|IAngel|
Member level 3
Hey all - (Hopefully) Quick question for you all.
I designed a very low power 2-stage opamp with the following specs:
Vsupply+ = 500mV
Vsupply- = 0V
When I test the open loop gain with fully differential small signal inputs (1uV), I get a gain of about 60dB max and 40dB min at the extremes of the CMR. I tested teh CMR by connecting the output to the negative input (UG buffer) and found that from about 80mV - 450mV the output voltage followed the input, aka CMR is from 80-450mV. So far so good. Now to test the stability, I biased the input AC signal at 250mV and swept the AC from 1Hz-1GHz. I found the Phase Margin to be about 45Degrees (where it should be) with the compesation capacitor I used, so the amp should be stable. Still so far so good.
Here is the problem then:
I want to use this amp with closed loop feedback as an amplified difference amp. I want to take a small DC signal that varies from about 260mV - 240mV in a linear fashion, remove the offset (somewhat) and amplify that 20mV difference to be very large (aka a large linear signal - maybe 200mV - gain of 100). Now, that gain I want is only 40dB and my amplifier can do much more, so I'm not worried about that - here's my problem:
Before I even bothered simulating my linear signal, I applied my same 1uV sine wave (same as I tested my open loop gain, phase margin, etc with) and I hooked up the amp as an inverting amplifier with feedback (a resistor from my signal to -ve input on amp, resistor from amp output to -ve input on amp) - the standard way as mentioned. To test its functionality, I hooked up a 1k resistor at the front of the input, and a 2k resistor in the feedback loop (Vout/Vin = 2K/1K = gain of 2). When I put my AC signal in at the same bias point as before (250mV), I get a severly compressed signal. This in no way works as an inverting amplifier. I tried at 0V bias as well and that didn't work either.
I then tried a voltage summer to see if that would work properly. Grounded +ve input, put 2 equal resistors between the input signals and -ve input, put a feedback resistor from output to -ve input. Didn't work at all either. With a 0V bias I get almost unity gain (1uV) otherwise, I get SERIOUS compression of the signals.
So: If my amp is working in open loop differential mode (with stability), it works as a UG buffer, I can pull all "working specs" from the buffer config and unity gain config, but it won't work in any other closed feedback situations...
Any ideas? I thought maybe my gain wasn't high enough, but I've read that 20dB should be enough for these applications (with a little error - error should decrease with higher gain). Then I thought that maybe my CMR was too low, so that the signals would only work in a VERY limited DC bias band, but unity gain buffer CMR showed my CMR to be almost rail-to-rail.
What could I be missing? I am running in the subthreshold region, but since you can make opamps out of BJTs, that shouldn't do it...not to mention the amp works in the ways mentioned... (plus I tried a 1.2V version with all the FETS deep in saturation...same problem)
Any ideas?
- btw: Spectre simulator, Cadence Analog Artist schematics, IBM 0.13u toolkit.
Thanks!
I designed a very low power 2-stage opamp with the following specs:
Vsupply+ = 500mV
Vsupply- = 0V
When I test the open loop gain with fully differential small signal inputs (1uV), I get a gain of about 60dB max and 40dB min at the extremes of the CMR. I tested teh CMR by connecting the output to the negative input (UG buffer) and found that from about 80mV - 450mV the output voltage followed the input, aka CMR is from 80-450mV. So far so good. Now to test the stability, I biased the input AC signal at 250mV and swept the AC from 1Hz-1GHz. I found the Phase Margin to be about 45Degrees (where it should be) with the compesation capacitor I used, so the amp should be stable. Still so far so good.
Here is the problem then:
I want to use this amp with closed loop feedback as an amplified difference amp. I want to take a small DC signal that varies from about 260mV - 240mV in a linear fashion, remove the offset (somewhat) and amplify that 20mV difference to be very large (aka a large linear signal - maybe 200mV - gain of 100). Now, that gain I want is only 40dB and my amplifier can do much more, so I'm not worried about that - here's my problem:
Before I even bothered simulating my linear signal, I applied my same 1uV sine wave (same as I tested my open loop gain, phase margin, etc with) and I hooked up the amp as an inverting amplifier with feedback (a resistor from my signal to -ve input on amp, resistor from amp output to -ve input on amp) - the standard way as mentioned. To test its functionality, I hooked up a 1k resistor at the front of the input, and a 2k resistor in the feedback loop (Vout/Vin = 2K/1K = gain of 2). When I put my AC signal in at the same bias point as before (250mV), I get a severly compressed signal. This in no way works as an inverting amplifier. I tried at 0V bias as well and that didn't work either.
I then tried a voltage summer to see if that would work properly. Grounded +ve input, put 2 equal resistors between the input signals and -ve input, put a feedback resistor from output to -ve input. Didn't work at all either. With a 0V bias I get almost unity gain (1uV) otherwise, I get SERIOUS compression of the signals.
So: If my amp is working in open loop differential mode (with stability), it works as a UG buffer, I can pull all "working specs" from the buffer config and unity gain config, but it won't work in any other closed feedback situations...
Any ideas? I thought maybe my gain wasn't high enough, but I've read that 20dB should be enough for these applications (with a little error - error should decrease with higher gain). Then I thought that maybe my CMR was too low, so that the signals would only work in a VERY limited DC bias band, but unity gain buffer CMR showed my CMR to be almost rail-to-rail.
What could I be missing? I am running in the subthreshold region, but since you can make opamps out of BJTs, that shouldn't do it...not to mention the amp works in the ways mentioned... (plus I tried a 1.2V version with all the FETS deep in saturation...same problem)
Any ideas?
- btw: Spectre simulator, Cadence Analog Artist schematics, IBM 0.13u toolkit.
Thanks!