[info] Frankenso - ECU shield compatible with OEM case / official

[mck1117]

I found a bug in Frankenso's analog inputs. The RC filter claims to be a 100 ohm resistor, with a 1, 0.1, or 0.01uF cap, giving a bandwidth of approximately 1khz, 10khz, or 100khz respectively. However, there's also a 10k resistor in series (before the diodes), so those bandwidths are really 1% of the desired values. Yeah, they're 10hz, 100hz, and 1khz.

Confirmed with ltspice:

This is from the channels that say it should be a 100khz knee, but is closer to 1500hz.

Here's a solution: The LPF should be on the output of the opamp, not the input. We should leave a decoupling cap on the input, but move the filter stage to the output of the opamp, using the output divider.

(ignore the extra resistor divider pointing upward, that was to check no-filter performance)


Here are the -3dB bandwidth options possible by exchanging the filter cap (C2) in this new circuit:

omit = 32khz
10n = 20khz
100n = 3.2khz
1u = 320hz

[kb1gtt]

Great feedback and thanks for pointing it out. Also great post, content with low levels of fluff.

Do we need more than 1kHz? I'm tempted to update the schematic, and leave the components. Most automotive analog signals are much slower than 1kHz.

I'm tempted to change the note to something like the below.
low pass filter knee at 1kHz or 10Hz with 10k series resistor.
low pass filter knee at 10kHz or 100Hz with 10k series resistor.
low pass filter knee at 100kHz or 1kHz with 10k series resistor.

[Rhinoman]

Do we need more than 1kHz? I'm tempted to update the schematic, and leave the components. Most automotive analog signals are much slower than 1kHz.

Whats the analogue sample rate?

[kb1gtt]

I recall it varies. I forget the exact rates. I believe we vary between 100mS to 10mS.

[Rhinoman]

Well then there is no point filtering at 1khz - you'll just get aliasing of out of band frequencies.

[kb1gtt]

The 1kHz is only on the altCAM and altCRANK inputs. Those can be used as alternative hall effects inputs, and have much higher cut off's. I think most folks use the MAX chip for hall inputs, but I also believe a couple folks have saved some $ and not used the max chip, using these alternative options instead. I recall they had some problems with the cut off and they may have changed the 10k to something else.

I'm a bit ignorant about the low level ADC hardware on this chip. I understand Chibios has a tick time which is used to collect an ADC reading. It is likely that the ADC is being done by some hardware at a much higher frequency, perhaps even filtered some how. Then Chibios simply grabs a result every so often. I really don't know the nitty gritty details about how Chibios is doing the ADC. I understand the concerns about anti-aliasing distortions (also known as fold back, or wrap around). I have not seen such inducted errors in the posted log files, so I don't think we are seeing it. Some day I should better learn the details how Chibios is doing the ADC, such that I know when potential problems could arise. So far I've not seen problems, so I haven't looked into it further.

[mck1117]

Fast ADC (MAP, MAF, TPS, Knock) is sampled at 10khz.
Slow ADC (everything else) at 20hz.

I believe this is how the ADCs work: a timer triggers each ADC, which takes a sample, then triggers an interrupt which is bubbled down to rusEfi. The ADCs can go up to around 1mhz, I think, but that's massive overkill for us, and impossible unless you have it DMA results in to a buffer.

As for actual bandwidth, the slow sensors (temperatures, battery voltage, etc) have analog bandwidths on the order of a few hz, so tens of hz is reasonable for both filter and sampling. The fast sensors could have mechanical and electrical bandwidths up around 1-2 khz, so I think a few khz analog and 10khz digital is a pretty sensible place to be.

So, on second thought, maybe the existing filter is totally fine, and should be left as is (but fix the notes on the schematic). The slow inputs have a ~15hz filter, and fast inputs have a ~1.5khz filter, both of which are reasonable.

[mck1117]

Oh and I cracked open my factory gen3 LS ecu to see what GM/Delphi did, and the cam/crank inputs on it have a 470 ohm series resistor, <1nf decoupling cap to ground, then directly in to a chip which is presumably integrated protection diodes/schmitt trigger/etc. The analog inputs have a 10-30k series resistor, similar 1nf cap, then disappear in to an internal layer .

[AndreyB]

174) D20 D19 D22 silkscreen seems to be inaccurate - D20 really means D19 etc?

[kb1gtt]

I think the issue is with the schematic. If it looked like this on the schematic, would this resolve the issue?

[AndreyB]

Oh, you mean the issue is that on the schematics the labels are visually very confusing? Yes, this would resolve the issue if we put the labels in a less confusing way.

[AndreyB]

Can we route some stuff to unused middle plug pins please? This would allow easier PnP Miata NA6 and make unused pins more useful in general.

GND to bottom of W36
GND to bottom of W35
+5v to bottom of W28
+12v to bottom of W27

See https://raw.githubusercontent.com/rusefi/rusefi_documentation/master/installations/MazdaMiataNA6_Frankenso_pnp/middle_plug_v0_1.jpg
See https://github.com/rusefi/rusefi_documentation/tree/master/installations/MazdaMiataNA6_Frankenso_pnp

[AndreyB]

Remember to update Frankenso Schematic -- filter notes - https://rusefi.com/forum/viewtopic.php?f=4&t=569&start=469
also

177) adjust mounting holes to allow 89661-32590 and 89661-32260 cases

[kb1gtt]

This keeps falling through the cracks. I finally remembered to tag it.

178) RESOLVED IN R0.5.5 Update AN input schematic to note "low pass filter knee at 1kHz or 10Hz with 10k series resistor"
179) RESOLVED IN R0.5.5 Update AN input schematic to note "low pass filter knee at 10kHz or 100Hz with 10k series resistor"
180) RESOLVED IN R0.5.5 Update AN input schematic to note "low pass filter knee at 100kHz or 1kHz with 10k series resistor"

[AndreyB]

Fresh batch of assembled boards has arrived. With two of them producing a bunch of lovely sparks during power supply testing, I am now always wearing safety glasses while doing 12 volts testing or soldering.

Stay safe my friends!

[AndreyB]

[md_sanci]

Hello all,
I am running frankenso for 3 weeks. Today i tried an autotune and the engine died. By reflex i did ignition off. The board shut down, ts disconnected. Ignition back on and ts would not reconnect.
I have restarted the ecu and ts. Now i have connection. The engine would not start any more.
Noticed tps died. Adc input zero.
Checked wiring till the ecu and sensor and they seem fine.

Tomorrow i will check pin config and opamps and inputs.

Did somebody had failures like this that adc channel just died? I would try last result to reroute it to another channel.

I have reloaded the msl file so i would presume that pin configs were resent to the controller.
Water temp looked ok so adc is working.

[AndreyB]

(moved to Frankenso topic since you say Frankenso?)

I am not aware of sudden burn or loss of ADC.

[Crazy Striker]

If you assembled the board yourself, identify the problematic channels and redo all the solder joints. I had a similar problem due to poor solder adhesion and cold solders

Gwendal

[kb1gtt]

Also with a multi meter connected to frankenso ground, you could provide a table of measured voltages and frankenso's claimed values. I suspect you'll find Crazy Strikers comment about quality of assembly / cold solder joints is applicable. Is this a DIY board? Pictures of the board might also help identify problems.

[AndreyB]

Nice authors of ibom tool have made a fix for us and now Frankenso IBOM contains MFG and VEND part numbers - https://rusefi.com/docs/ibom/frankenso_latest.html

[Old Grey]

Are there any plastic parts in the Frankenso kit that aren't oven temp proof. I assume all the THTs get put on after, but how about the USBs, coin cell holder, etc.

[mck1117]

USB and coin cell are probably safe. I've reflowed SMD USB connectors before, and many of them come with a small piece of tape on top so that they can be placed with a pick-n-place machine.

The big harness connector might be safe, but I'm not sure why you'd want to.

[AndreyB]

many of them come with a small piece of tape on top so that they can be placed with a pick-n-place machine.

oh, that's what it's for! do they need tape for a suction cup to work?

[mck1117]

oh, that's what it's for! do they need tape for a suction cup to work?

Yep, the pick-n-place machine uses vacuum and a little rubber suction cuppy thing to hold on to components. So anything that doesn't have a relatively uniform flat top needs a piece of tape or something.

Likewise, the vertical USB connectors on the microRusefi board had a little "fake" male USB connector plugged in to them, with a flat top on it so that a pick and place can grab on to it.

[AndreyB]

Code: Select all

185)                    change JP9 & JP11 footprint to match Wxx jumper footprint to facilitate assembly

[kb1gtt]

I don't think this is for Frankenso, but the brain board. What board is this applicable to?

[AndreyB]

this is about Frankenso, this is about op-amp channels which could be used when Frankenso has MAX9926

[Old Grey]

When you get the Frankenso pre made, can you see red dots of glue under the 0805s on the underside. I hear that 0603 and smaller will stay on the underside without glue with a 2 oven process - oven solder bottom side parts and re oven for the top side parts -, due to the solder surface tension, but I'm not sure if I should risk 0805s. Basically there is a lot of 0805s and not much space between the pads for glue. So do I try for glue, or hope they stay attached hanging on the underside without glue during the 2nd oven heat.

[mck1117]

When you get the Frankenso pre made, can you see red dots of glue under the 0805s on the underside. I hear that 0603 and smaller will stay on the underside without glue with a 2 oven process - oven solder bottom side parts and re oven for the top side parts -, due to the solder surface tension, but I'm not sure if I should risk 0805s. Basically there is a lot of 0805s and not much space between the pads for glue. So do I try for glue, or hope they stay attached hanging on the underside without glue during the 2nd oven heat.

I've successfully had 0603 and TSSOP-14 stay on the bottom of a board during a second reflow, while assembling microRusefi boards. I pasted, placed, and reflowed the back, then did the front, with the board propped up on stacks of more PCBs on the sides. Nothing fell off!