2 ?'s - Analog Mux chip & AD9850 DDS at low freq's

[Skimask]

2 questions... (1/2 a repost from another thread...but...ya know...)

1)
For my ASA project, I've settled on an ADG1406 16 channel analog MUX....I think.  Want to run it by anybody that is probably more in the know than me...or cares for that matter
This is for the curve tracer project I'm playing around with.  Need to switch a select resistor into the D.U.T. part of the circuit.  I was going to use a bank of relays.  A-takes up too much room, B-don't want to listen to the relays clicking.
The signal going into any of the inputs will run anywhere from 100mV to ~12v pk-pk, AC sine wave, roughly 20hz up to 100khz at most.  Resistors to be switched will run anywhere from 10ohms up to ~100k.  At the higher pk-pk input voltage settings, the lower value resistors will be 'locked out' so the chip doesn't get overloaded with too much current.

28 pin TSSOP, dual supply @ +15v/-15v (I'll be running it at +12/-12 at most), fairly low Ron at the switches, low enough for my app.
Page 7 of the datasheet (http://www.analog.com/static/imported-files/data_sheets/ADG1406_1407.pdf) shows it'll do 180mA per channel continuous, which is about double what I'll be feeding it, and even then, I'll be feeding a sine wave thru it, so shouldn't be hitting anywhere near the max current limit for any length of time.

2)
On the subject of those AD9850 DDS modules from ebay.  I'm going to assume I don't get one that counterfeit or otherwise "faulty"...
Google'ing the subject shows a lot of HAMs using these to generate high freq's (e.g. >1Mhz).  I'm looking to use one of these modules down near the audio range (e.g. 20-20k).
Is the output from these modules pretty clean at low freq's as well?

[ve7xen]

B-don't want to listen to the relays clicking.

o.O what's wrong with you!

On the subject of those AD9850 DDS modules from ebay.  I'm going to assume I don't get one that counterfeit or otherwise "faulty"...
Google'ing the subject shows a lot of HAMs using these to generate high freq's (e.g. >1Mhz).  I'm looking to use one of these modules down near the audio range (e.g. 20-20k).
Is the output from these modules pretty clean at low freq's as well?

I haven't measured, but the signals look pretty clean on the scope. I'd expect cleaner than they are at MHz frequencies, but maybe you want to add some more filtering; the eBay modules have a 2 (or maybe 3?) pole filter at around 15MHz IIRC. If you're using them for AF it might be worth adding some additional low pass. With a wideband spec of min 63dB SFDR at 1MHz I doubt it's necessary though, depending on your application. If you need cleaner, you probably need to go for a different technology or higher bit-depth DAC.

[Skimask]

o.O what's wrong with you!

Just another mechanical thing to go bad...not that solid state items don't go bad...but...ya know.

I haven't measured, but the signals look pretty clean on the scope. I'd expect cleaner than they are at MHz frequencies, but maybe you want to add some more filtering; the eBay modules have a 2 (or maybe 3?) pole filter at around 15MHz IIRC. If you're using them for AF it might be worth adding some additional low pass. With a wideband spec of min 63dB SFDR at 1MHz I doubt it's necessary though, depending on your application. If you need cleaner, you probably need to go for a different technology or higher bit-depth DAC.

The example app in the datasheet suggests a 5 pole filter at ~40Mhz and looking at the pics of the PCB itself, it looks pretty much like the same thing.  This would further suggest that the module would need further filtering at the lower frequencies.
I'll find out in a few weeks.

[Smokey]

I don't know if this is the same module you have, but it looks like the one I got off ebay at least.  There isn't much going on with that board.

[Skimask]

Yep, same thing but different.
I got the blue PCB, slightly more square-ish, likely the only difference between the 2.

[ve7xen]

The example app in the datasheet suggests a 5 pole filter at ~40Mhz and looking at the pics of the PCB itself, it looks pretty much like the same thing.  This would further suggest that the module would need further filtering at the lower frequencies.
I'll find out in a few weeks.

Now that I'm at home and looking at the actual device I think you're right, the datasheet 40MHz filter seems to be what they've implemented.

I have taken some quick scope captures at 1KHz for you. All with 1KHz test frequency, 20MHz bandwidth limit switched in, 1Meg unterminated, DC coupled, microcontroller powered off, and a bit of a ratsnest on the breadboard. CH1 is the 'SINA' filtered output from the eBay module, CH2 is the 'SINB' pre-filter output.

The FFT is obviously not that useful with an 8-bit scope, but what the hell, you can at least see some of the harmonics. I don't have time to rig it up to a soundcard right now for more useful measurements. I'm curious though, maybe I will get around to it later.

[Skimask]

Well, that's a lot better than what I've got now...which is exactly nothing
Looks like the bulk of the mess is about 40-50db down from the fundamental.  That's plenty good enough for me and should be easy enough to implement a lower freq filter on that output, along with a opamp to shift the signal levels around a bit.

[Skimask]

I don't suppose you've got time to post a pic of that same 1Khz wave with only ~1 cycle displayed?

[ve7xen]

Sure, I've still got the test setup sitting on the bench (now under a pile of other stuff ). All of these are from the filtered output. There's a bit of a glitch just before and after the zero crossing that I didn't catch before on the 'wider' capture. It doesn't occur every cycle, I'm not sure how periodic it is. I've given a close look at a single shot of that as well (20MHz filter off).

Not sure how much it matters with this device, but I'm using a much better power supply for it than last time, I think much of the hash on the unfiltered output before was due to the crappy switching supply I was using. On a proper linear bench supply now.

Finally while I had this powered up again I figured may as well plug it into the PC, so I did some analysis there too. This is not the best measurement technique, some gator clips strung across the bench to the PC's onboard line in, but it has much better dynamic range than my 'scope and gives a good idea of the harmonic products; the noise floor seems below that of my sound card.

[Skimask]

Well, that output is several orders of magnitude cleaner than what I'm using now, even unfiltered and with the glitch.

Which makes me wonder about that glitch....
Could it be that these modules are so cheap because they're using bugged AD9850 chips that somebody pulled out of a reject bin?
Digikey sells that chip for something like $23, whereas this module is only $5 if you get it from the right guy off ebay.

[ve7xen]

Definitely a possibility, though it seems odd to me that there would be a large supply of bugged AD9850s on the market these days, as the design is at least 10 years old. I'd expect if there were bugs in the silicon they'd be long fixed by now, though maybe there is some other parameter that Analog sorts for and these are indeed rejects?

Either way performance seems not bad at all for the price. I was going to suggest using one of the more modern 14 bit chips that use a lot less power and have much cleaner output, but at $20-30 a piece from any reputable distributor (and close to zero availability on the grey market) it's worth sacrificing to use these cheap modules instead.

[AlphZeta]

Well, you got what you payed for I have read somewhere that these chips might be from rejected batch. Or they were removed from circuit boards and then resold. Either way, I am not surprised that the performance is hit or miss depending on the source of the chip.

Anyway, I bought one a while ago and the waveforms weren't as clean as the pictured. But for what I use it for it is good enough.

[bingo600]

I have one of these also

The ocxo died on it , after a short time.
I was feeding it 5v , and it didn't say i shouldn't do that.

Maybe the ocxo was 3v3 ??


/Bingo

[ve7xen]

The ocxo died on it , after a short time.
I was feeding it 5v , and it didn't say i shouldn't do that.

Maybe the ocxo was 3v3 ??

Wouldn't surprise me. I've been running mine on 5V for some testing, but nothing too long term (few days continuously). AD9850 is only spec'd for a 125MHz clock at 5V though...

[ve7xen]

Thinking on this glitch a bit more, I wondered if it's not due to the high impedance of the test circuit and the switching of the adjacent square wave outputs. Low and behold, this is exactly the case. Adjusting the comparator input pot moves the glithces around, and adjusting it to its maximum (it will never trigger) removes them completely.

Given the performance measured seems in line with the 1MHz performance specs in the datasheet, I'd say these are either real, or good enough fakes. I don't have a spec-an to test them at RF frequencies of course, so perhaps they falter there...

[Skimask]

And still, at $5 a piece, can't beat it with a stick...

Well, you could...but...

[ve7xen]

I redid the frequency domain measurement with a better sound card to post on my blog. This is an emu 1212m, and with this FFT configuration and the input shorted at the end of the cable I used, the noise floor is flat at -130dB except for a -120dB spur somewhere around 30kHz. It's not much different, but confirms it's the output of the module and not the POS onboard sound we're measuring:

[ve7xen]

I learned one more thing today - the AD9850 can't tolerate 15V on Vcc! Oops. Glad it wasn't a project of mine hooked up at the time I made that stupid mistake.

[Skimask]

At least at ~$5 a piece, not much of a loss.

[Skimask]

Finally got around to playing with the cheap AD9850 DDS module from ebay.
Works just as good as I thought it would...easy to program, output is a LOT better than what I was using (5 bit DAC output from a PIC...didn't expect a lot from that anyway).

The glitch mentioned earlier...I saw that glitch when plugged into the SINA output, but not on SINB.  (Using a Tek 2246ModA 'scope)

For $4.91USD, I'm happy enough...

EDIT:  Also, did some rudimentary checking with the 'scope and some figuring...
If my math and 'scope reading is right, the -3db point on the SINA output is around 6.42Mhz.