DIYEFI.org Forum

I agree with Marco, there are pros and cons of having them as they are, or as they are proposed. The regulators are high gain devices and could get into RF oscillation issues when under load, with specific lengths of wire, ect. The potential sources of oscillation are just like ESD and EMI, they are more art than science. You often have to make those design decisions based more on gut feel and experience, than based on proven science. I can see how it's reasonable to expect that sags shouldn't exist. Copper traces should be sufficient to allow proper current flow, the chips should be stable over a wide temperature range, ect. If a potential source of sagging is identified, it should be addressable and corrected, however a track-able source of sagging hasn't been identified. At the moment, Marco's gut feel is that potential RF issues will induce more error, but other peoples gut feel is that there will be other sources of sag. The real issue is that either source of error is really gut feel. We don't have any real empirical data to back either theory. So might as well get people to stop arguing about it, and just do it.

Sorry Fred, but I didn't saw you decided between one scheme and the other. You understood the good reasons to make the change, and gave an accurate example (to use the big reg for vref), but you said your board was working ok when you used it. What I really overlooked is the "hard limitation" part of your post.

Then slacker came in and show a strong decision about this, which I don't have.

Just to give more data here, these are the circuits negatively affected by the actual architecure (they use pullups)
* CLT
* IAT
* MAT
* The extra temperature sensor
* TPS

And circuits affected mostly by a moving reference (no direct pulllup, and the signal is mostly linearized and stand either if VDD is 5 or 4.5v, unless you're hitting the end of the scale). I mean, if Vdd is 4.5v, the readings are almost okay until the signal gets higher than 4.4v or so.

* Vbat
* EGT
* MAP
* AAP
and maybe:
* O2
* MAF

Well, at least my bosch MAP sensor wasn't affected by changes in Vdd, but freecale's MPX could be a lot different.

The change made it to the schematic now, and you're right Fred, the last push was a long time ago. Most of the work happened when I was at the university. I can't push from there, and I lost the habit. Also, these last weeks were used mostly for uni stuff, smps testing, and setting up the p&h bench. I must push the changes today.

nitrousnrg wrote:Sorry Fred, What I really overlooked is the "hard limitation" part of your post.

No problem at all, you're forgiven! :-)

The change made it to the schematic now, and you're right Fred, the last push was a long time ago. Most of the work happened when I was at the university. I can't push from there, and I lost the habit. Also, these last weeks were used mostly for uni stuff, smps testing, and setting up the p&h bench. I must push the changes today.

viewtopic.php?f=45&t=1206 - try this. You can use your home machine with dyndns as the gateway. It's down right now, but cached versions probably exist, so look around.

Fred.

It turns out that the map sensors are proportional to their supply, so we should *definitely* always use the adv 5v for the reference. From AN1646:

Freescale wrote:Since the sensor output is ratiometric with the supply voltage, any variation in supply voltage will also proportionally appear at the output of the sensor.

And:

Freescale wrote:An additional consideration is that the power to the sensor and the A/D voltage reference should be tied to the same supply. Doing this takes advantage of the sensor output ratiometricity. Since the A/D resolution is also ratiometric to its reference voltage, variations in supply voltage will be canceled by the system.

So, there you have it, from the horses mouth!

Fred.