BluesPowerMini: Dual Shunt Regulator power supply

[doors666]

This is a monster dual shunt regulated power that I made. Its quite flexible in what it can do. it can use small diodes or the larger to220 ones. It has 25mm caps and a CRC filter at the input with 2*2W resistors, so a lot of voltage can be dropped here. Start with the best foot forward and give a relatively clean supply to the regulator.

All small transistors are 2sc and 2sa series, so 2sc2240, 2sc1815 etc and their 2sa pnp counterparts can be used. These are better than the bc series transistors (ok, you can shout now). The fets are IRF510 and IRF9510.

It uses three zener diodes to set the output voltage, so a lot of flexibility is there with that. Zener voltage is filtered with both a film cap and a electrolytic cap and then its buffered before using it to set the voltage. This build sets the voltage to about 16.5v and runs at about 300ma. Of course the heatsinks are not enough for that without any load.

It has 4 output points per dc rail, so more than one project can be powered from this, like a preamp, phono pre and headphone amp etc. Separate connection for earthing is also provided. The pcb is quite large at 4"x5".

I have also a BluesPowerMini coming out soon which is a much smaller board with less features and more suitable for a single project.

Some Pictures:

 

[doors666]

Since I have a LM317/LM337 dual regulated power supply also, I decided to do some comparisons in the simulator.

its basically designed for preamps, so I use the standard 100ma DC load with about 16.5v output, suitable for 18v opamps, also useful for other supplies and numbers will probably be similar for other voltages too. I will test only +ve regulator. AC Load will be about 10ma RMS (lets use 15ma Peak) at 1khz and 20khz. A preamp driving a 10k amp load with 2V RMS voltage needs only about 0.25ma AC and a little more for the cable capacitance.

Its configured for using a 18v transformer, Ripple is 3v peak to peak at 100hz, so the input voltage to the regulator will be about 25v.

I will run the shunt regulator at 195ma using a 3.6R resistor in the CCS. There is no equivalent parameter for LM317. Of course one can run the shunt regulator at higher currents and get better performance, with heatsinking and wasted power being the additional cost.

Things to measure:

1. Ripple at 100hz
2. Response at the AC Load frequency, at 1khz and 20khz.
3. Output Impedance.

Adjusted R8 on LM317 so that the voltage outputs are close enough and can be shown correctly in LTSpice. This would be done in real life using the pot in the LM317 supply.

in the first installment, I will use a basic Lm317 schematic and a basic shunt regulator, no CRC filter at the input. Only a 4700uf Cap with 35mR ESR.

The schematics used for both the circuits:

Results:

Response at 1KHZ

Ripple:

A Closer look at the ripple:

Now lets look at the fft for a 1khz load:

LM317 with 1khz Load:

Ripple at 100hz = -73.5db
Ripple at 200hz (2nd harmonic) = -81db
Ripple at 300hz (3rd harmonic) = -86.5db
Ripple at 900hz (9th harmonic) = -109db

Even the 9th harmonic is quite high.

1KHZ response = -87db
2khz response (2nd harmonic) = -114db
3khz response (3rd harmonic) = - 123db

Noise Floor = looks about -170db

Shunt Regulator with 1khz Load:
Ripple at 100hz- -162db
Ripple at 200hz (2nd harmonic) = -170db
Ripple at 300hz (3rd harmonic) = -176db

1KHZ response = -111db
2khz response (2nd harmonic) = -142db
3khz response (3rd harmonic) = - 146db

Noise Floor about -210db

The shunt regulator does a lot better for 100Hz ripple, 162-73.5=88.5db better.
For the 1khz response, its 111-87=24db better.
Noise floor is about 40db better.

Lets check the output impedance:

Both look quite similar upto 1khz, but after that, LM317 starts deteriorating and then there is a huge spike in the high frequencies, does not improve et al. For the shunt regulator, it stays very low till 100khz, where the graph ends. Comparatively, the output impedance for LM317 looks absolutely horrendous and I think this is something which additional input or output filters cannot improve, only degrade.

Now lets check the response at 20khz.

At 20khz, it looks worse than 1khz for both the regulators, but that is expected.

Ripple numbers are same for 1khz load or 20khz load for both the regulators.

LM317 Response at 20khz = -69db
Shunt Regulator respose at 20khz Load: = -96db

As you can see, it has become worse for both than at 1khz, but shunt regulator still has a lead of 96-69=27db at 20khz.

So the shunt regulator thumps the LM317 regulator on all parameters, and with big margins.

Next Lets see what happens if we add a CRC filter at the input of both the regulators. Tomorrow.....

 

[doors666]

Next, I add a CRC filter at the regulator input supply instead of the usual C filter. Since I used a 4700uf cap earlier, I didnt want to stray too far from that value, so I split that into two 2200uf caps, effectively reducing the total cap value. I used a 10R resistor for the filter. You can increase the resistance value and get better performance, then you will need more ac voltage and more heat dissipation and a larger resistor. With 10R and 100ma current, the resistor will drop only 1v. I use the same values for both the regulators.

Please note that the frequency response at 1khz and 20khz remains unaltered due to this. The output impedance also does not change, so you can look at the previous post's images for that. The only thing that is affected is the ripple voltage.

Ripple (Green is LM317, red is shunt regulator)

FFT - (upper is shunt and lower is LM317)

LM317 with 1khz Load:
Ripple at 100hz = -91.5db
Ripple at 200hz (2nd harmonic) = -105.5db
Ripple at 300hz (3rd harmonic) = -115.5db
Ripple at 900hz (9th harmonic) = -144db

Shunt Regulator with 1khz Load:

Ripple at 100hz- -175db
Ripple at 200hz (2nd harmonic) = -183db
Ripple at 300hz (3rd harmonic) = -181db

Ripple is not affected by the load frequency, so no point showing images for 20khz. Shunt regulator still wins, at 100hz its 175-91.5=83.5db. Shunt regulator's first harmonic is better than LM317's 9th harmonic.

 

[doors666]

I had not specified ESR for the output cap. This affects the output impedance. Adding a 200mR ESR to the 470uf cap on the output of both the regulators shows the following graph for output impedance. Now the huge notch is gone. Shunt still does a lot better, output impedance being about 2.5mR for shunt and 80mR for LM317 at 20khz. After 2Mhz, LM317 becomes better than shunt, way beyond the audio band.

 

[doors666]

Finally the board for the mini version of the BluesPower Shunt regulator pcb arrived. This one is basically a toned down version of its big brother. Its pretty much the same circuit that is being used in BumbleBee preamp.

Main differences with the BluesPower:
1. No CRC filter at the input.
2. Single zener to set the voltage. Blues power supports 3 zeners.
3. Less output pads. 2 per rail and 3 for ground (1 can be used for earthing). BluesPower supports 4 per rail and ground and another for earthing.
4. No buffer for the zener voltage, drives the gain stage directly.
5. Only film cap for zener filter, you can fit an electrolytic there if you want, but not both unless you use bottom of the board.
6. Second ccs doesnt have a split resistor and a cap, but just a resistor.
7. Doesnt support the flowery heatsinks. You need flat heatsinks. Might need insulation.
8. All of the resistors supported are 0.25w, so not really suitable for high voltage builds like for tube gear, those need 2w resistors that are supported by BluesPower.
9. This uses the more common BC550 and BC560 small signal transistors.
10. Single indicator LED, tied to the +ve rail only. I should have connected this to both the rails.

The power fets are at the edge of the board and you can use any flat heatsink. For larger currents, the TO-3 heatsinks are will be a good choice. You will need to insulate atleast one of them. In my build, the ccs fet dissipates less heat, so I have used a smaller heatsink for that and insulated that. The heatsinks are anodized anyway, so they are non conducting, but I insulated anyway just to be safe.

This build I am running at about 75ma. With the teensy heatsinks shown, it runs barely warm without load (which is the most heat dissipating), but keep in mind that I am in bangalore and its weather has been quite cool these past few days. This is suitable for two dual opamps like BBS preamp or CNC phono or any other one or two opamp project. Of course you can run it at higher currents too (e.g. for a headphone amp), you just need to provide suitable heatsinks. Apart from the power fets, nothing else will dissipate any significant heat.

Unfortunately, I ended up using different caps for the output by mistake, I have both ky and kze of same value in the same bag and just soldered them without checking. Need to change that.

BTW, neither of these regulators use any jfets, so that headache is not there. Although if you need matched output voltages for both the rails, you will need to match the zeners. Its quite easy to do that, just need a larger voltage supply than the zener voltage and a suitable resistor. You can also match them in circuit, solder two wires for zener and connect the zeners to them (just tie the wire to zener or use a breadboard) and measure output voltages till you find matching ones.

I used 16v zeners. These are 5% tolerance, After trying 5, I got 2 that had a voltage of 16.58v. These set the output voltage to 17.24v exactly. The output voltage is zener voltage + 0.65v. Perfect for those 18v opamps. Transformer is 18-0-18v. Both my dmm and oscilloscope failed to detect any ripple. I need to use soundcard and RMAA to measure it. I am going to replace the LM317+CRC power supply from my BBS preamp as well as in my basic B1 build.

Pictures:

Partly soldered boards for clarity

Completely soldered board, 9v battery is for comparison. I have the berg connectors for the led for the time being.

Simulations show ripple at about -159db. the load is 40ma DC and 10ma RMS AC at 1khz. output impedance at 1khz is 1.5 milliohm, at 20khz it is 2.2 milliohm, at 100khz it is 8.3 milliohm. That is about a 1000 times better output impedance than the standard 9v batteries which I believe have an output impedance of about 3 ohms.

I am going to make kits soon for this one. Let me know if anyone is interested.

 

[doors666]

Size comparison of the two supplies::lol:

 

[doors666]

Blues Power Mini Shunt Regulator Kit

Blues Power Mini Shunt Regulator Kit

Finally the kits for the Blues Power Mini builds are ready. Few more days for the Blues Power big shunt regulator kits.

The schematic, BOM, Build Guide etc and images are on the site here.

Some details on the parts:

Bulk caps ELNA RKD 2200uf 35v 125C
Mosfet for shunt elements IRF510 and IRF9540
CCS Transistors TIP31C and TIP32C

TO92 transistors Philips BC550C and Fairchild BC560B
0.1uf 63v epcos polyester caps

Split resistor cap for main ccs Nichicon VR 220uf 35v
Output caps These should not be too high in value, keep them at 470uf max. Dont put caps with too low ESR. Max cap dia is 10.5mm. I have the following options available. Please let me know which ones you want.

Nichicon PJ 330uf 25v 105C
Panny FC 105C 470uf/25v, 330uf/35v, 220uf/35v
UCC LXZ 330uf/25v 105C
ELNA RJJ 100uf 35v 105

You also need to specify the zener diode voltage. Your output voltage will be about 0.65v more than the zener voltage. Zeners have 5% tolerance. You need to match the zeners if you want same output voltage on both the +ve and -ve rail. I will provide 5 zeners so that matching ones can be chosen if needed. I have quite a few zeners, 7.5, 8.2, 9.1, 10, 12, 15, 16, 18, 20, 24 etc.

For setting the total ccs current, two resistor values are provided 8.2R and 10R. Each value has 5 pieces for matching the currents. The first will give you about 80ma of current and the second about 65ma. Leaving aside about 12-13ma for the shunt regulator and about 10ma for the shunt element, you will get about 40ma and 55ma max load current. These should be ok with upto 3 dual opamps, assuming you are not using a particularly current hungry one. If you need more current, let me know. Heatsink needs to be correctly sized though. Small commercial heatsinks like I used in the pic in the prev post (the larger of the two) can only handle the 10R 65ma. With 8.2R 80ma, in my setup, it was at 50C after couple of hrs of use with the BBS preamp as the load. Dual TO3 heatsinks will be a good idea for the 80ma or slightly higher currents.

Let me know if any one is interested or if you need help configuring it.

Kit Picture:

 

[doors666]

finally got a decent heat sink for this supply, quite cheap too. its the dual TO-3 heatsink that fits perfectly.

Now it running with a 6R current setting resistor. I just soldered a 22R at the bottom of the existing 8.2R, lazy guy that I am..... Its running at about 108ma shunt current now. The heatsinks are still quite cool, the temperature went up from about 26C to 41C, giving an increase of about 15C. I think I can easily use a 5.6R resistor with this, which will give me about 115ma total shunt current with a little bit more heat dissipation.

 

[doors666]

Finally managed to get some respectable heatsink for this bad boy. This one is for a 4*TO-3 heat sink and doesnt cost a fortune.

Running at about 215ma shunt current, temperature of the heatsink goes up by about 15C with no load. There is still some margin and current can be raised even further. Of course one can use even larger heatsinks and get some serious current through this.

I am using an 18-0-18 3A transfromer. The output voltage is about 16.9v using a zener and a diode to set the voltage, using these two the thermal effects of zener cancel out. For even better cancellation, both of these can be sodered in air rather than going through a pcb trace.

Kits coming soon.

 

[jls001]

215 mA is a lot!

Any chance this can be hot rodded further, say circa 600 mA like the Salas (with larger sinks, of course)?

 

[doors666]

if you are talking about 10v supplies, then it needs a 15v transformer, giving about 22v dc output.

Total heat dissipation will be 22*0.6=13.2W per rail.

CCS will drop (22-10) 12v at 600ma = 7.2W

Mosfet will handle 10v at 600ma = 6W.


With 15v supply, you need 18v trafo, giving 25v dc output.

Total heat dissipation will be 25*0.6=15W per rail, 30W for total supply. We are at small class A amp territory here.

CCS will drop (25-15) 10v at 600ma = 6W

Mosfet will handle 15v at 600ma = 9W.

so Lets assume our worst case per device is about 10W. lets say we limit our heatsink temperature to about 55C. The to-220 case to sink thermal resistance is about 1-1.5C/W (through mica and goop), and the junction to case is about 2C/W. Total junction to sink resistance is about 3.5C/W. So for 10W, junction will be 35C higher than the heatsink. With our heatsink at 55C, the junction will be at 90C, which is ok as most of these devices are listed at 150C junction temperature. 10W is also within the derated power dissipation at 90C junction temperature or 70C case temperature.

I have ignored the load current here as for preamp, it will be insignificant as compared to 600ma of shunt current. The input CRC filter supports 2*2W resistors, so you can drop about another watt there.

If we assume ambient temperature at 40C (summer, inside a chassis it could be worse), we have a margin of 15C. We also dissipate 15W, so we get 1C/W heatsink. Apply the usual margins.

As long as heatsinking is adequate it should be ok.

One can always use TO247 devices mounted off board. The devices are at the edge of the board, so that helps and the leads can be real short.

 

[jls001]

I completed building the Blues Power shunt regulated dual power supply.

Here's the board being populated:

Fully populated board:

Another view of the board showing the biasing resistor in the foreground:

Test fitting the heat sinks:

Test fitting the MOSFETs on one side of the heat sink:

The final assembly: the heat sink shown above turned out to be quite inadequate. I doubled the heat sinks but I was still unhappy with the heat build up. So I ended up using a much chunkier heat sink. The size of this heat sink is 8" wide x 2-3/4" x 1-1/4" (base to tip of fin).

Toroidal transformer used is 120 VA, 15-0-15V.

I intended this to power my Hypnotoad phono preamp as I've been using two cheap 12V adapters. Later I realised that there is sufficient power to comfortably power even my Kuartlotron buffer, so these are the two devices I'm powering now.

The rear showing AC power input (right) and two DC outputs:

I tried various voltages and was surprised to hear drastic changes in sound.

Initially I adjusted the output voltage to +/-14.3V. This produced a very forward-sounding sound stage. It also had lots of drive and flow in the music. It also made the music louder. I tried both my music PC and turntable and the results were the same. Though there was a nice flow in the music which wasn't there before, the presentation was too in-your-face for my liking, so I ended up adjusting the voltage to +/-10V. The music with this voltage didn't sound all that different from my earlier power supplies, so I increased the voltage to +/-12V, then finally +/-13V which seems to be the sweet spot as it got back the drive and flow while not sounding forward. So this is what I'm using now.

A quick note about the voltage: the Hypnotoad uses two different opamps so their optimal voltage ought to be around +/-15V, though I've tried 9V battery as well as 12V adapters to power it. The Kuartlotron buffer should have a supply voltage of +/-10V by design. I asked the designer of it was OK to go higher and he gave me the go ahead, but asked me to adjust the bias to half the new supply voltage. There's no heat build up in the Kuartlotron even with higher supply voltage.

Now that I was happy with voltage, it was time to play around with the current dissipated by the MOSFETs. The general consensus is that passing more current through the MOSFETs makes the music sound better.

I started off quite high - around 425 mA per rail. I increased the current to about 520 mA, then about 630 mA, then finally to about 800 mA by lowering the bias resistor in steps. AFAIK, 600 mA is where most folks who do shunt regs stop. So 800 mA is a lot of current. Till about 600 mA, the heat sinks don't even get warm but at 800 mA it is warm to the touch though one can still hold it almost indefinitely.

The improvement in the sound with this shunt reg is very substantial, so it is worth all the trouble.

I wish to thank its designer doors666 for hand holding me through the build and constantly helping me troubleshoot when I got into trouble. And I got into quite a few - I broke the legs of one IRF510 MOSFET due to misalignment while fitting it to the heat sink. And for some reason which I'm still yet to figure out, I blew three of the four power rail capacitors. I took that as an opportunity to fit much higher value capacitors as doors666 had cleverly left sufficient space.

The toughest part of the build was correctly fitting the MOSFETs to the heat sink.

All in all, a very fine shunt reg for your line level devices. Output voltages can be tuned to suit specific needs by correct sizing of transformer and choice of zeners and diodes.

 

[soundbuff]

Doors666,
This all quite overwhelming for a non -techie like me. I would however like to hear it if that is possible some day. My ears are no longer young but experienced they are. Thanks for sharing.

 

[doors666]

I wanted to reply to this a while ago but completely forgot about it. Glad you like the supply and its giving you good improvements in sound.

I started off quite high - around 425 mA per rail. I increased the current to about 520 mA, then about 630 mA, then finally to about 800 mA by lowering the bias resistor in steps. AFAIK, 600 mA is where most folks who do shunt regs stop. So 800 mA is a lot of current. Till about 600 mA, the heat sinks don't even get warm but at 800 mA it is warm to the touch though one can still hold it almost indefinitely.

Thats a lot of current. Even the sims agree that performance improves as you increase the current. Its because of the reducing output impedance.
You have good heatsinks, I am yet to get decent heatsinks. I have tried max about 300ma so far, so your build is better than mine.

The improvement in the sound with this shunt reg is very substantial, so it is worth all the trouble.

I wish to thank its designer doors666 for hand holding me through the build and constantly helping me troubleshoot when I got into trouble. And I got into quite a few - I broke the legs of one IRF510 MOSFET due to misalignment while fitting it to the heat sink. And for some reason which I'm still yet to figure out, I blew three of the four power rail capacitors. I took that as an opportunity to fit much higher value capacitors as doors666 had cleverly left sufficient space.

The toughest part of the build was correctly fitting the MOSFETs to the heat sink.

No problems jls.
The way I do it is that first I solder the mosfet, making sure the mosfet is fully in and not tilted in any direction and then solder it. If you make a mistake in this step, it will always be an uphill task. Then mark on the heatsink and drill the hole. Everytime I solder the mosfet, I make sure that its inserted correctly and not tilted, that way you wont face any misalignment issues.

The CCS mosfet is slightly blocked by the caps, I use a small screwdriver for this, or a screwdriver that bends 90 degrees near the tip can also be used.

Yeah those pannies bulged, those are still good caps that I used at 300ma as well as in my amps. I guess 4700uf is a bit too less for the high continuous currents. I suspect it might have issues with even around 500ma. I remember reading 10000uf per class A 1Amp. Some class A guys can shed more light on cap requirements for high current systems.

The system supports 25mm caps. There are two caps in CRC format per channel. If the R is shorted out, then the caps become 9400uf, though it will give slightly less ripple performance than the CRC filter, but it will be able to sustain much higher currents.

All in all, a very fine shunt reg for your line level devices. Output voltages can be tuned to suit specific needs by correct sizing of transformer and choice of zeners and diodes.

Thanx. It can also be made to work for headphone amps and tube pre etc, by using a different set of components.

Yeah, the pcb has space for 3 zeners, where you can use any combination of zeners, diodes, resistors (within limits of course) etc. You can also take the two end points to a very small perfboard (and solder it to the main board with small leads) and then make your own voltage reference e.g. a few LEDs for low voltage sources, a zener and a trimpot to tune the output voltage incase you want super tight output voltage matching and arent able to get it with just zeners etc...

 

[saikatbiswas82]

Is it possible to build a shunt power supply for high current requirements like 3-4amps or maybe more?

Sent from a handheld device. Some typos may creep in

 

[doors666]

Is it possible to build a shunt power supply for high current requirements like 3-4amps or maybe more?

Sent from a handheld device. Some typos may creep in

It can be done, but we are talking about some serious heat dissipation, Class A amp level.

What do you plan on driving from such a supply. What are the voltage levels (input and output).

You will also most probably need to mount the mosfets offboard (with short wires) and use to-247 devices instead of to220. The PCB also does not support so much capacitance, so you will need to provide rectified dc to the pcb and not use the recitifier thats on the pcb. The rectifer will also dissipate quite a lot of heat so will need to be a chassis mounted one. Voltage drop - 0.7v, current - 4A, heat per rectifier diode - 2.8W

If you can make sure that there will always be a load on the supply, the heatsink requirements will come down a lot. But then if you ever run the supply without a load for even 15secs, it will blow up.

It sure will be a fun exercise.

 

[omishra]

Is it possible to build a shunt power supply for high current requirements like 3-4amps or maybe more?

Sent from a handheld device. Some typos may creep in

Generally shunt regulators are good for preamplifier, buffers only where not much current required but highly low impedance power supply is required.

 

[cosmic_wayfarer]

This is a monster dual shunt regulated power that I made. Its quite flexible in what it can do. it can use small diodes or the larger to220 ones. It has 25mm caps and a CRC filter at the input with 2*2W resistors, so a lot of voltage can be dropped here. Start with the best foot forward and give a relatively clean supply to the regulator.

All small transistors are 2sc and 2sa series, so 2sc2240, 2sc1815 etc and their 2sa pnp counterparts can be used. These are better than the bc series transistors (ok, you can shout now). The fets are IRF510 and IRF9510.

It uses three zener diodes to set the output voltage, so a lot of flexibility is there with that. Zener voltage is filtered with both a film cap and a electrolytic cap and then its buffered before using it to set the voltage. This build sets the voltage to about 16.5v and runs at about 300ma. Of course the heatsinks are not enough for that without any load.

It has 4 output points per dc rail, so more than one project can be powered from this, like a preamp, phono pre and headphone amp etc. Separate connection for earthing is also provided. The pcb is quite large at 4"x5".

I have also a BluesPowerMini coming out soon which is a much smaller board with less features and more suitable for a single project.

Some Pictures:

Hi Newbie here!

I have built this shunt regulator, this is my first shunt, and i am pretty happy with it. I am currently powering my two-opamp line amp with this, I have a discrete preamp in the works for which this is intended. And I have to say there is a considerable improvement in the sound quality with this regulator.

I alway wanted to build a shunt regulator especially the one by salas from diyaudio.com, but couldn't because of those FETs. I couldn't find the FETs or even the replacements here in the local market and they are prohibitively expensive online.

So, Just wanted to thank the designer. I have use a 431 as the voltage reference instead of the zener since I had the former laying around. Attached is the schematic i used and some pics.