Programmable multi-voltage-threshold detector IC?

[Greg Robinson]

Hi everyone,

I'm currently working on a linear bench power supply design. I have a five-tap (single-winding) transformer I'm hoping to use for this project, similar to multi-tap transformers used in classic linear bench power supplies where the input voltage is switched to minimise power dissipation in the pass device at low voltages and high currents.

The obvious answer is of course multiple window detectors with some hysteresis to switch in the next tap when a minimum threshold is reached, but I was hoping for a slightly simpler and more elegant solution.

Is anyone aware of an IC with programmable multiple-voltage thresholds? I'm thinking there might be something that uses a resistor ladder to set multiple thresholds that trigger corresponding output pins.
One thing that came to mind was the classic LM3914 bar-graph IC, but unfortunately it does not suit my purposes because the taps are not all equal voltage, nor are they logarithmic, so the LM3915/LM3916 are out too.

If anyone can make any suggestions, it would be greatly appreciated!

[Cliff Matthews]

I was pondering the LM3914 tap-switch idea too. I get them for 0.25 each. http://www.ebay.com/itm/5pcs-LED-Display-Driver-IC-NSC-DIP-18-LM3914N-1-LM3914N-1-NOPB-Best-/140716254031 You have 10 outputs (>20 in cascade) is that not sufficiently granular to select one of the other 4 taps? (the first tap could be the default at startup, through just a rectifier diode to block when another higher V tap is selected).

[Ian.M]

Use a MCU with a built-in ADC?
Anything else will be far more complex and the code to select 1 of N outputs depending on the input compared to an array of thresholds can be knocked out in an hour or so.

[Greg Robinson]

You have 10 outputs (>20 in cascade) is that not sufficiently granular to select one of the other 4 taps? (the first tap could be the default at startup, perhaps through just a diode).

Thanks Cliff! That was something that had not occurred to me, I think that will suit me quite well.

Use a MCU with a built-in ADC?

Thanks for your response Ian. I want to keep all the control loops in the analog domain for now, I want to keep things fast with minimal complexity.

[Ian.M]

Most MCUs could sample an ADC and run through a chain of if() ....  else ... statements to do the thresholding faster than heavy duty relays can switch.
However if you really don't want to go down that route, use a multiple rail multiple output PSU supervisor IC, and SPDT relays that switch between current tap and the output of the relay for the next lower tap so you don't need extra logic for the windowing.

[Gyro]

I'm currently working on a linear bench power supply design. I have a five-tap (single-winding) transformer I'm hoping to use for this project, similar to multi-tap transformers used in classic linear bench power supplies where the input voltage is switched to minimise power dissipation in the pass device at low voltages and high currents.

You say it's a single winding transfomer, do you mean that it is an autotransformer, which does not have separate isolated primary and secondary windings? If so, it is unfortunately completely unsafe to use it for a bench power supply, galvanic isolation between mains and the output is essential. The only way you could use it safely would be to build in (not attach separately) a 1:1 isolation transformer.

Sorry if I am misunderstanding you, maybe you can clarify?

[Greg Robinson]

You say it's a single winding transfomer, do you mean that it is an autotransformer, which does not have separate isolated primary and secondary windings? If so, it is unfortunately completely unsafe to use it for a bench power supply, galvanic isolation between mains and the output is essential. The only way you could use it safely would be to build in (not attach separately) a 1:1 isolation transformer.

Sorry if I am misunderstanding you, maybe you can clarify?

Single SECONDARY winding, with multiple taps (windings connected in series internally), as opposed to multiple isolated (from each other) secondary windings.

[Gyro]

Ah, good. Adding that magic 'Secondary' word makes all the difference! 

Edit: No reason that they need to be muliple secondary winding, tap switching will be fine. Do note though that the power output capability of your transformer will go down on the lower tappings as the secondary winding wire will have a certain maximum rating that will correspond to the full load rating of the transformer when using all of the secondary winding. Any chance that one of the taps is a center tap? if so, you could maybe gain efficiency by switching between full wave rectification (low range) and bridge rectified (high range).

[Greg Robinson]

Edit: No reason that they need to be muliple secondary winding, tap switching will be fine. Do note though that the power output capability of your transformer will go down on the lower tappings as the secondary winding wire will have a certain maximum rating that will correspond to the full load rating of the transformer when using all of the secondary winding.

Yep, current rating is of course the same for the entire winding, that's easy, it's a bench supply, so will be current limited. Expecting the full VA rating of the transformer from lower voltage taps would be an incredibly noob mistake. Hope I'm beyond that sort of blunder!

Any chance that one of the taps is a center tap? if so, you could maybe gain efficiency by switching between full wave rectification (low range) and bridge rectified (high range).

No centre tap, so unfortunately that idea won't work.

Most MCUs could sample an ADC and run through a chain of if() ....  else ... statements to do the thresholding faster than heavy duty relays can switch.
However if you really don't want to go down that route, use a multiple rail multiple output PSU supervisor IC, and SPDT relays that switch between current tap and the output of the relay for the next lower tap so you don't need extra logic for the windowing.

Well this is going to be a fairly high voltage power supply, around 400V (don't worry, I repair, design and build tube guitar amplifiers for a living, and have built similarly high voltage variable supplies before - I know what I'm doing, but I expect someone to give me a stern talking to anyway now that I've revealed that... *sigh*), so I'm pretty much stuck with relays (or expensive SSR's or equivalent discrete circuit) for the tap switching.

[Cliff Matthews]

What are your thoughts on cascaded relay's in bar-graph mode? At one time I was thinking about MOSFET's in dot-mode but started to think about gate capacitance and having 2 on during tap transfer.. (needless to say, I flushed that idea). I have 8 old Hammond transfo's lots of bridges, and a bunch of P&B relays, so this is what my boiler-plate started to look like (one day maybe I'll get'er done..)

[Greg Robinson]

I would move the relays to before the rectfiers for better clearance times (DC can sustain current for a longer period than AC during switch opening, it's also more likely to cause contact pitting or welding), and just use a single rectifier. I'm also unsure why you've got unequal voltage tapping points (that's an issue I'm trying to work around, something I would have avoided if possible), but I'm sure you have your reasons. Overall your solution looks excessively complicated to me, although I don't see any fatal errors.

[Gyro]

Yep, current rating is of course the same for the entire winding, that's easy, it's a bench supply, so will be current limited. Expecting the full VA rating of the transformer from lower voltage taps would be an incredibly noob mistake. Hope I'm beyond that sort of blunder!

That's the problem on a forum, particularly in the beginners section, you can never make assumptions about what people know (hence the 'secondary' question too!). Better safe than sorry.

[Seekonk]

I like the idea that HP used in their early supplies, a phase triggered SCR as the pre regulator for the linear supply.