Hi,

I'm working on a project to move a scientific instrument from lab to a field portable device. Currently, I'm working on a boxcar integrator/averager to replace the oscilloscope in the device. A sample and hold circuit will be used to integrate a 5 to 50 ns wide pulse from a detector, and the sample and hold output of the detector will be digitized with a 10 bit ADC that operates at 10 kS/s.

The sample and hold circuit is pretty basic:
4066 bilateral switch --> 100 nF storage capacitor --> EL5144 100 MHz op amp (unity gain)

The switch is operated with a 100 ns time gate/window for integration. Also, I will probably add one or more op amps (EL5144) with a gain of 5 to match the input range of the ADC.

The circuit has been breadboarded, and it appears to work, albeit with some noise - see attached graph. I was about to solder components to a PCB, and then noticed the datasheet for the EL5144 states "sockets will add parasitic capacitance and inductance that will result in compromised performance." Is this going to be a major issue? And will the performance degrade further (from the breadboard)?

Also, all the IC's here are DIP package.

I apologize if this is a neophyte question. Thank you in advance for any input.

 

Sockets make for easy field replacement, but generally tend to introduce noise, as well as other unpredictable characteristics as they age.

You really have not given enough information to make an informed decision on what you are trying to accomplish. However, socketed DIPs should be reserved for the laboratory, and soldered in before being placed in use in the field.

That's my story and I'll stick to it.

 

I apologize for the lack of detail.

We work with fiber optic sensors where sensor molecule produce an optical response dependent on some parameter in the environment - e.g. pH, temperature, concentration of a specific chemical or toxin, etc. Most often, we choose sensor molecules where the intensity of the light varies with the parameter of interest.

We use a laser with pulse width is ~1 ns as the light source. The short pulse width is required for a long list of reasons. In short, the system is similar LIDAR or optical time-of-flight reflectometry.

The optically modulated light from the sensor molecules is routed through an optical fiber to a photomultiplier tube (PMT) that generates a current proportional to the incident light intensity. A single photon will generate a 10-60 mV electronic pulse (depending on gain), 1-2 ns wide with 50 Ohm termination. With the laser, the peak height of waveform is 50 mV to 1 V, occasionally higher. The waveform width of 10-25 ns. The resulting waveform is recorded with a 1 GHz scope. Then, software is used to numerically integrate the waveform, and computes a value (e.g. concentration) for the parameter of interest with integrated result and calibration data.

Also, a second PMT is used to precisely determine when the laser fires and triggers the oscilloscope. And, since pulsed laser have shot-to-shot fluctuations, the scope is set to average a desired number of waveforms (100-1000 shots). Also, noise is averaged out.

We also have a boxcar integrator, which would perform the tasks of the oscilloscope and numerical integration with the software. We just don't use it because the scope is versatile, you can see the waveform on the screen, no delay generators needed, etc. Unfortunately, the scope cannot be easily made portable and fast digitizers (500 MS/s) are too costly.

I looked around for a boxcar integrator/averager suitable for a field instrument, and could not find one; only units meant for a laboratory environment. So, I sat down and tried to figure out how a boxcar averager works, found a sample and hold circuit in a book, and tried it out.

Also, we intend to use microcontroller for the field instrument. Although the sampling rate is ~10 kS/s, it should be able to digitize the output of a sample and hold circuit.

 

Were you planning to do production units with sockets?, or just initial trial units. If just trial units, then it may not be an issue if the trial units work as anticipated (ie. you have enough parasitic 'headroom' in the socketed pcb's built). If some trial units don't work, and you reckon it is noise or delay related, then you know you are working close to the layout limit.

If you want to stick to sockets then, as well as potential problems as SgtW advised, you also risk performance headroom with mass production. Removing sockets reduces those risks.

Ciao, Tim

 

Just 1 unit now. I'm a postdoc in chemistry who's been given the task suited for an EE, but I won't let it defeat me (at least for now). If it ever goes to production, I would think that SMDs would be used and a real engineer would be involved.

 

Sockets also come in many qualities. Stay away from this type of socket.

And use this instead if you have to use a socket. But I agree the best thing is a solid soldered connection. Also get a set of NEW sockets. Not some old and used ones from a school lab etc

 

Right now, I'm thinking of a compromise: Solder the EL5144s to the PCB and using the sockets for the less critical components (4066 and 74AC14). And, use the higher quality (machined?) sockets.

 

The bottom picture is an example of a machined socket. The top picture is cheap hobbyist grade.

 

I've often found those cheap, hobbyist grade sockets won't even fit into a breadboard properly, and keep slipping or falling out. Recommend using the turned leg ones (the bottom image.)

 

I've often found those cheap, hobbyist grade sockets won't even fit into a breadboard properly, and keep slipping or falling out. Recommend using the turned leg ones (the bottom image.)

If you are talking soldering it in to the breadboard... huh?

We both agree about the sockets though. I've used the 1st when I was a lot newer, and desperate.

 

If you are talking soldering it in to the breadboard... huh?

We both agree about the sockets though. I've used the 1st when I was a lot newer, and desperate.

No I'm talking about plugging it into a plastic breadboard.

 

There's a thought, a socket into a socket.

 

There's a thought, a socket into a socket.

cant be TOO careful i suppose...

 

Sockets also come in many qualities. Stay away from this type of socket.

And use this instead if you have to use a socket. But I agree the best thing is a solid soldered connection. Also get a set of NEW sockets. Not some old and used ones from a school lab etc

I have the same advice as t06afre when it comes to the preferred IC socket type. The bellows type (shown first) have a tendency for the contacts to develop an oxidation layer which can compromise contact with the IC's pins. This oxidation layer often requires the IC to be removed from the socket and reseated to restore a decent contact.

hgmjr

 

despite my first comment, i want to say that this thread has been really informative to me. thanks to all who contributed