Triple Digital power supply Voltage display on one LCD display

[Arrow]

Hello all,
I'm building a triple digital power supply. The outputs are all isolated, but I like to use one big 40x4 LCD display to displays everything of these 3 supplies. As they are isolated I can not use a common ground. I was wondering how they do this in real life. I was thinking of optic communication. Do anyone have an idea?

greets.

[ve7xen]

It really depends on the topology of your power supply what the best way to handle this will be. I2C is a bit tricky to isolate because it's bidirectional on all ports. I'd probably do this by using a 2 channel SPI DAC + 3x optos (CLK, MOSI and CS) for each channel and running it at relatively low speed. But there are lots of other ways that will work.

[c4757p]

I2C is a huge pain in the ass if you're doing anything different from a single one-voltage, non-isolated bus. Out of all the various projects I've done, I think I had one occasion where I found I2C useful without some jury-rig. (Not counting the times I've used a single I2C chip, because it's not really a bus then, is it? Much easier in that case.) What I would personally do depends on whether I'm looking to save BOM cost or development time. If development time, I'd isolate SPI with optos. If BOM cost, I'd use a unidirectional slow software UART (just one opto per channel, a slow software version so I don't need separate UART transceivers per channel). Bidirectional if you need central control as well as display, and you still save three optos per unit. But it would take a good bit more time to write the code for that.

[M. András]

http://www.digikey.com/scripts/DkSearch/dksus.dll?site=us&lang=en&v=336&c=57&WT.z_supplier_id=336&WT.z_page_type=SP&WT.z_page_sub_type=SS&WT.z_oss_type=Category&chp=0 what about these? not exactly for bidirectional fairly fast/cheap

[c4757p]

Those are nice. In small quantities, a two channel digital isolator barely costs more than an opto, is much faster, and much easier to use. I amend my suggestion to recommend digital isolators unless you are going to be making a bunch of these (optos get cheaper in quantity). I've used them before, but I still don't always think of them.

[Arrow]

I2C is a huge pain in the ass if you're doing anything different from a single one-voltage, non-isolated bus. Out of all the various projects I've done, I think I had one occasion where I found I2C useful without some jury-rig. (Not counting the times I've used a single I2C chip, because it's not really a bus then, is it? Much easier in that case.) What I would personally do depends on whether I'm looking to save BOM cost or development time. If development time, I'd isolate SPI with optos. If BOM cost, I'd use a unidirectional slow software UART (just one opto per channel, a slow software version so I don't need separate UART transceivers per channel). Bidirectional if you need central control as well as display, and you still save three optos per unit. But it would take a good bit more time to write the code for that.

Why do I have to use slow UART? You mean the opto's are slow?

[Arrow]

http://www.digikey.com/scripts/DkSearch/dksus.dll?site=us&lang=en&v=336&c=57&WT.z_supplier_id=336&WT.z_page_type=SP&WT.z_page_sub_type=SS&WT.z_oss_type=Category&chp=0 what about these? not exactly for bidirectional fairly fast/cheap

These look great!!! and I did not know of this RF isolators.

[c4757p]

If you do it slowly it's easier to do it entirely in software using GPIO pins. If you're using a microcontroller with three UARTs you're paying too much for a microcontroller.

[M. András]

http://www.digikey.com/scripts/DkSearch/dksus.dll?site=us&lang=en&v=336&c=57&WT.z_supplier_id=336&WT.z_page_type=SP&WT.z_page_sub_type=SS&WT.z_oss_type=Category&chp=0 what about these? not exactly for bidirectional fairly fast/cheap

These look great!!! and I did not know of this RF isolators.

found these while looking for isolating spi/parallel adc -s from the main controller while able to keep it at max speed to solve my own ideas btw im courious about what the professionals say about these chips

[Arrow]

If you do it slowly it's easier to do it entirely in software using GPIO pins. If you're using a micro controller with three UARTs you're paying too much for a micro controller.

Like I said before it will be a triple power supply. Each supply will have their own uC. So in total it will be 4 uC. This last uC will be the master that will read the inputs and also control the display. I'm planning to use 1 single UART for all and use multiprocessor communication recognition code. So the master will broadcast the code for all, but only the assigned uC will talk back.
BTW the power supply will be 3x 0-50VDC /6A over the entire range. And it will be a mix supply. I used a switching power supply in the first stage. This SMPS supply delivered power to the second stage linear power supply. The linear power supply controls the switching power supply and controls in this way the Vdrop for the linear which will be a constant 5V. In this way I could decrease my series pass transistors to only 2 for the entire range. I preferred the linear PS for the noise, and I use the switching ps for the efficiency.


@M. András
yes, they look great, I have already put them in my digi-key cart ready to order.
thanks

[MikeK]

Why can't you run the outputs through scaling op-amps?

[ve7xen]

Like I said before it will be a triple power supply. Each supply will have their own uC. So in total it will be 4 uC. This last uC will be the master that will read the inputs and also control the display. I'm planning to use 1 single UART for all and use multiprocessor communication recognition code. So the master will broadcast the code for all, but only the assigned uC will talk back.

This can work, you'll still need two optocouplers (or isolators) per isolated supply. It's a bit trickier to do right though.

BTW the power supply will be 3x 0-50VDC /6A over the entire range. And it will be a mix supply. I used a switching power supply in the first stage. This SMPS supply delivered power to the second stage linear power supply. The linear power supply controls the switching power supply and controls in this way the Vdrop for the linear which will be a constant 5V. In this way I could decrease my series pass transistors to only 2 for the entire range. I preferred the linear PS for the noise, and I use the switching ps for the efficiency.

Sounds interesting. Posted it anywhere?

Why can't you run the outputs through scaling op-amps?

The three outputs are supposed to be isolated from one another (and mains earth).

[Harvs]

Like I said before it will be a triple power supply. Each supply will have their own uC. So in total it will be 4 uC. This last uC will be the master that will read the inputs and also control the display. I'm planning to use 1 single UART for all and use multiprocessor communication recognition code. So the master will broadcast the code for all, but only the assigned uC will talk back.

Just remember that if you're planning on joining all the reception lines into a bus it needs to be an open-drain type arrangement with common pull up (or something like that).  I.e. the digital isolator referred to I'm pretty sure drives in both high and low direction, so you can't just join the outputs together.

So you're building a switching pre-regulator.  Are you building a triple off-line isolated switching supply with independent isolated channels?  That's not exactly straight forward!

[ptricks]

It seems you might have  a use for something that I came up with a while back. I needed to do isolated communications in a small area where I needed to send from one board to the others and then have them send data back when it was their turn. I used IR LED and those IR detectors used to read things like tv remote controls. The sensors have all the circuitry needed and are cheap and reliable. Each board has a detector. The main control board pulses out the data using IR LED, the detectors are so sensitive that in a space, like inside a PSU case, you don't need to worry about line of sight to send /get data. After the sub board is triggered it can process the data and send it back using another IR LED.  It is a very flexible system and low cost. 

Essentially this is what  opto-isolator do but this makes it so that you can broadcast data from a main board using light to broadcast to multiple targets instead of having to wire up each sub board to the main board, a lot easier to design on a pcb as well since you can put the send/receive parts anywhere.

[jeroen74]

Just remember that if you're planning on joining all the reception lines into a bus it needs to be an open-drain type arrangement with common pull up (or something like that).  I.e. the digital isolator referred to I'm pretty sure drives in both high and low direction, so you can't just join the outputs together.

You could use a regular and cheap optocoupler for this. I don't think this whole system requires that much speed, something like 300 to 1200 baud should be sufficient. I would use the UART with the RX and TX combined into one open-drain line.

[Arrow]

Just remember that if you're planning on joining all the reception lines into a bus it needs to be an open-drain type arrangement with common pull up (or something like that).  I.e. the digital isolator referred to I'm pretty sure drives in both high and low direction, so you can't just join the outputs together.

Yes, I will put the pull ups for the master. The slaves don't need that, they are all on separate supplies &chip. ( the chip have separate supplies I/O). Good observation!

So you're building a switching pre-regulator.  Are you building a triple off-line isolated switching supply with independent isolated channels?  That's not exactly straight forward!

Why not?? please explain.

Essentially this is what  opto-isolator do but this makes it so that you can broadcast data from a main board using light to broadcast to multiple targets instead of having to wire up each sub board to the main board, a lot easier to design on a pcb as well since you can put the send/receive parts anywhere.

Nice design!! In the past I had also made a design for a robot with multi-comm. with 12 uC. I used the UART interface. Baud-rate was 230.4K , I  set all the UARTs on one bus, no need of pull ups or anything else, and the communication was perfect. Again because only 1 communication was allowed.

[Arrow]

What topology are you using? I'm using a switching pre-regulator as well 300W x2. I 'm in the process of testing a scaled down version 30W. I elected to use a flyback topology 3 converters interleaved 100W per phase.
[/quote]
There are three separated switching power supplies of 300W, connected to three separated linear PS. Am I missing something here? 

[Arrow]

You could use a regular and cheap optocoupler for this. I don't think this whole system requires that much speed, something like 300 to 1200 baud should be sufficient. I would use the UART with the RX and TX combined into one open-drain line.

You are right, but the testing of the digital isolators RF modulator are also nicer.   
When the times arrives I will evaluate the cost the space the package (smt or through hole) and the signal edge. 

[jeroen74]

Something like a TLP281 is quite small.

[Arrow]

Interleaved converters provide for ripple cancelations easing output filtering and capacitor stress on the input output. I'm using one linear regulator.

www.ti.com/lit/ml/slup231/slup231.pdf

What topology are you using forward,halfbridge etc..?

So , you are building a PS aswell, and you are using interleaved converters. great!!!
I'm using the forward conv. But the switching part is already done. I'm not worry about that. I am busy with the design for the linear part. Finding op amp that can handle rail to rail with single 60V supplies and 0V offset ect wish me luck !!!

[jeroen74]

Why no circumvent the RR problem with a negative rail and higher rail voltage?

[digsys]

If you don't need super high accuracy, I've used this Linear opto configuration many times.
Saves having to put a micro on every PS stage. There are a few types of optos around, just check specs.

[Arrow]

Why no circumvent the RR problem with a negative rail and higher rail voltage?

Yes, I do that, but then I have to connect the DAC en uC also to the negative supply and then I will end in an endless virtual circle. That's why I think 0V offset is best. I gonna check if I can send a schematic, it will be easier to understand. The schematic can change anytime it is still in the design stage. But let me give some specs. I'm using the simple LM317HVT regulator. First I was planning to use the TL783 it has a better Line regulation ect. but the datasheet tells me that it can not draw so much current unless it has a higher Vdrop ( I didn't like that) when the LM317HVT can draw higher current at lower Vdrop.

[jeroen74]

I once emulated that kind of circuit, by using an extra optocoupler of the same type as a temperature compensation element inside the feedback path. Without it, the output pulse shape was very temperature dependant (not good if you use PWM and filter this to get a DC voltage); with the new circuit the pulse shape was almost completely insensitive to temperature.

If you google for 'optocoupler temperature compensation' you should find the simple circuit.

[Arrow]

What do you think of this?
http://www.ti.com/lit/ds/symlink/opa547.pdf

One more thing, once I get all the opto's signal, I have to do a constant scan on the ADC to refresh the output voltage and the current. But I think it can be done.