Konnichiwa,
Well, I know what I would recommend.
If you use cascaded local single transistor AC only shunt using low noise BJT's and sensible value series resistors fed from a reasonably low noise Regulator (317 optimised will do) you can get mightily close to the final BJT's own noise with excellent AC behaviour as well...
Sayonara
peranders said:
Well, I know what I would recommend.
If you use cascaded local single transistor AC only shunt using low noise BJT's and sensible value series resistors fed from a reasonably low noise Regulator (317 optimised will do) you can get mightily close to the final BJT's own noise with excellent AC behaviour as well...
Sayonara
Take a look at the LT1019 series of references, they are very quiet and also extremely stable. Note that you can also use active LPF (sallen key ) after the reference if you need/want something even quieter - just be certain to use small resistors <1K set corner below 100Hz and make sure that the op-amp you use is <10nVrtHz and also low noise below 10Hz, like say op2177arm.. I use a 3rd order filter implementation with the final filter pole in the ouput and use a lag compensation network to set the op-amp dominant pole somewhat below the output pole. Very quiet and commonly used in ATE applications where we need very quiet dc for testing A/D and D/A chips.
I can post further details if anyone is interested...
Kevin
I can post further details if anyone is interested...
Kevin
)
Here's another article on the TL431 from the folks at TI -- it follows another discussion in Switching Power Magazine last year which discussed methods of compensation:
http://www.edn.com/contents/images/091505di.pdf (go to page 5 of the PDF)
http://www.edn.com/contents/images/091505di.pdf (go to page 5 of the PDF)
Konnichiwa,
I would think it obvious.
My complete solution uses schottky diodes, CRC with small first capacitor and optimised LM317 (optimised output and adj bypassing) and dual simple single transistor cleanup shunt.
Over the 40KHz bandwidth that matters for multibit DAC's that gives around 25uV RMS after the LM317. The cleanup shunts knock it down pretty much to transistor noise above a few Hz, so use a low noise transistor in the final position and < 0.2uV noise is on the cards above a few Hz, while very low rate jitter/phasenoise is obviously measurable, it's audibility is another story....
All that is done using very generic parts and with good AC behaviour due to minimised and insulated feedback loops and not using "heroic" overbuild designs, the people who I pulled this from call this approach "Finesse".
Sayonara
carlosfm said:
I would think it obvious.
My complete solution uses schottky diodes, CRC with small first capacitor and optimised LM317 (optimised output and adj bypassing) and dual simple single transistor cleanup shunt.
Over the 40KHz bandwidth that matters for multibit DAC's that gives around 25uV RMS after the LM317. The cleanup shunts knock it down pretty much to transistor noise above a few Hz, so use a low noise transistor in the final position and < 0.2uV noise is on the cards above a few Hz, while very low rate jitter/phasenoise is obviously measurable, it's audibility is another story....
All that is done using very generic parts and with good AC behaviour due to minimised and insulated feedback loops and not using "heroic" overbuild designs, the people who I pulled this from call this approach "Finesse".
Sayonara
Kevin -- could you post a pic with a clearer view of the axes (or just email me jack@tech-diy.com)
Some of these TL431 "Noise Fixes" cause other problems -- remember, it's an "opamp mit reference avec pass transistor" with reactive elements and .AC analysis is required.
Some of these TL431 "Noise Fixes" cause other problems -- remember, it's an "opamp mit reference avec pass transistor" with reactive elements and .AC analysis is required.
Konnichiwa,
The specific implementation and design is for a commercial product and will not be placed in the public domain.
You can find information on the principles involved on the website of wenzel associates under the topic "Finesse Voltage Regulator Noise!".
To get the kind of performace I showed you MUST optimise the circuit first in P-Spice and later in the real world with real parts.
My aim was merely to show that to get low noise needs nothing too spectacular or complex, especially when powering oscillators.
Using a passively and actively filtered buffered reference is where a good reference is needed is again obvious, however you really need to dig out low noise Op-Amp's to equal simple transistor shunts.
Sayonara
kimschips said:
The specific implementation and design is for a commercial product and will not be placed in the public domain.
You can find information on the principles involved on the website of wenzel associates under the topic "Finesse Voltage Regulator Noise!".
To get the kind of performace I showed you MUST optimise the circuit first in P-Spice and later in the real world with real parts.
My aim was merely to show that to get low noise needs nothing too spectacular or complex, especially when powering oscillators.
Using a passively and actively filtered buffered reference is where a good reference is needed is again obvious, however you really need to dig out low noise Op-Amp's to equal simple transistor shunts.
Sayonara
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.