ADR3625 Constant Current Source (Figure 80)

Hi,

From Page 7 of the ADR3625 datasheet, you can see that all GND pins must be tied together. This means that Iq will have to flow through the load in the circuit you reference, as it comes from the GND FORCE pin. The topology of the IC does not allow for any voltage difference between the GND pins and GND FORCE, and that would not decouple Iq from GND FORCE.

The ADR3625 topology is shown here in page 19. You can see that the topology is not quite like LT3045, particularly because the output is push-pull and there is a resistor divide in the path of the output sense line.

Again. Pins 3, 5 and 8 must be shorted together.

Iq vs. temperature is shown on page 9, this temp dependence is pretty consistent across individual parts.

Another way to approach this would be with LT3092, which is designed to be a two-terminal I-source, and is appropriate for the current values you want.

The real problem with trying to build an I-source like this is that the resistors need to be high-precision/low-drift as well. You might want to look at iDAC products that will output set currents over a large range and are quite accurate/low drift. Products that can handle 100mA include LTC2672, AD5770R, MAX5832 and MAX5113. ADN8810 is worth a look as well.

Selection Table for Current Source-Sink D/A Converters | Parametric Search | Analog Devices

Regards,

Brendan

Thanks for the idea about the current output dac, I would not have considered that!

I decided to put some numbers to the stability of these options... And now I see that I was going way overboard to consider using the ADR3625.

After going away and looking at this again, and doing an error budget, I came up with a configuration using the LT3092 that should be stable to ~400ppm (0.04%) for a 100mA output current within a lab environment (20-30 degC) after an initial calibration (Again, I am not interested in absolute accuracy but want stability over time/temp etc. better than 0.05%)

Using a couple of 0.1% 15ppm/C 0603 resistors in parallel as a Rshunt and using the same type for Rset, I believe the drift in resistances would only cause errors of ~ 50ppm due to self heating at 0.1A.

Then the main sources of error with the 3092 configuration are the Vos change with temperature and the Iset change with temperature of the LT3092 over the lab conditions (10degC) and self heating (3.3V input, 1.2V dropping across the shunt, 0.1A = 0.2W * 25C/W = 5degC)

The LT3045 appears to allow me to use an identical topology as the LT3092, however with much improved Vos and Iset change in temperature. In that case, I believe that I can achieve stability on the order of 100ppm using the LT3045 in a similar topology (TA04 from lt3045 datasheet). This would be well into the range of sufficient stability for my needs

Are there any other significant sources of drift I should be aware of for LT3045?
Any particular considerations for using it in this configuration (TA04) ?
Is there any long-term stability information available for LT3045?

Thanks,

Andrew

The LT3040 shares the current source architecture with the LT3045 and the LT3040 datasheet has a section called long term drift.   I show that in the image below. 

A bullet on the front page of the LT3045 datasheet say +/- 1% initial accuracy for the current source, and then the EC table says +/-2% worst case, but I agree limiting the temperature will improve the current source accuracy.  I've asked the designers about drift of these current source references in the past and they say they don't see any drift at burn in.   The Typical Application you are considering is fine. 

Hi Andrew, connecting with the engineer who supports LT3045. His answer is here.