| General > General Technical Chat |
| Low ESR Output Capacitor for REF5010 and REF5025 |
| << < (6/6) |
| David Hess:
--- Quote from: EPAIII on July 13, 2022, 12:15:30 am ---The latest posts have me seriously thinking about this approach. The REF chips are +/-0.05%. The 0.1% resistors suggested are above that level so that alone gives an accuracy hit. I would prefer 0.01% resistors or better, but then the price goes up. And I guess I can sort through a large batch of them with a meter of lesser accuracy because the goal would be matching values, not absolute ones. --- End quote --- That is the reason I mentioned the issue with the high voltage precision feedback divider. Stacking the precision resistors seems like a solution to this calibration problem, without even being significantly expensive. It takes up a lot more space compared to two resistors though. |
| magic:
HV and precision rarely go together in IC technology. Those APEX chips have 0.25mV/°C guaranteed offset drift, which is 0.01%/°C when amplifying a 2.5V reference and marginally better at 10V. In comparison, the cheapest grade of jellybean OP07 is guaranteed within ±0.25mV worst case at all temperatures, and then there are higher grades and better precision opamps (LT1001, OP77, OPA277, ...). The opamp will need to drive some discrete amplifier inside the feedback loop, you need >10x voltage gain to go from 30Vpp to 300Vpp. Then there is the question of how the output works. You can either have one opamp which controls the final output and uses a variable ratio feedback divider, or one opamp which amplifies always to 300V and taps on the feedback divider for getting lower voltages. Those will have high output impedance, so an active buffer will likely be necessary to drive the output, some sort of bootstrapped opamp configuration. The single opamp solution avoids the buffer, but it means that Vref is the lowest voltage the circuit can output. --- Quote from: EPAIII on July 13, 2022, 12:15:30 am ---Back to switching those resistors. Some years ago I came up with a circuit for a decade resistance circuit that allowed it to be used as a decade pot. I believe I used twelve matched resistors for each decade and two wafers on a rotary switch which effectively jumped across two of those twelve with the next, lower decade. The problem was that each successive decade needed resistors that were 0.2X those of the previous one so only one decade could use 10xxx resistor values. Then 20xx and 40x ones would be needed (example 12 each: 10.00K, 0.01%; 2.0K, 0.1%; and then 400, 1%). But will that work with an op amp? I think yes, but it should be tested. --- End quote --- It sounds like you invented the Kelvin-Varley divider. |
| David Hess:
--- Quote from: magic on July 13, 2022, 06:16:12 am ---HV and precision rarely go together in IC technology. Those APEX chips have 0.25mV/°C guaranteed offset drift, which is 0.01%/°C when amplifying a 2.5V reference and marginally better at 10V. In comparison, the cheapest grade of jellybean OP07 is guaranteed within ±0.25mV worst case at all temperatures, and then there are higher grades and better precision opamps (LT1001, OP77, OPA277, ...). --- End quote --- A separate precision operational amplifier would be required to null out the high voltage Apex part. This method is sometimes used to make a high bandwidth precision amplifier. |
| magic:
The inverting configuration is neat because it allows voltages higher and lower than Vref to be produced by a single amplification stage without further buffering, subject to finding a way to arrange and switch the feedback divider appropriately. The HV amplifier could be an Apex or a cheap discrete contraption. |
| David Hess:
--- Quote from: magic on July 14, 2022, 07:17:02 am ---The inverting configuration is neat because it allows voltages higher and lower than Vref to be produced by a single amplification stage without further buffering, subject to finding a way to arrange and switch the feedback divider appropriately. --- End quote --- The advantage of the inverting configuration as shown is that the low voltage precision operational amplifier only sees low voltage. In the non-inverting configuration, the low voltage precision operational amplifier has to follow the common mode voltage, which could be higher than its supply voltage, so bootstrapping or something else is required. There are all kinds of ways to do it in either configuration; I just showed that as an example of how a low voltage precision operational amplifier can be used to null out a high voltage operational amplifier. The accuracy of the inverting configuration is also not limited by finite common mode rejection. |
| Navigation |
| Message Index |
| Previous page |