High side high voltage MOSFET switching with 3V logic drive?

[Martin G7MRV]

Im not sure how to go about it. The PIC processor that would provide the PWM signal will be working at 3v (to match the rest of the radios logic - it also directly interfaces with the vintage logic circuits), the 80v is available within the radio anyway.

Ive been studying this idea here https://www.edn.com/design/analog/43...h-a-PWM-signal but with the idea of replacing the LM317 with a TL783, which can go up to 125v. I suspect though that i'd also need something between the op-amp and the reg to provide a sufficient swing?

I have very little experience of PWM other than as lamp dimmers or old radio-control!

[SiriusHardware]

Mmmm. An 80V squarewave in the vicinity of, even connected to, a receiver? Even low voltage PWM may generate a lot of unwanted RF hash. I liked the optocouplers idea better for that reason, at least all the control signals are then completely static.

[Martin G7MRV]

The way im interpreting that circuit, the PWM signal at 3v is immediately smoothed out to a varying DC level by the RC network on the input of the op-amp, so the control voltage out of the op-amp is just a varying DC level into the regulators adj pin? If thats right, then the PWM signal could be kept to a minimum by having the RC filter very close to the source pin on the PIC, possibly even a bit of shielding.

But yes I appreciate the point regarding interference. I suppose the only way to know for sure would be to build and test it.

The opto method is very clean and stable, but very complex and needs a lot more space, in an already very cramped radio. The PWM idea would be extremely compact and very few parts, but would it work? Would the op-amp need to be rated for a high voltage? I guess I could order a TL783 to try

[Terry_VK5TM]

Originally Posted by Martin G7MRV

My only problem with this approach is the need to go up to close to 70v with an 80v source (actual max measured voltage is 65v), so the LM317 isnt suitable,...

First page - LM317 -> 170v in -> 1.2 - 160v out.

The adjustable pot could be replaced with a FET driven by PWM (but put a fixed resistor in series with the FET).

[Martin G7MRV]

ah, thats something I didnt know! Well spotted Terry! So the 317 doesnt have reference to ground, so as long as the max input voltage isnt exceeded across the device, it will be ok.

So, the only issue then is driving the adjust, and as you say, a PWM driven FET should do that. If the PWM is first smoothed out to a varying DC level immediately at the output pin of the PIC it should be possible to control the risk of noise generation. But if not, then a potential divider chain could be selected by the PIC output ports, or even a digital potentiometer driven by the PIC over I2C? Actually, that sounds like a very good idea,

I dont have any high voltage transistors in my stock, nor a digital potentiometer, so i'd have to order these, and they will probably take longer than the optocouplers which are already enroute, so i'll continue to prototype the opto idea, but i'll try this regulator idea out as well afterwards.

On a plus point for the digital pot, these are made by Maxim - and I can get developer samples free!

Just throwing something into the pot, current mirror?

[Terry_VK5TM]

Slight oops on my part, I had a much larger app note that I was going to post that had other stuff in it hence the first page comment and now it appears I have lost that one - I blame it on the headaches this weekend.

Anyway, to return to subject, one of the diagrams was a series of paralleled transistors with associated resistors for voltage setting, each one being individually switched in depending on required voltage.

I'm not a particular fan of this type of arrangement as if none of the transistors is turned on, whether through fault or on purpose, almost the full input voltage can appear on the output. Better to have 1 permanent resistor and shunt that to get the desired voltage.

That would also get quite messy for 18 different voltages, so a variation might be to switch a set of 4/5 transistor/resistor combinations in various parallel sequences.

Re digital pot IC's - most don't have the current handling capability for this sort of application (well they never used to and it was specifically mentioned in some older datasheets).

Merlin's comment about a current mirror rings a bell about some sort of arrangement where a voltage was somehow injected into the adjust pin, I'll see if I can find it (I'm not referring to the -ve voltage option to make these things go to 0V, although it may be somehow related).

[Red to black]

One point I never saw mentioned wrt to the opto in series solution, is we generally connected the inputs (LED) in parallel and the output side (transistor) in series. apologies if this was taken for granted.

[Trigon.]

Perhaps the most compact possible solution with the minimum amount of wiring would be to use Maxim's MAX14752 high voltage multiplexers.

Since the maximum voltage to be switched is 65v (post#18), their 72v rating is good enough, but would need the 80v supply dropped (e.g. by 12v using a zener down to 68v).

The muxes are 8:1 so three would be required. There would then be two possible ways of selecting the required input. One would be to use 3 select lines commoned to all, plus 3 chip selects and mux outputs commoned. Or, requiring one less signal, 8 taps can go to the first mux, another 8 to the second mux with the remaining two taps plus the outputs of the first two muxes being four inputs to the third selecting the final output. (A,B,C to muxes 1&2, C,D to mux 3.) (This also would prevent a software bug accidentally connecting two taps together...)

Note that the MAX14752 select inputs can be driven directly from the 3v logic - so three chips, and a zener + supply cap! (eBay can supply 3 muxes for £16.05 with free delivery.)

The 'gotcha' is of course the chip packaging - the muxes are in 16 pin TSSOP so would need an adaptor board to make life easy. (5 for 99p from eBay/China or a liitle under a pound each from the UK.) Still a very small footprint though.

Cheers

[Martin G7MRV]

Thats an interesting device. And being a Maxim part, its something I can obtain free samples of to play with. Ive got the break-out boards for this size SMD already.

Driven off a MCP23008 8-bit port expander, with a further MCP23017 16-bit port expander to handle the other control lines in the project (which control the radios synthesiser), would leave me two spare ports on the expanders, meaning I only need two more I/O ports, so could get away with one of Microchips tiny little 8-pin PIC microcontrollers as well.

Im liking this idea! I'll get onto Maxim for some development samples.

So, I now have three very viable means of taking over control of the tuning voltages. Thanks to you all!

All thats left now, other than to build prototypes when the parts arrive, is for me to learn to program PICs in C! Thats where the real steep learning curve is!

edit - Maxim samples ordered - several MAX14752s, some 3v RS232 drivers and a few digital pots to play with!

[Martin G7MRV]

Sorry, thats wrong, I missed out one of the MAX14752s in my calculation! Two MCP23107s would do it, leaving me 6 ports free, plus whatever the PIC has. So I could use one of the smallest PICs still, so long as I have two I/O pins for the I2C, and use the interrupts on the port expanders to get my input controls. And I have some little 8-pin PICs in stock!

[Martin G7MRV]

The MAX14752s arrived today - oh my! 16TSSOP package about 5mm square, with 8 legs per side! If they work, it will make for an extremely small, compact control board! (if that is I can solder them!). Another aspect of course iw whether there is enough headroom in the 80v regulator to provide the current to work them, but there should be.

Ive ruled out the PWM controlled regulator method, due to the interference risks. So its going to be a toss up between the physically larger but very cheap optocoupler method, or the very small but considerably more expensive multiplex ICs.

Time to start prototyping...