Demanding Frequency Generation Problem

[TurboTom]

For an accurate experimental evaluation of blade vibrational modes of one of our micro turbine engine designs, I'ld like to observe the turbine wheel with a camera and illuminate it with a high speed strobe on the running engine. If engine RPM is stable enough, it may be sufficient to trigger the strobe from a free-running frequency generator. But I've got the idea to generate a trigger signal that's locked to the engine's rpm and then offset (by frequency) by a small amount so we could actually "observe" the wheel rotating with an off-the-shelf camera (or just the help of a mirror) even though the engine is spinning at more than 60,000 rpm. If you want to put it that way, it's the same principle a sampling oscilloscope employs.

If there's a way to sync an AWG to an external frequency, it would appear simple. But I don't know of any manufacturer of AWGs with integrated frequency counters that offer the option to use the measured frequency value as an input to one of the AWG channels. Then, the second channel could be configured to generate an offset frequency and the job would be done. So to speak, a "poor man's FLL".

External DC FM (VCO) and using an external phase comparator probably wouldn't help much since most of the AWGs digitize the external modulation input only with eight bits (well, over a small span it may do) and I'm not really sure if the "frequency offset" function for the second channel would work with the modulated frequency of the first.

Any suggestions how this could be accomplished with off-the-shelf gear?

Thanks,
Thomas

[mawyatt]

A good AWG should have good frequency and phase stability, so you could just trigger the strobe from the AWG set to slightly above Nyquist to cause the apparent rotation.

BTW looks like the turbo fan used in my Porsche 911 Black Twin Turbo?

Best,

[mawyatt]

60,000RPM is 1000PPS and this might be too quick for a conventional strobe, you may need to use a LED type strobe that can keep up. As an alternative you could use a technique where the strobe is triggered after multiple Nyquist periods and get a result of a rotating image. Using an AWG will be handy as you can adjust the frequency to get the effect you are looking for.

Best,

[AndrewBCN]

For an accurate experimental evaluation of blade vibrational modes of one of our micro turbine engine designs, I'ld like to observe the turbine wheel with a camera and illuminate it with a high speed strobe on the running engine.
...
If there's a way to sync an AWG to an external frequency, it would appear simple. But I don't know of any manufacturer of AWGs with integrated frequency counters
...

You just need a frequency meter + a programmable pulse generator, not an AWG with integrated frequency counter. Your strobe light just needs a pulse to trigger, not a specific waveform.

Apart from the problem of a strobe light capable of working at 1kHz (here I assume you'll just use some power white LEDs), the frequency meter + programmable pulse generator part can easily be handled by an STM32 MCU development board and an AD9833 module, total cost < $15 and I would throw in a small OLED display SSD1306 so you can get a frequency display with the turbine RPM and an idea of what the MCU is doing.

Oh, and as an input to the frequency meter I wouldn't bother with a hall sensor, just use an inexpensive microphone to measure the turbine RPM.

[TimFox]

For comparison (on flash rates), the General Radio 1531-A Stroboscope has a maximum flash rate of 25,000/minute = 417 Hz.

[TurboTom]

True, the general appearance of these micro turbine rotors isn't that different from a turbocharger rotor, only much bigger than the variety that you usually find on car or truck engines. The turbine wheel has a diameter of 165mm. These small turboshaft engines provide round about 200hp from a package no heavier than maybe 45kg (100lbs). See here if you like.

Conventional xenon flash lamps aren't suitable for this application but there are short arc lamps available that permit quite fast repetition rates so I could probably fire every third revolution of the rotor. LEDs are problematic due to the very short flash duration required so the picture doesn't get blurred. I'm not sure how much I can overload a typical white LED if I operate it pulsed (t < 1µs) and I'm also not certain of the decay time of the fluorescent used in those white LEDs. Short arc xenon flash lamps achieve microsecond pulses without difficulty, depending on the configuration of the pulse shaping network.

A free running trigger oscillator is somewhat a problem: The generator is stable, that's not the issue, but the engine may vary in rpm, depending on the load. Though it's governed, there's usually soome droop in the regulation loop, and the ability to observe transients would be a "nice to have" as well. That's why I'm looking for a way to sync/couple the trigger generator to the engine.

BTW, these engines are usually equipped with either a tacho generator or a specific (switched magnetic reluctance) RPM pickup. Their output signal isn't necessarily synchronous with the rotor frequency, but obviously in a fixed ratio to it.

[nctnico]

The HP3314A comes to mind. This has a phase-locked loop feature. A digital pattern generator with an external clock input might also work.

Other than that maybe FM modulation might work.

[David Hess]

The HP3314A comes to mind. This has a phase-locked loop feature. A digital pattern generator with an external clock input might also work.

A lot of old analog function generators support phase locking with variable phase.  And even without variable phase, the function generator could be phase locked and then a pulse generator triggered from the function generator to provide any amount of phase delay, which bring up using just a pulse generator.  Trigger it once per cycle, and then the delay sets the phase of the illumination as long as the speed does not vary too wildly.

[TurboTom]

I checked the HP3314A for its suitability in the manual but unfortunately, the PLL function is too basic. If I understood correctly, it only permits integer multilpiers or divisors (1~1999) and since there's a ratio between the RPM pickup's output and the rotor frequency of someting just below 2 (1.964) so a fractional N PLL would be required to get it right. Anyway, it was quite interesting to see how HP designed this half-analog / half-digital gemstone of the late 1970s / early 1980s.

I guess the simplest method to achieve my goal without constructing a lot of glue circuitry that needs debugging is just connecting a fast frequency counter to a PC, have the PC do the frequency shifting math and send the result to a generator that (hopefully) will convert this into the trigger signal without any update pausing or hickups otherwise...

I should find enough gear in my basement to give this a try I guess... 

[edpalmer42]

Can you program your AWG to output a pulse after a variable delay of 0 to 1 ms.?

60K rpm = 1 ms / revolution.  Figure out a way to provide one pulse per revolution to trigger the AWG.  The variable delay after that will allow you to 'rotate' the turbine and view it from all sides while it's spinning.

No PLL or frequency offset required.

Ed

[TurboTom]

Can you program your AWG to output a pulse after a variable delay of 0 to 1 ms.?

60K rpm = 1 ms / revolution.  Figure out a way to provide one pulse per revolution to trigger the AWG.  The variable delay after that will allow you to 'rotate' the turbine and view it from all sides while it's spinning.

No PLL or frequency offset required.

Ed

And that's exactly the problem. As I told before, there's no integer relation between my RPM signal and the rotor revolutional frequency. The delicate character of this high-speed machinery and the high temperatures at the "business end" make it at least uncomfortable to take a direct RPM reading there. I'ld much prefer to utilize the RPM signal that's available anyway and just process this in a way to provide a synchronous trigger signal to the strobe.

I may install an accelerometer on the engine casing since there's always some residual vibration at the fundamental frequency left, regardless how well the rotor had been balanced. But since all shafts, gears and even the bearings add frequencies of their own to that "melange", substantial filtering would be required to generate a half-way clean output suitable for triggering the strobe. Thus, I like the idea to use the anyway available RPM pickup signal much better.

[langwadt]

pick random mcu board, arduino if you like, stick rpm signal in to a capture timer, set strobe output with a compare timer set sligthly faster/slower than the captured timer