overdrive REF pin on LTC1655 DAC?

[dentaku]

I have an LTC1655 16 bit DAC connected to an Arduino that I've been experimenting with, using the internal 2.048V ref, but I'm wondering about this in the datasheet "The reference can be overdriven from 2.2V to VCC/2".
I'd like to get the full 0V to 5V out of the DAC instead of 0V to 4.096V.

I do have a simple TL431 2.5V reference but I'm wondering if I can just get away with a VCC/2 divider made with a 10 turn 100k trimpot for the time being so I can easily tweak it?
What would happen (anything bad) if I adjusted that voltage to below 2.2V or above 2.5V?

http://www.linear.com/product/LTC1655
REF (Pin 6): Reference. Output of the internal reference is 2.048V (LTC1655). There is a gain of two from this pin to the output.
The reference can be overdriven from 2.2V to VCC/2 (LTC1655). When tied to VCC/2, the output will swing from GND to VCC.
The output can only swing to within its offset specification of VCC (see Applications Information).

[Kleinstein]

The reference input is relatively low impedance. So a 100 K divider would not work well. Even a 10 K divider will see significant loading and might thus show drift. So just a divider is not that practical.

At low voltage an external reference would interfere with the internal one. So one could expect extra errors and would essentially not be able to go below the internal value. With a slight to high value the DAC  (or more accurately the output amplifier) would reach saturation from some point on. Close to the rails the amplifier can not deliver much power anyway.

[dentaku]

So, what would be the correct way to overdrive the REF input to 2.5V to get the full 5V out of his DAC?

Like I said, the only thing I have is a TL431 2.5V Adjustable Precision Shunt Regulator. Would that be useful for something like this?
I've been reading a bit more about this topic and it looks like most references are designed for exactly this purpose.
http://www.ti.com/product/tl431?keyMatch=tl431&tisearch=Search-EN-Everything
I'm not doing anything special or particularly accurate here, just messing with stuff and learning new things.

[Rerouter]

The inbuilt ref regulator turns off when the external pin is driven to a higher voltage. To meet spec, you need to be able to sourcr atleast 100ua into the ref pin.

In such a case i would strongly recommend buffering vcc/2 with an op amp and using that to drive the ref pin.

[Kleinstein]

When really using the TL431, one could just use a resistor from the +5 V to provide a positive current (e.g. close to 1 mA with a 3.3 K resistor) and have the TL431 as a shunt regulator at the ref input. A divider would need a buffer amplifier.

However the DAC internal reference seems to be significantly better than a TL431 - so not much sense in using it. The extra range from about 4.096 V to about 4.9 V is not that large.

[dentaku]

In such a case i would strongly recommend buffering vcc/2 with an op amp and using that to drive the ref pin.

I was thinking about doing that too. I try it and see how well it works.

[dentaku]

However the DAC internal reference seems to be significantly better than a TL431 - so not much sense in using it. The extra range from about 4.096 V to about 4.9 V is not that large.

OK, I'll see how close I can get the 5V. If it's too far away from 5V I'll just stick with the internal ref and only use values from 0 to 64000 in my code so that my maximum output is 4V then maybe amplify that output by 2 with an opamp and get a 0V to 8V swing.

I already did an experiment last night where I send MIDI notes into the Arduino and use the note numbers as a variable named currentNote.
I do ((currentNote * 64000) / 96) so that note 0 (lowest C)  gives my roughly 0V, Note 12 (one octave above) gives me 0.5V and the next octave is 1V etc. If I multiply the output by 2X I will have each octave give my 1V more than the previous one so it will span a whole 8 octaves. Most analogue synthesizers work using 1V/Oct so that's why I want it to work this way. It's also more useful than just the 5 octaves I was planning to use if I can get a full 5V out of this DAC anyway.
We'll see how well it works.

[Ian.M]

If you are going to use an OPAMP anyway why not simply add a 1.22x gain stage after the DAC, using a rail to rail OPAMP?

[dentaku]

If you are going to use an OPAMP anyway why not simply add a 1.22x gain stage after the DAC, using a rail to rail OPAMP?

OK, 5/4.092=1.22189. That's logical.
I could do that too if I wanted to get exactly 5V but after coming up with my full 8 octave scheme I'm thinking I might prefer that now, even though it might be a bit finicky.
Of course your suggestion would also mean that I only need a 5V supply for everything instead of also needing 5V plus at least a 9V supply for the opamp .

I'm going to set my DAC to 65535 and amplify it until it's 5V and see how close this DAC is at giving me 1V/OCT down in the first octave. It seems to be a bit off down in the low voltages but it's fins after that.

[dentaku]

I amplified it by roughly 1.22 and it works and gives me 5 octaves, so at least that's one way to do it only needing a single supply. 

Now I'm going to try the amplifying 0V to 4V by two to get 8 octaves method.
If I power the opamp from the VIN pin on the Arduino while it's plugged into a 9V power supply I have 8.4V to work with. In fact, that how I'm doing it now because the USB was only supplying 4.95V.

Here's what the output looks like sending MIDI notes C-2, C-1, C0, C1, C2, C3. I get 0V to 5V in 1V steps just like I wanted.

[Ian.M]

Hmm  The horizontals of those steps look rather fuzzy.  You may be getting breakthrough from the ripple on the from the PSU on VIN if there's excessive ripple or the OPAMP doesn't have a good enough PSRR.  That would caus an annoying vibrato effect on an analog synth.

Scope Vin and see how bad the ripple is. Also try a large time constant RC filter between VIN and OPAMP Vcc.  e,g. 47R in series, 470uF electrolytic || 0.1uF ceramic to Gnd, or even go up to 100R if you can stand the voltage drop and 1000uF if you've got the space. 

[dentaku]

Yup, it IS quite fuzzy and I already knew that but I'm just messing with stuff at the moment. Amplifying the output of the DAC is just making it worse too I'm assuming.
I was also thinking about adding some filtering on the output later but not too much because it adds a bit of glide between notes.

This is just slapped together on a little breadboard with wires everywhere so I wasn't expecting much. There's not even any proper decoupling yet.

Here's a screenshot of the output of the DAC with the opamp disconnected so obviously it's not 1V/Oct anymore. The yellow trace is the 8.4V coming out of the VIN.

[Ian.M]

On a breadboard that hash could be coming from*anywhere*.   AC couple VIN and up the gain and see if there's anything clearly periodic there from the PSU you can trigger on, then if so have the DAC output a steady 2V (midrange) and see if the noise on the output correlates with the PSU noise.

[dentaku]

Ok, I'll try that later today.

[leblanc]

I would just use matched resistors to provide Vcc/2 to an non inverting unity gain opamp, as a few people suggested.

Another less-conventional approach could be to use a switched capacitor IC (such as the LTC1043) in a "divide by two" configuration.

[dentaku]

I would just use matched resistors to provide Vcc/2 to an non inverting unity gain opamp, as a few people suggested.

Another less-conventional approach could be to use a switched capacitor IC (such as the LTC1043) in a "divide by two" configuration.

The problem I've found with that is the 5V from the arduino is more like 4.95V so I'll never get to 5V.
Using the 8.4V from the VIN pin to power an opamp and amplifying the output of the DAC gives me more room to work with.

[dentaku]

On a breadboard that hash could be coming from*anywhere*.   AC couple VIN and up the gain and see if there's anything clearly periodic there from the PSU you can trigger on, then if so have the DAC output a steady 2V (midrange) and see if the noise on the output correlates with the PSU noise.

I guess this thread is moving away from the original question...

There doesn't seem to be any noise coming from the 9V DC adapter I was using because the noise was the same when using a 9V battery instead.
last night it was either around 70kHz and then around 140kHz.
I noticed that it got much worse if I pushed the cable of the probe closer to my computer. Tonight I brought the scope upstairs where there's noting but a clock plugged into any outlet and I still found some spikes at the same frequency when looking at the Wavegen output set to DC or at the top of the square wave from the Probe Comp output.
The random part of the noise is definitely much less I use the little spring ground probe instead of the regular hook and ground wire with the alligator clip.

I guess the only other thing I can check is to plug it in at someone else's house to see if it's something to do with the electricity in this house. I've tried three different probes and they give me the same results.

I never thought about anything until you mentioned it. I've never used a digital scope before winning this DSOX1102G and I don't think I've bothered looking at a signal while set at 10mV/div.
What could there maybe be in the scope itself that causes spikes at 140 or 70 kHz?

Here's an image of just sitting the probe on the desk with the grabber and ground clip connected to exaggerate the periodic signal and another image with the accessories removed.
If I put my hand very close to the tip without touching it I can see the spikes again but when I move away I can't trigger on them.

[Ian.M]

My first suspicion would be switched mode PSU circuits in LED or Fluorescent lighting.   You'd only need one bulb or fixture without EMI suppression and it would radiate from your mains wiring throughout your whole house.   Unless you've got an underground bunker with no electronics with SMPSUs or electronic ballast lighting, EMI in the 21st century is virtually unavoidable.

Short of building your own benchtop Faraday cage EMI shielding box there's not a lot you can do, especially if you are in an urban neighbourhood so are a few yards way from the neighbour's stuff that you cant turn off.   If you do want to DIY a shield box, (e.g. from a large biscuit tin), one of the major problems is getting the scope probe leads out of it.  You can buy 15mm dia tinned copper hollow braid, and one good option would be to solder stubs of copper pipe just large enough to pass a BNC connector to your box, then cover the whole lead right back to the BNC plug with braid, with an O ring or similar holding it in good contact with the pipe stub and the BNC plug shell.