PSRR, power supply rejection ratio, is a pretty important parameter for a regulator.
This test was done with 21V input and 15V output in SPICE. The load was 100mA.
At the input was a signal 1kHz 300mVp-p.
The level of this signal was recorded at the output. The difference is expressed in dB.
The Super Regulator is outstanding according to my test.
LT1085 and LM7815 had somewhat poor PSRR at 1kHz.
LT1085 has a little better result at 100Hz
Result PSRR at 1kHz
142dB - Super Regulator, see image 1
125dB - My TL431 Regulator, see image 2
80dB - LM317
66dB - LM7815
62dB - LT1085
This test was done with 21V input and 15V output in SPICE. The load was 100mA.
At the input was a signal 1kHz 300mVp-p.
The level of this signal was recorded at the output. The difference is expressed in dB.
The Super Regulator is outstanding according to my test.
LT1085 and LM7815 had somewhat poor PSRR at 1kHz.
LT1085 has a little better result at 100Hz
Result PSRR at 1kHz
142dB - Super Regulator, see image 1
125dB - My TL431 Regulator, see image 2
80dB - LM317
66dB - LM7815
62dB - LT1085
Attachments
How about also:
I think you as a designer must take a holistic responsibility of ones designs and describe the complete set of characteristics parameters. Bringing one out of several is not serious.
Check these cars out... the blue car has 100 hp more than the red!
OK - so which one is the "winner"?
//
- Noise (vs freq)
- Output impedance (vs. freq)
- Reactive load sensitivity (vs. freq)
- Temperature stability
I think you as a designer must take a holistic responsibility of ones designs and describe the complete set of characteristics parameters. Bringing one out of several is not serious.
Check these cars out... the blue car has 100 hp more than the red!
OK - so which one is the "winner"?
//
Super regulator.
I see this one uses a AD825, a FET input op amp. This seems overkill to me, the two input signals are low impedance.
Furthermore the AD825 is 66dB open loop gain ( x2000 gain ), that is unusual.
Is there a good reason for these caracteristics ?
It seems to me, a NE5532 would perform about as well and is way less pricey.
In case it's open loop gain is too high, it only takes a feed back resistor ( in- to out ) to adjust it's gain at x2000, more or less.
I think i have used this power supply topology in the 1970s copying bluntly a design given to me, that worked so well, i did not look close into it.
I see this one uses a AD825, a FET input op amp. This seems overkill to me, the two input signals are low impedance.
Furthermore the AD825 is 66dB open loop gain ( x2000 gain ), that is unusual.
Is there a good reason for these caracteristics ?
It seems to me, a NE5532 would perform about as well and is way less pricey.
In case it's open loop gain is too high, it only takes a feed back resistor ( in- to out ) to adjust it's gain at x2000, more or less.
I think i have used this power supply topology in the 1970s copying bluntly a design given to me, that worked so well, i did not look close into it.
Back to the original post.
This TL431 has been my favorite PSU for many years.
This IC is a lot for little money: Inside it has a 2.5 volt voltage reference, an op amp and a low/medium power bjt. That is the heart of most regulators, like the super reg where you see these three seperate with better performance but way more pricey.
I had a go at this TL431 based regulator for a 32V 400mA PSU and a 48V 120mA PSU.
I use the same schematic, but prefer a CCS made of two bjts ( ring configuration ) with two resistors. This gives over 10 MegOhm dynamic resistance which is way more than needed, it is less temperature stable than the red LED ccs but acceptable here, where we do not need a very accurate current value.
My 48V PSU is about the same schematic, the only difference is one more TL431 and a resistor. The two TL431 share the 48v working each at 24v.
My 32V PSU.
I removed C1 the 22uF accross the TL431. This cap makes the regulator unstable, one can see it in simulation when applying transient loading.
To simulate this I add a voltage source at my 80 Ohm load resistor to make a 3V pulse, so I can see what goes on with the load going 400mA/440mA/400mA
With no cap, PSSR @1kHz 80dB
C2 the 22uF accross Vout Vref enhances this PSSR@1kHz to 102dB.
Without the cap, Vout is applyed at Vref by a resitor divider that has a ratio 2.5/32. With the cap, Vout is applyed at Vref direcly at AC. This PSSR enhancement at AC is consistent with 32/2.5.
My dBs are consistent with the original post where the PSU is 15V 100mA, while my PSU is 32V 400mA
Instead of the C2 22uF one can use a denoiser at Vref.
With a one transistor denoiser, I get PSSR@1kHz 146 dB.
We cannot add the two enhancements, it is one or the other. Using both, they would fight each other.
This TL431 has been my favorite PSU for many years.
This IC is a lot for little money: Inside it has a 2.5 volt voltage reference, an op amp and a low/medium power bjt. That is the heart of most regulators, like the super reg where you see these three seperate with better performance but way more pricey.
I had a go at this TL431 based regulator for a 32V 400mA PSU and a 48V 120mA PSU.
I use the same schematic, but prefer a CCS made of two bjts ( ring configuration ) with two resistors. This gives over 10 MegOhm dynamic resistance which is way more than needed, it is less temperature stable than the red LED ccs but acceptable here, where we do not need a very accurate current value.
My 48V PSU is about the same schematic, the only difference is one more TL431 and a resistor. The two TL431 share the 48v working each at 24v.
My 32V PSU.
I removed C1 the 22uF accross the TL431. This cap makes the regulator unstable, one can see it in simulation when applying transient loading.
To simulate this I add a voltage source at my 80 Ohm load resistor to make a 3V pulse, so I can see what goes on with the load going 400mA/440mA/400mA
With no cap, PSSR @1kHz 80dB
C2 the 22uF accross Vout Vref enhances this PSSR@1kHz to 102dB.
Without the cap, Vout is applyed at Vref by a resitor divider that has a ratio 2.5/32. With the cap, Vout is applyed at Vref direcly at AC. This PSSR enhancement at AC is consistent with 32/2.5.
My dBs are consistent with the original post where the PSU is 15V 100mA, while my PSU is 32V 400mA
Instead of the C2 22uF one can use a denoiser at Vref.
With a one transistor denoiser, I get PSSR@1kHz 146 dB.
We cannot add the two enhancements, it is one or the other. Using both, they would fight each other.
Last edited:
Back to the original post: TL431 based regulator.
It shows an output ripple 162nV peak to peak.....
This is possible ( calculation wise ) because of an assumed input ripple 150mV peak.....This can only be with huge reservoir caps ahead or a cap multiplier.
For my 32V 400mA, I plan on 2Vp-p at the input of the regulator. With the PSSR option 102dB, this will give below 20uV at the output.
To achieve 2Vp-p the needed reservoir capacity is ( Load current / Ripple p-p ) x 7000uF.....That gives (0.4 / 2)*7000 uF = 1400uF.
So, I’ll will use 2 x 1000uF 63V caps that should give me near 10 micro Volt peak to peak ripple at the output of this regulated PSU.
It shows an output ripple 162nV peak to peak.....
This is possible ( calculation wise ) because of an assumed input ripple 150mV peak.....This can only be with huge reservoir caps ahead or a cap multiplier.
For my 32V 400mA, I plan on 2Vp-p at the input of the regulator. With the PSSR option 102dB, this will give below 20uV at the output.
To achieve 2Vp-p the needed reservoir capacity is ( Load current / Ripple p-p ) x 7000uF.....That gives (0.4 / 2)*7000 uF = 1400uF.
So, I’ll will use 2 x 1000uF 63V caps that should give me near 10 micro Volt peak to peak ripple at the output of this regulated PSU.