• These commercial threads are for private transactions. diyAudio.com provides these forums for the convenience of our members, but makes no warranty nor assumes any responsibility. We do not vet any members, use of this facility is at your own risk. Customers can post any issues in those threads as long as it is done in a civil manner. All diyAudio rules about conduct apply and will be enforced.

RingNot: power supply for chip amps. Bare PCBs and/or assembled+tested units

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Member
Joined 2011
Paid Member
This is a power supply PCB which provides dual voltage outputs (± 28V) to power chip amps. It can power two LM3886 amplifiers, each amp operating at its maximum output power (68W @ 4 ohms, or 38W @ 8 ohms). It is suitable for either a two-channel stereo pair, or a single-channel pair of LM3886 amplifier chips operating in bridged mode. The PCB is designed and optimized for the Antek transformers AS-2222 (shielded) and AN-2222 (unshielded).

The board features both Soft Recovery diodes (30A, 200V) and full CRC snubbers, to completely eliminate oscillatory ringing in the power transformer secondaries. Conventional PSU designs, using standard silicon rectifier bridges and single capacitor "snubbers", produce unwelcome ringing in the secondaries. This is shown in the first two pictures below.

Photo 1 shows an Antek AS-2222 transformer driving a conventional PSU which uses a 25 A bridge rectifier (GBU2510) with single 100nF capacitors (acting as lousy snubbers) across each secondary. This arrangement stimulates oscillatory ringing in the secondary when the rectifiers switch off.

Photo 2 shows the same transformer, driving a RingNot PSU board. Oscillatory ringing has been completely eliminated, thanks to Soft Recovery diodes and full CRC snubbers across each secondary.

I am placing the RingNot design into the public domain; its complete schematic is shown below, and a .zip archive of the Gerber files for PCB manufacturing is attached. Anyone who so desires can download the Gerbers for free, and send them off to a PCB fab to have boards made. Bills Of Materials for DigiKey, Mouser, and Element14 are contained in the attached Builder's Guide .pdf document. The Builder's Guide contains a very detailed set of step by step instructions for stuffing and soldering the board.

I am also selling my (small!) remaining inventory of bare PCBs and of assembled and tested, finished units. Once they're gone, I have no plans to build or sell any more boards or finished units. If you want more than these, you will need to organize a Group Buy or something. I've got about a dozen bare PCBs and about a dozen fully assembled, finished units as of 22 June 2015. Yes the circuit boards are red. Really.

BUYING DETAILS

A lot of people here seem to really like PayPal so I have opened a PayPal account and can accept money via PayPal. If you decide to buy a RingNot, you will tell PayPal to send me a GIFT of 15.00 USD (for a bare PCB) or a GIFT of 50.00 USD (for a fully assembled, finished power supply). These PayPal prices include shipping, to anywhere in the world.

In addition to PayPal, I can accept USD-denominated checks. I actually prefer checks, so I offer a discount to buyers who choose to pay by check. Quoted prices include shipping.

Code:
      assembled PSU | pay by | pay by |
      ship to       | PayPal | check  |
      ==============+========+========+
      USA           | $ 50   |  $ 33  |
      --------------+--------+--------+
      CAN+MEX       | $ 50   |  $ 40  |
      --------------+--------+--------+
      rest of world | $ 50   |  $ 45  |
      --------------+--------+--------+
      
      
      bare PCB      | pay by | pay by |
      ship to       | PayPal | check  |
      ==============+========+========+
      USA           | $ 15   |   $ 6  |
      --------------+--------+--------+
      CAN+MEX       | $ 15   |  $ 10  |
      --------------+--------+--------+
      rest of world | $ 15   |  $ 12  |
      --------------+--------+--------+
To order: send me a PM with your shipping address and phone number (US Customs form requires a phone#). Tell me how many PCBs you want and how many fully assembled, finished units you want. Tell me whether you intend to pay by PayPal or check. I'll give you my PayPal ID and/or mailing address.

Please download and read the Builder's Guide .pdf before buying. It provides lots of details and answers plenty of questions.


THERMAL COMPARISON

The Aavid 5301 heatsink, which mounts all four diodes of the GBU bridge rectifier in the conventional PSU, has a quoted thermal resistance of 6.30 K/W. (link 1)

The Aavid 5771 heatsinks used in the RingNot power supply, mount only a single diode. They have a quoted thermal resistance of 25.90 K/W. (link 2), a factor of 4.1 greater. Fortunately they have a factor of 4 less heat input (one diode instead of four diodes), so we expect them to rise to approximately the same final temperature as the larger heatsink with full bridge.

I connected the conventional PSU to a two channel electronic load. The PSU's +28V output was connected to a load current of 4.0 amperes, and the -28V output was connected to a load current of 4.0 amperes. These represent a guardband beyond worst case load, since the transformer is only rated for 200 Volt-Amperes. The heatsink temperature was measured with a Type-K thermocouple (a) before the load was connected; (b) after 1 hour of operation with two 4.0 ampere loads. Temperature rise was 51.7 Kelvin.

Then I connected the RingNot PSU to the electronic load. +28V was connected to a load current of 4.0 amperes, -28V was connected to a load current of 4.0 amperes. Measuring the heatsink using the same thermocouple before and after one hour of operation, the temperature rise was 53.3 Kelvin. This measured result agrees with intuition: (1/4th) the power at (4X) thethermal resistance, gives the same final temperature.

The individual diodes on individual heatsinks, heated up to a final temperature approximately equal to the final temperature of the bridge rectifier on a single large heatsink. Just as the calculations predicted.
 

Attachments

  • Conventional_Secondary.png
    Conventional_Secondary.png
    33.7 KB · Views: 5,646
  • populated_view2.jpg
    populated_view2.jpg
    357.4 KB · Views: 2,287
  • populated_view1.jpg
    populated_view1.jpg
    131.8 KB · Views: 1,659
  • unpop_rear.jpg
    unpop_rear.jpg
    223.9 KB · Views: 4,835
  • unpop_front.jpg
    unpop_front.jpg
    243.5 KB · Views: 5,157
  • RingNot_Gerber_Files.zip
    25.3 KB · Views: 451
  • RingNot_Schematic.png
    RingNot_Schematic.png
    22.1 KB · Views: 5,562
  • RingNot_Secondary.png
    RingNot_Secondary.png
    31.7 KB · Views: 5,417
  • RingNot_PSU_Entire_Builders_Guide_One_File.pdf
    834.7 KB · Views: 882
Last edited:
  • Like
Reactions: 1 user
Member
Joined 2011
Paid Member
If you decide to send the RingNot Gerber files out to a PCB fab, I enthusiastically recommend that you visit the website pcbshopper dot com (link) to get a general idea of pricing. Tell the quote generator
  • Board dimensions: 8.4cm X 9.2cm
  • Layer count: 2
  • Soldermask color: (any)
  • Silkscreen: both layers
  • Surface finish: cheapest
  • Number of designs: 1
  • Quantity: 10
  • Ship To: (select your country)
  • Quote!
and then wait 60 seconds for all the quotes to arrive and to get sorted by price.

If you're in the US, I think you will be very pleased with the speed of your order if you happen to choose Elecrow for the fab and Shenzhen-DHL for the shipping. Here's an example. I had some RingNot PCBoards made there, and total delay from order to PCBs in my hands in the US, was six days. (!) That's very fast, faster than many domestic shops who charge a lot more. (link here on diyAudio)
 
Secondary ringing effect

Is the ringing in the secondaries mostly just bad form, potential cause of EMI, slight potential noise back on the mains, and somewhat "rough" on the transformer? Shouldn't actually get through the capacitor bank nor to the circuit (amp), unless it's through EMI, right?
 
Member
Joined 2011
Paid Member
Hi Boba, what an honor that your very first posting EVER, is right here!

Lots of people report their positive listening experiences with before-and-after testing. One of them is frequent diyAudio poster (and esteemed high end designer) John Curl. He postulates that the reason why his Vendetta Research preamp sounded better when he got rid of transformer ringing, was imperfections in the ground system. The ringing coupled into the audio ground, was his best guess. Others mention electrostatic (capacitive) coupling, electromagnetic (inductive) coupling, and plain old transformer behavior. If there's oscillatory ringing currents in secondary winding 1, these are coupled by plain old transformer operation, namely flux linkage in the core, to all the other windings. And of course EMI as you state.

Have a quick google for "RRSF" (reverse recovery snapback filter), which is/was a super popular snubber that garnered much effusive praise in the early 2000's. People reported very large sonic differences, after versus before. To be fair the most enthusiastic proponents of RRSF used vacuum tube equipment, where oscillatory currents in the HT secondary could easily zip thru the transformer core and appear on the 6.3VAC heater windings. I don't think you want ultrasonic oscillations on the heaters!
 
RingNot 2ndary on diyAudio PSU v3

Mark,

Thanks for the answer & the kind welcome!

Ok that makes sense to me, and is interesting.

So, comparing this RingNot to the diyAudio PSU v3.x, I guess the diyAudio PSU v3.x also has PCB circuits for the CRC snubbers and could be used like the RingNot if one were to use soft Recovery Diodes and the same CRC circuit? In which case the main pros of the diyAudio PSU being some additional flexibility, and 2 extra capacitor slots on the PCB...



BOBA
 
Is the ringing in the secondaries mostly just bad form, potential cause of EMI, slight potential noise back on the mains, and somewhat "rough" on the transformer? Shouldn't actually get through the capacitor bank nor to the circuit (amp), unless it's through EMI, right?

The ringing is caused by the parasitics of the transformer + rectifier circuit. The parasitic inductance of the transformer windings and routing set up a resonant circuit with the stray capacitance of the circuit. When the current in this resonant circuit is interrupted, as when the diodes turn off, the energy left in the resonant circuit will cause it to ring until the energy has been dissipated in the parasitic resistance of the circuit. The ringing tends to be in the 100s of kHz to low MHz range - so well outside the audio band.

The ringing itself is not really an issue. However, in some amplifiers, the ringing may couple into sensitive parts of the circuit where it can be demodulated and cause a DC offset shift. This means every time the diodes turn off, so at twice the mains frequency, there'd be an impulse injected into the circuit. The result would be an increase in the 100/120 Hz mains harmonic.

The frequency of the ringing is rather low, hence, the coupling tends to be rather small in most circuits. I strongly suspect the amount of coupling is heavily dependent on the layout of both the supply and the amplifier as well as the wiring between the two.

There are several ways to address this ringing. One can add a small capacitor (100 nF is usually a good starting point) across each secondary winding. This lowers the frequency of the ringing and effectively suppresses any coupling into the amplifier. Another approach is to design a snubber, such as the one in Mark's circuit. As shown in the scope shots, a snubber, when designed well, works very well in suppressing the ringing.
"When designed well" is the key. The way to design the snubber circuit is to first measure the parasitics of the transformer+rectifier circuit. That's fairly straight-forward to do experimentally. Once the parasitics are determined, the snubber resistor is calculated to set the Q of the parasitic resonant circuit to about 1. This app note from CDE describes optimum snubber design well: http://www.cde.com/resources/catalogs/igbtAPPguide.pdf
The point is that the snubber design will depend on the transformer chosen, which, I'm guessing, is why the RingNot design doc lists the Antek AN-2222/AS-2222 as the only approved mains transformer(s).

In my amplifiers, both tube and solid state, I have never been able to measure any difference between a supply with a snubber and one without. This with test equipment that's capable of measuring down to a -160 dBV (10 nV) noise floor without breaking a sweat! I, personally, add the 100 nF across the secondary for peace of mind and because the capacitor across the secondary technique works well with all transformers. That said, I also tend to optimize my PCB layouts such that coupling between circuits is minimized.
Whether any audible difference between the snubber types (or no snubber at all) can be reliably detected can be debated. I strongly suspect that, unless a significant difference in the residual mains hum can be measured on the output of the amplifier, there will be no discernible difference in double blind ABX listening tests.

Tom
 
Last edited:
Member
Joined 2011
Paid Member
Ringing frequency is (1/(2pi))*sqrt(1/LC) and can be rather high ... I've seen ringing at 3 MHz which is higher than top end of the AM radio band ... when L and C are both small, as they are on a toroidal transformer with a high current 6.3V secondary. Adding an external C in parallel with the secondary, reduces the ringing frequency. Picture 1 of post 1 in this thread, shows the effect of Cexternal=0.1uF upon a 200VA toroid with 22V secondary. Ringing frequency has been decreased to approx 90 kHz. (~ 4.5 cycles in 50 usec).

Snubber design is now very well understood at diyAudio and can be accomplished quickly using only a signal generator, resistor, and oscilloscope. "Quasimodo ExtraLight" (here in QM post #561) shows how. The great thing is, you don't need to measure the number of microhenries of inductance of your transformer, and you don't need to plug that number into a math formula. All you need is the transformer itself plus the gear mentioned above. Those who don't own and don't want to own a signal generator, can build up a little test jig called Quasimodo (search for it) or else a CheapoModo (search for it), to use with a scope, and accomplish the same result: Simple, No-Math Transformer Snubber. You can get an optimum snubber without ever knowing the values of the parasitic L or the parasitic C. Many, many diyAudio members have already done so, and posted their quite excellent results, here on the site.

I myself advocate setting the "Q" of the resonant circuit to 0.5, which corresponds to a damping ratio "zeta" of 1.0. (the relation is zeta=0.5/Q). And zeta=1.0 is what I recommend to beginners, in the Quasimodo design note. Why? Because I think zeta=1.0 gives the widest margin-of-safety of all possible target values. If well-meaning newbies make experimental mistakes which result in ±40% errors, zeta=0.6 and zeta=1.4 are relatively harmless results. However when you put a ±40% error bands around other targets, the endpoints become a lot less forgiving. Cornell Dubilier's suggested value for high frequency switchmode power supplies (Q=1, zeta=0.5) results in zeta=0.3 at the low end when error= -40%. This is way too underdamped and way too much overshoot for my taste, especially when we consider that this application is a leisurely 120Hz fullwave rectifier and not a 100kHz SMPS. There's no reason to fear overdamping; 90kHz to 3MHz overdamped sinewaves will settle out long before the 8.3 millisecond mains halfcycle is over.

Tom's half right - the snubber on the RingNot board was designed to be a "Universal Snubber" for all toroidal transformers with approx 200VA rating and approx 2x22VAC secondaries. However I calculated its component values before I ever touched or measured an Antek AN-2222 or AS-2222. (here's the link) Later, when I did lay hands upon an Antek 200VA 2x22VAC toroid, I verified that the Universal Snubber (on RingNot) did work extremely well, just as predicted. The other reason I don't recommend using RingNot with any other transformer is: my own resources. I decided I simply don't have the bandwidth to support any and every random transformer somebody might want to connect to a RingNot. Sure, it will probably work fine, and if the person really knows what they're doing, it probably will work flawlessly. But that's not a configuration I support. Besides, the Antek transformers are extremely well built (compare their leakage inductance to everybody else's; Antek kicks butt!). And they are very robust, with large margins of safety. And, at least for US customers paying US shipping, they're cheap. Thank you Nelson Pass for introducing me to Antek transformers.

Since RingNot is in the public domain, anybody can send it off to fab and get 10 boards for $37.00 or (as in post 2 above) 20 boards for $54 with only six day turn-around(!!). They can do this long, LONG after I've successfully gotten rid of my own meager pile of extra boards. Years later, I suppose.
 
Last edited:
Hi tomchr,

Just as one can support DIY Audio by buying a Universal Supply Board through the DIY Audio Store
Or maybe buy your board with a discount to diyaudio members ;)

I'm a bit surprised by the price of your boards, even if i can understand that there is an engineering cost included. The last board I've bought was a WaveIO USB to I2S card , cost was 99 €, two PCBs of modulus cost more than 100 € ... You can PM me an offer, I'm planning to build an amp this summer !
 
Member
Joined 2011
Paid Member
Just as one can support DIY Audio by buying a Universal Supply Board through the DIY Audio Store
Have you changed your mind since writing this:
The DIYA Universal Supply Board may include fancy snubbers, but then designers turn around and throw away the benefits of the first supply cap by implementing CRC filtering. ... Yes, that's contrary to the common DIY belief as well. My apologies for being data driven... :) :devilr:

~Tom
 
Member
Joined 2011
Paid Member
So, comparing this RingNot to the diyAudio PSU v3.x, I guess the diyAudio PSU v3.x also has PCB circuits for the CRC snubbers and could be used like the RingNot if one were to use soft Recovery Diodes and the same CRC circuit? In which case the main pros of the diyAudio PSU being some additional flexibility, and 2 extra capacitor slots on the PCB...
Yes, I agree. In fact if you study the schematic of the diyAudio Universal PSU V3, you will find that it includes a total of four snubbers; schematics below.

The circuits in the red circles are CRC snubbers directly across the transformer secondaries (identical to RingNot); their component values can be designed either using the Hagerman_with_math procedure, or the Quasimodo_no_math procedure. The circuits in the green rectangles are RC snubbers directly across the final output of the PSU. The official Bill Of Materials calls them "Output Snubbers" and recommends fixed, constant component values which do not change when different transformers are used. If you've put inductive, wirewound resistors on the board; or if you install explicit inductors to make CLC filters instead of CRC filters; these Output Snubbers may be beneficial.

_
 

Attachments

  • left_half.png
    left_half.png
    35.4 KB · Views: 1,494
  • right_half.png
    right_half.png
    59.4 KB · Views: 1,437
Member
Joined 2011
Paid Member
Pascal, I think it's a much better idea to choose the amplifier first, and after the amplifier is chosen, then select a suitable power supply. Some LM3886-based amplifiers use a single-ended power supply; if you choose one of these amps then RingNot is not compatible! RingNot provides dual (bipolar) output voltage: ±28V.

There are dozens of chip amp PCBs and kits to choose from, including a large number of them which are discussed in the diyAudio forums. You can use search terms to find postings which review the sonic performance of chip amp boards and kits, and which compare the various offerings.

chipamp.com, to name one example, sells amplifier PCBs for very reasonable prices. neurochrome.com sells amplifier PCBs for jaw-dropping high prices, but the boards do include some nice features. I myself like the balanced input feature using the special purpose IC from THAT Corp., and I also like the DC servo which reduces DC offset at the output. These are only two of the many chipamp vendors who sell many other offerings, which I suggest you begin to study and compare.
 
...
The frequency of the ringing is rather low, hence, the coupling tends to be rather small in most circuits. I strongly suspect the amount of coupling is heavily dependent on the layout of both the supply and the amplifier as well as the wiring between the two.
One always hears to "twist the heater wires together" in tube circuits (this is even seen in guitar amplifiers!), to reduce the induction of large 60Hz currents into other parts. With that in mind, this is clearly a good idea for the wires from EVERY transformer winding, and keep wherever else these currents go close to each other as well - have the twisted pair secondary winding go directly to the bridge (that's either chassis mounted or on the edge of the PCB to keep the traces short and loops small), and have the filter capacitors directly adjacent to the bridge to keep those traces short as well.

But I'd do this in addition to a snubber circuit or other suggestions, not instead.

In my amplifiers, both tube and solid state, I have never been able to measure any difference between a supply with a snubber and one without. This with test equipment that's capable of measuring down to a -160 dBV (10 nV) noise floor without breaking a sweat! I, personally, add the 100 nF across the secondary for peace of mind and because the capacitor across the secondary technique works well with all transformers. That said, I also tend to optimize my PCB layouts such that coupling between circuits is minimized.
Whether any audible difference between the snubber types (or no snubber at all) can be reliably detected can be debated. I strongly suspect that, unless a significant difference in the residual mains hum can be measured on the output of the amplifier, there will be no discernible difference in double blind ABX listening tests.

Tom
What type rectifier diodes do you use? Surely they're the modern, fast-recovery type. The venerable and ubiquotous 1N400x series works "well enough" at 60Hz, but has a big forward charge that makes for a long reverse recovery time - it's as if it were designed to make the stray inductance ring on the reverse cycle. I'd be surprised if it didn't show up in any electronic circuit in the same box. A passive snubber circuit can stop this, but with fancy low-charge rectifiers being a lot less expensive than they once were, one can go with the full belt-and-suspenders approach without spending much more money.
 
I provide state of the art performance. I also provide circuits that are well engineered, well designed, well tested, and well documented. This provides confidence for the builder that they will be able to accomplish the build and achieve the same stellar results that I measure in my setup. I charge accordingly. Should you disagree with my design philosophy or pricing, there are many other options, albeit, none at the performance level I'm getting.

Tom
 
Member
Joined 2011
Paid Member
Following the advice of post #2 above, I got PCB fab pricing quotes (good for today only) for ten RingNot PCBoards. They are sorted by price with the lowest price at the top.

Today, you can get ten RingNot PCBs shipped to your address in the USA, for as little as $16.05. Fast shipping costs a little more; I personally have had spectacularly good results with Shenzhen DHL, receiving finished boards six days after I ordered them (link) .

Ten boards for $34.90, versus one board for $50.00? Easy decision.

_
 

Attachments

  • 06_July_2015.png
    06_July_2015.png
    108.3 KB · Views: 1,246
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.