LM723 die pictures

[iMo]

None cracks in the passivation layer this time?

[Noopy]

No, nothing to see... 

[Noopy]

Lambda Semiconductors called their LM723 variant LAS723.






The edge length of the die is 1,9mm. The design is not comparable with any of the designs I have documented so far. Nevertheless, the arrangement of the function blocks is similar. Current sources are located in the upper left corner. The output transistor is integrated at the upper edge. There, however, the Z-diode has been omitted, which is usually connected in series to the output transistor. In the lower right corner there is a voltage reference block. But this is much more complex than on most other LM723.




The detail shows that this is a bandgap reference, not the more common temperature compensated zener diode. In the lower right corner, you can clearly see the typical arrangement of a large and a small transistor. The large transistor contains four of the emitter areas of the small transistor. The emitters are connected to a whole series of resistors which determine the temperature drift of the reference voltage. To be able to adjust this temperature drift exactly, the design offers many possibilities to adjust the resistor values. Contacts between metal layer and silicon can be moved and shorting bridges can be set or removed. In the lower left corner of this picture you can see the current mirror that supplies the bandgap transistors.


https://www.richis-lab.de/LM723_16.htm

 

[magic]

The detail shows that this is a bandgap reference, not the more common temperature compensated zener diode. In the lower right corner, you can clearly see the typical arrangement of a large and a small transistor.

Actually, in the bottom right corner, I can see the VREF contact pad and a zener diode right above it

But the rest is indeed a Brokaw cell bandgap reference. The zener clamps its output so it can't go too high above ground.

What's weird is that output voltage of the Brokaw cell is one diode below VREF (if I counted all the emitter followers correctly) so VREF is clamped to zener voltage plus diode drop, which is about the normal output voltage of 723 regulator. I suppose the zener has slightly higher breakdown and remains inactive except under transient conditions when the bandgap cell is too slow to react or whatever, otherwise it would be seriously weird...

[Noopy]

Actually, in the bottom right corner, I can see the VREF contact pad and a zener diode right above it

Yes there is a zener. 
I assume this zener diode is actually used as a pn capacitor. That would also explain why it is so big. And the locaction would be good for some capacitance.

But the rest is indeed a Brokaw cell bandgap reference. The zener clamps its output so it can't go too high above ground.

I don´t see this... 

What's weird is that output voltage of the Brokaw cell is one diode below VREF (if I counted all the emitter followers correctly) so VREF is clamped to zener voltage plus diode drop, which is about the normal output voltage of 723 regulator. I suppose the zener has slightly higher breakdown and remains inactive except under transient conditions when the bandgap cell is too slow to react or whatever, otherwise it would be seriously weird...

In my view the VREF output is connected to a voltage divider and this divider defines the multiplicator of the bandgap voltage. Like we have often seen in voltage references..

[magic]

Okay, PaintCAD^® again

Black is ground.
Pink is the zener and a compensation capacitor
Red/magenta is the bandgap NPNs and emitter resistors.
Orange is PNP buffers.
Yellow is the active load.
Green is load for the orange buffers and a first level shift.
Cyan is more level shift and active load (outside the view).
Blue is the VREF output darlington.
Violet is current limiting.
Brown is the feedback divider.

VREF voltage is Brokaw output voltage (red collector) + orange Vbe + green Vbe + cyan resistor drop + cyan diode - 2x blue Vbe.
Approximately it's red collector + 1x Vbe + cyan resistor drop, so red collector is a bit less than one Vbe below VREF. And the zener clamps this voltage.


edit
I'm not saying that this is not a bandgap reference, pretty surely the BG circuit wouldn't be here if it weren't regulating the output voltage.
But the zener is reverse biased close to breakdown and appears to really be a zener and not just capacitance.
It may be handling edge cases like startup, load transients, VCC transients, although I can't imagine what exactly it does

[Noopy]

Well done!

I agree with you. 

And it is a LM723 with a bandgap reference. 

[iMo]

LAS723' Vref is 2.5V only..

[magic]

This explains a few things, such as why a bandgap reference was used and why the ratio of feedback resistors doesn't seem large. I noticed it but somehow haven't thought about it long enough to realize the obvious...

The zener is obviously not close to breakdown, then. Still not sure why it's there.

Could it be they also had a 7V version which only differed in metal layer? (Probably not, bias current of the zener would be low and poorly controlled, unless some of the bandgap components were reused to make a load for it).

[Noopy]

Yes, that explains a lot.

I assume you are just on the wrong track. It's just a pn capacitor, not a zener diode. 

[Wolfgang]

Well done!

I agree with you. 

And it is a LM723 with a bandgap reference. 

Maybe similar (but not the same) than the SG3532 regulator. This also has a 2.5V reference.

[Noopy]

Well done!

I agree with you. 

And it is a LM723 with a bandgap reference. 

Maybe similar (but not the same) than the SG3532 regulator. This also has a 2.5V reference.

We should take a look inside the SG3532. 

[p.larner]

some pics of the infinion SG3532 would be good to compair.

[Noopy]

some pics of the infinion SG3532 would be good to compair.

I will do so. The question is which manufacturer and where to get it. It seems like the first one is from Silicon General. I will get one built by Microsemi.




But first we have a LM723 built by Samsung! No index is shown on this part. However, the datasheet specifies two variants. Index C is specified for an operating voltage range of 0°C to 70°C. Index I is specified for an operating voltage range of -25°C to 85°C. Index I is specified for an operating voltage range of -25°C to 85°C.




The dimensions of the dies are 1,4mm x 1,3mm. KA723 could be an internal designation. The circuit shows the division of the circuit parts known from many other 723 variants. Nevertheless, the design seems to be their own.


https://www.richis-lab.de/LM723_17.htm

 

[magic]

KA was the usual prefix of old Samsung parts. KA358, KA317, KA555, KA5532, etc.
You can still find those datasheets.

BTW, Samsung sold their semiconductor division to Fairchild (the new Fairchild, spun off from National) so sometimes the prefix can also be found on Fairchild branded parts. New Fairchild started as National's discrete transistors division, and IIRC all their IC products came from acquisitions - mainly aforementioned Samsung and Raytheon, giving a mix of KA and RC parts, until Fairchild mostly renamed them more conventionally.

[Noopy]

KA was the usual prefix of old Samsung parts. KA358, KA317, KA555, KA5532, etc.

Now that you have written this it seems familiar. 

[iMo]

It is interesting the wiring complexity of the LAS part vs the others (usually clones). I wonder what are the addons or diffs (except the bandgap) in the LAS part.. There is "the thermal overload protection" mentioned in the DS, for example, but it should be more, imho.

[magic]

Maybe your turn to post some schematic
It's a simple chip, only a little tedious due to that maze of wires.

There is clearly the differential amp to the left of the reference with its tail current sink nearby, and the whole left edge is bias generation. Output stage at the top, and not much is left.

[Wolfgang]

Well done!

I agree with you. 

And it is a LM723 with a bandgap reference. 

Maybe similar (but not the same) than the SG3532 regulator. This also has a 2.5V reference.

We should take a look inside the SG3532. 

Hi,

do you have a SG3532 chip at hand ?

Wolfgang

[Noopy]

I wouldn't expect something special in the LAS723 either. The bandgap makes the circuit bigger.

I hope to get a SG3532 built by Microsemi but not sure yet.

[Dan123456]

Woah… this stuff is so cool!!!

Absolutely beautiful stuff all 

Apologies if this has been covered somewhere before but, how do you guys do this?

I had never heard of this process before so just googled “delidding ics” (as closest thing I knew of was removing the heat spreader from CPU’s) and found an awesome video from Applied Science where he was using fuming nitric acid.

He said he wasn’t sure if it had to be fuming nitric or if it could just be good old “regular” 68% concentrated nitric?

I would love to give this a go as your results look so cool but don’t stock any RFNA / WFNA and don’t really like making it (68% is more than good enough for my uses! )

[Noopy]

You find a lot of information about decapping in a lot of places.

We have discussed the decapping here:
https://www.eevblog.com/forum/projects/decapping-and-chip-documentation-howto/msg2663778/#msg2663778

My website has a own section:
https://www.richis-lab.de/Howto.htm
(Google translator is your friend)

I´m from europe. We don´t get nitric or sulfuric acid not even in low concentrations. Because of that I started burning the epoxy. With my furnace the sucess rate is quite high. 

Read what we have written and after that feel free to ask whatever you want. 

[p.larner]

i got the infinion chip from ebay,its a ceramic job so hard to decapitate i guess,needs a lot less volts on the current sense so has more transistors in that section,or a darlington i guess.

[Noopy]

i got the infinion chip from ebay,its a ceramic job so hard to decapitate i guess,needs a lot less volts on the current sense so has more transistors in that section,or a darlington i guess.

Is it really infineon? I would assume it´s linfinity?

I now will get one from Silicon General. Looking at the naming this could be the first one.

Ceramic is easy to decap. Just knock with a screwdriver against the material between the two ceramic parts and it will fall in pieces. ...ok a little experience helps... 

[magic]

He said he wasn’t sure if it had to be fuming nitric or if it could just be good old “regular” 68% concentrated nitric?

I would love to give this a go as your results look so cool but don’t stock any RFNA / WFNA and don’t really like making it (68% is more than good enough for my uses! )

If you don't feel like reading the whole thread linked by Noopy, and I'm not even 100% sure if I ever described it in detail:

Pour 1~2ml into a 5ml beaker, drop the chip, cover with watch glass or at least HDPE bottle cap (for vapor reflux, reduces fuming and loss of acid).
Boil gently for a few minutes. Nitric acid remains transparent, allowing you to watch progress, only turns green from dissolved copper.

68% will eat exposed aluminium bonding pads, but won't reach under the passivation glass in short time. FNA passivates aluminium, hence the preference.

edit: Hopefully you know it already, but NO₂ is poison and stinks ugly like chlorine, so don't be an idiot, do it outside