I thought if I used my table of known values I could estimate it by putting a value into the pullup resistor, and this should line up with the voltages I had measured.

Eventually I got this table out of it by using 210ohms as the pullup resistor.

180C 0.232V - Predicted
175C 0.254V - Predicted
170C 0.278V - Predicted
165C 0.305V - Predicted
160C 0.336V - Predicted
155C 0.369V - Predicted
150C 0.407V - Predicted
145C 0.448V - Predicted
140C 0.494V - Predicted
135C 0.545V - Predicted
130C 0.603V - Predicted
125C 0.668V - Predicted
120C 0.740V - Predicted
115C 0.817V - Predicted
110C 0.914V - Predicted
105C 1.023V - Predicted
100C 1.15V
90C 1.42V - Predicted
85C 1.59V
80C 1.74V
75C 1.94V
70C 2.10V
65C 2.33V
60C 2.56V
58C 2.68V
57C 2.70V
56C 2.74V
55C 2.78V
54C 2.80V
50C 2.98V
49C 3.06V
47C 3.18V
45C 3.23V
43C 3.36V
40C 3.51V
37C 3.67V
35C 3.75V
30C 4.00V

As before, the formula in HPTuners is here:
https://www.hptuners.com/documentation/files/VCM-Scanner/Content/vcm_scanner/defining_a_transform.htm?Highlight=defining a transform

Specifically:

Quote

The Function fields define how the numerical values associated with the input are converted to their corresponding output values. In the sample screen above, the function is converting values in the range 0-5 (volts) to values in the range 0.62-1.10 (Lambda).

To define the function, you just need two known input values and their corresponding output values. Consult the documentation for the device whose output you are transforming.

Once you have these values, the number to put in the first box can be calculated with the following formula:

(Input2 - Input1) / (Output2 - Output1)

If we used the max and min values from the example above, you would get:

(5 - 0 volts) / (1.10 - 0.62 lambda) = 10.41666

To determine the value of the second box, you simply plug either of the two known input and output pairs into the resulting equation and solve for the missing value.

In our example, if you insert the 0 volt input and it corresponding output value, you get:

0/10.4166 + X = 0.62.

So, X = 0.62.

In my case I used 50C and 150C, given the sensor is supposedly for that.

Input 1 = 2.98V
Output 1 = 50C

Input 2 = 0.407V
Output 2 = 150C

(0.407-2.98) / (150-50)

-2.573/100 = -0.02573
2.98/-0.02573 + 47.045 = 50

So the corresponding formula should be:

(Input / -0.02573) + 47.045 = Output.

 

If someone can confirm my math it'd be great. Supposedly you can pick any two pairs of the data to make this formula.

[GTSBoy]

28 minutes ago, Kinkstaah said:

(Input / -0.02573) + 47.045 = Output.

Nah, that is hella wrong.

If I do a simple linear between 150°C (0.407v) and 50°C (2.98v) I get the formula

Temperature = -38.8651*voltage + 165.8181

It is perfectly correct at 50 and 150, but it is as much as 20° out in the region of 110°C, because the actual data is significantly non-linear there. It is no more than 4° out down at the lowest temperatures, but is is seriously shit almost everywhere.

I cannot believe that the instruction is to do a 2 point linear fit. I would say the method I used previously would have to be better.

[Kinkstaah]

God damnit. The only option I actually have in the software is the one that is screenshotted.

I am glad that I at least got it right... for those two points.
Would it actually change anything if I chose/used 80C and 120C as the two points instead?

My brain wants to imagine the formula put into HPtuners would be the same equation, otherwise none of this makes sense to me, unless:

1) The formula you put into VCM Scanner/HPTuners is always linear
2) The two points/input pairs are only arbitrary to choose (as the documentation implies) IF the actual scaling of the sensor is linear.

then

3) If the scaling is not linear, the two points you choose matter a great deal, because the formula will draw a line between those two points only.

[GTSBoy]

46 minutes ago, Kinkstaah said:

I am glad that I at least got it right... for those two points.

No, your formula is arse backwards. Mine is totally different to yours, and is the one I said was bang on at 50 and 150.

I'll put your data into Excel (actually it already is, chart it and fit a linear fit to it, aiming to make it evenly wrong across the whole span. But not now. Other things to do first.

[Kinkstaah]

2 hours ago, Kinkstaah said:

I feel I should re-iterate. The above picture is the only option available in the software and the blurb from HP Tuners I quoted earlier is the only way to add data to it and that's the description they offer as to how to figure it out.

The only fields available is the blank box after (Input/ ) and the box right before = Output. Those are the only numbers that can be entered.

[GTSBoy]

38 minutes ago, Kinkstaah said:

I feel I should re-iterate. The above picture is the only option available in the software and the blurb from HP Tuners I quoted earlier is the only way to add data to it and that's the description they offer as to how to figure it out.

The only fields available is the blank box after (Input/ ) and the box right before = Output. Those are the only numbers that can be entered.

Yeah, that's fine**. But the numbers you came up with are just wrong. Try it for yourself. Put in any voltage from the possible range and see what result you get. You get nonsense.

** When I say "fine", I mean, it's still shit. The very simple linear formula (slope & intercept) is shit for a sensor with a non-linear response. This is the curve, from your data above.

Look at the CURVE! It's only really linear between about 30 and 90 °C. And if you used only that range to define a curve, it would be great. But you would go more and more wrong as you went to higher temps.

And that is why the slope & intercept found when you use 50 and 150 as the end points is so bad halfway between those points. The real curve is a long way below the linear curve which just zips straight between the end points, like this one.

You could probably use the same slope and a lower intercept, to move that straight line down, and spread the error out. But you would 5-10°C off in a lot of places. You'd need to say what temperature range you really wanted to be most right - say, 100 to 130, and plop the line closest to teh real curve in that region, which would make it quite wrong down at the lower temperatures.

Let me just say that HPTuners are not being realistic in only allowing for a simple linear curve.