Metal Enclosure for Sensitive TIA Circuit

Hello All,

I have a prototype TIA circuit with high-gain. I have the circuit laid out on a 4-layer PCB with the input and output pins of the TIA free of any inne r-layer ground copper pours in order to reduce the parasitic capacitance be tween these pins and ground. I need the low-noise performance between 0-50k Hz out of this circuit, and in order to do this, I placed the circuit in a metal enclosure in order to shield it. What I've found is that if I don't g round the circuit to the metal box, I get noise peaking at frequencies high er than 20kHz. Grounding the circuit to the box eliminates this peaking.

This circuit will be part of a bigger circuit, so I was planning on includi ng a guard-ring in the layout that would surround this TIA circuit that I w ould solder on a metal can that would act like the metal box I used for my prototype. I was wondering if anyone had knew of any "shields" that I could purchase somewhere that would accomplish this, or if anyone had any layout suggestions that could possibly mitigate this noise issue?

Reply to
George Trio
Loading thread data ...

On Tue, 7 Jul 2015 11:04:19 -0700 (PDT), George Trio Gave us:

snip

formatting link

Reply to
DecadentLinuxUserNumeroUno

What's wrong with connecting the board ground to the box? You want to ground everything early and often.

A ground screw every 2 inches is not too many, and if there's a lot of GHz RF around, (e.g. your cell phone), one every half inch is a nice number.

Make sure the lid of the box makes a good connection to the body. Powder-coated and anodized cases usually need a bit of Dremelling to get that to happen.

On your real board, put one of those nice Laird shields

formatting link

over top of your TIA section, and don't relieve more of the ground plane than you absolutely have to. Don't put sensitive traces over a power plane, and don't run them on the unshielded side of the board.

Even one single via pad sitting over a mildly bouncy supply plain can ruin your whole day. (Happened to me once.)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

t on a 4-layer PCB with the input and output pins of the TIA free of any in ner-layer ground copper pours in order to reduce the parasitic capacitance between these pins and ground. I need the low-noise performance between 0-5

0kHz out of this circuit, and in order to do this, I placed the circuit in a metal enclosure in order to shield it. What I've found is that if I don't ground the circuit to the metal box, I get noise peaking at frequencies hi gher than 20kHz. Grounding the circuit to the box eliminates this peaking.

ding a guard-ring in the layout that would surround this TIA circuit that I would solder on a metal can that would act like the metal box I used for m y prototype. I was wondering if anyone had knew of any "shields" that I cou ld purchase somewhere that would accomplish this, or if anyone had any layo ut suggestions that could possibly mitigate this noise issue?

Keep things grounded (if possible). Maybe the 20 kHz is electrostatic pickup from the florescent lights? (I've got a TIA with battle scars.. the first pcb I had to grind off the entire ground plane with a dremel.. lesson learned.)

How much gain?

Metal covers are expensive. we buy some little ones from Orbel Corporation. part number G-0750SA1125-0200XC but they cost ~$2.50. Seems like a lot for a little piece of metal.

Make sure there are no through holes piercing your shield on the ground sid e.

George H.

Reply to
George Herold

out on a 4-layer PCB with the input and output pins of the TIA free of any inner-layer ground copper pours in order to reduce the parasitic capacitanc e between these pins and ground. I need the low-noise performance between 0

-50kHz out of this circuit, and in order to do this, I placed the circuit i n a metal enclosure in order to shield it. What I've found is that if I don 't ground the circuit to the metal box, I get noise peaking at frequencies higher than 20kHz. Grounding the circuit to the box eliminates this peaking .

luding a guard-ring in the layout that would surround this TIA circuit that I would solder on a metal can that would act like the metal box I used for my prototype. I was wondering if anyone had knew of any "shields" that I c ould purchase somewhere that would accomplish this, or if anyone had any la yout suggestions that could possibly mitigate this noise issue?

ide.

My TIA has a gain of 15Megs.

Reply to
George Trio

Thank you for the response!

Reply to
George Trio

Thank you for the app note!

Reply to
George Trio

Dr. Hobbs,

I have another question for you, in regards to the comment about not relieving more of the ground plane than necessary; what kind of issues would relieving more of the ground plane than necessary bring up?

Reply to
George Trio

Heh heh heh. I wish I had you at my last corporate job. It would have been fun to rank you up against the guys who consistently made the system work, correctly and with low noise, and watch the fur fly.

"Ground everything early and often" is an excellent prescription for systems that (a) don't have things hanging off of long cables, and (b) don't have a mix of noisy current consumers and sensitive low-level sensors all living in something at the end of said long cable.

In that case, "ground very judiciously" is the correct watch phrase.

--
Tim Wescott 
Wescott Design Services 
 Click to see the full signature
Reply to
Tim Wescott

You want the ground again right at the output of the first amp. Otherwise it's more of an antenna (capacitance-wise) and has more inductance too.

George H.

Reply to
George Herold

The only reason not to GND the box would be if there is an expectation of t aking an electrostatic blast.

When the box is un-GND'd then his TIA output node is connected to every oth er node in the ckt via the coupling capacitance from output to box and from there to ckt node, so 2x ckt-box capacitance in series. It doesn't take mu ch of a box to see 100pF coupling.

er.

Reply to
bloggs.fredbloggs.fred

"Phil" to colleagues. (The fancy sig block is for SEO purposes.)

A slot in a ground plane is actually a kind of antenna. (Google 'slotline'.) When you make ground currents take detours, it adds inductance in series with your grounds, and that inductance will talk to the outside world.

So keep your high-Z nets very short, and relieve the ground plane only by the pad/trace dimension plus a couple of times the dielectric thickness between the trace and ground. (This reduces fringing capacitance, which usually dominates in PCB geometries.)

Using a bootstrap, where you drive an island of 'ground' so that it follows the summing node, is a big win. You can do that without relieving the ground plane at all--you just put the bootstrapped pour on the level above the ground plane.

Say your stack is

L1: signals L2: ground plane L3: supply plane / noisy digital stuff L4: signals / noisy digital stuff

You an make an island on L2 or L3 for your bootstrapped 'ground'. Make a somewhat larger island on L3 or L4, heavily via-stitched to L2 around the edges. That'll give you better performance than a relieved ground plane, and won't let in all the cell phone signals in the neighbourhood.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

"Ground everything early and often" applies to things that I do, which are quite different from what you do. I very rarely drive motors, for instance. The OP was talking about putting a TIA on a board, not some big distributed industrial control thing.

VFDs and cell phones are sort of the poster children for bad EMI actors, and require different sorts of solutions.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

Fun, thanks! I've done a driven shield for down probe signals. I never thought about on the pcb.

Tim W. A good book, blog, app note, on ground would be much appreciated by all.

For low frequency stuff, choosing that one ground point is important. (and mostly obvious, (thankfully), right where the signal goes in.)

George H.

Reply to
George Herold

I know. I was just tweaking you, and my former coworkers in absentia (because I've been a happy spectator to the collision when a guy who can't be budged from "ground everything early and often" shows up).

It's actually quite a challenge when you have a mix of bad actors and sensitive parts -- what seemed to work best was to establish very good commons between, e.g., a detector and its amplifiers, then keep that ground as isolated as possible from the ground of the honkin' big motors

-- all the way back through 100 feet of cable to a bolt in the central box, if necessary.

In my case it was a big distributed mil/aerospace control thing -- but yes.

Fortunately for everyone's sanity, about the time that I was getting senior there, we started digitizing video. This allowed the really sensitive electronics to have their own localized ground that could be flapping madly with respect to the box's ground, as long as it was flapping in perfect synchrony with everything around it including the references to all the amplifiers and ADCs.

--
Tim Wescott 
Wescott Design Services 
 Click to see the full signature
Reply to
Tim Wescott

Thank you very much for the response, Phil!

Reply to
George Trio

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.