Unknown Circuits - die pictures

[Noopy]

Let´s start a new topic for unknown circuits.
UICs - Unknown Integrated Circuits? 




This component contains the letters WE, the character sequence F61489 and the numbers 2384. 2384 could be a date code, which would then refer to the year 1984. WE probably stands for Western Electric. This is also supported by the integrated circuit. F61489 could be the designation of the component. 






The die has several testpads. There are large contacts on the edges, which were probably mainly used to tune the circuit. There are also many small contact areas within the circuit that are only suitable for special tests. A clear division into two parts is recognisable in the structures. The right part is smaller but contains very similar areas. The image is available in full resolution: https://www.richis-lab.de/images/TBD/01x03XL.jpg (24MB)




The only labelling can be found on the left edge. BKA, JRB and TM could be the initials of the developers. M7062D appears to be an internal project designation.






Some test structures have been integrated in the bottom left-hand corner and at the bottom edge, which make it possible to check the alignment of the masks against each other.




There are resistance strips on each side, whose resistance values can be set with a few testpoints and fuses.




The upper section contains structures that are typical for switched capacitor filters. The functionality of such filters is described in more detail with the U1001 (https://www.richis-lab.de/phone01.htm). The repeating structures of several opamps can be clearly recognised. These opamps are connected with different capacitances, which are represented by individual rectangles.

The filters indicate that this is indeed a Western Electric component. Western Electric was very active in the telecommunications industry. It therefore seems very likely that, like the U1001, it is a transmit and receive filter for digital transmission links.




The control logic is integrated in the lower section of the module.


https://www.richis-lab.de/TBD01.htm

 

[RoGeorge]

First! 

[Noopy]

This component comes from an internal collection of Philips or NXP. There is no datasheet for the designation TDA8445. 8618 is probably a date code.




There are some remnants of bondwires in the package. It is probably not the first die to be inserted into this package.




On the left-hand side, three identical and one very similar circuit section are arranged one above the other. Four output stages can be seen at the bottom edge.

This picture is also available in a higher resolution: https://www.richis-lab.de/images/tbd/02x03XL.jpg (51MB)




At the top edge of the die there is the character string R0185C1. But that string cannot be assigned.




Here you can see one of the repeating blocks. It is immediately recognisable that this is a bipolar process. Two layers of metal were used. In the centre is a large block of small transistors connected in parallel. The strips in the right-hand area represent I2L logic, as described in more detail in the CA3161 (https://www.richis-lab.de/logic22.htm).




There is no information on the type designation TDA8445. However, Philips sold a TDA8446. It is a video signal switcher that can be used to process different video standards. It is not only the name that suggests a similarity. The three identical circuit blocks of the TDA8445 could be the RGB inputs, while the fourth, similar block could be the Y input. At the bottom edge there are three identical and one slightly different output stage. This would match the outputs of the TDA8446. However, the TDA8445 is missing two pins compared to the TDA8446.


https://www.richis-lab.de/TBD02.htm

 

[amyk]

Some Chinese websites I could find claim that TDA8445 is a TV "field output" (vertical) IC.

[Noopy]

Some Chinese websites I could find claim that TDA8445 is a TV "field output" (vertical) IC.

I have found this site but I assume that's a mistake. These "field output integrated circuits" are power devices.

[RoGeorge]

There are some remnants of bondwires in the package. It is probably not the first die to be inserted into this package.

Could that be just a malfunction in the bonding machine?  Asking because I assume it won't be easy to remove a die without damaging the packaging, or is it?  Then, how would the bonding machine know to avoid the bond leftovers on the pads?  Or maybe it was made with a manually bonding machine?

I know nothing about chips manufacturing, is package-reusing a common practice for IC prototypes?

[Noopy]

I think we can be sure that this is an prototype. With that in mind I assume it was bonded by hand.
I assumed there was a die in the package before this one because a bad bond process probably would have left some residues on the die but we don't see anything on the bondpads.
You can remove a die from such a package. Of course that's just something for development purposes but as I said I assume that is a prototype device.

 

[magic]

The outputs are totem poles driven by the digital circuitry, so this is not an amplifier. Surprisingly, it looks like the logic drives high and low side separately, so shoot-through prevention needs to be included in the logic.

Pinout so far:
17, 18, 1, 2: analog(?) inputs, the circuitry here should be easy to figure out
3, 4: outputs
5: GND
6, 7: outputs
8: VCC
9: seems to act as output enable, maybe a logic supply too
...
15: looks like a logic supply

[magic]

Actually, this looks like some DAC or DDS kind of thing. The "analog inputs" on pins 17, 18, 1, 2 really are (low power) outputs, each driven by an ordinary class AB NPN/PNP emitter follower pair and level-shifted 0.7V down by a diode and a cascoded constant current sink. These outputs are controlled by simple RC4558-style opamps with capacitive feedforward around the PNP differential stage and a cascoded second stage. The circuitry seems oriented towards speed rather than precision.

Noninverting inputs are biased by the potential on pin 13, while pin 14 controls an open collector NPN which temporarily pulls pin 13 to ground. Pin 15 supplies the opamps, not logic and pin 16 might be some sort of output stage shutdown. Inverting inputs are connected to on-die feedback resistors and variable current sources controlled by the logic - that's what the arrays of transistors hidden under upper layer metal appear to be. There are also provisions for the logic to disable any opamp and drive its output into either positive or negative saturation.

Pins 12, 11, 10 go into the logic or some analog-looking circuitry in the middle of the logic, under pad 13. Pin 9 does appear to be logic power supply, pin 8 is power supply for the digital outputs on pins 7, 6, 4, 3 and nothing else, apparently. Pin 5 is ground.


Do we have any masochist I2L expert here able to explain what this control circuit is doing?

[Noopy]

Thank you magic! Very interesting!

I found no Philips part with a functionality similar to what we see here. But of course I didn´t check every Philips part.  Perhaps it was just a prototype which never went into series production.

Do we have any masochist I2L expert here able to explain what this control circuit is doing?

No, that´s not my business. 

[magic]

It really is a triple channel DAC.

Next to each transimpedance amp there is an array of 31 NPNs: emitters grounded, bases driven in parallel, collectors connected in groups of 16, 8, 4, 2, 1. For each group there are three NPNs controlled by the logic which route the group's current either to the TIA directly, or to the TIA through an inverting Wilson mirror (3 PNPs between the TIA and the array), or to a dummy emitter follower where nothing happens. Pretty sure we have seen this kind of DAC architecture before.

By the way, I2L logic outputs are open collector so they shouldn't be able to drive the NPN current steering switches. It seems that in this chip the logic is "grounded" through a diode-connected Darlington pair, shifting its potential two diode drops above real ground. What appears to be inverter gates driving the switches, actually are ordinary emitter followers working in forward mode and feeding current from logic virtual ground into the bases of the switches.

The apparent fourth channel is another story. The TIA is a dummy - it simply outputs the same quiescent DC voltage as other amps, there is no signal current going into it. It looks like pin 17 is a reference voltage for the other outputs. There is no Wilson mirror on this channel, in its place there is a fifth opamp which drives the bases of all current sink arrays in such way to produce a particular output current from the fourth DAC. The fourth DAC is also modified, it is single-ended only and has groups of 4, 10, 4, 2, 1 By controlling this DAC, the logic forces the opamp to set different levels of current per one cell of the array. Hence gain of the main DACs can be varied.


I found only three inputs to the chip. They go to the analog section near pad 13, which is simply a bunch of comparators. Comparator outputs go to the logic, so these are digital inputs. I doubt that they are fast enough to feed complete raw data to all DACs in real time. I think it's some special purpose chip, which receives simple commands and generates more complex output. I suppose it could be all sorts of weird things, from an OSD generator to a fully integrated simple video game.

I didn't find any obvious oscillator or other clock source

[Noopy]

Next to each transimpedance amp there is an array of 31 NPNs: emitters grounded, bases driven in parallel, collectors connected in groups of 16, 8, 4, 2, 1. For each group there are three NPNs controlled by the logic which route the group's current either to the TIA directly, or to the TIA through an inverting Wilson mirror (3 PNPs between the TIA and the array), or to a dummy emitter follower where nothing happens. Pretty sure we have seen this kind of DAC architecture before.

So these are 5Bit-DACs. Now it´s clear what the transistor arays are used for. Yes, this architeture sounds quite common.

By the way, I2L logic outputs are open collector so they shouldn't be able to drive the NPN current steering switches. It seems that in this chip the logic is "grounded" through a diode-connected Darlington pair, shifting its potential two diode drops above real ground. What appears to be inverter gates driving the switches, actually are ordinary emitter followers working in forward mode and feeding current from logic virtual ground into the bases of the switches.

You are right. That is really strange. 


 

[magic]

It's a 6-bit DAC because there is a sign bit too

The sign bit flips all current sinks of the given DAC channel between the noninverting or the inverting path. This is unexpected, because it means that for negative outputs the signal passes through a PNP Wilson current mirror. I don't know how fast it is, but lateral PNPs are somewhat infamous for being slow so maybe it's not a very fast DAC and maybe not for video. But for what? I don't know...

I looked at the logic of the DACs and it's not too complex. The image below shows one bit. Starting from the left, we have three ordinary noninverting emitter followers which drive the three current steering switches (inverting, noninverting, null). Next is a pair of common emitter transistors with direct mutual feedback - a flip-flop. Right side disables the null output, left side disables the inverting and noninverting outputs. The FF is controlled by a two-transistor, single-ended to differential buffer: the right transistor drives one side of the FF, the left transistor is a slave inverter which drives the other side. POR is power-on-reset, which asynchronously forces all bits of all DACs to null when the digital power supply is too low. LATCH disables the buffer and causes the FF to hold its current state - this allows all DACs to share one data bus because their latch lines are separate. The final transistor is data input from the bus to the buffer.



We have five such bit drivers per DAC. The sixth, almost identical circuit reads the sign bit from the bus and disables all inverting or noninverting switches accordingly. The remaining two bits control the TIA (enabled, saturated positive, saturated negative). They have their own, independent latch signal which is shared by all DACs and each DAC takes these two bits from a different positions on the bus, so the TIAs are controlled independently with one bus word.

The fourth DAC (gain control) is connected to the same bus and has its own latch line, which is shared with a small circuit that drives the digital pins. Upper four bits go to DAC4, lower four bits go to the pins.

The bus is driven by eight similar circuit blocks arranged in a column right next to the DAC logic. There are links between consecutive blocks and I think it may be an 8-bit shift register. We have so far covered about 75% of all the logic, so the reminder can't be too complex. I'm starting to think that the chip is nothing more than a DAC with a simple serial input interface like SPI or similar...

[Noopy]

Good job! Thank you!

Philips had the TDA8444 with six DACs controlled by I2C. In the datasheet they are talking about a I2L-compatible process.
"Each DAC can be programmed separately by a 6-bit word to 64 values, but VMAX determines the maximum output voltage for all DACs."
Sounds like the TDA8444 is quite similar.

[magic]

Interesting. Part numbers are not too far apart and I just noticed that TDA8445 has an open collector transistor under logic control, connected to one of the I/O pins. I²C is old enough, so maybe that's what it is. I will see if I ever get bored enough again to decode this interface logic

The DACs are a little different, though. Bipolar not unipolar, output amplitude is variable digitally in a few preset steps, and an analog input controls the DC bias instead. The TIA outputs don't seem capable of reaching close to ground.


Hmm, do you have some I²C controller which could scan this chip for any addresses it might acknowledge?

[Noopy]

The TDA8444 and TDA8445 are related but not the same. 

I'm trying to avoid software wherever it is possible, a I²C controller is already software.