The SLB (Smooth Like Butter) Active Rect/CRC/Cap Mx Class A Power Supply GB

[xrk971]

Edit Jan 23, 2020: Connection diagram for single trafo and dual rail SLB with M2X amp
The SLB (Smooth Like Butter) Active Rect/CRC/Cap Mx Class A Power Supply GB

Edit June 16, 2019: Error on Q5 orientation on Single Rail (SR) board. Please install Q5, BD139 with pin 1 to the square pad (as customary) but this will have front of part in contact with HS1. Ok to use smaller stamped aluminum finned heatsink here for back mounting. More info here:
The SLB (Smooth Like Butter) Active Rect/CRC/Cap Mx Class A Power Supply GB

Edit June 12, 2019: in an effort to reduce my logistical overhead and efforts to pack and ship as track - I will be handling the GB2 via my Etsy shop. The SLB can be found here:
Smooth Like Buttah SLB Class A PSU | Etsy

I have been working on the development of a new power supply for Class A amps that really is quite a big leap forward from current designs used for so long. It utilizes an active bridge rectifier provided by the LT4320 and MOSFETs, followed by a precisely engineered CRC filter designed to match the load, and finally followed by a capacitance multiplier. The advantages that such a power supply will have include:

1. Almost no heat dissipation or loss through the rectifier
2. Improved voltage headroom since no 0.6v drop across the silicon rectifiers
3. Smoother inherent ripple from the rectifier since switching is active and time phased to occur at zero crossing
4. More compact since fewer and smaller bulk caps are needed
5. Lower overall cost since fewer large bulk caps are used (and we all know this is one of the most expensive parts of a Class A amp aside from the case)
6. Most importantly, improved performance relative to a traditional CRC supply with lower ripple and higher current capability.
7. The SLB provides a smooth and slower (1 second) ramp that may reduce speaker turn on thump. However, current in rush from the transformer being turned on is not prevented and a suitable soft start circuit such as the Soft as a Feather Pillow (SFP) SSR soft start can be used for a soft start.

You may say that the downside is that it’s more complicated - but for those of you going through the trouble of using 8 x MUR TO220 diodes with their requisite heatsink, insulator pad, bolt, screw, insulated bushing, snubber film cap, and tight spacing that makes assembly difficult - I am not so sure anymore.

Here is a photo of a single rail SLB portion of the amp (we integrated it as part of the amp for testing). You can see how small it is when using SMT variant of the active MOSFETs (mounted underside). Those are 15,000uF 50v bulk capacitors. Compare that to how usual Class A CRC or CRCRC supplies are using 22,000uF or 33,000uF caps. The large 3pin Molex connector allows us to place the cap Mx pass transistor anywhere we like and allows snappy assembly/disassembly from the chassis or heatsink:

Jhofland, Aksa, Vunce and myself have been working on a single rail prototype of this PSU which we are calling the SLB. The name says it all - Smooth Like Butter. Jhofland has been leading the layout and electrical design, Vunce and myself are doing the verification builds and testing of the performance and listening to the resulting music, and Aksa has been instrumental in providing insight and guidance, plus overal cap Mx topology design. It’s been a great collaboration and I feel that we have arrived at something really special and useful.

The design requirement was that it can flow 5 or more amps and have ripple in the single mV range (we are looking for about -60dB to -70dB ripple rejection) with a simple hookup of a secondaries from a power trafo. You then get clean mV ripple (sine like and not sawtooth) with a small adjustable output voltage range based on how much drop you want across the cap multiplier.

Regarding capacitance multipliers, we are testing 3 different topologies: Darlington, LTP differential (similar to the MrEvil design), and a new complementary feedback (CFP) pair with BJT’s. Work is on going but we are getting close to a downselect.

Here is the measured ripple (1.4mV rms) from the Darlington cap Mx with 4.35A load (37.5vdc output) with an 8ohm resistive load:

Under a real SE Class A amplifier load, the ripple is a bit higher at about 6mV rms with the Darlington topology, but still very respectable given a 4.35A current:

Our simulations show that the CFP topology should be even better. The LTP differential topology will be tested soon (tonight perhaps).

I thought it is now a good time to share the initial developments to get everyone ready for the SLB PSU. I hope that this compact little supply will be a useful new tool in your toolbox of Class A amps. It certainly provides some of the cleanest power Inhabe ever measured or heard.

Edit - here is layout:

Update May 5, 2019: Verification build test with 4.4A and 35.4v out and 3.1v drop gives about 1mVrms ripple with the SLB (with R17/18 replaced by jumper):

Here is the SLB vero build (v1):

Update May 6, 2019: Schematic and BOM of v2 (May 3, 2019) Production version of SLB here.. Pricing of SLB is $22.50ea (2mm thick FRP, 2oz Cu, ENIG finish, blue mask):

Edit May 6, 2019: Here is the SLB-SR (single rail version). BOM and Schematics here. Pricing on SLB-SR is $18ea (also 2mm thick FRP, 2oz Cu, ENIG finish, blue mask):

Edit May 9, 2019: Here is the pereformance of the SLB in single rail form as used on an actual SE Class A amp with 37.4v rail and 4A current. The top red trace is for a 10R value on R117 and bottom yellow one is 1.5R for R117:

Edit May 23. 2019: Production PCBs arrived.
Dual Rail

Single Rail

Edit June 5, 2019: Md_Stryker made a Mouser shopping vart/BOM for SLB v1.2 here. Please do not modify it - copy to your own project first!
The SLB (Smooth Like Butter) Active Rect/CRC/Cap Mx Class A Power Supply GB

Edit Jan 23, 2020:
Connection Digram for a No-Hum M2X PSU with SLB and single trafo:

Connection Diagram for no-hum Alpha Nirvana:

 

[SemperFi]

Interesting. But why can the bulk caps be smaller? The ripple voltage is still 50/60Hz and it is still the load current with the frequency that determines the ripple for a given capacitance. ?

 

[xrk971]

Ripple current is smaller if the PSU is correctly engineered with proper ESR ratings observed. For a CRC filter on a PSU, one does not want to use low ESR caps or too big of caps. Low ESR caps and high capacitance caps will cause excessive charging ripple currents and degrade life of cap. Ripple voltage is reduced mostly by use of a capacitance multiplier - or really should be called a “ripple eater”. A smaller value CRC will have higher ripple voltage than a larger CRC, but the large value CRC will do so at the cost of higher ripple currents which is bad for lifetime. We then use a ripple eater to attenuate the ripple by circa -60dB to -70dB.

 

[baswamin]

XRK,

Mark me down for couple of boards.

Thanks
Balaji

 

[tubesguy]

Is the intent to have just a single-rail component, or is this just proof of concept?

 

[Mark Johnson]

Looks to me like a couple of ferrite beads, installed in the right places, would do quite a bit of good. For the scope traces presented here you'd want to select "low" frequency ferrite cores not the cellphone frequency cores. It's a cheap and easy experiment which can be done on non-modified PCBs, i.e., not requiring relayout. If you like what you see then you can weigh the pros and cons of relayout versus retrofit.

 

[pinnocchio]

Very interesting!

 

[xrk971]

Is the intent to have just a single-rail component, or is this just proof of concept?

Proof of concept was single rail. Final here for GB will be dual rail. The proof of concept also used an all SMT implementation of the active bridge rectifier for compactness. However, for a general purpose DIY PSU, we have decided to go with all through hole. This means that qnty 8 x TO220 MOSFETs will be used, but since they will not be dissipating any heat really, no heatsinks (or associated insulator pads, bolts, bushings, nits, etc) will be required.

Looks to me like a couple of ferrite beads, installed in the right places, would do quite a bit of good.

I have a lot of noise pickup from something in my lab. The low level hash that makes that fuzzy trace is there even when I switch the PSU off. But perhap some ferrite beads may indeed help.

 

[xrk971]

Here is the initial schematic a dual rail CFP cap Mx version.

We will be using the STP110N8F6 ultralow Rdson n channel MOSFET:
https://www.mouser.com/datasheet/2/389/stp110n8f6-956395.pdf

 

[Mark Johnson]

I'd lay out a suitable wattage resistor in series with the emitters of the high power pass transistors Q10 and Q12. Just in case your CFPs on your board, earn their reputation for unwanted oscillation. A good way to choose the resistor value is to pretend you are connecting "N" number of CFPs in parallel to spread the heat load across N devices. Then these emitter series resistors effectively operate as ballast resistors, which guarantee equalized current-sharing among the N parallel CFPs, despite each transistor having a slightly different VBE. The emitter resistors prevent thermal disaster via current-hogging.

Whatever emitter resistance you would use for N=3 parallel CFPs, to prevent current-hogging, just use 1/3 that resistance in a single CFP. Bada bing, oscillation is tamed.

You can build a test board with the resistor jumpered out, to see how badly the circuit oscillates with no resistance.

 

[xrk971]

Were did you see an oscillation, or is that just an assumption of the parade of awfuls to come. I have tested a variant of the CFP cap Mx before and had no issues with oscillation.

 

[xrk971]

Here is what the board layout will look like. Note the exclusive use of through hole parts. The pass transistors have the option of on board underhung mounting or externally mounted via flying leads on 90deg Molex locking quick connects.

 

[meanie]

Subscribed !

Count me in GB

 

[Mark Johnson]

CFP has a widespread and universal reputation for unwanted oscillation. Take a look at the Tektronix power supplies that use CFP, they've got lots of oscillation countermeasures. Then ask yourself: is Tektronix the kind of company which bothers to expend board area and parts cost, to solve a non-problem?

Me, I've experienced CFP oscillation when I foolishly used (a) no ballast resistors; at the same time as (b) BoHunker large beefy output transistors with lots of phase shift (low fT), simultaneously with low current complementary devices in the CFP, having high fT. I don't know whether this combo oscillates every single time, or only every time somebody doesn't bother to prototype it in a debug and tweak period into their project schedule. No data on that.

Evaluate the odds, then take your chances and roll the dice. Try not to remember the junior engineer's lament, "There's never time to do it right but there's always time to do it over".

 

[jwjarch]

Shall I officially start the list? baswamin and meanie, feel free to edit.

SLB Group Buy List
baswamin - 2 boards?
meanie - board(s)?
jwjarch - 2 boards

 

[meanie]

SLB Group Buy List
baswamin - 2 boards?
meanie - 2 boards
jwjarch - 2 boards

 

[xrk971]

Thanks for self organizing the interest list Jwjarch!

 

[SemperFi]

You should try a bunch of different MOSFETs. The gate drive on those monsters you've selected will create tons of HF hash. No gate resistor even! The driver/MOSFET circuit should be treated as a high frequency circuit (due to fast edges and high current spikes on gate drive)
This circuit is a great idea if you have low voltage supply and high current demand where the voltage drop over the didoes make a difference.
If you can make for a smooth turn on/off of the MOSFETs then it may be better for audio as well.

 

[xrk971]

Definitely will test the circuit to make sure it’s clean and not oscillating. For the SMT prototype, I found that it was exceptionally smooth (SLB) without artifacts at the MOSFET switch on/off. This probably has to do with how smart the controller is in timing the gate switching with the zero crossing of the mains input. The ripple after the cap multiplier is so smooth that it looks like a sine wave rather than a sawtooth.

For reference, here is data from an initial test that I did with the active bridge and CRC but no cap multiplier. You can see a sawtooth at 120Hz but there are no spikes ornringing artifacts. This was using a 25V 250VA trafo to make 3.8A continuous at 32v into an 8ohm resistive load. This was giving 390mV rms ripple.

The manufacturer’s recommended circuit doesn’t call for a gate stopper resistor even (perhaps they are internal to the chip). They are trying to make the package as easy to use as possible and then only required component aside from the said MOSFETs, is a 1uF bypass cap across the output pins.

 

[SemperFi]

Well there may be no issues at all, but fact is gate drive is full of HF hash. Unless there is some resistance in the chip and there may be. You may have high frequency EMI radiating from the MOSFETs, but the magnitude and frequency may be of no real issues. The rectified sines are nice, and zero volt switching is probably the reason as you say, but at zero volts on the drain the gate capacitance is at it's most, it is huge! But there is no miller, so perhaps not so bad after all. I would treat the gate drive as a HF circuit, but as you have experienced, there may be no real world problems to worry about. But why not experiment a little.
The gate drive send and return paths should be as tight as possible, and it doesn't hurt to have a resistor to slow things down. Those should be SMD, 1206 is easy enough to work with and small enough.
These are the loops I mean, see picture.