A couple of weeks back I posted a teardown of the Samsung SPH-A660 phone and featured the Qualcomm RF receive chips. This is the brother chip the RF transmitter RFT6100
From Qualcomm literature “the RFT6100 device is a direct conversion chipset that integrates all the upconversion and modulation functionality necessary for CDMA-mode phones operating in cellular and PCS bands. This transmit chip consists of two sets of I/Q modulators, one for cellular and one for the PCS band. The baseband I/Q input from the MSM directly modulates the cellular or PCS carrier derived from the respective LO generation circuit.
Both the cellular and PCS output drive their own VGA with a gain control range of 85 dB. A final cellular driver amplifier provides a modulated RF output.
To accommodate split-band and filtering, the PCS VGA drives two output amplifiers that can be selected independently or simultaneously via an input selectable switch. All RF outputs have fully integrated 50-ohm matching networks. Integrated on the RFT6100 is the receiver phase locked loop (PLL), the transmit PLL and the entire transmit VCO including resonant components. The VCO drives the LO generation block, which in turn generates the required local-oscillator signal for all CDMA bands of operation.”
So a lot going on there – I think the two I/Q modulators are the rectangular blocks at the bottom of the die (either side of the 4 square inductors) but can’t positively identify the other blocks referred to in that description.
The top metal looks a bit funky because I over cooked the depot. You can see where the acid has etched and removed all the metal from the bond pads (Normal as the acid that dissolves the plastic mold also likes to etch copper and aluminum) but has wicked into the tracks a fair bit removing metal tracks away from the bond pads.
However all over the die the top metal has been etched, even regions well away from the bond pads which indicates the depot acid has penetrated the passivation film (Normally a silicon oxynitride, or silicon nitride/silicon oxide layer.) This highlights a weakness (pinholes) in the passivation probably on the top metal edges as it coats the thick top metal.
Here are the die marks – HG11-V2689-1H from 2003
And as with the other Qualcomm chips they like to give credit to a whole cast of characters involved in the design. In this case I wonder what the separator is for between KG,KB,JD,AS,BW, JZ and AT,CG, JF,DK,TS,KZ,EK – enquiring minds want to know, so if any of you should read this, please let me know in the comments
Visually what dominates this die photo is of course the on-chip inductors. No less than 12 are used here. On chip inductors are typically used very sparingly, as they take up a lot of room and because silicon is semi-conducting and the dielectric and metal films are quite thin it is difficult to produce a good inductor with high Q-factor. In RF circuits where you need good 50 ohm matching networks they are indispensable hence there use here.
The dimensions of inductors need to vary to change inductance value, and frequency response, but what I don’t get is why the design needs to be so different on the same chip. On this one chip there are 4 very different inductor designs.
1. An octagon with 3 turns all in top metal
2. A square spiral with 8 turns in top metal
3. Similar square spiral with 6 turns in top metal. This one has a polysilicon ground shield – this is to minimize capacitive coupling to the substrate, the pattern in the polysilicon is there to prevent eddy currents from forming.
4. The most complex, 4 turns in two levels of metal and a polysilicon ground shield.
Images taken with dark field illumination using 5x and 10x objectives.