DC Blocker (mains) for high power amp

I've been asked by a friend to make him a DC blocker for his Krell monoblocs - a pair of brutes with a maximum power consumption of 6kW, and for which they recommend a dedicated 20A spur!

Having perused the many, oft-conflicting threads, I'd like to use a simple blocker circuit with large elctrolytic caps and antiparallel diodes as per the much-discussed Bryston circuit, but using much larger caps (up from Bryston's 0.33mF/6.3V to 1.0mF/16V).

An externally hosted image should be here but it was not working when we last tested it.


These are rated for a maximum ripple current of 14 Amps, and it is my intention to use a pair in parallel to give >20 Amps headroom, plus a similarly robust bridge rectifier.

If I've understood it correctly Bryston use their caps in parallel assuming no significant reverse voltages (<1V) will be encountered; while Rod Elliott and others use caps in back-to-back series to protect against any reverse voltage scenarios.

The disadvantage of the latter approach is that you need 4 x the capacitance to achieve the same ripple current handling, which becomes hopelessly expensive and outsized in the case I'm trying to get my head around.

So, would it be possible to protect the caps from reverse voltage by using a suitable diode after the blocker to drop reverse voltage above a certain threshold to Earth - since it shouldn't be present unless a fault state exists? If so, which kind (transil, avalanche, single high power rectifier?).

I want to create a safe and extremely robust solution, but don't have the skills to model this, nor the test equipment to do more than PAT type safety stuff. I'm aware of the need to properly rate, fuse, insulate and earth every last aspect.

Many thanks for any input.
 
Bryston show 33000uF = 33mF = 0.033F
1mF is less, not more !

Look at effective capacitor impedance to the "normal" current flowing into the Krell transformer.

At 50/60Hz you can calculate the impedance using F=1/[2PiRC]
R in this case becomes impedance, Xc = 1/[2PiFC] where Pi= 3.14, F = 50 or 60, C = Farads

Once you have impedance and normal current you can calculate the voltage drop across the blocking capacitor. This MUST be LESS than the diode starting to turn on voltage. Use 500mVf for a Power Diode/Rectifier (don't use low voltage diodes). Two in series adds up to 1000mV. The capacitor Voltage drop at your mains frequency when passing normal current must be less than 1V

Expect to use 5V or 6V or 10V capacitors of ~0.02F to 0.1F to be able to run a big Krell without the diodes passing. The reason for the very low voltage rating is to get the capacitor package/s down to a small size that will fit in the space available and not cost more than the PSU.
The Power Diodes prevent the capacitor seeing overvoltage during start up and during abuse. The diodes remain cold.

During start up, or abuse incidents, the current flowing becomes massive. In this situation the diodes pass, but should not heat up significantly, because the duration should be very short.

BTW,
I bought a batch of 50 (lower cost) 3300uF 10V. I series/parallel 6, 8, or 10 to suit the normal current. i.e. I use the ESP sch. Or rather ESP appeared after (copied this from this Forum) Peranders and myself.
 
Last edited:
Thanks Andrew, and excuse the decimal place confusion, I actually intend to use 0.1F 16V caps - each is rated for a ripple current of 14A at 100Hz, and ESR of 0.01 Ohms.

Using your calcs I get a worst-case scenario at 20A current (unlikely!) of a 0.64V drop across one cap, and half that across the parallel pair I intend to use.

Which is great, and thanks, good to know my estimates were about right, but do you then consider that the diode bridge implementation is sufficient to protect the caps against reverse voltages?

Much obliged!
 
Member
Joined 2011
Paid Member
Cornell-Dubilier publishes quite a nice Aluminum Electrolytic Capacitor Application Guide. I have snipped out a little piece of it, image below.

They discuss reverse bias voltage (opposite to the polarity of the capacitor), and strongly recommend that you limit reverse bias to 1.5 volts or less.
 

Attachments

  • Cornell_Dubilier_AEappGUIDE_pdf.png
    Cornell_Dubilier_AEappGUIDE_pdf.png
    59.9 KB · Views: 1,237
Personal opinion, power transformers must be built considering and being able to stand the possible DC component found in mains.

Forgetting to do so is unrealistic and this kind of solutions, if actually needed, are kludgy at best.

Better do it right from the beginning.
 
uhm.. what dc component in the mains?

if you have a dc ground loop existing between the pole transformer, your structure's ground, or whatever other loads are present.. this must get fixed asap.

dc doesn't go through transformers, find out where its coming from.
 
Administrator
Joined 2007
Paid Member
uhm.. what dc component in the mains?

More correctly asymmetry of the pos and neg cycles due to non resistive loading of the mains at large. This causes the average to shift away from zero... which can be interpreted as a DC shift.

(and you need at least 30 amp capability for the 6kW mentioned in post #1. If they draw that much then I would imagine you need a 30 or 35 amp ring main or a dedicated 30 amp spur from a consumer unit)
 
Cornell-Dubilier publishes quite a nice Aluminum Electrolytic Capacitor Application Guide. I have snipped out a little piece of it, image below.

They discuss reverse bias voltage (opposite to the polarity of the capacitor), and strongly recommend that you limit reverse bias to 1.5 volts or less.

Thanks Mark, that was my primary concern, still is.


Personal opinion, power transformers must be built considering and being able to stand the possible DC component found in mains.

Forgetting to do so is unrealistic and this kind of solutions, if actually needed, are kludgy at best.

Better do it right from the beginning.

Lovely idea, but toroidals saturate easily in the present of DC offset - millivolts are enough, and the bigger they are, the harder they fall... I also tend to think Krell may know a thing or two about specifiying amp components...:wrench:


uhm.. what dc component in the mains?

if you have a dc ground loop existing between the pole transformer, your structure's ground, or whatever other loads are present.. this must get fixed asap.

dc doesn't go through transformers, find out where its coming from.

As the nice man sez below - it's an AC waveform distortion due to the presence of commonplace devices like thermostats on the same domestic circuit. Worth sorting, because even if you don't have buzzing transformers, it has a significant impact on efficiency.


More correctly asymmetry of the pos and neg cycles due to non resistive loading of the mains at large. This causes the average to shift away from zero... which can be interpreted as a DC shift.

(and you need at least 30 amp capability for the 6kW mentioned in post #1. If they draw that much then I would imagine you need a 30 or 35 amp ring main or a dedicated 30 amp spur from a consumer unit)

The 6kW figure is a maximum rating - i.e. a massive transient at high volume into hyper-inefficient speakers in some theoretical amp designer's universe where hearing loss isn't an issue...; they draw 200W on idle, and about 500W on normal musical programs according to their owner.

Thanks to everyone for their input.
 
Member
Joined 2011
Paid Member
If I've understood it correctly Bryston use their caps in parallel assuming no significant reverse voltages (<1V) will be encountered; while Rod Elliott and others use caps in back-to-back series to protect against any reverse voltage scenarios.

That's right. Bryston guarantees, by design, that no significant reverse voltage (< 1.5V) occurs across the electrolytic capacitors in their mains blocker.

How do they do this? They guarantee, by design, that no significant surge currents flow, which would place the bridge diodes in their high current, high Vfwd regime.

And how do they do THAT? They use a soft start circuit. A triac in series with the transformer primary is controlled by either (i) a soft start IC designed to control AC motors in washing machines {part number TDA1085C (datasheet)} in their early power amp products; or else (ii) a microcontroller {part number PIC16F818} in their more recent power amps. This triac gets fired at just exactly the right phase of the incoming mains waveform, to minimize surge current ("ELI the ICE man"). Then the controller gradually ramps up the conduction angle, across many many cycles of the AC mains waveform.

These components are easy to spot on the Bryston 6B-SST schematic (LINK). The triac is reference designator Q3, part number BTB-24800B (datasheet), and the microcontroller is reference designator IC1. To see the earlier version with the washing machine controller IC, look at the schematic of the Bryston 3B.

Naturally, YOU can implement soft start any way you wish; you don't have to use a triac. Douglas Self's book and Bob Cordell's book discuss other approaches to soft-start, which don't employ triacs.
 
Thanks again Mark. The Krells necessarily incorporate their own soft-starts, so that aspect vexes me less than my vague grasp of the what-ifs of various potential fault states, and what I might be able to do to try to divert them - should it be technically feasible.

I should mention that from the safety POV it's my intent to use a thoroughly earthed alloy enclosure, to insulate its contents thoroughly, and to (belt & braces!) fuse any circuit I may build upstream and down...
 
Note that paralleling components like fuses or diodes may not double their current handling capacity. Because the two components will not be exactly identical, one will carry a little bit more current, so it will heat up more and it's resistance will drop, so then it carries a little bit more current.
 
Member
Joined 2011
Paid Member
The Bryston-style mains DC blocker circuit (Fig.1 below) uses two capacitors, each of "C" farads. Its total capacitance is 2C and it and applies 2 x Vdiode of reverse bias to each capacitor.

The Westhost/Elliott-style mains DC blocker (Fig.2) uses two capacitors in series. Its total capacitance is 8C, and it applies 1 x Vdiode of reverse bias to each capacitor.

Interpolating between the two, we can imagine a third mains DC blocker circuit (Fig.3). Its total capacitance is 4C, and it applies 1 x Vdiode of reverse bias to each capacitor.

Fig3 has the same low-stress (1 x Vdiode) reverse bias as the Fig.2, but needs only half the total capacitance. However it does require either discrete diodes, or else two bridge rectifier modules.

_
 

Attachments

  • fig1.png
    fig1.png
    7.9 KB · Views: 1,019
  • fig2.png
    fig2.png
    7.4 KB · Views: 1,652
  • fig3.png
    fig3.png
    9.7 KB · Views: 980
Member
Joined 2002
Paid Member
I think Mark was referring to the total volume or bulk of the capacitors, not the effective capacitance. But I don't see the performance equivalence between fig 3 and fig 2. You still have to choose the capacitors so that their loaded voltage drop is less than a diode drop. If you need 4C for each Vbe in fig 2, you still need it in figure 3.
 
The bridge rectifier can be wired two ways.
+ to - linked.
Connect in and out to ~ and ~
This gives two diode drops in both directions. As shown in Mark's left and middle drawings and in ESP & Bryston's drawings.

or

~ to ~ linked AND + to - linked.
In and out are at two adjacent tabs (either ~ to -, or ~ to +)
This gives one diode drop in each direction. It also parallels the diodes for more current passing capability. eg use a 5A diode bridge which has a peak rating of ~ 30A and paralleled allows upto 60Apk to pass until the fuse blows.
I use a set of 4off 1n5402 which seems to have plenty of current capability.
 
Last edited:
Member
Joined 2011
Paid Member
Fig.3 contains two single-diode-drop DC blockers in series. The one on the left chops off the first Vfwd of DC bias, and the one on the right chops off the second Vfwd of DC bias. The net effect is to remove 2xVfwd of DC bias, which is the same DC blocking performance as Fig.1 and Fig.2

diyAudio member AndrewT is correct that a single 4-diode bridge rectifier module can indeed supply all 4 diodes of Figure 3. I've drawn this out in Fig.4, below.

Our Original Poster, diyAudio member MrFatBarSteward, hypothesizes that capacitor price is approximately proportional to capacitance value. If so then the capacitor portion of the cost is just the sum of the red C's in each design:

  • Fig.1: total capacitor cost = 2 x C
  • Fig.2: total capacitor cost = 8 x C
  • Fig.3: total capacitor cost = 4 x C
  • Fig.4: same as Fig.3

diyAudio member mirlo suggests that the capacitors in Fig.3 need to be twice as big as those in Fig.1, since the mains current times the capacitive reactance must be less than 2xVfwd in Fig1 but only 1xVfwd in Fig3. I'm of two minds on this point; it's obviously true during massive current surges, when the exponential behavior of the diodes kicks in. But how true is it over the long term of normal operation? Maybe that's a saving grace. Or maybe it's just fooling yourself by setting aside a margin-of-safety in one case while gobbling it up in another. I'm of two minds.

diyAudio member destroyerOS suggests it's wise to spend the extra money and obtain the extra peace-of-mind, by using the extra safe Fig2 circuit. Maybe he's right. Maybe the extra cost (about the price of two tickets to the cinema?) is small while the peace-of-mind is large.
 

Attachments

  • fig4.png
    fig4.png
    28.1 KB · Views: 676
Last edited:
Hear hear Destroyer in post 13. This subject has been done to death on this forum. How many ways can you possibly connect a bridge and two caps? Get over it!

Wait - do different DC blockers have different sounds? I sense a 200 page thread coming on here.
 
Last edited:
I don't understand the problem.

Get a big *** low noise rectifier, mount it in a small box, attach large *** capacitors, have a 20A IEC inlet, and a hardwired short cord out terminated with 20A IEC.

http://www.partsconnexion.com/product4795.html

That's some bridge you've linked! I can see the benefit when it's working at 50-60 cycles/S in a PSU, but for brief moments of startup conditions in a power filter is there any beenfit from using such low noise diodes?