Start-up Time in the Real World!
Posted: 29th January 2018
On occasions our customers ask us for the start-up time on an oscillator. No doubt they have legitimate reasons for wanting to know this and the question seems simple at first glance. They probably need to calculate the delay between system boot-up and valid data transmission.
There is no problem in IQD providing this data but we do need to clarify precisely what the customer is asking for. In particular, if they are comparing our products to our competitors’ products, you can appreciate that we need to compare like with like. For example, do they mean start-up from power on, or start-up from output control function?
If the power to the oscillator is applied for the first time, then the oscillation circuit must start-up ‘from cold’ and this can take a few milliseconds. If they are using the oscillator’s output control function, then there are two possibilities for what is happening when the oscillator is disabled. One option is that the internal oscillator circuit is left running and the output buffer is disabled; this gives very quick start-up time in the microsecond range but the current draw when disabled is not much lower than when enabled. The other option is that the internal oscillator is stopped when the device is disabled; this gives a very low current draw when disabled in the micro or nano amp range, but does mean that the start-up time is longer.
As a rule of thumb, older oscillators have an output buffer disable function whilst newer and smaller oscillators have an internal oscillator shutdown function. This is because newer devices in the electronics industry are more power conscious, in particular small sized hand held battery powered devices.
It is probably worth a moment here to talk about terms used in the industry. Looking across all frequency control product manufactures and vendors datasheets you will find the output control function described as ‘Enable Disable Function’, ‘Tristate Function’, ‘Standby Function’ and a few other terms with no consistency as to which phrase means what. At IQD we use the term ‘Tristate’ to mean that the output buffer is disabled but the internal oscillator is still running and ‘Standby Function’ to mean that the internal oscillator has been disabled.However, just because two frequency control product companies use the same terms, it does not follow that they both mean the same thing!
When do you consider the oscillator to have completed start-up? When the frequency is within the specification on the data sheet, perhaps ±50ppm of the nominal frequency? Or when the amplitude of the output signal has reached a limit, perhaps 90% of the final amplitude? Or the time for the frequency to be within some ppm from its final frequency, perhaps ±20ppb of the frequency after 1hour?
Generally speaking, if the datasheet does not clearly define the measurement, then it is reasonable to assume that the figure given relates to the first of the options above i.e. that the frequency is within specification. However, if this figure is important to you then it is wise to be cautious and clarify this with your frequency control product supplier. If you are purchasing an Oven Controlled Crystal Oscillator (OCXO), then it is likely that the figure given is either the second or third option above.
Finally some considerations for your technical team. Start-up time can be affected by temperature, obviously this is particularly true of OCXO products. Measurement of start-up time can be affected by reflections and slugging on the signal between the oscillator and the measurement system. When we make this measurement we are using ideal conditions i.e. presenting the rated oscillator output load condition to the oscillator. If this is not the case on your PCB then the results you get will be different.
Accurate measurement of start-up time on a standard oscillator will require a high bandwidth oscilloscope to ensure that the rising edges are correctly timed and again, a well matched impedance between the oscillator output and the oscilloscope input is necessary to avoid slugging the waveform.
Start-up time is a function of frequency but data sheets show a single figure to cover a full range of frequencies. This means that the figure on the data sheet is the worst case figure and it will have some considerable safety margin within it. Measured data may be significantly better than the data sheet.
We do not recommend using counters to make this type of measurement. Modern gap free measurement counters are very good and do have a function to measure the time delay between triggers on two inputs. However using a counter you will not be able to see whether your trigger event is occurring at a relevant amplitude or whether there are some runt pulses occurring before stable oscillation is achieved.
If all else fails, IQD have a huge stock of frequency products at our head office along with our well-equipped test laboratory. We are very happy for you to define the start-up time measurement as you would like it done and we will make the appropriate tests and send you actual results. This way you can be 100% sure that you are dealing with accurate data.