Summary: HMC829LP6GE enters a 'frozen' state when restarted at high temperatures (but well below the 85C max operating temperature), the temperature depending on the level of the input 30MHz reference .
I am currently using Hittite’s HMC829LP6GE to generate a 3.6GHz LO using a 30MHz sinusoidal input reference in one of our products. The datasheet specifies an input reference level of 1V-3.3V peak to peak. The HMC829 works well at room temperatures. However, when the temperature is increased and the power of the HMC829 cycled, the HMC829 will lose lock and enter some kind of ‘frozen’ state; an output will appear at ~2.7GHz and the device is unable to be reset in firmware, even when it is cooled down to room temperature. The reference level will also jump in level down and then back up again by around 5dB (we know it is the HMC829 causing this because we replaced it with a resistor of similar input impedance and there was no level jumping). If the device is switched on at room temperature and then heated to these high temperatures, it will work fine; its only on restart that it enters the frozen mode.
Originally we only had about 1V pk to pk for the reference, which worked fine at room temperature, but this restart problem started to occur at around 40degC. When we increase the reference to ~2.5V pk to pk, it will restart perfectly at 50degC, sometimes at 55degC and nearly never at 60degC. It will always work at these temperatures if switched on at a lower temperature. The more we increase the reference the higher the temperature can go, but we are unable to increase the reference any further in our circuit. This problem is consistent across a number of boards of the same product.
I had a theory that maybe the chip was getting too hot, but it raises temperature by only 6degC when operated at room temperature so it is probably not reaching 85C when ambient is 60C. We have laid out the HMC829 with the bottom pad soldered to a ground pad with many vias leading to the bottom and internal layer ground planes of the PCB, so heat transfer shouldn’t be that much of an issue; it should be below the 85degC operating maximum.
We’ve had another problem with this chip and a few others, such as the HMC704. Sometimes, the initial program of the device is not consistent. That is, when we restart the device and program it, it may move to a random frequency output or drift in frequency and becomes unprogrammable. We have fixed this in one product by putting a large resistor in series with the SPI clock line and putting a capacitor to ground from the clock line in another product (the same fix didn’t work in both products). We think this may be to do with the HMC and Open mode configurations, and the device going into the wrong mode, but looking on the oscilloscope it seems we are correctly setting the mode we want by only raising the CLK line before starting any programming. The signal into the devices also looks very clean when measured on an oscilloscope without the modifications. Anyway, this is more of a side problem that we have fixed, the temperature problem is of greater concern. However, it wouldn’t surprise me if they are related because they are both to do with the start-up of the device.
Are there any records or suspicions that the HMC829 does not work as well at high temperatures? Is there something in firmware that may change how the HMC829 behaves? Any help you can give us on these problems would be greatly appreciated.