Amplifier design pre-consultation consultation

Greetings!

I'm a research scientists with the Johns Hopkins University, and I'm working on a set of designs for an X-ray detector, and trying to spec out various methods for obtaining the data we need. One of the designs is a system based on diode arrays + amplifiers + ADC system. I've already got a good handle on the detector end, and the data acquisition system, but I'm stuck on the amplifier system.

We've used commercial amplifiers in the past, but they would likely be overkill for our situation, and end up quite pricey on a cost/channel basis. Given our specifications, I'm wondering if the optimal solution would be to pay for a consultant to develop and test a design specifically for our application, and then take that design and punch out the number of boards that we would need.

Our generic needs seem to be fairly modest, 100-250kHz bandwidth with a gain of 10^7, but as always the devil is in the details. Naturally, we want the lowest noise possible so that we can measure signals at the nA or sub-nA level.

So, here's the question. Are the specifications and schematic sketch shown here:

formatting link

adequate for a professional to provide a consultation estimate? Would the amplifiers be simple enough that a 2nd year EE student could manage the design, or are we talking about skirting the bleeding edge?

I'm never contracted a consultant before, so should I expect a consulting price tag of $1000? $10000? I'm working with a budget that's higher than a hobbyist, but not quite corporation level.

I would also be happy to discuss specific amplifier design ideas. Given the capacitance of the detectors in question, I would imagine that a very low voltage noise opamp is the way to go, or perhaps a JFET front end. The BF862 looks pretty good, and it's relatively high capacitance wouldn't matter much compared to the diode.

Thanks, Kevin

Reply to
ktritz
Loading thread data ...

You don't say what part of Johns Hopkins you work at, but go talk to the particle and nuclear experimental physicists. Diode detectors followed by amps followed by A/D and triggers are their bread and butter. Google "silicon strip detectors" and find somebody who has worked in it more recently than me (1980's!).

Tim.

Reply to
Tim Shoppa

That certainly is a good suggestion, though often their designs are more suited to high speed pulse shaping/counting. Also, I'm actually located offsite at the Princeton Plasma Physics Laboratory, though I do have a few colleagues on campus that might be able to point me in the right direction.

Another option is the Applied Physics Lab, but I have a feeling that unless you have an actual collaboration with them, getting their engineers involved is a 10k or higher proposition. My budget might be a bit constrained compared to what they are used to.

Kevin

Reply to
Kevin

Ok, Kevin, can't see your post and won't see replies (gmail account?) but let me comment by tacking on to Tim's post:

I assume that's 100kHz to 250kHz, right?

Basically yes. You'd have to add things like: Production volume? How is this power-supplied? What environment EMI-wise? $40/ch is quite realistic but only for large production volumes, of course. Not if you have to do small boutique runs for circuit boards.

Much of this will have to shake out during the initial design phase, a fixed bid isn't quite feasible here.

Not manage, let him/her do it. But a 2nd year student will need consulting help unless he/she has tons of ham radio or hobby project experience.

If you want a complete design with layout and all, that will be five digit Dollars. Since you are at a university why not engage the help of more students? Good ones will be dying for meaningful hardware projects. Sure, they'll get stuck here and there and for that case you should line up a consultant. That's what even many industrial clients do. They sign up with me and call me only when they get stuck. Then they are only billed for the hours I helped them but the bulk of the work was done in-house. An upside is that this way they keep core expertise in-house, IOW by the end of the project there will be people who know the stuff inside out.

And there's always this newsgroup :-)

One would have to sit down and scope out what's out there. Chance are, at this frequency you can beat the JFET with an opamp. That would be followed by more amps to get the desired gain.

Technical comments: The cables lengths are a problem. Mind the surroundings, there will usually be lots of noise sources. Switch mode supplies, PFC or variable frequency drives in elevators etc. All this operates smack dab in your band of interest. 3ft to the diodes is going to be tough. Same for the 50ft to the ADCs. Why that long? Can you do a digital link instead? if not you might want to consider fiberoptics or modulate in onto a carrier somewhere in a quite corner of the RF spectrum. 100kHz-250kHz will be one hellacious noise bucket unless the installation is on a remote island or completely shielded.

Can you use a mail domain other than Google? They worked up a bad reputation because of spam and some folks here have that blocked.

--
Regards, Joerg

http://www.analogconsultants.com/
 Click to see the full signature
Reply to
Joerg

Ok, switched to my Verizon account, hopefully this will work for everyone.

Correct, thanks.

I think the estimated production run is on the document, anywhere from ~250-800 channels. When I've investigated parts and PCB manufacture, it seems like I could probably get by with ~$20 in parts, and ~$5-10 for the PCB. The assembly is where I have no information.

The EMI environment is pretty ferocious actually, so shielding and grounding will be very important.

I'm actually a scientist stationed at a national lab (PPPL), so I don't have that much of a connection with the engineering department at Johns Hopkins. I could try and forge a connection. This newsgroup has been pretty valuable for ideas and component suggestion. In our immediate group, I probably have the most knowledge and experience, and that is pretty slim as it is. I have some access to the engineers here at the national lab, so I might try and have them assist in the design.

Generically, the circuit from the Linear Systems design note:

formatting link

26,D16998

looks like it would work for our application. They have a 1M feedback resistor, but are also speccing a high bandwidth than we need. Of course, I realize that there's quite a distance between a circuit in an AppNote, and a realized PCB design that actually works.

Unfortunately, there is not much to be done about the cables. The detectors have to be inside of a vacuum chamber, and the electronics are not generally vacuum compatible. One of the options I'm considering is a vacuum compatible front-end, but that would severly restrict the available components.

The ADCs are located further from the machine to get the computers and other associated hardware out of the radiation environment. EMI shielding will be of utmost priority, and we do have a fair amount of flexibility with the chassis, so I could build it out of 1/4" copper if need be.

I had toyed with the idea of trying to do this with a vacuum compatible ASIC which would incorporate an amplifier, multiplexed ADC, and digital output right near the detector head, but my guess is that would break our budget. I'm also not sure if we could get the required bandwidth out of such a system.

Reply to
Kevin Tritz

Yes, thanks, that works great.

Makes life easier. A lot. Last time I had to deal with noise down to about 5Hz and that is no fun at all because of not well defined 1/f noise-knees.

Ok, I thought that was channels per system. The last really small prototype run I did was 40 channels (four per board, so 10 boards) and it came in just under $3k total for assembly (non-RoHS). But the next one would be under $2k since the stencils and programming will be re-used. I think $40/channel can be done at 800.

Then I'd really consider moving at least part of the amp right up to the diodes. Or shield/diff the heck out of it but that will not be easy. Often EMI efforts cost more time than the actual design.

Ah, Princeton. Even back in the 80's when I was studying for my masters we were always looking for outside projects. Sometimes as course projects where we had to complete two mandatory ones or just as paid work. I built a lot of RF stuff back then to augment my beer/food/parachuting budget. Later HW projects became scarce at our own institutes and students would almost start fist fights over who'd get in. Many went outside academia for that, even for their big masters project. I don't think finding someone should be a problem. The tough part will be to find a student with at least some practical know-how. A ham radio license is usually a pretty good indicator, if the student has built some stuff from scratch for their hobby or for others.

That's the way it is usually done, plus follower amps for more gain. Phil Hobbs wrote a great article about the topic:

formatting link

He can be found here in the newsgroup quite often.

A word of caution: The LTC6244 is non-stock at Digikey for all versions. Usually not a good sign. But there are others.

Ok, depends on how much of a vacuum and whether contamination by the electronic box is a concern. Potting and/or local pressurizing might be an option but I am not an expert for vacuum situations. 3ft of cable in a noisy environment is no small feat. The photodiode is only a weak current source, almost like a whisper at a rock concert.

There are ways to do it. The low-tech way with shields will make for a bulky and pretty stiff cable. You'll likely end up with as many twin-ax cables as there are channels in a system, plus maybe a large metal conduit for them. Basically similar to aircraft wiring.

Modulation or FO would both increase the BOM budget and R&D expenses while reducing cables costs and providing better noise margins. It's just one of those compromises that have to be weighed and pondered.

Another option is to place the ADCs on board and pipe the data over serially. You'll reach Ethernet speeds but 50ft are easy for that. Lots of work though. What receives the data? Does that card already exist in a shape where a change is not feasible anymore?

I am sure it could but unless you can get almost free IC design help plus a MOSIS MPW run or something like that it'll break the budget, big time. Longterm this might be the best avenue. And the most expensive in design.

--
Regards, Joerg

http://www.analogconsultants.com/
 Click to see the full signature
Reply to
Joerg

=A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0... Would

0

Everybody wants to get a bit of business that the Burr-Brown (now Texas Instrument) FET-input OPA656 hogs.

formatting link

Linear Systems has published some great application notes but that may not be one of them

I did quite a lot of work at Cambridge Instruments (UK) on their electron microscopes. They operate with a "chemical vacuum" - o-ring seals and no baking-out - and the only real issue for electronics in the vacuum chamber was heat dissipation in the absence of convection. Good wide thermal conduction paths to structural metal-work mostly worked pretty well. If you need a physical vacuum close to the detector you might get away with baffles and differential pumping ...

Four (or more) layer boards with buried ground and power planes are surprisingly insensitive to external fields. Joerg has publicly advocated burying signal lines as strip-lines in inner layers (between ground planes) though this does make it difficult to get characteristic impedances over 50 ohms. Fanatics have been known to use semi-rigid coaxial cable (or conformable coaxial cable which relies on soaking the outer braid with solder) for maximal screening on cable links. The bonus is that the coaxial connections are good up to a few GHz (18GHz with SMA connectors. when I last looked).

Overkill if you don't need a really good physical vacuum right up against the detector.

-- Bill Sloman, Nijmegen

Reply to
bill.sloman

I suspect he's more worried about outgassing than the effect on the electronics.

Do you have any info on that? I wouldn't think that ceramic hermetic parts or chip passives would be much of a problem. It's probably possible to get big "hybrid" packages with solder-on lids.

Speaking of connectors- the housing, cable and connectors could

*easily* dominate the cost of this system-- part cost, lead time and even assembly cost. Even without the 1/4" copper (shudder) the OP suggested. [Lead might be better (if you can keep it cold enough), but it's not RoHS. ;-) ]

Best regards, Spehro Pefhany

--
"it\'s the network..."                          "The Journey is the reward"
speff@interlog.com             Info for manufacturers: http://www.trexon.com
 Click to see the full signature
Reply to
Spehro Pefhany

Problem is that defense went COTS to a large extent and not many parts are available in ceramic anymore. Those that are usually break the bank in terms of BOM budget.

That's where creativeness comes in. Looking into potting compounds and such. If you can combine amp and detectors the problem may be reduced to one connector at the system side and that can be a cheap shielded multi-pin type. It doesn't have to be a Lemo.

[...]
--
Regards, Joerg

http://www.analogconsultants.com/
 Click to see the full signature
Reply to
Joerg

I wonder how feasible it would be to chop the photodiode signal in the vacuum, then AC amplify it outside. You should be able to reduce the circuitry in the vacuum to a bunch of diodes, which can be sourced for vacuum use*, at the cost of the chopped version not being the most elegant.

You _would_ get some of your chopper drive signal bleeding through to your return signal, which would come out as a DC bias after amplification and rectification. I have no clue if this would be a killer for you or not.

  • I know that 2N2222 dice are regularly used in a vacuum at 77K for temperature sensing.
--
Tim Wescott
Wescott Design Services
 Click to see the full signature
Reply to
Tim Wescott

Epoxy packages and FR4 boards are fine in a chemical vacuum. I'm sure that they do outgas to some extent, but we had moving parts in our vacuum chamber and they had to be minimally lubricated with some kind of fluorcarbon grease, which also outgassed.Every metal surface in the vacuum chamber carried a monolayer of what turned out to be some kindof carbon compound - anyplace that got hit by a lot of electrons would eventually develop a visible black spot of insulating gunge. In my stroboscopic electron microscope the "off" electron beam hit a spot about 0.1mm from the 0,1 mm diameter hole, and initially we had to pull out and clean the apperture every day because the gunge would charge up and push the beam off-axis.

We solved the problem by using a thin film apperture (where the metal doing the blocking was only a few microns thick). We still got our spot of gunge, but the parked beam kept the metal hot enough that gunge was mainly graphite, and conductive.

Vacuum tight electrical connectors are available - not all that easily - and they certainly aren't cheap. In the late 1980's you could buy vacuum rated BNC sockets (with an O-ring seal), but SMA/SMB sized stuff tended to be improvised.

-- Bill Sloman, Nijmegen

Reply to
bill.sloman

OP

The long cables are likely to be a real problem, and expensive.

The OP will need at least one vacuum leadthrough per photodiode channel if the amplifiers are located outside the vacuum chamber with long cables. Let's consider that a "sunk cost" that will be incurred no matter what, unless vacuum compatible electonics can be found (and that implicitly contains lots of vacuum leadthroughs, it's just built into the component price).

Why not build all of the electronics inside a pressurised (1 atmosphere *) welded or soldered metal can that goes inside the vacuum chamber with vacuum seals to connect to the photodiodes. Amplify and digitise the signal inside this can, and send the data either over a single high speed digital link, or an optical link as others have suggested. The cost of these circuits will easily be paid for by the cheaper cable. If the chamber needs to be baked, then put thermal insulation around the components inside the sealed metal can, so that the can gets hot but the components do not. If you need really good insulation, use a Dewar (stainless steel thermos flask). If the electronics dissipates too much heat and needs to be kept cool, or if you want to bake the system for days, then you could run a couple of small diameter metal water pipes from the lab through the necessary vacuum seals, through the electronics can and back out again, to water cool the PCB.

  • Having an air pipe to vent the inside of the can to atmosphere might also help in diagnosing leaks - if you are using a helium mass spectrometer to find the leaks then you could squirt helium into the can when you want to know if the can is the source of your slow leak...

Chris

Reply to
chrisgj198

Olin College of Engineering's SCOPE program might be a good fit.

formatting link

Kevin Gallimore

----== Posted via Pronews.Com - Unlimited-Unrestricted-Secure Usenet News==----

formatting link
The #1 Newsgroup Service in the World! >100,000 Newsgroups

---= - Total Privacy via Encryption =---

Reply to
axolotl

Kevin -

Having looked at other requirements you've mentioned (Vacuum compatibility, cabling, noise) you don't need just an EE to design a pre-amp for you. At the rate I see requirements appearing, your proposed scheme with all its cabling will make the preamp engineering and construction costs less than 1 percent of the total price of instrumentation. Most of your cost is going to go into cables and connectors!

You need someone with broader experience in experimental design, construction, integration, and data acquisition. An EE could certainly be part of this team, but it would be one with experience in integrating such experiments into your particular environmental requirements.

When I was in the strip detector business, the preamps were mounted at the detector, and then there were massive bundles of differentially driven twisted pairs in ribbon cables ("Twist-N-Flat") running hundreds of feet to the big electronics. This was pretty "hot stuff" and in fact the trigger and A/D converters were massive racks cooled by chilled water. (Now, I also know chilled water is also something common to plasma experiment cooling!)

But that was being done with essentially late 70's technology - electronics had evolved rapidly over the 80's and far more integrated solutions were coming available. By the early 90's, designs had realized that all that cabling wasn't optimal - instead integrated solutions with preamp, trigger, A/D all mounted within inches of the detector were being done. Again, google "silicon strip detector" to see the compromises and solutions.

Tim.

Reply to
Tim Shoppa

Although, back in the 70's I remember jumping up and down when I discovered the uA733. Tons of bandwidth, modest consumption. I used to do all this stuff with Harris HA2540 and similar chips but they got hot and were very expensive. However, later the ceramic DIP version and the LCC disappeared from the marketplace :-(

--
Regards, Joerg

http://www.analogconsultants.com/
 Click to see the full signature
Reply to
Joerg

Yup, the 733 used to be hot stuff. Nowadays you can get a real opamp with g=10 at 1.8 GHz, or a sub-dollar 8 GHz MMIC. Makes it almost too easy. Hell, a coke at Zeitgeist costs 2 bucks.

John

Reply to
John Larkin

MMIC are great but I often can't use them because of channel to channel gain tolerances. Most imaging systems must remain within +/-0.5dB. Factory trim is frowned upon and pretty much off-limits in this here office. Opamps are really cool but they can hardly touch the uA733 in terms of cost. 35-40c a pop.

Don't they have Russian River IPA for around $10/pitcher? Much better deal and hugely better tasting :-)

--
Regards, Joerg

http://www.analogconsultants.com/
 Click to see the full signature
Reply to
Joerg

Interesting, really high gain is best accomplished in stages, then you can put very low power front ends in vacuum. The current levels are a bit challenging as well. My normal reaction to seeing that high of a gain specification is "What went wrong in the design?" I am not sure that it applies to this situation. Vacuum without conductive heat relief is difficult, the water pipe suggestion should be considered.

Reply to
JosephKK

There are specific constraints that set the level of current. Basically, we are trying to image the edge of a fusion-grade plasma with high spatial (< 1cm) and time resolution (100kHz-250kHz). The diodes measure the X-rays from the plasma, but the x-rays of this energy only travel through a vacuum, so the diodes need to be in the vacuum with the plasma. It's a high vacuum environment (~10-8 torr), so the components need to be "clean" (little to no outgassing). There could be some conductive heat relief by heatsinking to the vessel wall, though I would have to be careful of the grounds.

Given the constraints on the device, things like in-vessel water cooling, or a sealed can at atmospheric pressure just aren't possible, or rather they would not be allowed unless the managers of the device believed that this diagnostic was absolutely critical. There would be too much of a risk that a leak would cost run-time on the device.

The lower the current we can measure, the better we will be able to see the real "edge" of the plasma, where the signal goes to zero. And the spatial resolution and distance to the plasma limits the size of the aperature and detector that we can use.

It seems like the task is a bit more daunting than I had initially hoped, but the suggestions and information in this thread have been quite valuable.

These are the vacuum feedthroughs I was hoping to use:

formatting link

which could handle 80 single-ended signals (with a good shield). If I had to go differential, the cabling and feedthroughs would likely be too much to manage.

The vendor that provides the diodes has an in-vacuum amplifier for a different model diode array, but the gain is a bit low, and the noise specs don't seem very impressive:

formatting link

Reply to
Kevin Tritz

The high gain op amps will not have a low impedance output. Be sure that the ADC doesn't couple back into the gain stage, i.e. effect the op amp by presenting a time varying load as it converts. In simple English, make sure the ADC uses a buffered design.

I have an amp design I got from the Yale physics department via a website that looks pretty good. What they did was use some diode clamping to insure that if the sensor gets whacked, the amplifier can recover quickly.

Large-Area, Low-Noise, High Speed, Photodiode-Based Fluorescence Detectors with Fast Overdrive Recovery S. Bickman, D. DeMille Yale Univeristy, Physics Department, PO Box @08120, SPL 23, New Haven, CT 06520

Reply to
miso

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.