oscilloscope features for ultrasonic flow meter development

[radhaz]

I'm interested to hear some opinions/experiences from people in the know. I have been tasked with finding a scope that would be appropriate for developing an ultrasonic time of flight flow meter. I have some basic assumptions such as the ability to capture/display/export waveforms, record tx/rx events.

[coppice]

Since you need to resolve to somewhere in the 5ps to 50ps range for most time of flight applications, jitter is usually the critical parameter determining an oscilloscope's usefulness. If you are measuring by timing the zero crossings, LeCroy's new 99GHz oscilloscope should do a reasonable job. You might struggle to get budget approval for one of those, though. 

[georges80]

5ps to 50ps for ULTRASONIC time of flight - huh??

cheers,
george.

[pascal_sweden]

If you are measuring by timing the zero crossings, LeCroy's new 99GHz oscilloscope should do a reasonable job. You might struggle to get budget approval for one of those, though. 

1.000.000 USD

[coppice]

5ps to 50ps for ULTRASONIC time of flight - huh??

If you are measuring flow rates, you need to detect the small changes in the time of flight due to the movement of the fluid. The usual approach is to measure the upstream and downstream times of flight, and use the difference between these two measurements to detect the speed of flow. The resolution you need depends on the fluid (sound in water travels around 5 times as fast as sound in gas, so you need 5 times better time resolution to achieve the same quality of measurement results), the slowest flow rate you need to measure, and the accuracy you are looking for. In general you are looking for resolution in the 5ps to 50ps range for a liquid like water, and a bit less resolution for most gases.

[Wuerstchenhund]

LeCroy's new 99GHz oscilloscope

Actually, it's a 100GHz scope 

[VintageTekFan]

If you are measuring flow rates, you need to detect the small changes in the time of flight due to the movement of the fluid. The usual approach is to measure the upstream and downstream times of flight, and use the difference between these two measurements to detect the speed of flow. The resolution you need depends on the fluid (sound in water travels around 5 times as fast as sound in gas, so you need 5 times better time resolution to achieve the same quality of measurement results), the slowest flow rate you need to measure, and the accuracy you are looking for. In general you are looking for resolution in the 5ps to 50ps range for a liquid like water, and a bit less resolution for most gases.

There are sensor mounting techniques that allow a little fudge on that number - like bouncing off a pipe's opposite wall (double the distance).

coppice, I just ran some numbers, and I'm coming up 3 orders of magnitude off from what you said - can you run us through how you figured that out?  I feel I may be missing something.

[coppice]

LeCroy's new 99GHz oscilloscope

Actually, it's a 100GHz scope 

I thought it was three 33GHz bands stacked up? Are they actually 33.33GHz bands, or something?

[coppice]

coppice, I just ran some numbers, and I'm coming up 3 orders of magnitude off from what you said - can you run us through how you figured that out?  I feel I may be missing something.

• The speed of sound in water is about 1500m/s
• A typical consumer type water meter sends the sound over a U shaped path that is about 12cm in total, with about 8cm along the bottom of the U where the actual doppler occurs.
• The time of flight from transducer to transducer is about 80us. The flight time in the section which changes with flow rate is about 53us.

If we want 1% resolution/accuracy at the minimum operating flow rate, and we have 5ps resolution, our minimum flow rate will be that which causes a 500ps change in the flight time. So the minimum flow rate is where the 53us is changed to 53.0005us (or 52.9995us). At this point the speed of sound is shifted from 1500m/s to 1500*53.0005/53 = 1500.0141509434 m/s over the doppler section of the path.

So, the minimum flow rate we can measure to 1% resolution is 14.15 mm/s. The is the kind of goal the industry currently has, but most designs are a factor of several away from this, especially for short term measurement. For longer term measurement averaging generally helps the overall result.

[nctnico]

I'd investigate on how those flowmeters work. For sure the readout can't cost $100k+ so there must be a good way to make measurements with several $k worth of equipment.

[Marco]

What's the U bend for? To locally speed up the flow?

Any way, why not first try to work with a sampling scope? You can buy a 11801x/CSA803x with a SD24 sampling head for under a thousand easy. That has low enough jitter to see sub 5ps shift in a pulse's edge, should be enough to diagnose the critical parts of the circuit. If you really need something real time in the end then you can just flog it again and rent a couple 10s of thousands worth of scope.

I'd investigate on how those flowmeters work. For sure the readout can't cost $100k+ so there must be a good way to make measurements with several $k worth of equipment.

A measly pic can measure delay with 3.5 ps resolution (50ps single shot). But unless you build it all from theory and have it work straight away in practice you'll probably need a scope at some point to diagnose where you screwed up.

[coppice]

What's the U bend for? To locally speed up the flow?

Most consumer type ultrasonic flow meters are settling around an acoustic path that goes from a transducer to a 45 degree reflector in the middle of a straight pipe, a long the pipe for maybe 8cm to another 45 degree reflector, which sends the signal out of the pipe to a second transducer. This makes a sound path shaped like a U with a long flat bottom section. This is not ideal, as there is a lot of non-laminar flow around the reflectors, but most other chamber shapes seem to have greater issues that the simple U shaped path.

Any way, why not first try to work with a sampling scope? You can buy a 11801x/CSA803x with a SD24 sampling head for under a thousand easy. That has low enough jitter to see sub 5ps shift in a pulse's edge, should be enough to diagnose the critical parts of the circuit. If you really need something real time in the end then you can just flog it again and rent a couple 10s of thousands worth of scope.

If the 100GHz scope were a few thousands dollars its definitely what you would like for this work, but obviously its not what you are going to pick in the real world. A sampler is a PITA, for a couple of reasons, although its probably the only practical tool for capturing the full resolution of the signal. When flow is turbulent you would really like to be able to capture individual measurements in great detail. You'll probably have to live with what you can get from a general purpose scope around the lab. Most of these meters run from a small battery, so they only sample a few times per second. That means a sampler will take a long time to get a detailed picture. You can implement a test mode that samples rapidly, but then you aren't looking at real operating conditions.

I'd investigate on how those flowmeters work. For sure the readout can't cost $100k+ so there must be a good way to make measurements with several $k worth of equipment.

A measly pic can measure delay with 3.5 ps resolution (50ps single shot). But unless you build it all from theory and have it work straight away in practice you'll probably need a scope at some point to diagnose where you screwed up.

The Microchip CMTU module has been around for a while, and i know some ultrasonic meter makers have played with it. I don't know any that have found it a good choice for a production meter. I would be interested to hear from anyone who has.

[Marco]

Most consumer type ultrasonic flow meters are settling around an acoustic path that goes from a transducer to a 45 degree reflector in the middle of a straight pipe, a long the pipe for maybe 8cm to another 45 degree reflector, which sends the signal out of the pipe to a second transducer. This makes a sound path shaped like a U with a long flat bottom section. This is not ideal, as there is a lot of non-laminar flow around the reflectors, but most other chamber shapes seem to have greater issues that the simple U shaped path.

Why not just go from the "bottom" of the pipe to the "top" upstream or downstream? Sure your "beam" will cross multiple flow speeds, but eh it's still monotic with volumetric flow AFAICS. You'll need math and calibration to get an estimate of volumetric flow either way.

[coppice]

Why not just go from the "bottom" of the pipe to the "top" upstream or downstream? Sure your "beam" will cross multiple flow speeds, but eh it's still monotic with volumetric flow AFAICS. You'll need math and calibration to get an estimate of volumetric flow either way.

Do you mean a beam passing diagonally across the pipe? That topology is used a lot for measuring in large industrial pipes. It doesn't obstruct the pipe at all, and it can be implemented as a clamp on device. This makes it attractive. However, it doesn't scale all that well to the small pipes in a  consumer meter. If the bore is only 20-30mm a diagonal path would have to be very shallow to get much displacement down the tube. I've seen a lot of weird and wonderful topologies tried in the last few years, but the U shaped path with 2 45 degree reflectors really seems to be where the industry is settling.