This single rail power supply is an improvement over R21 power supply add-on module. It has an active rectifier, RF filter and super-regulator on a single compact PCB (120 x 70 mm). Performance is comparable to R21, except for the PSRR at high frequency, where this new supply maintains > 60 dB PSRR far into MHz range. It doesn’t have, for now, negative voltage counterpart.
It works, as is without any changes, from 10 to 60 V output voltage for a version without RF filter. Max. output voltage for a version with RF filter is 50V. Output current is up to 10 A (20 A short term), with dropout voltage of 0.15 V for 10 A load. Headroom voltage for 10 A load is 0.5V.
PSRR is greater than 110 dB @ 100/120 Hz and supply self noise is about 2 µV.
Measured performance:
1 oz copper is fine for dynamic loads up to 10A. For max performance and high current capability, 2 oz is better. All measurements were taken on 2 oz PCB version. Take a note that it may be cheaper to order 2 oz + ENIG finish PCB than just 2 oz and standard finish one, at least when JLCPCB is in question. It depends on other orders in the momentary queue, so you’ll have to check.
Parts
Worst shortages are behind us and I tried to reduce number of different semiconductors anyway. Everything is available but main heatsink is available in limited quantity. I may need to find another one soon.
C2 should be, depending on required voltage, biggest capacitance type that physically fits the PCB. Diameter up to 40 mm is OK. In the BOM, I’ve put the biggest available with 63 V rating.
For output voltages up to 30V, use at least 33.000 – 47.000 uF capacitors. Rule of thumb is that to have max.1 Vpp ripple at the voltage regulator input, we need 10.000 uF per every A of output current. As less input ripple is better, using large capacitors is better.
At 50-60 V output voltages, it would be good to replace R8 & R9 with 4K7 & 3K9 values, to avoid high dissipation and temperature.
More details about required changes at 50-60V output are at posts #245 and #246.
For output terminals, Fast-On connectors or terminal blocks with 5 mm, 5.04 mm or 6.3 mm pitch can be used.
Shared Mouser cart for one click order:
https://hr.mouser.com/ProjectManager/ProjectDetail.aspx?AccessID=b48c6947bd
Building
Proposed assembling order from the R21 build guide still applies in general. Add parts by height order, starting with diodes. However, as there are several more SMD parts, best method would be to put all SMD components first, except opamp, and then proceed with TH components. All SMD components are situated on the PCB B side. Place opamp as last or after most TH components surrounding it, has been soldered.
Murata RF filter is part to be mandatory soldered by hand. Don’t try to use hot air tools! Upper cover will melt. How do I know? 🙂
Common Q7/Q8 heatsink is required only above 30V output. At 50-60 V output, you can attach token heatsinks to Q5 & Q6 as well.
I’ve build 3 samples for verification purpose, so everything should work reliably.
Start-up behavior
Contrary to the R21 module, which was designed to ignore any amount of capacitance downstream, this regulator is not designed with driving capacitors bank weighting 10 kg in mind. Though, any decoupling capacitance in powered circuits is fine. This enabled to have regulation from the start and output voltage has slow ramp-up or soft-start, reaching nominal output only after several seconds.
Transformer secondary voltage
There is little or no difference in required transformer AC voltage, compared to a regular CRC or CLC supply. Voltage regulator part can work with full performance at only a fraction of single V drop. However, there is mains voltage variation and reservoir capacitor ripple we have to account for. So, usual voltage drop across regulator should be 1 - 1.5 V. Active rectification provides some 1.5 V higher rectified voltage than a diode rectifier. If replacing a classic power supply with this one, you are good.
For new builds you can use transformer with 1 V more at the secondary.
EDIT
08/12/24 - Added shared Mouser cart for easy parts ordering
10/12/24 - project.zip replaced with a new one because of BOM correction (just one 1800 uF capacitor less)
20/12/24 - Shared Mouser cart and Excel BOM updated with correct C6 (220 uF/35V) part. 'Wrong' one can be used as well.
12/01/25 – BOM update with correct trimmer resistor that has inline pins. ‘Wrong’ one can be still used with little contacts bending. Used the opportunity to switch LT4320 for a lower priced one as well.
Do not use BOM inside the project zip file. Always use the latest date separate BOM or link to the shared Mouser cart.
https://hr.mouser.com/ProjectManager/ProjectDetail.aspx?AccessID=b48c6947bd
11/02/25 - Recommended changes required for operation above 50V at posts #245 & #246
It works, as is without any changes, from 10 to 60 V output voltage for a version without RF filter. Max. output voltage for a version with RF filter is 50V. Output current is up to 10 A (20 A short term), with dropout voltage of 0.15 V for 10 A load. Headroom voltage for 10 A load is 0.5V.
PSRR is greater than 110 dB @ 100/120 Hz and supply self noise is about 2 µV.
Measured performance:
- 110 dB PSRR @ 100/120 Hz at 0.3 V headroom and 2 A load
- 2 µV total noise @ 2 A DC load or 15 nV/rtHz noise density
- 0.002% load regulation @ ΔI = 5 A (< 0.5 mV output voltage change for the 5 A load increase at 25 V output)
- 0.00008 %/V line regulation @ ΔV = 30 V (< 0.5 mV output voltage change for the 30 V input voltage rise at 25 V output)
- 1.4 mΩ output impedance @ 20 kHz (including output PCB tracks resistance!)
1 oz copper is fine for dynamic loads up to 10A. For max performance and high current capability, 2 oz is better. All measurements were taken on 2 oz PCB version. Take a note that it may be cheaper to order 2 oz + ENIG finish PCB than just 2 oz and standard finish one, at least when JLCPCB is in question. It depends on other orders in the momentary queue, so you’ll have to check.
Parts
Worst shortages are behind us and I tried to reduce number of different semiconductors anyway. Everything is available but main heatsink is available in limited quantity. I may need to find another one soon.
C2 should be, depending on required voltage, biggest capacitance type that physically fits the PCB. Diameter up to 40 mm is OK. In the BOM, I’ve put the biggest available with 63 V rating.
For output voltages up to 30V, use at least 33.000 – 47.000 uF capacitors. Rule of thumb is that to have max.1 Vpp ripple at the voltage regulator input, we need 10.000 uF per every A of output current. As less input ripple is better, using large capacitors is better.
At 50-60 V output voltages, it would be good to replace R8 & R9 with 4K7 & 3K9 values, to avoid high dissipation and temperature.
More details about required changes at 50-60V output are at posts #245 and #246.
For output terminals, Fast-On connectors or terminal blocks with 5 mm, 5.04 mm or 6.3 mm pitch can be used.
Shared Mouser cart for one click order:
https://hr.mouser.com/ProjectManager/ProjectDetail.aspx?AccessID=b48c6947bd
Building
Proposed assembling order from the R21 build guide still applies in general. Add parts by height order, starting with diodes. However, as there are several more SMD parts, best method would be to put all SMD components first, except opamp, and then proceed with TH components. All SMD components are situated on the PCB B side. Place opamp as last or after most TH components surrounding it, has been soldered.
Murata RF filter is part to be mandatory soldered by hand. Don’t try to use hot air tools! Upper cover will melt. How do I know? 🙂
Common Q7/Q8 heatsink is required only above 30V output. At 50-60 V output, you can attach token heatsinks to Q5 & Q6 as well.
I’ve build 3 samples for verification purpose, so everything should work reliably.
Start-up behavior
Contrary to the R21 module, which was designed to ignore any amount of capacitance downstream, this regulator is not designed with driving capacitors bank weighting 10 kg in mind. Though, any decoupling capacitance in powered circuits is fine. This enabled to have regulation from the start and output voltage has slow ramp-up or soft-start, reaching nominal output only after several seconds.
Transformer secondary voltage
There is little or no difference in required transformer AC voltage, compared to a regular CRC or CLC supply. Voltage regulator part can work with full performance at only a fraction of single V drop. However, there is mains voltage variation and reservoir capacitor ripple we have to account for. So, usual voltage drop across regulator should be 1 - 1.5 V. Active rectification provides some 1.5 V higher rectified voltage than a diode rectifier. If replacing a classic power supply with this one, you are good.
For new builds you can use transformer with 1 V more at the secondary.
EDIT
08/12/24 - Added shared Mouser cart for easy parts ordering
10/12/24 - project.zip replaced with a new one because of BOM correction (just one 1800 uF capacitor less)
20/12/24 - Shared Mouser cart and Excel BOM updated with correct C6 (220 uF/35V) part. 'Wrong' one can be used as well.
12/01/25 – BOM update with correct trimmer resistor that has inline pins. ‘Wrong’ one can be still used with little contacts bending. Used the opportunity to switch LT4320 for a lower priced one as well.
Do not use BOM inside the project zip file. Always use the latest date separate BOM or link to the shared Mouser cart.
https://hr.mouser.com/ProjectManager/ProjectDetail.aspx?AccessID=b48c6947bd
11/02/25 - Recommended changes required for operation above 50V at posts #245 & #246
Last edited:
PSRR
It is a little better than with R21 at lower frequencies, but RF filter contributes to big difference at HF. Filter effect starts to show above 25 kHz. As regulator PSRR is falling at 6 dB/octave slope and filtering is increasing by the same, we get flat PSRR from that point. Checking filter datasheet, PSRR could remain at 60 dB up to GHz frequency.
Measurements are a little wiggly above 100 kHz, as voltage regulator input, at those frequencies, presents an almost short circuit to the signal source.
It is a little better than with R21 at lower frequencies, but RF filter contributes to big difference at HF. Filter effect starts to show above 25 kHz. As regulator PSRR is falling at 6 dB/octave slope and filtering is increasing by the same, we get flat PSRR from that point. Checking filter datasheet, PSRR could remain at 60 dB up to GHz frequency.
Measurements are a little wiggly above 100 kHz, as voltage regulator input, at those frequencies, presents an almost short circuit to the signal source.
Noise
Self generated noise from the supply is only 2 uV in the 20 Hz – 20 kHz bandwidth. With ordinary transformer and no additional mains filtering, you can expect total noise to be less than 10 uV. In example, using crap transformer and everything close and open on the bench, I’ve measured < 5 uV.
Self generated noise from the supply is only 2 uV in the 20 Hz – 20 kHz bandwidth. With ordinary transformer and no additional mains filtering, you can expect total noise to be less than 10 uV. In example, using crap transformer and everything close and open on the bench, I’ve measured < 5 uV.
Transient response
Measurements were taken with high and fast changing load current, with rise time of 10 us, which corresponds to fastest possible audio signal. 20 A transients are ‘nothing’ to this supply.
I’ve included output measurements with FFT analysis to check for harmonics and output modulation by a real amplifier load. First, single 1 kHz tone to check for harmonics introduced by supply. Next, 31 tone with pink power distribution, which is similar to the real music signal.
There is nothing of concern. Output modulation by load, for any power supply, is a simple product of load current and supply output impedance at specific frequency, and harmonics are a product of supply non linearity. We have here harmonics at residual levels of used equipment (LNA and ADC distortion).
I’m sure that everyone has heard of power supply or capacitor sound. Power supply can affect powered circuit output by noise and by not keeping its output constant at any load and frequency. Powered circuit has its own isolation factor between supply and output (PSRR) but it may be low for some circuits.
This supply is in practice pretty close to the ‘ideal power source’.
Measurements were taken with high and fast changing load current, with rise time of 10 us, which corresponds to fastest possible audio signal. 20 A transients are ‘nothing’ to this supply.
I’ve included output measurements with FFT analysis to check for harmonics and output modulation by a real amplifier load. First, single 1 kHz tone to check for harmonics introduced by supply. Next, 31 tone with pink power distribution, which is similar to the real music signal.
There is nothing of concern. Output modulation by load, for any power supply, is a simple product of load current and supply output impedance at specific frequency, and harmonics are a product of supply non linearity. We have here harmonics at residual levels of used equipment (LNA and ADC distortion).
I’m sure that everyone has heard of power supply or capacitor sound. Power supply can affect powered circuit output by noise and by not keeping its output constant at any load and frequency. Powered circuit has its own isolation factor between supply and output (PSRR) but it may be low for some circuits.
This supply is in practice pretty close to the ‘ideal power source’.
Attachments
Output impedance
Output impedance includes PCB tracks to the output and solder + Fast-On connectors resistance. There are two measurements. One at the copper direct at output terminals and another on top of soldered Fast-On connectors.
Output impedance includes PCB tracks to the output and solder + Fast-On connectors resistance. There are two measurements. One at the copper direct at output terminals and another on top of soldered Fast-On connectors.
No, it can’t be done that way and I don’t see any reasonable way to add RF filter to the R21 module.
Thank you.
Unfortunately, final PCB is not rectangular but with rounded edges. 😛
Unfortunately, final PCB is not rectangular but with rounded edges. 😛
Very nice work as usual Tombo. Are you going to do a Neg regulator - or can 2 of these Pos regulators be tied together to give a dual +/- DC supply?
Separate negative voltage version is not to be expected soon or maybe anytime. Instead, I may make dual rail (+/-) version on a single PCB. I don’t need another dual rail supply at the moment, but that can change after several months.
Two identical supplies were tested as a symmetrical +/- dual rail supply for my amplifier and everything performed perfect.
Two identical supplies were tested as a symmetrical +/- dual rail supply for my amplifier and everything performed perfect.
Last edited:
Tombo, what is recommended max capacitance value that could be added after the reg output. Thinking about a dual +/- dual 24VDC supply for class A amp.
Ordered boards (in green, 1 oz copper). If anyone has trouble ordering such stuff please send me a PM. EU only.
Have also ordered pcb's for this new PSU.
Tombo, could you update this thread for the important design considerations when selecting the mains transformer for a given output voltage - say use 24VDC as an example ( a common rail voltage for many Pass class A amps). Also include the criteria to observe when adjusting the trimpot for the required output voltage and any other important points that should be observed when setting up your new PSU.
Also addressing my question in post #15 above would help many builders of the PSU.
Many thanks.
Gary.
Tombo, could you update this thread for the important design considerations when selecting the mains transformer for a given output voltage - say use 24VDC as an example ( a common rail voltage for many Pass class A amps). Also include the criteria to observe when adjusting the trimpot for the required output voltage and any other important points that should be observed when setting up your new PSU.
Also addressing my question in post #15 above would help many builders of the PSU.
Many thanks.
Gary.
Whatever is required by amplifier design for local decoupling is fine, even if it’s 10.000 uF. There is no need to use additional capacitors at supply output. Regulator part doesn’t have any stability problems with large capacitors at the output, but very large capacitance unnecessary increases thermal load on MOSFET during start-up.
Just as in normal power supplies really. Normally normal regulated power supplies do not benefit from extremely large value output capacitors. For the ultra low drop out voltage as exhibited one may use 1:1 AC for DC voltage ratio for low voltage application so 12V AC for 12V DC isn't it? Will work out to about 22V I guess but always with 22,000 µF to keep it simple and depending on the transformers VA rating and load current too of course. Probably too difficult to describe for all possible situations especially if the builder uses 10,000 µF or the like. Likely to be determined per situation by the builder.
The challenge goes both ways as a too high voltage conflicts with the PSU being and uLDO and unnecessary loss of power to useless heat. For instance an 18V AC transformer for 12V DC 3A output voltage with a 0.15V dropout voltage uLDO may not be the best choice.
The challenge goes both ways as a too high voltage conflicts with the PSU being and uLDO and unnecessary loss of power to useless heat. For instance an 18V AC transformer for 12V DC 3A output voltage with a 0.15V dropout voltage uLDO may not be the best choice.
Last edited:
Will add today some guidance about selecting transformer voltage. For class A, with good toridal transformer at 1/3 nominal load, we will get DC voltage at about secondary AC x 1.27 and add 1.5V on top of that for active rectification. Total voltage should be some 1.5 – 2 V more than what you want at the supply output.