Who wants to build and sell varios for a french company?

[mstrens]

So for a user that want to use OXS with one sensor (e.g. a current sensor) it does not make sense to use the additional board.

Of course it does make sense.. Specially with a current sensor.. That is the reason why I made an extra board for mine.
If you only have a vario and 2 resistors for the voltage sensor, then it doesn't make sense to have an additional board as you can solder all to the arduino board and just shrink tube it.. That is how I did my vario.. more or less like Rainer did his.

João

I think you misunderstood what I said.
Of course you can always use an additionnal board but it seems me important to foreseen that (like now) you can build an OXS without additionnal board when you need only part of all OXS feature. E.g. if you only want a current sensor (without vario, without additional regulatoer, ...) it is much smaller/lighter to just connect the 3 pins from the current sensor to the arduino. For a only a vario (like Rainer did) you only have to solder 2 wires on top of the 2 straight connection).
If we change the code in OXS as Mike proposed, so uning the internal UART from the 328P, the code will be lighter and easier to maintain but the drawback is that it would not be possible to use it without adding about 5 or 6 components to invert the signal. That would then always required an additionnal board and this seems me not the best solution.
Better to keep the complexity in the software (when possible) then making the task of the builder more complex.

Supporting in OXS the 2 solutions (with and without inverters) is even worst because both codes have then to be implemented and to be maintained.

[jhsa]

yeah, I probably misunderstood what you said, sorry..
But I do build a board for the current sensors to support the connections. And they are a bit bulky as well..

João

[ReSt]

Supporting in OXS the 2 solutions (with and without inverters) is even worst because both codes have then to be implemented and to be maintained.

So no problem when you can use software serial (as the original vario did), because you can invert the signal via a paramter of software serial

Reinhard

[andrewju]

While trying to maintain simplicity of the solution, aren't we impacting potential functionality / precision of the OXS?

I personally like Mike's advise:

Keep to the Pro Mini, but add the hardware on the addon board to allow the hardware serial to drive the SPort 'properly'.
Consider adding a 4.096V voltage reference to allow for accurate and repeatable analog measurements.

This should give us more memory for the OXS code and more accurate measurements... I would definitely vote for soldering a few more components to achieve this.

If needed, one could still use current SW in order to simplify to build in occasions where only some basic functionality is needed. Otherwise, the 'v2.0' SW will require an add-on board with a few more components on it in order to include all the benefits above. In my opinion, it is just fine...

[mstrens]

While trying to maintain simplicity of the solution, aren't we impacting potential functionality / precision of the OXS?

I personally like Mike's advise:

Keep to the Pro Mini, but add the hardware on the addon board to allow the hardware serial to drive the SPort 'properly'.
Consider adding a 4.096V voltage reference to allow for accurate and repeatable analog measurements.

This should give us more memory for the OXS code and more accurate measurements... I would definitely vote for soldering a few more components to achieve this.

If needed, one could still use current SW in order to simplify to build in occasions where only some basic functionality is needed. Otherwise, the 'v2.0' SW will require an add-on board with a few more components on it in order to include all the benefits above. In my opinion, it is just fine...

Please note that adding a voltage reference is not always a good solution.
I would be good to measure e.g. a battery voltage but it would not really help (on the opposite) when the sensor is a ratiometric sensor that must be powered by the main Vcc because it takes to much mAmp. Such a sensor (like many current sensors) provides a signal that is proportional to Vcc. If they can't be powered via the voltage reference, then the measurement (e.g. the current) becomes worst.

[jhsa]

Can't the reference voltage be 5V? then the current sensor could be powered from it. 4.096V is enough for the sensor but it would make the calculation a bit difficult? Or am I missing something??

João

[andrewju]

it would not really help (on the opposite) when the sensor is a ratiometric sensor that must be powered by the main Vcc because it takes to much mAmp. Such a sensor (like many current sensors) provides a signal that is proportional to Vcc. If they can't be powered via the voltage reference, then the measurement (e.g. the current) becomes worst.

I'd like to understand if I get this correctly. The external sensor will be powered by the same reference voltage, and if this voltage is less than 5V (i.e. 4.096V as suggested by Mike), we will get less precision from that sensor?
On the other hand, if Vcc is about 5V, we will get better precision from the sensor even if Vcc itself will not be very precise?

[mstrens]

Can't the reference voltage be 5V? then the current sensor could be powered from it. 4.096V is enough for the sensor but it would make the calculation a bit difficult? Or am I missing something??

João

The issue is that normally a voltage reference is foreseen to be stable but can't deliver only a few milliamp.
Probably that the current sensor needs more millamp that what can be provided by the voltage reference.
It must be checked in the datasheets.

The advantage of a 4 volt voltage reference is that the reference is still stable when it is powered by a NiMh battery.
If you use a 5 volt voltage reference and you power it with 4.8 volt (or lower) it is not working anymore (out of tolerance).

On the other side, 4 volt is too much for those that use OXSwith a single lipo cell. In this case they could use a lower value.

Anyway, I am not sure that this is really needed for our application.
We do not really 1% precision in our measurement.
When you look with a precise voltmeter how big are the voltage variations of a battery when you activate a servo, ... It is much more than 1% or even 5%.
Once calibrated, I do not see a real need using something else than the internal voltage ref from the Arduino.
Ok, it requires some calibration once but this is even less work than soldering an additional voltage reference.
Anyway a calibration is useful to to take into account (as much as possible) otherkind of errors (offset on ADC, non linear ADC, tolerance on resistors)

[mstrens]

it would not really help (on the opposite) when the sensor is a ratiometric sensor that must be powered by the main Vcc because it takes to much mAmp. Such a sensor (like many current sensors) provides a signal that is proportional to Vcc. If they can't be powered via the voltage reference, then the measurement (e.g. the current) becomes worst.

I'd like to understand if I get this correctly. The external sensor will be powered by the same reference voltage, and if this voltage is less than 5V (i.e. 4.096V as suggested by Mike), we will get less precision from that sensor?
On the other hand, if Vcc is about 5V, we will get better precision from the sensor even if Vcc itself will not be very precise?

Some sensor are ratiometric. It means that they works like a voltage divider with 2 resitors.
The sensor needs some power supply.
If the power supply is not stable, it has an impact on the signal provided by the sensor.
E.G. take a 4 elements Nimh pack and says that his voltage can vary from 6 volt to 4.6 volt 4.8.
Imagine you have a bidirectionnal current sensor that provide Vcc/2 when current = 0.
When powered by a battery at 6 volt, the sensor will give a signal equal to 3 volt.
When powered by a battery at 4.8 volt, the sensor will give a signal equal to 2.4 volt.
The arduino measures always in 1024 steps from what it get has voltage reference.
If voltage reference is constant (say 4 volts) , you see that the ADC measurement will be different for the current equal 0. The measurement is wrong.

If you say to Arduino to measure using the battery voltage as reference, the ADC will for both battery voltage returns 1024/2 steps. Then there is (in principle) no error.

If you measure an absolute value (e.g the voltage of a battery) you need an absolute reference.
But some sensors do not provide an "absolute" value.

[MikeB]

A solid voltage reference is, in my opinion, always a good starting point. I picked 4.096V as this should be fine when powered from a 4.8V battery. If you want to be flexible, for a single lipo, use a 2.56V reference.
Note, this doesn't have to be connected to the AREF input, it could be connected to an ordinary analog input. This way, you can use it to calibrate the VCC at any time.

Mike.

[andrewju]

If you measure an absolute value (e.g the voltage of a battery) you need an absolute reference.
But some sensors do not provide an "absolute" value.

This (and the good example before that) is understood. Thank you very much!

What you're saying is that there's no benefit of having an absolute reference for some sensors. But on the other hand, there's no harm too, right? If that's just a reference voltage on a separate analog input (as Mike suggested), you could always take the VCC or any other voltage as a reference for some calculations, and use an absolute reference for other calculations where it is beneficial. This will add a little bit to the overall complexity of the HW, but it should be manageable and should help to provide better precision in some cases...

I'm sometimes criticized for the "if it can be better - it has to be better" approach... So maybe I shouldn't be listened here...

[jhsa]

we would lose an analog port though.. And some arduinos (pro mini) only have 5?? The originals included

João

[mstrens]

If you measure an absolute value (e.g the voltage of a battery) you need an absolute reference.
But some sensors do not provide an "absolute" value.

This (and the good example before that) is understood. Thank you very much!

What you're saying is that there's no benefit of having an absolute reference for some sensors. But on the other hand, there's no harm too, right? If that's just a reference voltage on a separate analog input (as Mike suggested), you could always take the VCC or any other voltage as a reference for some calculations, and use an absolute reference for other calculations where it is beneficial. This will add a little bit to the overall complexity of the HW, but it should be manageable and should help to provide better precision in some cases...

I'm sometimes criticized for the "if it can be better - it has to be better" approach... So maybe I shouldn't be listened here...

I initially thought that Mike proposal was to link the Voltage reference to the arduino pin Vref.
The advantage is that you do not loose an analog input.
Still this is difficult to achieve on a arduino pro mini because this CPU pin is not part of the available pins on the left or right side connectors.
The alternative that Mike proposes is to connect the external voltage reference to an analog pin in order to measure it using arduino Vcc as reference.
With a basic calculation you can find back the right value of VCC that you can use for other voltage measurements.
There is another advantage: you do not need a voltage divider when you want to measure a voltage that can be as high as VCC.

Still, in most cases, I think you can get a very accurate measurement just using
- or the existing arduino internal voltage regulator (but you must calibrate it making some measurements with a voltmeter because there could be some large tolerance from arduino to arduino)
- or a good general voltage regulator that provide a stable VCC.

[MikeB]

The 'good' voltage regulator only works if the supply voltage is high enough. If running from a 4-cell NiMh, this is not guaranteed, and you can also get significant voltgae drops when servos move, even from a 5V BEC, due to wiring resistance.

My PRO-MINIs (V5) have all 8 analog inputs on pins, 4 at one end.

Mike.

[jhsa]

But then if it uses an analog pin, if someone needs ro read a 6 cell pack, it is not possible to have current anymore cos there aren't available pins. and that is with the arduino version that has 7 of them already.

[Kilrah]

Then don't measure 6 cells... Or take a 2nd oxs for it.
The arduino has limited resources, there's simply no way to do everything and do it well by staying on the current platform.
Compromises have to be made, and they will obviously have to favor some use cases over others.

Sent via mobile

[jhsa]

Yes of course.. I think I would favor the current way of measuring.. but im not a developer so, I just go along I guess.. or just keep using the very old version that I have been using until now, which works better for me than the new ones. I just can't calibrate the current sensor with the new way of doing it.. but that is another topic..
And how can you use 2 arduinos with the hub protocol by the way?

João

[Kilrah]

With hub you can't, need sport for that.

Sent via mobile

[jhsa]

Aha...

[mstrens]

I also prefer the option 1, more in the spirit of the Arduino system.

If nobody else wants to do the job, I can design the PCB if there is a good schema and a complete list of components, including the links to their datasheets.

@jpparisy
Did you had a look at the the schema and implementation I prepared (see in this thread).
Could you continue the job?

Thanks in advance

[davx]

But then if it uses an analog pin, if someone needs ro read a 6 cell pack, it is not possible to have current anymore cos there aren't available pins. and that is with the arduino version that has 7 of them already.

In fact, you can get up to 8 analog pins on pro mini: A0 to A7

Like this one:

[jpparisy]

I also prefer the option 1, more in the spirit of the Arduino system.

If nobody else wants to do the job, I can design the PCB if there is a good schema and a complete list of components, including the links to their datasheets.

@jpparisy
Did you had a look at the the schema and implementation I prepared (see in this thread).
Could you continue the job?

Thanks in advance

Yes, I've seen your schema, but it's not what Bertrand asked for: he wanted somebody to build and sell a barometer used as variometer and, optionally, a differential pressure gauge to be used with a Pitot tube as air speed indicator. I don't find any of this in your schema, which seems instead a hub for various sensors.

Of course, I can complete your job, let me just some times to do that.

[jhsa]

In fact, you can get up to 8 analog pins on pro mini: A0 to A7

You are absolutely right. It seems that I can't count.

[mstrens]

In fact, you can get up to 8 analog pins on pro mini: A0 to A7

You are absolutely right. It seems that I can't count.

When you use a component that need I2C bus (baro sensor MS5611 or a differential pressure 4525DO) 2 analog pins (A4 , A5) are reserved, so it remains 6 for other purposes. If you do not need I2C, then you have access to 8 Analog pin (for voltage, current, temperature, ...

[mstrens]

Yes, I've seen your schema, but it's not what Bertrand asked for: he wanted somebody to build and sell a barometer used as variometer and, optionally, a differential pressure gauge to be used with a Pitot tube as air speed indicator. I don't find any of this in your schema, which seems instead a hub for various sensors.

@jpparisy,

Thanks for your collaboration.
My question was just to know if you could do it or if I had to look further for somebody.
For me it is absolutely not urgent, so take your time.
I just think that it is important that I can explain:
- why I came to this schema and this implementation
- which connections can be changed if it makes the job easier.
I propose to do this by private message when you have time.

Originally Bertrand asked for somebody to build and sell an OXS with only "vario + main batt voltage" so in fact a combination of an arduino pro mini, a GY63 (it is already a board with a MS5611 and some components) and 2 resistors.
I know it takes already some time to solder those few components.

I just took the opportunity to propose to make a board that would make this job easier but that could also make the job much more easier when somebody want to use other/more features than OXS can support.
In someway, I misused the thread and it can explain some confusion.

Please note that OXS does not yet (officially) support a differential sensor in order to measure airspeed. The version is currently in test.

In fact the board that I proposed (see schema) is not to be considered as a hub because nearly all components are soldered on the board (all voltage dividers, filter for PWM, 4525DO, and if possible voltage regulator). The external components are just:
- the (one or two) GY63 (baro sensor). It is already a small board that has(have) just to be put on top of the proposed board (and connected by 4/5 pins)(Il already noticed that I forgot one connection)
- the hall effect sensor, the current sensor and a temperature sensor because it is much more convenient that those sensors can be located at some distance from OXS it self. Each of those sensors have to be connected to OXS by just 3 pins (Vcc, Ground and signal). So I have foreseen that each of them could be soldered to a JR cable plugged on the proposed board. So it becomes very easy to add them or to change them (or e.g. for the current sensor, just plug one sensor instead of another one if you want to measure another range of current).

If you look at the proposed implementation, you see on top a kind of free zone. It is the 4525DO (which is a "extra large" DIL component). Under there are no connectors because that is the place reserved to put on top the GY63.

[KAL]

A next step could be to put the MS5611 directly to the board ...
(just a thought)

Klaus

[Erni]

The problem with the unstable, or unknown Vcc could be solved if you use the internal "Secret Voltmeter", like this:

http://provideyourown.com/2012/secret-a ... y-voltage/


And regarding the missing debug information (Serial.print) if you use the harware UART to SPort, you could use this TinyIsp.

http://forum.arduino.cc/index.php/topic,123388.30.html

It is developed by Coding Badly from the Arduino foum.
What it does besides programming the chip is that it provides debug information via one of the programming lines.
That means that you always are able to get debug info regardles of the chip you are programming.
This is especially handy when programming the tiny's, but in this case also a 328.

[MikeB]

The problem with the internal bandgap reference is it isn't that accurate. A Mega64 specifies it as 1.15 to 1.35 volts, while a Mega328 specifies 1.0 to 1.2.

Mike.

[andrewju]

The problem with the internal bandgap reference is it isn't that accurate. A Mega64 specifies it as 1.15 to 1.35 volts, while a Mega328 specifies 1.0 to 1.2.

Is it different from chip to chip, or it changes over time even in the same chip?
I mean, if it differs from one chip to another but remains the same within one chip, we could probably calibrate it when a new OXS is built...

[mstrens]

The problem with the internal bandgap reference is it isn't that accurate. A Mega64 specifies it as 1.15 to 1.35 volts, while a Mega328 specifies 1.0 to 1.2.

Mike.

Is it not so that not accurate means that there can be quite big variation from one chip (one device) to another one.
I believe that it is quite stable for one specific device. Then if you make some easy calibration using a voltmeter and changing a parameter in the OXS config file, I think it is not so bad for our purpose.

Do you have a different meaning about this?