Hi Salas,
Thanks for your reply.
Your circuit does look simple enough to build. Still a number of questions:
Can I substitute 2SK170 with another JFET? I use RS Components most of the time now, since they have a local office in my town and delivery is fast (sometime same day delivery and no shipping fee). I'd hate to have to order this from Mouser (they have it in stock), it takes long time and they charge a shipping fee unless the order is over US$100.
As for the LEDs, I assume any pain jane green LED will do?
As for the 10VDC in, I assume I can use a center tapped transformer after rectification and filter caps? I assume the bottom of the ripple must be above 10VDC, correct?
Am I correct that I can use a 9-0-9VAC 50VA transformer for this task? I currently use soft recovery schottky rectifiers so as to minimize voltage loss after rectification.
Is it possible to reduce this 10VDC (ie: lower drop out voltage) so less heat is produced from the power MOSFET?
What is a 4 wire Kelvin connection? Can you explain a bit more or point me to a link so I can do some reading?
If I want to use this circuit for lower current regulation (ie: 5V 400mA or less), what do I need to change? For one I won't be needing remote sensing as this will be close to the circuit I want to power. I guess I can use a less expensive MOSFET instead of the IRFP9240 in the diagram?
Sorry for so many questions...
Thanks for your reply.
Your circuit does look simple enough to build. Still a number of questions:
Can I substitute 2SK170 with another JFET? I use RS Components most of the time now, since they have a local office in my town and delivery is fast (sometime same day delivery and no shipping fee). I'd hate to have to order this from Mouser (they have it in stock), it takes long time and they charge a shipping fee unless the order is over US$100.
As for the LEDs, I assume any pain jane green LED will do?
As for the 10VDC in, I assume I can use a center tapped transformer after rectification and filter caps? I assume the bottom of the ripple must be above 10VDC, correct?
Am I correct that I can use a 9-0-9VAC 50VA transformer for this task? I currently use soft recovery schottky rectifiers so as to minimize voltage loss after rectification.
Is it possible to reduce this 10VDC (ie: lower drop out voltage) so less heat is produced from the power MOSFET?
What is a 4 wire Kelvin connection? Can you explain a bit more or point me to a link so I can do some reading?
If I want to use this circuit for lower current regulation (ie: 5V 400mA or less), what do I need to change? For one I won't be needing remote sensing as this will be close to the circuit I want to power. I guess I can use a less expensive MOSFET instead of the IRFP9240 in the diagram?
Sorry for so many questions...
Last edited:
1. 2SK117, 2SK222
2. Yes, generic green around 2.1Vf you will find easily I guess.
3. Any trafo and rectification scheme that can give 10V DC at 3A without sagging.
4. At such DC current level some trafos lose nominal voltage, it will show on bench. Don't know just from nominal spec for specific makes. Try and see.
5. No, don't reduce under 10V, it loses spec, imagine that Vgs will be 4V alone.
6. Its the remote sensing 4 wire arrangement with its proper name. Google for Kelvin connection. I have it exactly as it should be connected on that schematic I showed. ''Remote load'' symbolized by a far away resistor are your +/0 DC input connector points on your digital machine.
7. Roughly 2V/R1=ICCS will give you any current setting. Shoot for leaving at least 100mA extra for the shunt not to starve. Fine tune by actually measuring that voltage drop across R1, because vgs with current levels through Mosfet is changing, plus they got different from batch to batch, and the LEDs Vf will be about, not spot on. Don't forget Vsquared/R1 so to size the R1's wattage at about 3x for any setting. You will use remote sensing if you want best subjective and objective performance. Because even cm of wire will add enough nH, and R especially influential for error sense at such current. We look for mOhm Zo performance here, that's why we build such stuff.
2. Yes, generic green around 2.1Vf you will find easily I guess.
3. Any trafo and rectification scheme that can give 10V DC at 3A without sagging.
4. At such DC current level some trafos lose nominal voltage, it will show on bench. Don't know just from nominal spec for specific makes. Try and see.
5. No, don't reduce under 10V, it loses spec, imagine that Vgs will be 4V alone.
6. Its the remote sensing 4 wire arrangement with its proper name. Google for Kelvin connection. I have it exactly as it should be connected on that schematic I showed. ''Remote load'' symbolized by a far away resistor are your +/0 DC input connector points on your digital machine.
7. Roughly 2V/R1=ICCS will give you any current setting. Shoot for leaving at least 100mA extra for the shunt not to starve. Fine tune by actually measuring that voltage drop across R1, because vgs with current levels through Mosfet is changing, plus they got different from batch to batch, and the LEDs Vf will be about, not spot on. Don't forget Vsquared/R1 so to size the R1's wattage at about 3x for any setting. You will use remote sensing if you want best subjective and objective performance. Because even cm of wire will add enough nH, and R especially influential for error sense at such current. We look for mOhm Zo performance here, that's why we build such stuff.
Hi Salas and all,
Thanks to all the advice and brain storming ideas, I've now made tweaks to my LT1085 circuit and managed to reduce noise to a point that the sound is the same as my reference. The reference device is an Apogee Big Ben reclocker, with SMPS replaced with a LPS. The LPS is powered by a Belleson super regulator.
I replaced the R2 resistor in my LT1085 circuit with 2 green LEDs. R1 has been tweaked to provide 4.6mA of current into the LEDs. I learnt from another post by Christer that LED has the lowest noise when there is about 5mA of current (0.3-0.4uV of noise).
This small change reduced the output noise/ripple from 2mV to 0.8mV. And the sonic improvement is equaled to my Big Ben.
I also changed one of my PS filter caps from a Panasonic FC to an Elna Silmic II. This further improves the sound - better foucs, ambience, detail, depth.
It would appear that a well tuned LT1085 is really not bad at all. At least it is comparable to my super regulated reference.
Thanks to all the advice and brain storming ideas, I've now made tweaks to my LT1085 circuit and managed to reduce noise to a point that the sound is the same as my reference. The reference device is an Apogee Big Ben reclocker, with SMPS replaced with a LPS. The LPS is powered by a Belleson super regulator.
I replaced the R2 resistor in my LT1085 circuit with 2 green LEDs. R1 has been tweaked to provide 4.6mA of current into the LEDs. I learnt from another post by Christer that LED has the lowest noise when there is about 5mA of current (0.3-0.4uV of noise).
This small change reduced the output noise/ripple from 2mV to 0.8mV. And the sonic improvement is equaled to my Big Ben.
I also changed one of my PS filter caps from a Panasonic FC to an Elna Silmic II. This further improves the sound - better foucs, ambience, detail, depth.
It would appear that a well tuned LT1085 is really not bad at all. At least it is comparable to my super regulated reference.
Use Output inductors to remove the noise (ie powder-metal toriod inductors - yellow). I assume that the regulator voltage input is noisy? Use a LC filter to attenuate the input noise. If its still not quiet enough you can always go with a coupled inductor filter. You wind a parallel pair (ie Bifilar). One winding is connected between your voltage input and the load (or regulator) and the other is connected to the voltage input a low Esr ceramic cap and ground. Noise is coupled between the two windings and is filtered out with the cap consuming the noise. If the noise still isn't attenuated enough you can add a center tap to the filter winding and use two caps, to create a second order filter using the same inductor core.
Coupled inductor filter whitepaper:
http://www.hamill.co.uk/pdfs/ciabfbb_.pdf
If your input noise is load and its your load that is creating the noise then use a LC filter on the regulator output, or use the coupled inductor method. Remember Inductors are your friends!
Coupled inductor filter whitepaper:
http://www.hamill.co.uk/pdfs/ciabfbb_.pdf
If your input noise is load and its your load that is creating the noise then use a LC filter on the regulator output, or use the coupled inductor method. Remember Inductors are your friends!
Is it possible to use this circuit for 3 to 3.5 amps?
Dear Mr. Salas,
I am wanting to build a high quality regulator for the 5 volts rail of a an ATX computer supply so this thread and your contribution are very interesting.
To be safe I am wondering if you would tell me what would be required, if anything, to use this supply? I am told the computer needs around 3.5 amps at turn-on but settles to a bit below 3 Amps while playing music.
I have a large heatsink for the MOSFETs - large enough for a 50 watts amplifier.
Any help you would care to offer would be greatly appreciated.
I will alos be using the regulator for the 12 volts rails - would you be willing to tell me what to change to effect this? I am assuming the string of LED's would need to be longer - but would there be more required than that? There would be two (one for the P24 and the other for the P4 connection) and their current requirements are significantly less, less than 0.5 Amp.
Thanks,
Dear Mr. Salas,
I am wanting to build a high quality regulator for the 5 volts rail of a an ATX computer supply so this thread and your contribution are very interesting.
To be safe I am wondering if you would tell me what would be required, if anything, to use this supply? I am told the computer needs around 3.5 amps at turn-on but settles to a bit below 3 Amps while playing music.
I have a large heatsink for the MOSFETs - large enough for a 50 watts amplifier.
Any help you would care to offer would be greatly appreciated.
I will alos be using the regulator for the 12 volts rails - would you be willing to tell me what to change to effect this? I am assuming the string of LED's would need to be longer - but would there be more required than that? There would be two (one for the P24 and the other for the P4 connection) and their current requirements are significantly less, less than 0.5 Amp.
Thanks,
Follow that schematic. Its scalable. Just use 7V more Vin than Vout and your current is set by 0.65/R1. Target at least 100mA more than what the system in mind asks at peak consumption. For 12V add leds, maybe six in total.
How so? Can you name a commercial super regulator that is better than Belleson?
How about this quite inexpensive item from Texas Instruments -- TPS7A3001unmodified, in its completely natural state:
Attachments
Is the photo depicting inherent noise of the regulator?
Belleson has two characteristics that these inexpensive regulator cannot provide: high current and high noise rejection. I was using their 1A version back then and now their 2A version. The 2A version can support 3.35A easily and can be modified to support 4.6A. The noise rejection is -110db from 20Hz to 40KHz. The same cannot be said for monolithic IC regulators.
I plan to measure the Line Rejection and Output Impedance of the Belleson. In the interim, the chart on noise speaks for itself.
FWIW, the test setup for noise is the same as the 1995 TAA articles, save that the AP SYS2722 has a much lower noise floor vis a vis the System One which was used in 1995. It is helpful, however, to still use the 100X balanced amplifier and get the measurements out of the weeds.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.