Need help with LM5085

[Un]fortunately I do not have the equipment required to carry out practical tests.

You say that the MCPCBs work OK and I wonder if they are providing benefits in terms of the cladding acting as some form of ground plane but then that would assume that the ground planes on your new boards are not functioning as required or they are in some way sub-optimal. Without reading back did your single layer boards perform badly as well?

I don't understand why the changes you have made to the way the current sense resistor is mounted affect the circuit in any specific way at least not anything, if at all, that would make sense. In terms of incorrect output voltages/operating frequencies and output voltage changing with input voltage the best I can suggest is that there is some sort of 'mode locking', sorry for the term, or 'mode change.

You will have seen that the self-oscillating circuits depend on the gate delays. Perhaps your in circuit delays are sufficient such that the circuit attempt to stop operating as fixed on time and moves into self-oscillation. There is mention in the data-sheet that particular delays must be held below some value. Otherwise as switching edges change their relative location with VIN changing it may be that at a particular input voltage, or in a range thereof, one of the edges is occurring at a point where some other 'decision' is being made by the IC and it up sets that. Rather than 'regulating' the circuit might choose to lock to that edge and such behaviour can happen over quite a wide range before things decide to become unstuck and work 'normally' again.

I'd still be inclined to try the small RC filter on the ISEN pin but one other suggestion would to be to look at the feedback path and connection. I do not know if this would be possible or even 'ideally' achievable on your present boards but would it be feasible, if this makes sense, to add a 'twisted pair' connection, wire-wrap or enamelled copper wire, from the IC's feedback pins directly to the output across the output filter capacitor?

I'm afraid, barring other 'unknown' things it is back to noise and layout again which might mean re-doing the boards but I would imagine you would be wary of trying without a guaranteed result and you are stuck for time. If I had the software I would be happy to try something out for. That might mean grabbing hold of and learning GEDA or KiCad, ooo and there is Eagle but that has its board size limited. I run Ubuntu. You might notice that when it comes to circuit diagrams I am a bit of a pedant when it comes to neatness. The same also applies to PCB layout and I do have a fair bit of experience of laying out SMPS circuits...

Genome.

The first MCPCB, the long one worked from the first minute, surprising but with the values calculated by the WEBENCH. It was a little sensible at the output but this is because the WEBENCH picked high values for the feedback resistors. When replaced to values recommended in the data sheet, 1-20K worked. This was a 12V/1.5A setup.

Because it worked I decided to stick with the LM5085 for other applications and the other board was made. Far from being with an optimal layout (same as the first one in fact) this one made some problems. It was 8.8V/2.5A same as the one I made last. But lowering the clock to 170kHz and replacing the 10mohm Rsen with a 5mohm brought it to work.
There is a glitch, but I am not sure if it is from the circuit or some software bug. This step-down powers a battery charger and 2 strings of LED which are PWM dimmed. There is a certain point where if the dimmer pot is turned very slowly there is a decrease in the intensity. A tiny point, really. Again, not sure the problem is with the step-down.

The third board, 15V/1.5A was from the beginning assembled with low clock and 5mohm Rsen and it works ok.

I couldn't in my worse dreams immagine the 2 sided board will behave like this. The ground plane on the bottom layer is almost one piece, except for a small track which goes to both step-downs and shuts them off by use of a small MOSFET (which by the was was removed during the tests).

The interesting part is that MCPCBs tend not to like MOSFETS, LED drivers and etc., looks like a capacitive sensitivity problem.
Weird, but the LM5085 with the poorest layout works on them.

It may be that the ground plane creates some capaitance, also maybe vias creating some noise as I saw in a Power Point they sent me. All this is not a problem with the MCPCB.

Anyway, I had to move ahead with this project, at least for now. So I installed today the original circuit, with the original MOSFET and the other components with the old values used on the MCPCBs, with only one change: The SMT Rsen was replaced with a TH type, soldered with one leg kept a little longer than needed.

Both circuits work ok, too bad I can't use the TH resistor as a permanent solution, as I can't solder anything on the bottom because the board is installed with a thermal silicone pad directly on an aluminum surface.

Beats me if I understand what makes this resistor to the circuit, but it works.

I am tired of this IC, really, 2 weeks fighting with it with no logical explanation is too much.
I find it completely unreliable.

What I am working on is a prototype, so here I can live with this "stitch" but I need to find a permanent solution.
I looked for alternatives at Linear and TI, but it looks like National have the largest range of buck converters.

I don't want to sound masochist, but one alternative can be the LM222678, what they call 4'th generation switcher.
It works at constant 500kHz, but integrated MOSFET and sense resistor, holds up to 5A from 4.5 to 60V input and very low parts count.
The immediate thought was reducing the external factors able to create mistakes, only the diode, inductor and feedback resistors are needed, sorry the in/out caps too.

The biggest problem is that while this can be a solution for the 8.8V circuit, the problem is with the 15V, as it is very close to the input's lower limit.
The 5085 has a very smooth switch when Vin is close or lower than the Vout and Vout the MOSFET is kept switched and Vout follows Vin.

The 222678 is not capable of doing this.

Nasty experience, if I had the room for big inductor and large capacitors I could use the good and old LM2576.

Do you have any experience with some other buck converter that works more or less at the same parameters?

The 5085 is like a young sexy blonde, very attractive but after a while you discover that is a little stupid and reacts in the most unexpected way.
So I guess that I should keep looking an older brunette, more experienced and more reliable.

I'm afraid I largely grumble about with basic things rather than 'esoterics'.

You want a basic or maybe general building block to cope with many situations and it needs to be 'small', within [non]specific definitions of 'small' and it needs to 'fit' where the same might apply.

Plus I assume you want it to work. Some people are really picky.


It seems, in some cases, you are having to duplicate 7085 circuits to gain different output voltages within any particular application. Perhaps in others you are processing those voltages to achieve some other form of regulation that would not be fixed. You mentioned a PIC controlled boost converter. I'm sorry, there is a lot of information buried in the thread.

Tentatively... how would things look, dare I say at 'system design level' if you had one converter block capable of operating from [mumble] 5 volts in to 60V in [figures subject to change] and providing, up to, four fixed output voltages of which one would tightly regulated and the others should 'track well' based on two of these...

https://www.ferroxcube.com/prod/assets/efd10.pdf

One would be a transformer and one would be a coupled inductor and hopefully standardised. Otherwise you would have a 'simple' and low tech forward converter hopefully operating down in the doldrums of 100KHz.

As I say I am tentatively picking things out of my hat.

Do you have a 'wish' list based on your range of requirements and what has to live with what. Perhaps in table form.

Don't forget, this is tentative.

Genome.

I'm afraid I largely grumble about with basic things rather than 'esoterics'.

Click to expand...

When things are not working as expected sometimes we need to go down to basics. There are people that are facing the problem and deal with it and others that prefere the easy way around, bypass the problem. However, sometimes when you are facing tight schedule you have no option. Personally I attempt always to solve the problem. I believe that learning the hard way from mistakes is practically the best education. This is specially true if I need to use a certain part/circuit in many applications, after all when you gain experience with it is easier to use it as a "block" module changing it to the different neds. Of course when I bump into some problematic part, such as this 5085 after a while you loose confidence and it is better to find something that works rather than dealing with an 8 components circuit for 2 weeks. It is cheaper to make a new PCB and will take less time.

You want a basic or maybe general building block to cope with many situations and it needs to be 'small', within [non]specific definitions of 'small' and it needs to 'fit' where the same might apply.

Click to expand...

It seems, in some cases, you are having to duplicate 7085 circuits to gain different output voltages within any particular application. Perhaps in others you are processing those voltages to achieve some other form of regulation that would not be fixed.

Click to expand...

Absolutely right. Let me in a few words explain what I do. My company designs LED illumination systems for automotive applications, both internal and external. Not the regular rear lights for trucks, but for special vehicles, including peripherial illumination, headlights, internal illumination and etc. These are usually emergency vehicles or mobile command/communication units.

This is a very demanding environment, not to mention the mechanical, thermal and electrical stress and the huge ammount of standards they need to meet. Definetly heavy duty stuff, somtimes we wonder why a simple lamps needs microprocessor, can-bus communication and lots of logics, not to mention human factors engineering and only at the end also photometric performance.

The last series of several models we were requested to design need battery backup, but the size of it very small. So we needed a step-down, charger and step-up. both the step-down and the battery powered step-up need to power some 4x1W LEDs. Simple, but make it work at high temperatures and then pass EMC and electrical tests including surges and spikes as high as 160V for 500ms simulating cranking, 5 repetitive 100V/50ms pulses, 250V induced spike and 1 hour at 28VDC with a +/-7V sinus modulation simulating a battery disconnection. Oh, and they are required to work down to 15V and this + all the above without any observed disturbance (let's remember that LEDs are visible and very fast).

Sooo.... believe me that the step-down is only the part visible above the water of an iceberg.

One would be a transformer and one would be a coupled inductor and hopefully standardised. Otherwise you would have a 'simple' and low tech forward converter hopefully operating down in the doldrums of 100KHz.

Click to expand...

Of course, size is a big problem, so is heat. Large parts such as inductors are a problem when subjected to vibration, capacitors became the weak link in terms of working hours specially at high temperatures ( you know, there are people still telling customers legends about LEDs working 100K hours, but nobody mentions that a good capacitor at high temperature will keep 1/10 of this in the best case - not that at these temperatures the LED will last that much, but this is not the point).

So higher frequency DC-DC converters, voltage mode such as our step down or current mode as LED drivers allow the use of smaller inductors and capacitors, in most cases ceramic types which are preferable if compared to electrolytics.

Size, yes, the space available is tight, always tight. Thickness is even more demanding.
If you look at the PCBs you might get an idea of the required size.

We also use LED DC-DC drivers, usually buck types, with constant current for most applications. The off-line types such as the SUPERTEX are capable of working from 8-450V !!!. They solve a lot of protection problems. By the way, I also work with a certain type for most applications, changing parameters according to the LED string size and current. I am using the Supertex HV9910 a lot.

But from time to time we need to use a combination of voltage mode step-down with linear constant current drivers on the LED strings, such as in the present project.

Regarding the regulation part, or how regulated I need the step-down to be, well, not at extremes, the LEDs have further regulation from the constant current driver (linear type) and the controller part, if there is such is also lowered to 5V with an additional regulator.

Plus I assume you want it to work. Some people are really picky.

Click to expand...

Well, yes, otherwise I can always use the LM5085 :lol:

You mentioned a PIC controlled boost converter. I'm sorry, there is a lot of information buried in the thread.

Click to expand...

No, the PIC is used as a controller for all the step-down, charger IC, and step-up, it monitors the incoming voltage, decides if to make input cut-off below 15V, switches off the step-down and turns on the step-up, and so on. I never tried a PIC controlled DC-DC altough I read several application notes on this subject.

Tentatively... how would things look, dare I say at 'system design level' if you had one converter block capable of operating from [mumble] 5 volts in to 60V in [figures subject to change] and providing, up to, four fixed output voltages of which one would tightly regulated and the others should 'track well' based on two of these...

Click to expand...

Hmmm.... something like this?
http://cds.linear.com/docs/Datasheet/8027fa.pdf

Looks like exactly what you are talking about - but you better seat before checking the price.

But yes, this would be the idea, an easily adjustable circuit within a certain power (volt/amp) range. Easy to say, but who knows what tommorow some custumer will ask for? For most applications something like the Linear LTM8027 module if it was at a reasonable cost could provide a nice solution.

Do you have a 'wish' list based on your range of requirements and what has to live with what. Perhaps in table form.

Don't forget, this is tentative.

Click to expand...

I will prepare it.

Thanks,

Emanuel

Whoops, as is the case with my 'bright' ideas this one seem to have fallen flat on its face

However, forgetting about the 'lateral' thinking, or lack thereof, going back to what you are doing, the method not the IC, since you mentioned Linear Technology then,

http://cds.linear.com/docs/Datasheet/4444fb.pdf
http://cds.linear.com/docs/Datasheet/44445fc.pdf

Synchronous Buck driver operating up to 100V on the high side Mosfet.

Add in a control IC, I'll have to find something suitable because their examples seem slightly complex, and a little bit of extra thinking and you would have something approaching an answer..

Genome.

Genomerics said:

Whoops, as is the case with my 'bright' ideas this one seem to have fallen flat on its face

Click to expand...

No, why?

I think your idea is indeed what I need - a flexible circuit easily adaptable to a reasonable range of output voltage and current.

I mentioned that I done the same with the LED driver, after overcoming several issues with it now life is easy, I have a certain PCB layout and a simple copy&paste is enough, with no worries if it will work or not.
The thing with the step down is maybe that I needed this setup for the first time and my first experience with it was as shown in the thread.

But I do need to find a solution, after all all I did so far is finishing the prototype, I have no intention to use the improvisations for the production.

Emanuel

Ideas may be 'good' but when it comes to implementation and doing sums sometimes you discover it was not good in the first place or will not work or, relatively speaking, it will just be too complex and big.... That would apply to what I was thinking of. Going back to what you have got and using building blocks to implement the 7085 but avoid its foibles should let you achieve the goal of something 'universal'.

We are looking at 7V-100V in with switching frequencies 200KHz-500KHz

http://cds.linear.com/docs/Datasheet/4444fb.pdf
**broken link removed**
**broken link removed**
AC/DC and DC/DC Power Supply - PWM Controller - UCC3803 - TI.com
http://www.ti.com/lit/gpn/ucc3803
http://www.datasheetcatalog.org/datasheet/texasinstruments/ucc3805.pdf

and conceptually you end up with,



Synchronous rectification should/might save you some power. Nice N-channel Mosfets with lower RDSon and higher voltage ratings. A 'proper' PWM control IC with 'proper' fixed frequency operation and the general separation of control and power devices should help with noise concerns.

7.5V - 100V operation @200-500KHz and pick your current level.

It would make the output inductor more complex but it would be possible to pick off low power windings from it for this part and any ancillary control functions you may need.

Genome.

Looks interesting, how stable will it be over temperature extremes?
The MOSFET for most of my applications can be a dual type in a SO-8 package, plenty types available.

What about the low voltage max. ratings of the UCC3803 and LTC4444, they will need a separate regulator?

I can assemble a prototype in a few days for testing.

Emanuel

I can assemble a prototype in a few days for testing.

Click to expand...

You really don't like your hair... ;-)

The UCC3803 and LTC4444 would be 'started' up from the input supply assuming you can guarantee the Under Voltage Lockout, UVLO, of the LTC4444,



The UCC3803 will get there quicker,



There may be 'sequencing' issues involved.

There are ways and means of bringing things up quickly from the input supply and then switching that off to lower dissipation at which point a secondary source of power should be available from somewhere else. If it is available anyway then better to use it. Otherwise if the unit has to be self contained then yes, as suggested, it would get its 'auxiliary' from the output inductor via an extra winding.

That would mean a 'special' component.

Of course if you can afford to 'waste power' then it might be run continually off the input supply. That would be 'sums to be done' based on what operating conditions you might expect.

In terms of temperature extremes I assume you are not talking about 'qualification' of the components in terms of 'commercial' 'industrial' 'automotive' 'mil spec'. They come, as you will know, in the various flavours.

In terms of regulation then the voltage reference that will be used is the one in the UCCX803,



They do not specifically quote a 'temperature coefficient' for the reference, unless it is somewhere else in the data sheet. Otherwise you get a 1000Hr dV figure at TA of 125C of 5mV or 0.1%.

In terms of 'weak' links then the Zetex current sense device is a bit of an unknown both in terms of bandwidth and its overall performance.

It is in there because I assume you do not wish to start putting 'arbitrary' sense resistors down in the ground return or associated places. As far as loop compensation is concerned then its 'problems', if they exist, might be handled. That is a 'guess'.

I can assemble a prototype in a few days for testing.

Click to expand...

After the 'basic' design is sorted.. :| ?

With apologies if required.

Genome.

Perhaps something like this?



No values, sort of, and probably not 100% in itself. There is also the issue that the UCC380X is going to start up first and probably/possibly try to turn on the upper LTC4444 driver at which point 'suddenly' nothing will happen since its 'boot' capacitor will be empty. However if the UCC380X's duty cycle is not 100%, and it will not be which unfortunately looses you input volts at the low end of the range, then that one might/should solve itself...

So, let the blither begin.

Zetex current sense thing. Q1, Z1, R1 protect it against over voltage. RSNS is, of course, the current sense resistor. Output gets sent down to the ISNS pin of the UCC380X on R2. Q4 and R10 provide 'slope compensation' from the oscillator ramp on RC of the UCC380X.

Start up supply is from Q3, R5 and Z2. Once running the circuit takes power from an auxilliary winding on the main output filter inductor, LPB, which will switch off Q3 via R3, R4 and Q2. Unfortunately there are various diodes in the way to avoid loading certain points and reverse bias of the Q3 BE junction..

R6, R7, R8, R9, C1, C2, C3, kitchen sink are general feedback components around the error amplifier in the UCC380X.. R11 and C4 set its switching frequency.

A1 might work out to be a couple of transistors.. It depends on how the LTC4444 deals with cross-conduction. The rest is decoupling and a couple of gate drive resistors, because I like them.

Genome.

Hello All,

Sorry for posting this old thread, but I had the same with this LM5085, would not start. A snubber solved it perfectly, I use 2n2 and 100 Ohms from the switching node to ground. Now I am stuck with EMI around 60Mhz...

Good luck,

Arian