Here's the sim I did with LTSpice, using the LT1117 version. The problem of increasing the C4 value is the increase in size, which would demand going SMD if you want to make an add-on with the noise eliminator parts. Actual measurements of completed circuit would solve the puzzle on why using LT317 model the PSSR improves so much, going down to -148dB between 80HZ and 300Hz. That curve improves with a small RC filter before the regulator. Certainly recommended.
Attachments
So this actually works, to some degree. Without it Vout varies by around 1V between 30C and 75C. With it I managed to get it within about 280mV between 30C-75C. For BC817-25 the diode resistor value seems to be somewhere around 2K. So far.
Curious if I can get it even tighter, I need to add a large value potentiometer to adjust Vout more easily. Adjusting the diode resistor changes Vout.
2K translates to about an even split of current between denoiser BJT and diode at about 11Vout or so.
edit: at 1.75K it's around 210mV 30C-75C
Managed to get it more or less spot on. Some under 10mV variation, with 1.55K resistor value, but this makes for a new issue, which is not so bad. There isn't much current going through the denoiser BJT. About 2.4mA measured. And 7mA going in the diode. For a total of 9.4mA through the CCS resistor.
AC still looks good in simulation, I don't think it changed much in reality. I might measure it at some point.
These are the exact values I used:
R2 was not so high, but over 1meg in my case. I think C4 should be lowered to 10uF for such a high value.
So at this point I'll clearly increase the CCS current, some 15-18mA total, I want to still have some 6-8mA through the denoiser BJT. Not sure if this will change R5 value, or Vout variation over temperature. I guess I'll find out.
This is pretty good, solves last issue with this circuit, and for only two more parts (and some extra current, maybe), although in practice it might be a resistor in series with a pot + the diode.
edit
I tested on the latest CCS NoNoiser PCB I made, removed some parts, shorted others, worked fine. But this circuit deserves its own board.
Another important aspect is that this direct coupled circuit was always stable for me, everytime I tried it on any PCB.
AC still looks good in simulation, I don't think it changed much in reality. I might measure it at some point.
These are the exact values I used:
R2 was not so high, but over 1meg in my case. I think C4 should be lowered to 10uF for such a high value.
So at this point I'll clearly increase the CCS current, some 15-18mA total, I want to still have some 6-8mA through the denoiser BJT. Not sure if this will change R5 value, or Vout variation over temperature. I guess I'll find out.
This is pretty good, solves last issue with this circuit, and for only two more parts (and some extra current, maybe), although in practice it might be a resistor in series with a pot + the diode.
edit
I tested on the latest CCS NoNoiser PCB I made, removed some parts, shorted others, worked fine. But this circuit deserves its own board.
Another important aspect is that this direct coupled circuit was always stable for me, everytime I tried it on any PCB.
Last edited:
Has anybody built a Denoisator using Diego's suggested a long time ago using fixed voltage regulators? If not, why not?
You could potentially do it but performance doesn't seem so great. But I guess you'd be able to use a red LED, or even higher Vdrop LED (max 3V or so)
You can think of LM317 as a sort of 1.25V fixed output regulator.
You can think of LM317 as a sort of 1.25V fixed output regulator.
But to use a red led all the resistors values around U6 and U7 would have to be re-calculated. Most probably the whole circuit.
In my case I want to modify an existing pcb from an MM preamp kit.
It has two pairs of fixed regulators per channel (7X18 > 7X15). I would replace the 7X18 by the 7X15, and add a Denoisator after them. It would be four Denoisators in total.
It has two pairs of fixed regulators per channel (7X18 > 7X15). I would replace the 7X18 by the 7X15, and add a Denoisator after them. It would be four Denoisators in total.
That 100R to ground trick has pretty bad performance overall. I wouldn't use it myself. But at least in simulation seems a bit better than stock. Some random 7815 model I found:
Simulation does show the diode is not required, so just getting some of the current from the CCS seems to work. I'll try it in practice see how it goes.
Thanks, Trileru.
I simmed the model here, and got close results. Less than 30dB improvement.
The original Diego suggestion got 50dB pssr improvement. But I wonder if there are other adjustments to work it out to make stable.
I simmed the model here, and got close results. Less than 30dB improvement.
The original Diego suggestion got 50dB pssr improvement. But I wonder if there are other adjustments to work it out to make stable.
But if you alter that resistor value, the output voltage increases. 100R was what Diego had used too.
Can someone point me to a schematic of a 5VDC 1.5A supply to power a DAC? It is the initial OVERSHOOT in output voltage that I am concerned about. Thanks.
What about this version?
Output 12.5v. Seems to be stable?
Can you make it misbehave? Or improve?
Output 12.5v. Seems to be stable?
Can you make it misbehave? Or improve?
Tested at higher current, 19mA total through the CCS, worked out at similar ratio, 6mA through the denoiser BJT and 13mA through the new resistor. It works, some 20-30mV drift across 30C-75C.
Worked the same for the discrete regulator, had to use IR +red combo. Red + blue was way worse. With IR in CCS and red as Vref worked at 6.5mA through denoiser and 12.5mA through resistor with same 20-30mV variation.
There is this extra current cost but it works. I can get 6-7mA through the denoiser with small temperature drift.
edit: the series diode did help me measure the pot value in circuit, with DMM.
Last edited:
This addon should work in place of the original LM317 resistors (and Cadj where the case). If tantalum caps are not microphonic. And output capacitor might have to change in some cases.
It should work from 5Vout to 30Vout, that tantalum package is common for 10-22uF with 35V rating.
Pots are for adjusting LED current, Vout and temperature compensation. There's a 1206 resistor in series with compensation pot, at 30Vout that adds up to some 0.4W, and those pots are 0.1W max rated. There would be a total of 0.6W dissipated power by this module, at 30Vout. Some thermal goop could be used to thermally couple the LED and BJTs.
This is just a concept atm, I might make and test it at some point.
It should work from 5Vout to 30Vout, that tantalum package is common for 10-22uF with 35V rating.
Pots are for adjusting LED current, Vout and temperature compensation. There's a 1206 resistor in series with compensation pot, at 30Vout that adds up to some 0.4W, and those pots are 0.1W max rated. There would be a total of 0.6W dissipated power by this module, at 30Vout. Some thermal goop could be used to thermally couple the LED and BJTs.
This is just a concept atm, I might make and test it at some point.
I may be able to reduce the wasted compensation current by adding an extra diode in the base of CCS BJT, like this:
D5 seems like it could fit in the 1.25V across the CCS. A Schottky has some 100mVdrop in those conditions. My IR LEDs have some 1.07Vdrop and simulation indicates it should work out without AC being affected much.
Some 3-4dB loss is acceptable. Applying D5 makes for about an even split of current between R1 and denoiser BJT (I tested this on my discrete test board I made). Which would be a good improvement.
This worked for my discrete supply but used a regular diode since I had the room. With a regular diode the compensation current drops to 2-3mA through the resistor but LM317 seems to take AC performance hit with a regular diode. Way too high Vf.
edit:
We could be able to recover some of the lost AC performance with an extra 10uF cap but I'm not sure it's worth it. Does seem to help the 1N4148 diode version, but I'm not sure I'll test with 1N4148. It would allow for least compensation current, 2-3mA.
D5 seems like it could fit in the 1.25V across the CCS. A Schottky has some 100mVdrop in those conditions. My IR LEDs have some 1.07Vdrop and simulation indicates it should work out without AC being affected much.
Some 3-4dB loss is acceptable. Applying D5 makes for about an even split of current between R1 and denoiser BJT (I tested this on my discrete test board I made). Which would be a good improvement.
This worked for my discrete supply but used a regular diode since I had the room. With a regular diode the compensation current drops to 2-3mA through the resistor but LM317 seems to take AC performance hit with a regular diode. Way too high Vf.
edit:
We could be able to recover some of the lost AC performance with an extra 10uF cap but I'm not sure it's worth it. Does seem to help the 1N4148 diode version, but I'm not sure I'll test with 1N4148. It would allow for least compensation current, 2-3mA.
Last edited:
Actually D5 subtracts from D4, which means I can use a red LED with a 1N4148 to base of BJT, which is the same stage from my discrete supply, which means it should be able to be compensated with low current. This gets rid of the IR LED and replaces it with a red one. I must test this.
Ok this works so so good! I got it in the 20-30mV range 30C-75C. The resistor value came out at 13.38Kohm which means 0.85mA going through it and 9.4mA going in denoiser BJT.
Red LED works fine, DC checks out. I guess red LED can be installed in previous CCS versions, the AC coupled ones, with the same 1N4148 diode trick. It drops some of the LED Vf raising CCS BJT base so it makes a sort of virtual IR LED as Vf. But also cancels out a good bit of overall negative tempco of the circuit, so the compensation trimming is very light.
It all came together quite nice, pretty sure this is the final version of my take on Elvee's Denoisator circuit:
AC performance doesn't seem affected, much.
But could bring it back to original levels with a 2.2uF between Vout and base of CCS BJT.
Red LED works fine, DC checks out. I guess red LED can be installed in previous CCS versions, the AC coupled ones, with the same 1N4148 diode trick. It drops some of the LED Vf raising CCS BJT base so it makes a sort of virtual IR LED as Vf. But also cancels out a good bit of overall negative tempco of the circuit, so the compensation trimming is very light.
It all came together quite nice, pretty sure this is the final version of my take on Elvee's Denoisator circuit:
AC performance doesn't seem affected, much.
But could bring it back to original levels with a 2.2uF between Vout and base of CCS BJT.
Is there a way to assemble this with fixed resistors? Using 3 trimpots doesn't sound very practical