galvanic isolation of sensors

ould use an ADuM1401 or 1402 at $2.50/ instead and power the whole thing op= en loop from a transformer with an 11V P-P secondary and a pair of diodes t= o make a half-bridge rectifier. =C2=A0You could drive all the transformer p= rimaries in parallel.

Nope. Ground one side of the transformer secondary. Drive a square wave i= nto the primary, such that you've got say +11V P-P on the secondary. With = one side grounded, this makes the other swing +/- 5.5V. Use diodes from th= at into bulk capacitors and you've got +5 and -5 rails. Ish. They'll droo= p some with load and have some ripple, but there's a very good chance that = your signal conditioning electronics on the isolated side don't care.

If you were to use something like a VersaPak (6 windings on a common core) = you could parallel 2 windings for the primary, series two windings for seco= ndary_1, and series two more windings for secondary_2. Then you could driv= e the primary from some kind of hefty driver like a paralleled triple logic= buffer (say a NL37WZ17) on a +5V rail (with an AC-coupling cap, of course)= and you've got two isolated power supplies. Create the square wave for th= e buffer from a PWM on the microcontroller; you can common the same square = wave source to drive all 5 buffers you'd need to make all 10 supplies. 5V = P-P doubles through the transformer to 10V, which after copper and diode lo= sses probably gives you about +/- 4.2V or so as isolated rails. You'd be l= ooking at little bitty ferrite transformers and driving them in the 100 kHz= sort of range, not the big laminated iron monstrosities you need at 60 Hz.

All this is a bit of work, and is based on the assumption that you're reall= y trying to drive the BOM cost down. If your time's worth more than the sa= y $5 or so per channel you could save, then I'd look to simpler solutions.

--=20 Rob Gaddi, Highland Technology --

could use an ADuM1401 or 1402 at $2.50/ instead and power the whole thing = open loop from a transformer with an 11V P-P secondary and a pair of diodes= to make a half-bridge rectifier. =A0You could drive all the transformer pr= imaries in parallel.

wave into the primary, such that you've got say +11V P-P on the secondary. = =A0With one side grounded, this makes the other swing +/- 5.5V. =A0Use diod= es from that into bulk capacitors and you've got +5 and -5 rails. =A0Ish. = =A0They'll droop some with load and have some ripple, but there's a very go= od chance that your signal conditioning electronics on the isolated side do= n't care.

) you could parallel 2 windings for the primary, series two windings for se= condary_1, and series two more windings for secondary_2. =A0Then you could = drive the primary from some kind of hefty driver like a paralleled triple l= ogic buffer (say a NL37WZ17) on a +5V rail (with an AC-coupling cap, of cou= rse) and you've got two isolated power supplies. =A0Create the square wave = for the buffer from a PWM on the microcontroller; you can common the same s= quare wave source to drive all 5 buffers you'd need to make all 10 supplies= . =A05V P-P doubles through the transformer to 10V, which after copper and = diode losses probably gives you about +/- 4.2V or so as isolated rails. =A0= You'd be looking at little bitty ferrite transformers and driving them in t= he 100 kHz sort of range, not the big laminated iron monstrosities you need= at 60 Hz.

lly trying to drive the BOM cost down. =A0If your time's worth more than th= e say $5 or so per channel you could save, then I'd look to simpler solutio= ns.

That's very cool. I might try throwing one together and seeing how it performs before I decide whichway to go. Thankyou :)

Simon

So, just in case anyone is interested in critiquing the current state of af= fairs, I've put the schematic

I'd be interested to know what (if any) benefits moving to a 4-layer board = (laid out as signal | split GND1/GND2 | Analog GND | Signal ) might bring, = as opposed to the current 2-layer design. The cost/board is an extra $2 or = so, but if it preserves the signal integrity better, it'd be worth it.

Note that these aren't PCIe boards [grin] That's just a handy-dandy cheap i= nterface for my plug-in modules. Note also that this is using the ADuM5401 = to provide 3:1+5V/GND isolation, as well as an Si8430 to provide an additio= nal 3 signal paths. The ADC (a MegaAVR48) is completely isolated from the h= ost "bus".

Cheers Simon.

affairs, I've put the schematic

(laid out as signal | split GND1/GND2 | Analog GND | Signal ) might bring, as opposed to the current 2-layer design. The cost/board is an extra $2 or so, but if it preserves the signal integrity better, it'd be worth it.

interface for my plug-in modules. Note also that this is using the ADuM5401 to provide 3:1+5V/GND isolation, as well as an Si8430 to provide an additional 3 signal paths. The ADC (a MegaAVR48) is completely isolated from the host "bus".

A couple things-- your layout of the crystal should be better.. don't run signal traces under the crystal.. do put a ground plane on the other side.. and don't run the capacitors to a connector that goes off the board then back to ground- run them to each other and/or straight back to the MCU ground pin (or a ground plane) and nowhere else.

Also, consider what might happen if one or more channels are negative railed- it may affect the other (working) channels since it will be beyond the -300mV level and current can spill over. I assume you've checked that 221K is okay for an input resistor under whatever sampling conditions you have.

Oh, and a pet peeve of mine- when you create a new op-amp symbol, take the few seconds to make a symbol that looks like an op-amp not just a block with wires. Same with diode arrays, transformers etc.

Best regards, Spehro Pefhany

affairs, I've put the schematic

rd (laid out as signal | split GND1/GND2 | Analog GND | Signal ) might brin= g, as opposed to the current 2-layer design. The cost/board is an extra $2 = or so, but if it preserves the signal integrity better, it'd be worth it.

p interface for my plug-in modules. Note also that this is using the ADuM54=

"The Journey is the reward"

eff.com

Thanks for that -I'll make the changes :)

Simon