What sample rate to feedback output voltage in dc-dc converter 1

[bpelec]

I am designing a full-bridge dc-dc converter that will use peak current mode control and switching at 100 kHz. The control loop will have two feedback signals: the current through the primary transformer coil and the output voltage.

Here's the problem... Due to the nature of the application, there is no physcial path I can use between the control circuitry and the voltage output. As a result, I need to remotely/wirelessly measure the output voltage.

My thought at the moment is to use a short range RF chip such as those made by Xenics. To make this work, I would have to sample the output voltage using an ADC, transmit the digital signal using the RF chip, and then reconvert the digital signal to analogue using a DAC.

Questions:

1) Am I mad?
2) What sample rate would I need to use on the output voltage in order to make the control loop work? The RF chips I have seen can sample at about 100 kHz or a bit more. This is clearly not enough to characterise switching harmonics in my system, but would it be enough to regulate the output voltage?

 

[IRstuff]

Can't you run that through your circuit simulation?

TTFN

 

[VisiGoth]

1) You are not mad. You need to consider latency through your communication system relative to the control loop bandwidth.

2) Look at sampling synchronous with the chop rate, and you should find the "best" time to do the sampling. You may find that certain phases will have great difficulty with repeatable and accurate measurements. It is after all a "power" supply.

Some advice.

1) Sample at least 8X higher that you loop control bandwidth
2) Sample synchronous with your chop rate, even integer ratios are OK, like 2/3.
3) Make sure your communications latency is taken into account into your control laws and control systems model.
4) Make sure you do not have jitter in your com system that is great enough to affects your control system.
5) Make sure you do not have a clock synch problem in your comm system. (do not assume that 100 kHz at the remote sensor is 100 kHz with no phase slip at the receiver).
6) Have a very good control systems model
7) Have a very good control systems model
8) Have a very good control systems model
Well, you get the point on the last item I hope. It will answer all of your questions, except the fidelity of making measurements during switching intervals.

 

[itsmoked]

I personally would rather run open loop... Look at the load and figure out what it will be doing. I mean if your load is a big fat resistor why would you need any regulation? If the load has a few different load modes can you not monitor the current at the supply and make fixed step changes in the voltage that testing would describe? Running wireless voltage feedback of a 100kH switcher does sound mad to me... Have you been bit by a squirrel that was acting strangely?

 

[OutToLunch]

here's an idea - you could convert the output voltage to an 8 bit (or however many bits you desire) digital signal and transfer those bits in parallel through opto couplers then put that through a D2A coverter and then back into your control loop.

 

[bpelec]

Thank you for all of your replies!

IRstuff, I am building the circuit up in OrCAD and I almost don't even want to try and make that kind of circuit simulate - I think it would be pretty painful. I am planning on builing a bread board version...

VisiGoth, thank you for all your suggestions and points. I will do a more detailed study into model and my loop bandwidth.

itsmoked, unfortunately my load is quite variable... I am inclined to agree with you that it is a bit mad, and may well not work, but bare in mind I do already have current feedback from the primary side of the transformer... How did you know about the squirrel?

OutToLunch, the physical separation between the control circuitry and the voltage output is around 20cm. As a result I won't be able to use opto-couplers. Thank you for the idea though...

Right, I think I need to go and do some calculations...

I will post again if anything comes if this...

BPELEC