I have an industrial application I'm trying to design a solution for. Using a base 24vdc power supply I need to provide 40vdc - up to 2A. That part's not difficult - plenty of pre-engineered designs to chose from. I'm looking at the TPS43060EVM-199 as my starting point now as it looks like a pretty clean solution to my very inexperienced eye.

The catch is, for my application I need to pulse the voltage down to 32 volts for signaling purposes. The ideal output should look like a square wave. The pulses will be about 8 microseconds of duration and spacing between them for transmitting bits of data.

The above power supply, using its dedicated controller, uses a simple resistor divider to determine the output voltage. So it's trivial to switch a second resistor in and out with the divider to achieve a different set point. So on paper this looks ideal.

Here's the part I'm not sure about. As is typical in these power supply designs there are capacitor banks for both the input and output to filter the power and provide a clean output. My concern is will the output capacitors impact the desired pulses and result in more of a delay in change resulting in something closer to a sine wave? Is it possible to accomplish what I'm trying to do using a single power supply or do I need a separate 40 volt and 32 volt and simply switch between them?

 

Yes, you can do all of that.

 

The quick and simple answer is Yes.

 

The pulses will be about 8 microseconds of duration and spacing between them for transmitting bits of data.

It depends on how much load you have on the power supply. Assume you need 2us rise and fall time.
ΔV=It/C
where C is the output capacitor on the power supply.
so if C = 100uF, the current must be greater than I=CΔV/t = 400A to get it down to 32V in 2us.
So switching the feedback resistor isn't going to work.
Switching between two supplies might also be problematic if there are decoupling capacitors further downstream.

 

Using a base 24vdc power supply I need to provide 40vdc - up to 2A. ..................

The catch is, for my application I need to pulse the voltage down to 32 volts for signaling purposes.

You will be pulling 4A from the 24V supply.

You have 24V. You are going to boost that up to 40V@2A. Then drop it back to 32 volts for 8uS.
I see several options.

1)Make 40V and 32V from 24V. Then switch from 40 to 32 and back.
2) use a 32.6V supply and diode to keep power from flowing backwards. 23.6-0.6=30V Then with a MOSFET pull up on the output to 40V.

3) Stand a 8V supply on top of a 32 supply. The 8V supply can be turned on/off.
4) etc.

 

Another option is to make a 36V power supply, then transformer-couple your data using a transformer with a 4V secondary. This doesn't work with NRZ data, though. You'd need to have it Manchester coded or something similar. Your transformer would also need to withstand 2A DC on the secondary without saturating. It's a non-trivial solution, but it's a non-trivial problem!

 

Just a thought. . . your power supply isn't going to discharge very much in 8us. How about simply switching it off for 8us to signal?

 

Thanks to all who replied. Some clarification and response:

I believe I did say "up to" 2A. What I did not say was typical use will be in the range of 10ma to 200ma @ 40VDC. My maximum that I want to design for allows for up to 2A. So the circuit needs to accommodate the full range basically from 0-2A.

It depends on how much load you have on the power supply. Assume you need 2us rise and fall time.
ΔV=It/C
where C is the output capacitor on the power supply.
so if C = 100uF, the current must be greater than I=CΔV/t = 400A to get it down to 32V in 2us.
So switching the feedback resistor isn't going to work.
Switching between two supplies might also be problematic if there are decoupling capacitors further downstream.

Shouldn't be any further capacitors to worry about but - you're telling me this isn't going to work which is what I expected.

Just a thought. . . your power supply isn't going to discharge very much in 8us. How about simply switching it off for 8us to signal?

I'm thinking that would involve a fair amount of testing to get that dialed in - and there'd probably be differences depending on the number of devices connected and the current. I'm being a little ambitious in this initial design - I don't have much in the way of test equipment, and I'm probably going to order pre-assembled PCB's, so I'm looking to start with something that has a good theoretical chance of working first time out.

You will be pulling 4A from the 24V supply.

You have 24V. You are going to boost that up to 40V@2A. Then drop it back to 32 volts for 8uS.
I see several options.

1)Make 40V and 32V from 24V. Then switch from 40 to 32 and back.
2) use a 32.6V supply and diode to keep power from flowing backwards. 23.6-0.6=30V Then with a MOSFET pull up on the output to 40V.
View attachment 284272
3) Stand a 8V supply on top of a 32 supply. The 8V supply can be turned on/off.
4) etc.

Option 1 I understand - that's certainly the safest choice and what I'm going to fall back on.
I don't quite understand your option 2 - even if there's a typo or two in there I don't understand the math .


I think what I really want is a dual voltage boost converter that I can switch between. It seems inefficient & wasteful to have two sets of input components - but I can't find any reference designs that do this. There's probably a reason for that - the required capacitors/inductors for supplying 40V are different than for 32V. Which means...I'll probably quit trying to be creative and just use the two separate boost converters and switch between them.

The only other option, or question I didn't ask, is can I switch some component or combination to drop 8V from the 40V? The spec allows me a tolerance of 7-9V for the pulse, so if there's some variance between a 5ma load and a 1.5A load that's acceptable as long as it is contained within that 8V +/-1V target.

 

I think what I really want is a dual voltage boost converter that I can switch between. It seems inefficient & wasteful to have two sets of input components - but I can't find any reference designs that do this. There's probably a reason for that - the required capacitors/inductors for supplying 40V are different than for 32V. Which means...I'll probably quit trying to be creative and just use the two separate boost converters and switch between them.

Technically, I actually need three power supplies since I also need 3.3V to run a microcontroller which is going to trigger the pulse switching and read the replies. With all that in mind, if I'm intending to cram all this onto the same PCB, is it desirable (or not) to use an external clock input to synchronize the switchers?

 

I worked on several projects where we sent power and data on the same wires. Here is an example of a power source and a load with long wires in between.
V1 on the left is the source of power, C1 is a typical capacitor on a power supply.
The two inductors L1, L2 pass DC through and block high frequency data.
V2 injects a high frequency signal onto the wires.
U1 on the receiver side detects the data.
The voltage at R1, the load, is about the same as the source power.
In some cases, data was sent both directions.

I don't know if you can change the 40/32V requirement or the timing requirements. There are ICs on the market that do this job.

 

I worked on several projects where we sent power and data on the same wires. Here is an example of a power source and a load with long wires in between.

I don't know if you can change the 40/32V requirement or the timing requirements. There are ICs on the market that do this job.

I understand this concept - but no that's not an option. I'm working with existing listed external devices and I'm implementing (or attempting) their published protocol.

I think I'm stuck with the dual power supply model and I'm laying out a board accordingly. Now...I understand the diode + MOSFET switch solution. But I'm wondering if I can use a pair of MOSFETS and switch between them to eliminate the diode. If I use a fast enough dual driver...would that improve anything? Or not worth trying?

 

Could you post the specification for the data protocol?

 

an
Could you post the specification for the data protocol?

Exact specs - no I can't. They're proprietary and I've signed an NDA. It's a very basic method where the controller sends voltage pulses as I've described (dropping 8V to send a bit) and then reading pulses of increased current draw from the field devices for replies. The spec shows using a current sense resistor - I'm modernizing it with an Allegro hall effect sensor. I can tell you it operates at a lightning fast 1200bps. Yes you read that correctly - this is a 40-year old protocol.

At this point, I've used TI's WEBench to create two simple power supplies per my requirements. I think. I'm using a pair of LM5022 designs per their specs - they even come with PCB layouts which helps as I'm a total n00b to this. My background is software - I need the functional hardware so I can work my magic . My first attempt simply mirrors the two circuits on one PCB, getting a little creative with the copper pours, and in my totally unqualified opinion might actually work. At least for a prototype.

What I'm struggling with now, as I have no explicit schematic to copy, is how to connect in the NMOS switch to connect the 40V to the output and bypass the 32V. Because I'm already using them as part of the LM5022 designs I'd like to use the https://www.ti.com/product/CSD18543Q3A N-MOSFET to switch the 40V and the https://www.smc-diodes.com/Products/Discrete/Schottky-Rectifiers/SBRD10200 diode on the 32V line. I will have a microcontroller that will drive the MOSFET as needed - but I'm sure/guessing I need some type of gate drive to go between them? Along with some supporting passive devices. Can you give me some guidance on this?

 

I think you would want high side switching between the two power supplies using P channel power mosfets. Diodes would not be necessary.

 

I imagine a PWM that makes a well-regulated 40V and a somewhat regulated 32V. Or use two PWM to get the 40 and 32.
There needs to be an isolated Gate Driver on the P-MOSFET. Simple Zener 15V G-S to limit the voltage swing. Current source to pull down to turn on the Gate.

This project assumes the load is resistive and has no capacitor on it.

I see you want to use a N-MOSFET. You will need an isolated gate driver like this, and a floating 15V supply to run the gate driver.

 

Though on your boost power supply.
I took your boost 24 to 32 supply. Used a 1:1 coupled inductor. Many companies make them.
Going from 24 to 32 is adding 8V. By doing that twice we get another 8V that could set on top of the 32.