LM2585T-ADJ experiences?

[Nigel Goodwin]

I'm currently playing with the LM2585T-Adj as an up-converter (to convert 12V to 24V), and I designed a small PCB for experimenting, and added a choice of fixed resistors, or a variable pot with limit resistors top and bottom.

First one I built up using a 560uF output capacitor, as per the datasheet, and fixed resistors - this worked fine, and gave 23.6V out - near enough for the fixed values.

However, 560uF are pretty large, so I thought I'd try a smaller one - a 100uF as used on the input - so I built another one up, this time adding the variable pot option. Also, because it seemed a 'nice' thing to do, I stuck a 0.1uF across the 100uF on the output (just as the 100uF on the input has). The unit then didn't work at all, all you got out was the input voltage minus the diode drop?.

After much testing, and finding nothing wrong, I tried the previous fixed unit, but added the extra 0.1uF to that as well - at which point that unit refused to work as well.

So I removed the 0.1uF's, both units now work perfectly, and I adjusted the variable one to exactly 24.0V.

Anyone know why adding the 0.1uF across the output capacitor stops them working?.

 

[crutschow]

Not good design practice but some regulator feedback loops actually depend upon the small ESR of the output electrolytic capacitor for stability and proper operation.
That would seem to be the case for the LM2585.
You could try adding a small resistance in series with the 0.1uF cap to see if that allows it to work.

 

[Ian Rogers]

I had an issue with an LM2675 where a 0.1uf on the input caused a problem... The 0.1uf destabilised the 52Khz operation and the LM2675 blew off the board... Just after it sang a high pitched tune....

Suffice to say, I no longer use them... It may have just been the ON semiconductor types because I'm sure I used them on the official ST ones...

 

[eTech]

I'm currently playing with the LM2585T-Adj as an up-converter (to convert 12V to 24V), and I designed a small PCB for experimenting, and added a choice of fixed resistors, or a variable pot with limit resistors top and bottom.

First one I built up using a 560uF output capacitor, as per the datasheet, and fixed resistors - this worked fine, and gave 23.6V out - near enough for the fixed values.

However, 560uF are pretty large, so I thought I'd try a smaller one - a 100uF as used on the input - so I built another one up, this time adding the variable pot option. Also, because it seemed a 'nice' thing to do, I stuck a 0.1uF across the 100uF on the output (just as the 100uF on the input has). The unit then didn't work at all, all you got out was the input voltage minus the diode drop?.

After much testing, and finding nothing wrong, I tried the previous fixed unit, but added the extra 0.1uF to that as well - at which point that unit refused to work as well.

So I removed the 0.1uF's, both units now work perfectly, and I adjusted the variable one to exactly 24.0V.

Anyone know why adding the 0.1uF across the output capacitor stops them working?.

View attachment 120429

You could replace the 560u cap with 5 or 6 100u caps in parallel. It would also lower the ESR as well.

eT

 

[simonbramble]

I think crutschow is correct. The loop response of the controller probably relies on a small ESR of the output capacitor. If you think that the ideal capacitor presents a 20dB/decade roll off on the loop response, (and a worst case 90 degrees phase lag with it) adding an ESR to this gives a high frequency zero that flattens off this roll off (and gives a 90 degree phase lead). The control loop is relying on this phase lead. If you want the benefits of a low ESR (better ripple on the output), but without the phase lag problems presented by an ideal capacitor I would experiment with a capacitor across the top feedback resistor. This will give a phase lead. Now... what value... with the low value feedback resistors you would need something like 10nF. That will boost the phase at 430Hz

Failing that, use the LT3757. This will more than adequately do the job, works with ceramic output caps and will probably be more efficient and smaller

 

[simonbramble]

You could replace the 560u cap with 5 or 6 100u caps in parallel. It would also lower the ESR as well.

. I don't recommend this. Putting many in parallel with give you the same problem as a single low ESR cap

 

[eTech]

. I don't recommend this. Putting many in parallel with give you the same problem as a single low ESR cap

Why is that? The controllers rely on small ESR.

 

[simonbramble]

The problem stated in the original post was that the controller stopped working when a low ESR capacitor was placed across the original high ESR capacitor. Therefore the net ESR of the 2 capacitors has been reduced and the controller stops working. From the posts above, we are assuming that the controller needs an output capacitor with a relatively *high* ESR in order to maintain loop stability. Putting several capacitors in parallel will parallel up the ESR of these capacitors giving an overall low ESR... and you are back to the original problem: the controller does not want to work with a low ESR output capacitor.

 

[Nigel Goodwin]

Not good design practice but some regulator feedback loops actually depend upon the small ESR of the output electrolytic capacitor for stability and proper operation.
That would seem to be the case for the LM2585.
You could try adding a small resistance in series with the 0.1uF cap to see if that allows it to work.

Thanks for that, I'll just leave it out - I was just interested why it would stop it working? - it's not like I'd even made a space for it on the PCB (I just soldered it across the bottom of the output capacitor). This is only a learning exercise, with a view to building it on a larger board as part of a project - and I woud have added space for the extra capacitor on that board, now I won't

 

[Pommie]

Does it need a minimum current to operate properly?

Mike.

 

[eTech]

The problem stated in the original post was that the controller stopped working when a low ESR capacitor was placed across the original high ESR capacitor. Therefore the net ESR of the 2 capacitors has been reduced and the controller stops working. From the posts above, we are assuming that the controller needs an output capacitor with a relatively *high* ESR in order to maintain loop stability. Putting several capacitors in parallel will parallel up the ESR of these capacitors giving an overall low ESR... and you are back to the original problem: the controller does not want to work with a low ESR output capacitor.

Putting 100uf capacitors in parallel should be OK as long as the total ESR and capacitance fits the requirements. However, ESR might not be the problem (or maybe its actually higher with the 0.1u cap than it should be). I'm curious now...guess I have to build one of these.

 

[Nigel Goodwin]

Does it need a minimum current to operate properly?

Mike.

I tried it with, and without, a load - actually a big 48 or 50 ohm resistor (can't remember which, it's in a box) used as a dummy load - one of the intended applications takes about 0.5A at 24V, and the resistor simulates it. Without the load, and without the 0.1uF, it still outputs 24V (on a meter) but you can see it's squegging on a scope.

 

[crutschow]

maybe its actually higher with the 0.1u cap than it should be

How could a capacitor in parallel increase the ESR?

 

[eTech]

How could a capacitor in parallel increase the ESR?

Sorry...mis-spoke...it wouldn’t.
My point is...I think the .1u cap is affecting the switching characteristic of the controller and is not really related to ESR.

 

[rjenkinsgb]

I've been looking at this & at the data sheets I can find, though they only have a block diagram rather than full internal circuit.

The IC appears to have an oscillator that is completely self contained and cannot be affected by anything other than too low a supply voltage.
The output looks to be gated off by either overcurrent on the switch pin, over voltage on the sense pin, or the comp pin being pulled low.

There is no way from anything I can find that output load effects should be able to completely kill the switch output?? It makes no sense at all.


When the switch output turns on, there is no current path from that to the output cap; that connection only exists after the switch turns off and the diode conducts.

Even if some interaction between the output cap and feedback sense pin temporarily shut it down, it should keep re-trying and give a possibly erratic output, not no output at all. There is something very screwy with that..



Lateral thinking mode -
What's the capacitance of the power diode? Could that be high enough so the initial current through the diode capacitance, in series with the ceramic cap, cause an instant overload every cycle as the switch pin turns on?

>google<

Wow...
The MBR340 capacitance with zero bias is 500pF, according to the data sheet; I think that is what's causing the no-output effect, the current surge is just too high.

 

[Nigel Goodwin]

I've been looking at this & at the data sheets I can find, though they only have a block diagram rather than full internal circuit.

The IC appears to have an oscillator that is completely self contained and cannot be affected by anything other than too low a supply voltage.
The output looks to be gated off by either overcurrent on the switch pin, over voltage on the sense pin, or the comp pin being pulled low.

There is no way from anything I can find that output load effects should be able to completely kill the switch output?? It makes no sense at all.


When the switch output turns on, there is no current path from that to the output cap; that connection only exists after the switch turns off and the diode conducts.

Even if some interaction between the output cap and feedback sense pin temporarily shut it down, it should keep re-trying and give a possibly erratic output, not no output at all. There is something very screwy with that..



Lateral thinking mode -
What's the capacitance of the power diode? Could that be high enough so the initial current through the diode capacitance, in series with the ceramic cap, cause an instant overload every cycle as the switch pin turns on?

>google<

Wow...
The MBR340 capacitance with zero bias is 500pF, according to the data sheet; I think that is what's causing the no-output effect, the current surge is just too high.

As it happens, I did actually use MBR340's (the schematic is taken off the datasheet) - as I had to order diodes for it, it made it simpler to order the one off the diagram rather than study datasheets.

I'll have to try using a different diode next time I build one up (I got ten boards, from JLC PCB).

 

[rjenkinsgb]

Most schottky recs seem to be high capacitance .

The only similar axial one I've found so far that's significantly lower is a STPS3150RL. That's 3A 150V, just under 200pF and around 33p from Farnells.

These look like a good possibility - fairly low forward voltage and 40pF maximum capacitance.

https://uk.farnell.com/on-semiconductor/mur420g/diode-power-rectifier-uf/dp/1651054

 

[alec_t]

I'm finding no data for the UPL1V561 output cap in the datasheet. Seems to be obsolete. Is that the actual type you used?

 

[crutschow]

.I think the .1u cap is affecting the switching characteristic of the controller and is not really related to ESR.

Well, the small capacitor can affect three parameters; the total capacitance, the ESR, and the series inductance. The change in capacitance is negligible, so that leaves ESR and the series inductance, take your pick.

 

[rjenkinsgb]

The ESR of the output cap is irrelevant -IF the diode in use has low junction capacitance.

It just happens that the diode in the circuit has ludicrously high capacitance and (if my theory is correct) it is that which causes the shutdown, if the rest of the current path from the IC switch pin to ground pin is low enough impedance:


There is a current sense circuit in the IC, in line with the switch. An excess current trips a latch and turns the switch off again for the rest of that oscillator cycle.

With the high capacitance diode and low impedance in the rest of the output circuit, that is apparently happening at every switch cycle, so no output is produced.

With a conventional fast recovery diode in place of the schottky one, the output cap type should have no effect.


Anyway, I now know what to use if I ever need a varicap diode for a medium wave tuner!
Having looked at dozens of power schottky data sheets last night, most are in the 400 - 1000 (or more) pF range with zero bias, dropping rapidly with a few volts applied...
[I used 1N4007 as varicaps in VHF gear for years. They have as good specs in that regard than some expensive commercial varicaps].


Conventional 3A - 5A fast recovery diodes are around an order of magnitude lower capacitance than the schottky ones.

It would be interesting to see the IC switch waveform on a scope.. I'd not be surprised if adding one or two ferrite beads on the schottky diode leads would allow it to work, by reducing the rate of change of the switch on edge current through the diode capacitance?