Re: [time-nuts] TADD-1 Design (was Re: Stepping up the output of an OCXO
Christopher Hoover wrote: Agreed, that that regulator doesn't have sufficient margin for the entire TADD-1. I was suggesting using the low-noise high PSRR linear reg only to establish the bias level. There are indeed better parts. This one looks pretty good for an integration solution (not include three C's): http://datasheets.maxim-ic.com/en/ds/MAX6126.pdf This app note, which filters the MAX6126 with an additional pass transistor and opamp, shows how to get down to 6 nv/sqrt(hz): http://www.maxim-ic.com/appnotes.cfm/appnote_number/3657 Chris A simpler solution should be possible. Since both the input and the outputs are transformer coupled, the circuit ground plane could sit at half the supply voltage. There need be little dc current flowing into or out of the circuit ground plane so a simple resistive divider should suffice to bias it at the desired voltage. If all RF loads within the circuit are returned to the circuit ground plane then most of the resistive dividers used to set opamp input dc levels can be eliminated. This complicates testing in that either a floating dc input supply would be necessary or differential probes with adequate common mode rating would be required. Bruce ___ time-nuts mailing list time-nuts@febo.com https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
Re: [time-nuts] TADD-1 Design (was Re: Stepping up the output of an OCXO (Dr Bruce Griffiths))
The power supply noise may also limit the performance. This is my major concern with the design. Any noise on the supply rail goes into the first stage via the bias network, and is transferred at whatever gain to the output. I noticed that when I used a particular bad wall wart to run one of my TADD-1's, I saw power supply noise on the outputs. I hacked one of my TADD-1's to decouple the bias from the power supply, and I added a wide-band bead in front of the regulator. Those changes helped considerably. (I also tossed out that bad wall wart.) Using a high PSRR reference for the bias would be worth investigating. There are a number of these on the market, intended for RF apps. E.g., TPS794xx family. http://focus.ti.com/docs/prod/folders/print/tps79401.html There's a good treatment of biasing single-supply op amps here: http://www.analog.com/library/analogDialogue/archives/35-02/avoiding/index.h tml -ch ___ time-nuts mailing list time-nuts@febo.com https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
Re: [time-nuts] TADD-1 Design (was Re: Stepping up the output of an OCXO (Dr Bruce Griffiths))
Christopher Hoover wrote: The power supply noise may also limit the performance. This is my major concern with the design. Any noise on the supply rail goes into the first stage via the bias network, and is transferred at whatever gain to the output. I noticed that when I used a particular bad wall wart to run one of my TADD-1's, I saw power supply noise on the outputs. I hacked one of my TADD-1's to decouple the bias from the power supply, and I added a wide-band bead in front of the regulator. Those changes helped considerably. (I also tossed out that bad wall wart.) Using a high PSRR reference for the bias would be worth investigating. There are a number of these on the market, intended for RF apps. E.g., TPS794xx family. http://focus.ti.com/docs/prod/folders/print/tps79401.html There's a good treatment of biasing single-supply op amps here: http://www.analog.com/library/analogDialogue/archives/35-02/avoiding/index.h tml -ch Chris Yes the 1/2 supply biasing does raise some issues, however the inband component of the power supply noise will be attenuated by the low source impedance. This will off course couple power supply noise into the source. Motorboating used to be common in valve radios when the high voltage power supply electros dried out. Using one of Wenzel's active power supply noise filters for each opamp would help considerably, in reducing the power supply noise and it only uses one npn transistor plus a handful of passive components: http://www.wenzel.com/documents/finesse.html The TI regulators are, from my perspective extremely noisy, one can do much better at least for higher output voltages. They also don't have a high enough output for the TADD-1. Lower power supply noise is possible using a buried zener reference with appropriate filtering as the regulator reference element. Input reverse polarity protection with low dropout is possible using a power mosfet and a few passive components. This avoids blowing fuses if the supply polarity is reversed. No heatsinking is required as the mosfet power dissipation is low. The problems of instability and coupling of power supply noise into the signal path were also present when single supply discrete transistor amplifier were the norm. Quite elaborate bias divider decoupling circuitry was used to minimise the effect. Bruce ___ time-nuts mailing list time-nuts@febo.com https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
Re: [time-nuts] TADD-1 Design (was Re: Stepping up the output of an OCXO
The TI regulators are, from my perspective extremely noisy, one can do much better at least for higher output voltages. They also don't have a high enough output for the TADD-1. Agreed, that that regulator doesn't have sufficient margin for the entire TADD-1. I was suggesting using the low-noise high PSRR linear reg only to establish the bias level. There are indeed better parts. This one looks pretty good for an integration solution (not include three C's): http://datasheets.maxim-ic.com/en/ds/MAX6126.pdf This app note, which filters the MAX6126 with an additional pass transistor and opamp, shows how to get down to 6 nv/sqrt(hz): http://www.maxim-ic.com/appnotes.cfm/appnote_number/3657 other circuit suggestions This avoids blowing fuses if the supply polarity is reversed. Or a PZT. We use a lot of them in the power circuit for hot plug sockets, like on disk shelves. (A fuse is cheaper wrt to the BOM, but the warranty costs kill you.) The problems of instability and coupling of power supply noise into the signal path were also present when single supply discrete transistor amplifier were the norm. Quite elaborate bias divider decoupling circuitry was used to minimise the effect. Yeah, at some point a bipolar design becomes easier, right? -ch ___ time-nuts mailing list time-nuts@febo.com https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
Re: [time-nuts] TADD-1 Design (was Re: Stepping up the output of an OCXO
Chris Christopher Hoover wrote: The TI regulators are, from my perspective extremely noisy, one can do much better at least for higher output voltages. They also don't have a high enough output for the TADD-1. Agreed, that that regulator doesn't have sufficient margin for the entire TADD-1. I was suggesting using the low-noise high PSRR linear reg only to establish the bias level. There are indeed better parts. This one looks pretty good for an integration solution (not include three C's): http://datasheets.maxim-ic.com/en/ds/MAX6126.pdf This app note, which filters the MAX6126 with an additional pass transistor and opamp, shows how to get down to 6 nv/sqrt(hz): http://www.maxim-ic.com/appnotes.cfm/appnote_number/3657 That circuit is only for about 3.2V output, you can do better for higher output voltages using the same technique with a buried zener reference. The TADD-1 has a 9V supply and requires a 4.5V reference. A 10V supply with a low dropout low noise regulator would be better as one could then obtain an output of slightly more than 1V rms instead of slightly less. other circuit suggestions This avoids blowing fuses if the supply polarity is reversed. Or a PZT. We use a lot of them in the power circuit for hot plug sockets, like on disk shelves. (A fuse is cheaper wrt to the BOM, but the warranty costs kill you.) The problems of instability and coupling of power supply noise into the signal path were also present when single supply discrete transistor amplifier were the norm. Quite elaborate bias divider decoupling circuitry was used to minimise the effect. Yeah, at some point a bipolar design becomes easier, right? -ch Yes, an isolation amplifier consisting of a cascade of 3 stacked common base stages with transformer coupled inputs and outputs should achieve an input output isolation of 120dB or so at 5-10 MHz with a total power supply requirement of around 20mA @ 24V per amplifier independent of the load (open circuit, 50 ohms or short circuit). 6 such amplifiers would required about 120mA @ 24V, considerably less power than a fully loaded TAD-1. Having the power supply current independent of the output loading is also advantageous. The phase noise floor should be -174dBc/Hz or lower with 13dBm input and output. The amplifier delay tempco should also be very low (~1-3 ps/K). This configuration could be redesigned for 12V operation but would required transformer coupling the output of the input common base cascade pair to a pushpull common base output stage. Power requirements would then be about 60mA @12V per amplifier or 360mA @12V for 6 amplifiers independent of the the individual amplifier load impedances. Performance would be degraded slightly with a phase noise floor a few dB (1-2) above that of the equivalent 24V amplifier. Bruce ___ time-nuts mailing list time-nuts@febo.com https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts