As Mark said.
Parasitics can influence parameters like Ft, but certainly not the noise
Parasitics can influence parameters like Ft, but certainly not the noise
CCS in schematic at post #3052.
The difference it makes at 100 Hz 1000 Hz on PSRR
R1=220 Ohm
Q1 collector load = CCS dynamic impedance = 1 Meg Ohm ( typical )
No CCS Q1 collector load = 1.8 K Ohm ( typical )
To make it simple ( but not simplistic.)
220 // 1000000 VERSUS 220 // 1800
About Zilch !
PSRR at 100 Hz 1000 Hz ......+ 1.14 dB
The difference it makes at 100 Hz 1000 Hz on PSRR
R1=220 Ohm
Q1 collector load = CCS dynamic impedance = 1 Meg Ohm ( typical )
No CCS Q1 collector load = 1.8 K Ohm ( typical )
To make it simple ( but not simplistic.)
220 // 1000000 VERSUS 220 // 1800
About Zilch !
PSRR at 100 Hz 1000 Hz ......+ 1.14 dB
Last edited:
But it's not only about PSRR. If you need more PSRR just add a capacitance multiplier. Or make the CCS NoNoiser.
Here's some of my previous measurements, CCS Denoisator, on the circuit you mention vs regular Denoisator
PSRR (input and output traces) BC337
120dB, not stellar but pretty good.
Here's the noisefloor
Here's the PSRR for the regular Denoisator:
And the noisefloor of regular Denoisator, at least according to my measurements
And since you were also talking about Diego's circuits, here's his Dienoiser PSRR performance
And noisefloor
Here's some of my previous measurements, CCS Denoisator, on the circuit you mention vs regular Denoisator
PSRR (input and output traces) BC337
120dB, not stellar but pretty good.
Here's the noisefloor
Here's the PSRR for the regular Denoisator:
And the noisefloor of regular Denoisator, at least according to my measurements
And since you were also talking about Diego's circuits, here's his Dienoiser PSRR performance
And noisefloor
You can also hotrod your CCS Denoisator BJT at 20mA to get some 3dB extra PSRR:
And noisefloor at 20mA:
I never even tried 20mA on regular Denoisator, but pretty simple to do with the CCS.
And noisefloor at 20mA:
I never even tried 20mA on regular Denoisator, but pretty simple to do with the CCS.
Forgot to mention that these measurements were made with LM317-N regulator, from TI
https://www.ti.com/product/LM317-N
The N version has better stock PSRR. Seems about double the price of regular LM317, worth it imo
https://www.ti.com/product/LM317-N
The N version has better stock PSRR. Seems about double the price of regular LM317, worth it imo
What is it ?
I do not need these ways to improve PSRR
My PSU regulator designs do:
PSRR -120 dB without denoisering. Using a TL431 plus 2 capacitors plus 6 resistors plus 4 BJT
PSRR -165 dB adding an Elvee 1 transistor denoisering. Using 1 BJT plus 2 capacitors plus 4 resistors.
-165 dB proven stable: Gain margin 12dB, phase margin 60°.
Last edited:
Are you sure? What was measured PSRR of those designs?
I had no problems pushing simulation of regulator design to 200 dB PSRR.
Measured performance of real circuit was something entirely different. 😉
@mchambin
Output impedance. Let alone noisefloor.
If you really want even less parts you can omit the two LM317 stock resistors and short the 220uF cap and get this performance, if you can live with its tempco.
Real measurements for the right side circuit:
And if we part count considering a Cadj application with a regular status LED (but it's IR in this case) then it's only two BJTs a 3.3nF ceramic and a 1N4148 diode extra. And not sure the 3.3nF cap is required, I just added it for good measure but worked without it.
Output impedance. Let alone noisefloor.
Can we see your measured noisefloor?
If you really want even less parts you can omit the two LM317 stock resistors and short the 220uF cap and get this performance, if you can live with its tempco.
Real measurements for the right side circuit:
And if we part count considering a Cadj application with a regular status LED (but it's IR in this case) then it's only two BJTs a 3.3nF ceramic and a 1N4148 diode extra. And not sure the 3.3nF cap is required, I just added it for good measure but worked without it.
Last edited:
Check the voltage at the LM317 Output of the right circuit, the one with CCS and R19 R20 C20 removed.
Come on. There is NO voltage regulation. Regulation is lost, removing Adj R19 R20 from the usual way to use LM317
Change the load to see what is the voltage regulation. Change R26 to see what happens to the output voltage.
I am talking about the two circuits at post 3068.
Do not confuse the left circuit that is OK, with the right circuit ( the one with CCS ) that is IMO wrong.
Change the load to see what is the voltage regulation. Change R26 to see what happens to the output voltage.
I am talking about the two circuits at post 3068.
Do not confuse the left circuit that is OK, with the right circuit ( the one with CCS ) that is IMO wrong.
It is a terrible way to set the LM317 Adj pin voltage. There is no serious voltage regulation.
What do you mean? Here's some 60mA and 600mA (RMS) active loads.
And a resistive load, 100mA vs 1.5A
And a resistive load, 100mA vs 1.5A
Last edited:
I was trying to see if I can get a better tempco for this circuit and I noticed this arrangement does seem to make it better. But the resistor value should be tweaked. I put it on the 2nd and 3rd circuit tied to CCS BJT emitter.
For circuit number 3 all three lines are superimposed on the graph. For 25C/50C/75C.
Although this isn't visible with the LM317-N spice model.
For circuit number 3 all three lines are superimposed on the graph. For 25C/50C/75C.
Although this isn't visible with the LM317-N spice model.
Last edited:
Actually it seems to also work with the LM317-N model, tried it with LT317A one and it also works, but with the Nexperia BC8x7 models. Doesn't seem to work with stock BC3x7 ones, at least for denoiser position.
Not sure if this is an artifact of the spice models or would work out in practice.
Not sure if this is an artifact of the spice models or would work out in practice.
Last edited:
Also works with the MPSAx6 models I have.
It does seem to eat a good portion of the total current for some configurations, some 4mA in this case. AC doesn't suffer much.
At first I thought it just leeches some of the current from the denoiser but it does something extra. I tried setting the same resulting current for the denoiser, from the CCS resistor, and tempco doesn't improve.
This doesn't work with the stock BC3x7/BC8x7/ZTXx51 models. Are they missing thermal stuff from the models?
Nexperia BC8x7 model does specifically mention it includes thermals.
It does seem to eat a good portion of the total current for some configurations, some 4mA in this case. AC doesn't suffer much.
At first I thought it just leeches some of the current from the denoiser but it does something extra. I tried setting the same resulting current for the denoiser, from the CCS resistor, and tempco doesn't improve.
This doesn't work with the stock BC3x7/BC8x7/ZTXx51 models. Are they missing thermal stuff from the models?
Nexperia BC8x7 model does specifically mention it includes thermals.
Last edited:
Following closely what Trileru is developing, I bring another interesting alternative to try, as opposed to the CFP option and the direct coupled option. It keeps the original parameters of the regulator intact and reflects an important PSRR, while allowing the initial connection overshoot to be tamed more easily (with the convenient choice of a second capacitor, marked as C4).
A parameter sweep was performed in relation to R2. The sweep was between 470 and 2200 ohms, in steps of 100 ohms. In general, good PSRR range is observed, with minimal output voltage overshoots after the first connection. Since the circuit was mostly tuned for approximately 16 volts output, there are no visible overshoots at that voltage. Only very small surges as we move away from that 16 volts it was tuned for. Anyway, the circuit seems to adapt quite well to a wide range of operation. We just need to physically implement it and know its real performance. One way to adjust the overshoot would seem to be by increasing the value of C4. A small increase in C4 would seem to be necessary for other operating points, without major impact on other parameters of interest, very different if we adjusted only C1 (in which case a significant increase would be necessary, also impacting other parameters).
