Re: 1510 -- servo, or not ?
Posted: Sun Jan 11, 2009 8:38 am
Thanks for posting here freqband.
I've spent a lot of time lately with the 1510/1512.
If you're going to be using this for instrumentation my recommendation would be to just use a Cgain and eliminate the output servo. I do recommend a stepped resistor gain control however for precise repeatability. I posted a spreadsheet to calculate gain resistors which can be used to create custom step sizes if needed: http://www.picocompressorforum.com/foru ... f=12&t=142
If it were for a recording application and one wanted to eliminate Cgain I would opt for an input servo using a very low offset op amp. We can discuss that one later if you think it applies. The output servo does reduce the 250uV (max) * gain + 5mV DC error term, but it does nothing to reduce clicki-ness during switching. An input servo helps reduce both clicki-ness and output Vos. But it requires an op amp which is several times more expensive than a THAT1510 for very low Vos applications.
Consider taking a couple of things from the single-supply drawing posted here: http://www.picocompressorforum.com/foru ... ?f=6&t=191
1) Use the T-bias scheme in which bias resistors of 1K5 to 2K meet and then connect to a 22K (or so) to ground. (Split supplies.) With a low impedance source the noise advantage is swamped by the source, but the common mode input impedance (on the right-hand side of the coupling caps) is much higher and improves LF CMR a lot because it reduces the input C matching requirement. If your application is split supply you will not need four input caps, only two.
Split the Rstop resistor as shown in the single-supply drawing. I was glad to see that in the drawing you posted.
Not shown on the single-supply drawing (or any THAT drawing to date) are added protection diodes in addition to the bridge. Add two 1N4148 diodes from the Rgain pins back to the inputs. The cathodes connect to the input, the anodes to Rgain. One diode extends from pin 8 to pin 3, the second from pin 1 to pin 2. These parallel the existing internal rev Vbe protection diodes and, along with 10 ohm series resistors make it pretty much bullet proof to phantom faults in which one input is grounded and then the second shorted. Don't ask me how I know this.
Consider using a 1512 for lower noise. The maximum gain is 6 dB less but the 1646 adds 6 dB back. But, at lower gains, the 1512 will be much quieter.
Here's a split supply version:
Your gain resistors and Rstop will be different. The 100R at the output isn't required if a 1646 follows.
I've spent a lot of time lately with the 1510/1512.
If you're going to be using this for instrumentation my recommendation would be to just use a Cgain and eliminate the output servo. I do recommend a stepped resistor gain control however for precise repeatability. I posted a spreadsheet to calculate gain resistors which can be used to create custom step sizes if needed: http://www.picocompressorforum.com/foru ... f=12&t=142
If it were for a recording application and one wanted to eliminate Cgain I would opt for an input servo using a very low offset op amp. We can discuss that one later if you think it applies. The output servo does reduce the 250uV (max) * gain + 5mV DC error term, but it does nothing to reduce clicki-ness during switching. An input servo helps reduce both clicki-ness and output Vos. But it requires an op amp which is several times more expensive than a THAT1510 for very low Vos applications.
Consider taking a couple of things from the single-supply drawing posted here: http://www.picocompressorforum.com/foru ... ?f=6&t=191
1) Use the T-bias scheme in which bias resistors of 1K5 to 2K meet and then connect to a 22K (or so) to ground. (Split supplies.) With a low impedance source the noise advantage is swamped by the source, but the common mode input impedance (on the right-hand side of the coupling caps) is much higher and improves LF CMR a lot because it reduces the input C matching requirement. If your application is split supply you will not need four input caps, only two.
Split the Rstop resistor as shown in the single-supply drawing. I was glad to see that in the drawing you posted.
Not shown on the single-supply drawing (or any THAT drawing to date) are added protection diodes in addition to the bridge. Add two 1N4148 diodes from the Rgain pins back to the inputs. The cathodes connect to the input, the anodes to Rgain. One diode extends from pin 8 to pin 3, the second from pin 1 to pin 2. These parallel the existing internal rev Vbe protection diodes and, along with 10 ohm series resistors make it pretty much bullet proof to phantom faults in which one input is grounded and then the second shorted. Don't ask me how I know this.
Consider using a 1512 for lower noise. The maximum gain is 6 dB less but the 1646 adds 6 dB back. But, at lower gains, the 1512 will be much quieter.
Here's a split supply version:
Your gain resistors and Rstop will be different. The 100R at the output isn't required if a 1646 follows.