Hi there,

We are busy with a cost saving exercise on one of our design. Part of the change is moving from a dual AD640 log amp set to a single AD8310.

The AD8310 gives a RSSI slope variations of 23.8 mV/dB at 10 MHz to ~ 22.8 mV/dB at 400 MHz (Figure 16). Our application is at 60±5 MHz, so this does not have much of an effect. However, the datasheet does specify the total variance of the slope of 20-26 mV/dB over temperature and frequency. How much would this change over temperature for this the smaller frequency range and how much of the change is over different the batches of components and not over temperature?

We are increasing the video slope in the design (to 37 mV/dB) and I would like to know how much temperature compensation is required, if any. We did not use extra temperature compensation on the AD640 design, but that may have been an oversight. The secondary question would be, do I need a trim pot in the slope increase circuit to allow for batch variation or how much would it change if I left it at a fixed scaling.

* If extra temperature compensation is required, what is the change in slope (negative / positive, etc).

Thanks for your support.

Regards,

Johann

Hello Johann,

I'll start off by saying I don't think we have distribution data over temperature and frequency for the AD8310. Having said, I think we can piece together a pretty good idea of how much change you can expect over each variable and estimate a total change.

I think the least variation you can expect is over frequency. Take a look at Figure16 in the datasheet. At 60 MHz over 10 MHz bandwidth there is hardly any slope change - there is fractions of a mV change from 55 to 65 MHz. So let's just go ahead and rule that out as a dominant contributor for the time being.

The next biggest contributor is the temperature. Take a look at Figures 3 and 6. The biggest slope change is at cold. Vout varies by about 1 dB at the low input powers which is 24 mV. This causes about 0.4 mV/dB decrease in slope at -40 degrees C. Visually, you can see the error at 85 degrees C is about have that of -40 degrees C. So this should translate to a 0.2 mV/dB increase in slope at 85 degrees C. So if you had a nominal slope of 24mV/dB, it will vary +0.2/-0.4 mV/dB.

The biggest contributor is part-to-part variation of the slope which can be seen in Figure 17. There’s a nominal slope of 23.66 mV/dB is varies from 22 mV/dB to 25 mV/dB. Figure 17’s population size is a bit limited (roughly 90 DUTs tested). With a bigger population size this distribution may widen a bit.

So you can see with a bigger population size, a few non-typical DUTs whose temperature variation is wider than +0.2/-0.4 mV/dB (this is not unexpected) and the variation over frequency, you can see why the slope is specified from 20 mv/dB to 26 mV/dB as the min and max. At 60 MHz, I would expect the min and max to be close to 22 mV/dB and 25 mV/dB, respectively. I would say at worst, you'll see 0.5 mV/dB more at either the extremes, due to temperature variation.

Hope this helps,

Joel