0
\$\begingroup\$

I am a software engineer and have limited knowledge of electronics. I am using a Raspberry PI to control a AD5360 ADC eval board (http://www.analog.com/en/products/digital-to-analog-converters/da-converters/ad5360.html#product-overview).

I need to attenuate +-5 volt output form the DAC to +-50 Microvolts on 16 independent channels.

I am thinking that I need an on chip attenuator. I am still prototyping so and part needs to be used on a bread board or with pins that can be soldered to a solder board. I need a fast response time because I change voltage on each of the 16 channels between 256 to 2048 time each second, and a high level of precision. If required I could build them, however because I need 16 I would prefer to buy them. I have tried to use a simple voltage divider however, this introduces too much variance at the microvolt range.

I am open to other ideas if they can solve the problem more efficiently or are more cost effective.

\$\endgroup\$
  • 1
    \$\begingroup\$ You just need a 100,000:1 attenuator on each output. It can be passive, just consisting of a few resistors. Why do you say a simple voltage divider introduces too much variance? It shouldn't do if implemented correctly. \$\endgroup\$ – Kevin White Mar 22 '17 at 14:05
  • 1
    \$\begingroup\$ you're a bit vague with the specifications here. 'Fast' is not a speed specification, 'too much variance' is not a tolerance, 'a high level of precision' is not ... you get the general idea . You will end up with a simple voltage divider, perhaps you don't know why yet. \$\endgroup\$ – Neil_UK Mar 22 '17 at 14:06
  • \$\begingroup\$ Common ground referenced? impedance? how will you verify? and be aware of thermal noise resistance and bandwidth. \$\endgroup\$ – Sunnyskyguy EE75 Mar 22 '17 at 14:15
  • \$\begingroup\$ 50 microvolts isn't a lot of volts at all. Do you have instruments capable of measuring this low and have you considered that extraneous noise might dominate the outputs? \$\endgroup\$ – Paul Uszak Mar 22 '17 at 14:42
  • \$\begingroup\$ What's this connected to that's capable of detecting microvolts? Is it by any chance biological? \$\endgroup\$ – pjc50 Mar 22 '17 at 15:45
0
\$\begingroup\$

I'd still go with voltage dividers, just use 1% (or even 0.1%) precision resistors. Note that "typical" resistors in hobbyist kits are usually 5% precision. And then buffer each voltage divider output with precision op-amp, because if you take the output straight from voltage divider to load then your voltage will definitely jump all over the place. So build 16 of 10MΩ/100Ω voltage dividers, and connect simple op-amp followers to them. Input resistance of the op-amp is very high, so it will not affect the output voltage of the divider, and then output resistance of the op-amp is quite low, giving enough current for your load.
As for op-amp selection, look for "precision op-amps": TI, Linear, others. "2048 time each second" is not quite fast. It's in the middle of sound spectrum, so any op-amp should be capable of working at that frequency.

\$\endgroup\$
  • \$\begingroup\$ Are 0.1% resistors necessary? Taken in the round, it's unlikely a hobbyist can build and test a circuit to that accuracy. Most 'scopes and DMMs are at least an order of magnitude less accurate. I would suggest 1% is okay and cheaper. \$\endgroup\$ – Paul Uszak Mar 22 '17 at 14:39
  • \$\begingroup\$ @PaulUszak I was going with requested "high accuracy". Of course, 1% may be enough, we don't have the required accuracy spec. I will edit my answer, thanks. \$\endgroup\$ – Volodymyr Smotesko Mar 22 '17 at 15:12
2
\$\begingroup\$

If your output impedance specification is reasonably high (say 10 ohms) then JUST using resistors is going to be your best solution. They add little noise and drift.

For example, you could use a 1M:10 ohm divider, using 0.1% or 0.05% resistors. Or lower accuracy if that meets your (unstated) requirements.

Added noise due to the resistors will be down around the theoretical minimum of 0.41nV/sqrt(Hz), so over a 10kHz BW only 41nV RMS.

You do need to take proper care preventing errors due to thermal EMFs if you expect to get high accuracy with only 50uV full scale, but that's a construction detail and outside of the scope of this answer as it applies to any possible method of producing the voltages. In general you would minimize asymmetry, minimize material choices that result in high uV/°C EMFs and keep everything isothermal.

If you need lower than 5-10 ohms you may have to add amplifiers, but that will greatly increase noise and could add large errors depending on the choice of amplifier.

As far as 'speed', such a divider will be limited by the DAC output- it would settle in nanoseconds with an ideal step input (microseconds at most)- so your kHz should not really be a worry if the DAC is up to the task. If you were looking at really high speed (above hundreds of kHz) you would frequency-compensate the divider by matching the ratio of 1/capacitances across the resistors.

\$\endgroup\$
0
\$\begingroup\$

+1. Opamps are unnecessary. DAC has pretty good on-chip buffers, so 100k/1ohm dividers will provide low output Z. 10k/0.1ohm could even be used...

\$\endgroup\$
  • \$\begingroup\$ It's a little tricky getting precision resistors much lower than about 10 ohms. If 1% or 10% accuracy is okay, then 1 ohm or 0.1 ohm is just as good as 0.1% 10 ohm. The OP has not deigned to enlighten us on the matter of accuracy. \$\endgroup\$ – Spehro Pefhany Mar 22 '17 at 15:27
  • \$\begingroup\$ 0.1 Ohm would be impractical because the resistance of PCB traces and solder joints would kill the accuracy. \$\endgroup\$ – Volodymyr Smotesko Mar 23 '17 at 11:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.