What's inside a low power isolated DC-DC converter brick ?

[ealex]

Hello

I'm wandering what's inside something like this: http://www.tme.eu/ro/Document/328aba69d00d623be59164448a0186de/AM1S-Z.pdf - it's unregulated and will start outputting something form ~ 1-2V input voltage

I could not find anything useful on the net - or did not use the correct search word.

The reason i want to find out - I want to generate some LCD bias voltages for some old style LCD's that have only row and column drivers on the board. Some of them need high negative bias voltages 25V and over, so the classic MC34063 negative voltage generator does not work. The bias current is small 10-20mA.

Do they have some a simple Royer converter / push-pull oscillator and the output voltage is just depending on the turns ratio ?
Like ( from mikrocontroller.net )


Or is there something more like : ( from http://danyk.cz/iz_odv_en.html)


Or dedicated IC / bare die stuck to a small transformer like Figure 3 from http://www.ti.com/lit/an/slyt211/slyt211.pdf


Mike had some destructive tear-down of some DC-DC modules, but i don't remember seeing anything that would point to a schematic

Thanks,
Alex

[Jeroen3]

Most likely some custom silicon with the first schematic.
You can build one yourself easily: www.ti.com/product/SN6501

[T3sl4co1l]

There's kind of three kinds:
- Free running oscillator: no current limiting, no regulation, dirty power in/out, requires minimum load.  Typically <3W.
- Flyback, primary regulated (primary pulse height, or auxiliary winding sense): current limiting, poor regulation, dirty power in/out, requires minimum load.
- Flyback, resonant, or other: optoisolator regulated, current limiting, good regulation, not as dirty power in/out, no minimum load.

The latter is most like a general purpose power supply, just with other-than-mains input voltage.  They're also the most expensive, but often the most worthwhile (e.g., RECOM and others' low capacitance, reinforced insulation modules, good for medical isolation, high voltage gate drivers, etc.).

Beware that almost all DC-DC (and AC-DC) converter modules are dirty as sin.  If they carry EMC ratings (FCC or IEC specs), they're almost always meaningless.  I've only seen one module that passed, once cables were attached: a medical grade module (15W, low isolation capacitance, better filtering than usual).  The small ones (like the cubic 5-10W ones from XP Power, CUI and etc.) actually have no filtering onboard: you're expected to provide it.  Line-side common mode filtering is typical, but you may find a secondary side CMC is needed as well, plus a suitably ingenious combination of Y caps between line, ground and output.

There are also HVDC to LVDC modules -- these work with an external rectifier or PFC stage, which can be common to many such modules.  This allows you to share the AC input circuit, while generating a custom combination of output voltages.  They're usually expensive, and needless to say, you provide your own filter externally as well.

Tim

[T3sl4co1l]

Are Y caps requiried on the output for earth ground conection?

No.

You will probably find them necessary, because almost all converters generate excessive common mode voltage (that is, between the AC mains / DC / primary side and the DC output side).  A very quiet, low-isolation-capacitance design may need very little, if any (which is characteristic of most medical grade AC-DC and DC-DC modules).

The output can also be earthed directly (galvanically as well as at RF), in which case you have grounded SELV, and instead of reinforced insulation, you need only basic insulation (because, mains fault current is shunted to ground, away from possible user contact).

Oh and, by "Y cap" I mean a capacitor which carries agency approval for SELV-to-GND, mains-to-GND and mains-to-SELV connections.  I also somewhat mean the capacitances used in those locations, for electrical filtering purposes.

Tim

[T3sl4co1l]

I wouldn't think an LED driver would need anything at all, as it doesn't need to be touched.  It can be double insulated.

Saying it can use ordinary (X type, or no agency approval at all) types is basically saying it can be at line potential.  You could still build such a circuit with functional (rather than basic or reinforced) isolation, but why?

Tim

[bktemp]

The reason i want to find out - I want to generate some LCD bias voltages for some old style LCD's that have only row and column drivers on the board. Some of them need high negative bias voltages 25V and over, so the classic MC34063 negative voltage generator does not work.

34063 can operate at up to 40V. Using 5V supply voltage it is possible to generate up to -35V.
Because of the voltage drop of the buildin darlington transistor, using an external PNP transistor improves efficency. Some 34063 datasheets have an "Voltage Inverting With ExternalPNP Saturated Switch" example circuit.

[ealex]

thanks for your replies.

my main application is under 3W ( 30V at 10mA = 300mW )
i'll be using a fpga to drive the display -  i could replicate the basic SN6501 push-pull switching, including dead-time then have some basic on/off regulation based on these app-notes:
"Common Analog Functions Using an iCE40 FPGA" http://www.latticesemi.com/view_document?document_id=45822
"Creating An ADC Using FPGA Resources" http://www.latticesemi.com/view_document?document_id=36525

i could wire a lvds input as a basic comparator -> on/off control on the switching outputs.
i don't think the LCD will mind some ripple. that way i can set the voltage with an external voltage divider ... or generate a reference with another PWM signal from the fpga ...

i'm already using a MC34063 and i need to add some sort of enable / disable - the LCD will go crazy if the proper power up / down sequence is not used.

i'm trying to see if i can use some of the already available IO's instead of an external chip.

[Buriedcode]

If your requirement is +/- 35 @ 10mA, why not a voltage multiplier/charge pump tapping from the switching node of a boost converter?  As small SOT23-5 boost converters are ubiquitous these days, and cheap, you could boost to 35V, then use a couple of diodes and caps to get ~-33V at a reasonable current.  I believe you can also use the output of that as feedback so its pseudo-regulated.  I do have a 3 transistor negative voltage generator I made many moons ago that can output ~20mA @ -40V.  Wasn't particularly efficient but used cheap parts, more specifically a cheap 100uH inductor.

For some panels I've used the max749, which has the advantage of not only being efficient, but its output can be set by a voltage divider, as well as a digital up/down pot.  I used the micro interface to set the digital pot before enabling.

 A royer push-pull is a nice simple circuit but often requires custom magnetics so makes it a bit of an arse, regulation is often done with a buck converter on the input (current-fed) controlled by the output voltage, often used for very low ripple 'clean' supplies, but IMO overkill for an LCD panel.  There are small transformers available for such purposes, but I haven't found a reliable source for low quantities.

[bktemp]

I have used this circuit for generating a negative voltage for a small LCD. The enable signal is level shifted to logic levels using a transistor.
Instead of connecting R13 to GND, a filtered PWM signal can be used to adjust the output voltage by 0-3.3V.

What type of LCD do you want do use? 25V sounds like a 320x240 LCD. The voltage should be regulated and maybe needs a temperature compensation.

[ealex]

hello

the display is a EW50107FLYU - 320x240 px, monochrome, transflective, led back-light.
it has a bunch of LA6324 and some passives besides the driver IC's.
i could not find any documentation about required current on VEE, just some pdf describing basic timing and how to connect it to a LPT port - enough to get it working

do you know what the LCD needs for the VEE voltage ? ripple / current / etc ?

i also have a 480x240 blue-white display with the same configuration and voltage requirements ( but with CCFL bl )

MAX749 is too expensive for just a personal project( i can get it from farnell for 2x the price of the LCD ... ).

For a custom transformer - i have a E13-... core ( need to measure ), including coil former that i could use.

one of the main problems i have with the setup i am using now: when stopping the FPGA VEE will remain on -> the LCD tends to do strange things when this happens.

i'll try making a simple push-pull fixed duty cycle driver from the FPGA, to see how/if it works.
i will add a capacitor in series with the drive signals, so DC levels will be stopped, see the attached image

[Buriedcode]

Most 320x240 LCD's require a drive voltage of about -25V, with a voltage divider between VEE and GND producing Vo of ~-20-22V.  They almost always specify Vo has VDD-V0, and I'm not sure why they recommend having a voltage divider between VEE and VDD, as it just wastes and extra bit of current, and it never needs to go above 0V.

The reason I suggested a boost converter, or a buck with a coupled inductor is that you can use fairly cheap power controllers with built in switches that will do the regulation for you, and will work with reasonably low inductance.  A constant duty push pull driver is all well and good but you'll need to know the inductance.  Also for <30mA charge pumps can easily be added to boost converters, 5 ->15V ->-15V -> -30V.  Just requires 4 diodes and 4 caps.

LCD's need power sequencing, I believe they can handle VDD before VEE, but don't apply the negative drive voltage before logic power and signals, and make sure you disable VEE before you stop the driving signals, then remove VDD.  This is because without the driving signals which are all periodic, the full VEE may be applied to areas of the glass for extended periods as opposed to normal operation where they are applied briefly (a swift kick of voltage to quickly turn on/off pixels, multiplexed), this can damage the display.  Look at any LCD datasheet and you'll find a power sequencing diagram. 

Here's something I found in my LTspice folder... from 2009... I vaguely remember it was a 3 transistor boost converter for low current that I modified for negative voltage, uses two zeners for regulation, eff. ~ 65%, but good up to 20mA.  Hardly pretty but its sitting here happily driving a monochrome blue 320x240.

[bktemp]

LCD's need power sequencing, I believe they can handle VDD before VEE, but don't apply the negative drive voltage before logic power and signals, and make sure you disable VEE before you stop the driving signals, then remove VDD.  This is because without the driving signals which are all periodic, the full VEE may be applied to areas of the glass for extended periods as opposed to normal operation where they are applied briefly (a swift kick of voltage to quickly turn on/off pixels, multiplexed), this can damage the display.  Look at any LCD datasheet and you'll find a power sequencing diagram. 

It is not only the LCD itself that das not like DC levels, it is also the driver ics. They often latch up if Vee is pesent without Vdd. Worst case is if Vee is turned on before Vdd, because the ic will be in latch up and if Vdd can supply enough current, all the magic smoke will escape.

The current consumption on Vee is typically in the lower mA range, but for some older driver ics it may be up to 20mA depending on the technology and display size.
The ripple voltage should be as low as possible, but for most LCDs up to a couple of 100mVpp will not degrade the optical performance.
Except from the power sequencing (turn Vee on after Vdd and shut down before at the same time with Vdd), it should be well regulated, because the voltage will affect the contrast. Therefore you need to adjust the voltage for best contrast and if the display will be used at a wide temperature range, you also need to adjust the voltage over temperature (around -0.2%/°C).

[ealex]

I've done both mistakes ... driver latch-up when VEE got left on and VCC was disconnected.
also, VEE on then the fgpa froze / was flashed -> VEE applied for too long on one of the lines - the display has a "weak" line. for now i'm treating this display as a test unit - not letting  the magic smoke out is the only constraint.

that damage gave me the idea to make VEE "safe" ( railroad signaling done in AC -> a stuck low / stuck high will not present a good signal ).

i could use something simpler, for a test - a TL431 configured as a shunt regulator -> it will ensure that -25V is not exceeded.

i'll make a short test with a simple push-pull when i get some time, i will add here my results.

[bktemp]

I don't think a shunt regulator is a good idea:
It will waste a lot of power, because the power supply needs to be designed for the maximum current consumption and the shunt regulator must dissipate all the excessive current whenever the load draws less current.
Most LCD circuits use either an inverting boost converter or a chargepump followed by a linear regulator for generating a regulated LCD voltage.
If you want to use an unregulated push-pull/forward converter I would use a LM337 or a simple transistor+zener diode regulator instead of the shunt regulator.

Many LCDs have an enable pin (often called DISPOFF\). Using this pin the Vee power supply does not need to have an enable pin. Instead it can be connected to Vdd. This ensures Vee is only present when Vdd is turned on.

[ealex]

found some info about the display:
https://www.scribd.com/document/275928564/320x240-LCD-Display

The driver chips are:
4x LC79401 http://www.pacificdisplay.com/ics_app%20notes/sanyo/LC79401.pdf
3x LC79430 http://www.onsemi.com/pub_link/Collateral/ENA2123-D.PDF

The #DISP_OFF signal is tied to VCC -> i can't use it.
I will go the MC36063 way, as i already have the parts available.

I'll check how i can add an "enable" signal to it