Simple multi-channel serial ADC (8-ch)?

If you intend to interface into a microcontroller they why not design in one of the microcontrollers that has 8 channels of A/D built in. Very simple internal parallel interface. Examples are SiLabs C8051F126 with 8 channels of 12-bit A/D, Atmel ATMega16 with 8 channels of 10- bit A/D or OnSemi MC68HC908GP32 with 8 channels of 8-bit A/D. Of course there are many others.

- mkaras

Joerg, I'd employ a controller such as the AVR Tiny26 with 8 differential or 11 single ended 10 bit channels. Since you need a dozend of them just hook them together with an SPI or an USI. Linear Tech has a a bunch of multichannel DACs with the LTC2600 : 8x 12 bits for 9$ as example.

Rene

Unfortunately there isn't going to be a uC on those cards. Also, I found that many built-in converters are noisy no matter how clean the supply pins are kept.

I wanted to avoid a uC on there because it wouldn't be needed for any other purpose. They are too slow for the analog stuff I am doing on the boards. The ADCs on uC also tend to be noisier than stand-alone chips. Of course I could set up the uC to do autoscan and chips like the MSP430F2xxx have that feature in hardware. All you need to do is set one bit in a register upon initialize. Beats me why stand-alone ADCs don't have that. That way the host wouldn't have to talk, just enable CS via the correct address on the bus and then read the stream.

Thanks, that could be a good one. I am currently leaning towards the TLV56xx series. It all depends what's in stock because that is an indicator of popularity, and sometimes of longevity as well.

Micros are sometimes cheaper than standalone ADCs. You don't need to use them for anything else.

How fast do you need? Is 500KSPS too slow for you? How about 1MSPS?

That's true. You have to be careful in layouts.

Some have interval timer based conversions and auto averaging.

Max1202 - simple SPI interface and internal reference. No waits - data comes out as you shift the command in.

But they need to be flash programmed and are still more noisy.

The ADC can be slow, a few kHz at the most. Not critical. What I meant is that I can't use a uC for any of the control gear tasks. That would have justified the hassle of in circuit programming. Unless you buy gang programmers many manufacturers do not offer file dumps. They give you a nice and cozy design environment but, for example, TI does not offer a feature where you connect an EZ430 pod or the like and just download a file right from the OS or from a web browser. You have to do that from the IAR console. Not such a good thing in this case since the engineers at my client are all specialized in optics and process control.

Even then they are noisier. Bond wires only provide so much in conductivity and because form factor is a marketing concern they often spring for a single VDD connection, two at the most.

That would almost suffice. But chips like the MSP430F2012 even have a nice SAR converter in there with autoscan. I still don't get why there are no stand-alone converters with this autoscan feature. All the ones I checked so far still needed an "invocation speech" to them via an SPI write or were plain out of stock everywhere. The latter is a huge red flag.

Thanks. But according to its datasheet (page 11) you must still clock a control byte into DIN. Seems just like all the others. Man, why do they do this?

Also, that vendor isn't really my favorite. Got too many black eyes WRT availability. Digikey only has 12 of these in SSOP, not a good sign and not enough for even one unit in our case.

BTW, I clicked on their site for SSOP: Leadtime 16 wks. Ouch!

Yes, I understand your concern. Programming environment is one reason I am avoiding the MSP. At least with ARM, JTAG is possible. For our particular project, we need to be able to JTAG one micro from another.

If flash is reliable, then I would not care about the programming part. I am not scared by the initial programming, but the possiblity of re-programming in the field makes me nervous.

For example, flash is very likely the problem for the AT91Sam7S256 failing in the field. Failure is Failure, even if you can re-program it.

I told them that, and certainly others did. All they'd have to do is maybe hire an intern who can write a simple download routine so you can dump data via SBW into the device without the design suite open. Preferably without even needing to have a clue of what's happening. Just by selecting a file, connecting the cable and hitting "Program Now". It's not rocket science. But so far, nada.

The topper would be a wee sub-program that allows one-pin programming so you could connect a little IR photodiode and then "beam it over". With good encryption to avoid data errors in ambient light. Look, ma, no cable! While it's no big deal to write such code for the MSP430 the PC part would be a challenge, at least for an analog guy like me.

Ouch, that would not be good.

But you are assuming that the bootloader is reliable, I would not rely on that unless the bootloader is in hardware rom, not flash and not eeprom.

If the spec is open, someone would write it for you. Problem is that most micro makers are protecting their programming interfaces like trade secrets.

The original one is hard-coded in but I don't remember whether that ain't just a write-protected flash area.

Yep, and that is a rather silly business behavior. For TI it has already cost them sales in my case. I had bugged them numerous times about some more data about their MSP430 devices, then gave up on it. Went analog, actually.

Just like a vendor who refused to disclose the schematic of a laser controller, a schematic I only needed to find out what the plant parameters would be. So, I designed it out. Slam bam, door shut. Even if they came back now and repented it would be two days too late.

We use Texas TLV2543 a lot. 11 channels, 12 bit. Yes you need to send a command to set up what channel you want to read, but the SPI gets the result of the last conversion as you clock in the next command, so its only one 16 bit transfer to read the last result and set the next one up.

I can't use that one because the digital folks decided to go 3.3V. But TI has similar ones that go down to 3V or even 2.7V. I guess I'll have to use one of them, looks like no other choice.

Still I think it would be smart to market an SPI AD converter that has it's own little clock, totally asynchronous from the SPI bus. Just a simple RC or some cheap resonator. Then you could run that at a sample rate that is low enough for part tolerance margins. Now, instead of getting a first value that was acquired who knows when you could be sure that it'll be rather recent, max x clock cycles ago. And it'll be there almost instantly.

Hi Jörg,

I love the ADuC702x 12 bits resolution and not bad at all. Try it and love it.

Marte

Hi Jörg,

again, why should a manufactor do this. Most people will use µC for this operation. The Core for this is easy to write, every µC has RAM for a FIFO, a Timer... and they are cheap...

Marte

But then I have to learn a new uC ;-)

And around $8 for just the ADC function would be a hard pill to swallow. Plus when you look at Digikey stock is rather low for that series. I guess I'll go with one of those SPI varieties from TI. Don't know what National would have, can't see it after they managed to mangle up their web site.

Because I believe there is a market. Here's the current situation: There is a host system, a powerful computer. But the connection to the rack I am developing for this client will be via a USB bridge. Sure, we could add another layer of complexity by having a uC somewhere in between and I seriously thought about that. But if there was an auto-scan converter none of this is needed. Of course, then there is my favorite "fix" for situations like that: Build a poor-man's autoscan. Figure out what the ADC wants to hear, generate that somehow with a digital concoction and trigger this from ye olde 74HC4060. Works, done it before, has caused goose pimples in design reviews.

Other scenario (that one is coming down the pike here soon): A mote. It only needs to measure one thing, once in a while, then send that info off to somewhere via a one-way serial link. Why should this need a uC and have to undergo all that costly EMC agency rigamaroo?

Hi Jörg,

I find it very easy to learn. With winarm package no additional cost, only $30 Minikit with serial download cable or simply samples, ready to start with... Only the I2C I didn't understood clearly.

the ADuC7022 may not be the cheapest solution, Atmel may have cheaper ones, or Reneas, or... The Analog MicroConverters do really have excelent AD-units. That's what I know, this is what you didn't guess, me too, but we tested it.

Marte