Andrew, I see where you are coming from and I too thought the same way, but having seen Eva's tests when he first put them up I have to say I have realised my thinking was wrong.
Don't forget that current flows through diodes, not voltage, so when you look at the voltage waveform, the 90 degree lagging current peaks at the voltage zero crossing. So by blocking the first bit of current in this manner you block the first bit of flattening.
Don't forget that current flows through diodes, not voltage, so when you look at the voltage waveform, the 90 degree lagging current peaks at the voltage zero crossing. So by blocking the first bit of current in this manner you block the first bit of flattening.
Follow-up Questions
I have a couple of quick follow-up questions:
1. mlloyd1's schematic shows two polarized electrolytic caps, low voltage, connected in parallel, but anti-phase. I've also seen schematics where the caps are connected in series, to form a non-polar electrolytic. Is one method preferred over the other?
2. Most of the SCR-type dimmers are going to operate from Line-to-Neutral. If the loads I wish to protect are all effectively Line-to-Line (240V, US), will I even "see" a DC offset?
3. Can someone suggest a technique for measuring DC offset in a residential environment?
Bob Stern said:
I have a couple of quick follow-up questions:
1. mlloyd1's schematic shows two polarized electrolytic caps, low voltage, connected in parallel, but anti-phase. I've also seen schematics where the caps are connected in series, to form a non-polar electrolytic. Is one method preferred over the other?
2. Most of the SCR-type dimmers are going to operate from Line-to-Neutral. If the loads I wish to protect are all effectively Line-to-Line (240V, US), will I even "see" a DC offset?
3. Can someone suggest a technique for measuring DC offset in a residential environment?
weinstro said:
The Cornell-Dublier article says the parallel connection method is useless; you have to connect them in series.
However, you only need electrolytics if you are using the conventional method of huge DC blocking capacitors rather than the PS Audio Humbuster method I described. In the latter case, the capacitors can be small enough that you do not need electrolytics.
weinstro said:
Eva's advice seems sound: Measure the output of a low pass filter connected across the AC line. She used a 100K 1W resistor and a 22uF non-polar capacitor and then measured the DC voltage across the capacitor.
Hi,
if your two phase supply (110 + 110) is from a single phase transformer then both phases will have the same distortion, provided the loading on both phases does not cause one of them to become unbalanced.
So the answer is a conditional YES, the DC from a balanced dual line will have no DC.
Richie,
the transformer coupled PSU uses an inductive element, the transformer, but once it is up and running I believe it behaves near enough as a resistive load. Could that be right? If not then the power companies would have made them illegal and would have invented a new system to adjust voltage to the consumers that would allow them to charge for the electricity we use.
if your two phase supply (110 + 110) is from a single phase transformer then both phases will have the same distortion, provided the loading on both phases does not cause one of them to become unbalanced.
So the answer is a conditional YES, the DC from a balanced dual line will have no DC.
Richie,
the transformer coupled PSU uses an inductive element, the transformer, but once it is up and running I believe it behaves near enough as a resistive load. Could that be right? If not then the power companies would have made them illegal and would have invented a new system to adjust voltage to the consumers that would allow them to charge for the electricity we use.
The parallel connection of caps to make a bipolar one is indeed useless, however in the schematic kindly offered by mllyod1 the max reverse voltage one of those caps will see is 2x Vdiode which it will easily survive.
Andrew, you have a good point about the startup vs running load characteristic. I do concur with what you said now that I think more about it. I guess the real way forward is simply to build one of these diode-based ones and test it. I have an electric heater but it is dial controlled, I will have to see if I can make it do nasty things to the mains.
Andrew, you have a good point about the startup vs running load characteristic. I do concur with what you said now that I think more about it. I guess the real way forward is simply to build one of these diode-based ones and test it. I have an electric heater but it is dial controlled, I will have to see if I can make it do nasty things to the mains.
some people have built the first circuit i posted, which used series caps. it has worked successfully for us.
i found the second circuit which used parallel caps while looking at the schematic (freely available on the web) for a successful, reliable amplifier that has been marketed for years. i have never built that one and i don't know anyone who has. however, having had the amp in my home, i was able to verify that the transformer buzzed when the circuit was bypassed but stopped buzzing when the circuit was enaged in the amp.
mlloyd1
i found the second circuit which used parallel caps while looking at the schematic (freely available on the web) for a successful, reliable amplifier that has been marketed for years. i have never built that one and i don't know anyone who has. however, having had the amp in my home, i was able to verify that the transformer buzzed when the circuit was bypassed but stopped buzzing when the circuit was enaged in the amp.
mlloyd1
richie00boy said:
Hi Mlloyd and others,
the cap in series very effectvely blocks the DC.
On that basis any cap will work BUT is it safe?
A single polarised Electrolytic or parallel inverse connected pair will work but the designer MUST ensure the electrolytic NEVER sees reverse voltage exceeding the manufacturer's limit. This is often about 1V to 1.5V reverse.
In normal operation the DC blocking cap SHOULD be designed to have less than the diode volts drop of peak forward voltage and as a consequence will also have a similar reverse voltage. Again the Electrolytic should survive indefinitely.
However normal operating current is only one senario. Another important senario is the start up condition. The peak current pulse needed to magnetise the toroid will be many times the normal running current and during this peak the forward voltage across the DC blocking caps could be many times the 1V or so that they normally see. To cover this short term eventuality the diodes start to conduct to bypass the blocking caps. Now we have a diode or diode pair with a forward voltage due to the start up peak current that will have a peak forward voltage that will probably exceed 1V per diode.
You now have a failure mechanism designed into the start up sequence.
The solution is to use a properly proportioned bi-polar capacitor or pair of back to back series connected polarised capacitors.
As far as I know you cannot easily buy large bi-polar caps to suit conventional power/integrated amplifiers. You need about 10mF 10V per kW of transformer but a series pair needs to be 20mF+20mF per kW to ensure the peak voltage across the cap is significantly below the diode voltage when the amplifier is supplying maximum output.
I hope this helps avoid exploding caps.
the cap in series very effectvely blocks the DC.
On that basis any cap will work BUT is it safe?
A single polarised Electrolytic or parallel inverse connected pair will work but the designer MUST ensure the electrolytic NEVER sees reverse voltage exceeding the manufacturer's limit. This is often about 1V to 1.5V reverse.
In normal operation the DC blocking cap SHOULD be designed to have less than the diode volts drop of peak forward voltage and as a consequence will also have a similar reverse voltage. Again the Electrolytic should survive indefinitely.
However normal operating current is only one senario. Another important senario is the start up condition. The peak current pulse needed to magnetise the toroid will be many times the normal running current and during this peak the forward voltage across the DC blocking caps could be many times the 1V or so that they normally see. To cover this short term eventuality the diodes start to conduct to bypass the blocking caps. Now we have a diode or diode pair with a forward voltage due to the start up peak current that will have a peak forward voltage that will probably exceed 1V per diode.
You now have a failure mechanism designed into the start up sequence.
The solution is to use a properly proportioned bi-polar capacitor or pair of back to back series connected polarised capacitors.
As far as I know you cannot easily buy large bi-polar caps to suit conventional power/integrated amplifiers. You need about 10mF 10V per kW of transformer but a series pair needs to be 20mF+20mF per kW to ensure the peak voltage across the cap is significantly below the diode voltage when the amplifier is supplying maximum output.
I hope this helps avoid exploding caps.
Follow-up Q on dc blocking cap selection
When I attempt to calculate appropriate numbers for the capacitor value, I'm left with some curious results.
In my application, I have a torroidal 2kVA isolation transformer configured with a balanced secondary output. Rated max current for this transfomer is 16.7 amps at 120V, 60Hz.
I've understood from the prior discussions that as long as the voltage drop across the cap is less than the forward voltage drop of the diode string, then the diodes shouldn't conduct. Assume I can tolerate a voltage drop across the caps of 4.8V.
I'm using the formula:
C = I/(2*PI*f*V)
At 60Hz, I get cap values of 18,500uF. Ripples values for caps of this rating at 6.3Vdc have ripple values of <6 amps, which is well below the 16.7A rated current. This implies that I need several caps to get the overall rating I think I need, and these caps aren't small. The other concern is that there isn't much margin left for managing dc offset, the original intent of the circuit.
Is there something flawed with my assumptions, here? What is the typical insertion loss with a real, working dc blocking circuit?
mlloyd1 said:
When I attempt to calculate appropriate numbers for the capacitor value, I'm left with some curious results.
In my application, I have a torroidal 2kVA isolation transformer configured with a balanced secondary output. Rated max current for this transfomer is 16.7 amps at 120V, 60Hz.
I've understood from the prior discussions that as long as the voltage drop across the cap is less than the forward voltage drop of the diode string, then the diodes shouldn't conduct. Assume I can tolerate a voltage drop across the caps of 4.8V.
I'm using the formula:
C = I/(2*PI*f*V)
At 60Hz, I get cap values of 18,500uF. Ripples values for caps of this rating at 6.3Vdc have ripple values of <6 amps, which is well below the 16.7A rated current. This implies that I need several caps to get the overall rating I think I need, and these caps aren't small. The other concern is that there isn't much margin left for managing dc offset, the original intent of the circuit.
Is there something flawed with my assumptions, here? What is the typical insertion loss with a real, working dc blocking circuit?
Re: Follow-up Q on dc blocking cap selection
First you need to use ohms law to determine the max impedance you can stand to develop 4.8V at full current, this is
R = V / I = 4.8 / 16.7 = 287 mR
Then you put that value into
C = 1 / (2 * pi * f * z) = 1 / (2 * 3.141 * 60 * 0.287) = 9242 uF
So back to back caps of 18,485 uF would be needed
You seem to have arrived at the right answer, was just confused by your use of V and I in the cap formula.
I think your reasoning is sound and is one reason why I have looked for alternative designs for DC filters.
weinstro said:
First you need to use ohms law to determine the max impedance you can stand to develop 4.8V at full current, this is
R = V / I = 4.8 / 16.7 = 287 mR
Then you put that value into
C = 1 / (2 * pi * f * z) = 1 / (2 * 3.141 * 60 * 0.287) = 9242 uF
So back to back caps of 18,485 uF would be needed
You seem to have arrived at the right answer, was just confused by your use of V and I in the cap formula.
I think your reasoning is sound and is one reason why I have looked for alternative designs for DC filters.
Hi,
think back to when and if the diodes might conduct.
They conduct when the voltage drop across the caps exceeds the sum of the forward voltage drops of the diodes.
This can occur during any peak current pulse that causes the capacitor voltage to rise.
If I were to blindly design this DC blocker I would select a dual diode pair with a forward voltage drop of 1.4V or so and limit the cap voltage drop to about 1.2V.
I would also calculate the peak current at maximum load 2000VA/120V*root2.
Using the formulae you have both correctly quoted I end up with 45mF+45mF back to back.
Oops, time for a rethink since even 16V caps of this total capacity are not going to be small.
I pose the opening question again, when do the diodes conduct? At start up, but could I tolerate occassional conduction when the transformer delivers peak load?
Now some investigation, what current does the transformer actually draw (thanks Wuffwaff) when operating normally, how often does it exceed this by 10%, or 50% or 100%. Armed with these results and revisiting the dual diode pairs I might discover that quad pairs of diodes might be acceptable if the cap rating is still comfortably in excess of this and that the peak operating current is just 1Apk and only exceeds +100% of this i.e.2Apk 23 times an hour.
Re do the numbers and see if the DC blocker becomes more acceptable.
If smaller high voltage capacitors are initially installed, a simple method of determining cap voltage drop patterns is simply to continuously measure the drop while the transformer is operating. A pair of 10mF 63V caps would be a good starting point and a DMM or better still an analogue meter to measure the real volts drop. Take care with insulation and safety precautions since you are monitoring small voltages supperimposed on mains AC voltages. Don't go killing our posters.
I hope this helps separate the chaff from the grain
think back to when and if the diodes might conduct.
They conduct when the voltage drop across the caps exceeds the sum of the forward voltage drops of the diodes.
This can occur during any peak current pulse that causes the capacitor voltage to rise.
If I were to blindly design this DC blocker I would select a dual diode pair with a forward voltage drop of 1.4V or so and limit the cap voltage drop to about 1.2V.
I would also calculate the peak current at maximum load 2000VA/120V*root2.
Using the formulae you have both correctly quoted I end up with 45mF+45mF back to back.
Oops, time for a rethink since even 16V caps of this total capacity are not going to be small.
I pose the opening question again, when do the diodes conduct? At start up, but could I tolerate occassional conduction when the transformer delivers peak load?
Now some investigation, what current does the transformer actually draw (thanks Wuffwaff) when operating normally, how often does it exceed this by 10%, or 50% or 100%. Armed with these results and revisiting the dual diode pairs I might discover that quad pairs of diodes might be acceptable if the cap rating is still comfortably in excess of this and that the peak operating current is just 1Apk and only exceeds +100% of this i.e.2Apk 23 times an hour.
Re do the numbers and see if the DC blocker becomes more acceptable.
If smaller high voltage capacitors are initially installed, a simple method of determining cap voltage drop patterns is simply to continuously measure the drop while the transformer is operating. A pair of 10mF 63V caps would be a good starting point and a DMM or better still an analogue meter to measure the real volts drop. Take care with insulation and safety precautions since you are monitoring small voltages supperimposed on mains AC voltages. Don't go killing our posters.
I hope this helps separate the chaff from the grain
Andrew T et al:
A valid question, to be sure ...
I guess I should have posted a safety warning with that circuit since it installs on the primary side of the transformer.
I'm not the original designer - I found this post back on netnews in the rec.audio group MANY years ago (before the web) and thought it presented a clever solution to the problem. So far, I have never been able to trace the origination. So, I have no insight into the designer's thoughts about parts selection, etc.
Some respected designers on this board who have/had commercial products in the field have commented that one should choose a capacitor for this circuit that has adequate ripple current handling capacity. I seem to recall perhaps even some manufacturer recommendations were made as well. Having experienced a cap or two exploding in my day, I think that is great advice.
I know of at least one manufacturer (who offers VERY long product warranty periods) having a similar circuit (also posted earlier in the thread) installed in products that are famous for reliability. Unfortunately, I no longer have my notes about the particular components and specs for the parts that manufacturer chose to use. Maybe someone who has one or has access to such products could comment or upload some pictures.
And finally, most of us using amplifiers with "monster"power supplies also use some type of "soft start" circuit on the primary side of the transformer (usually series resistance of some sort or maybe even a triac) to limit currents as the rail capacitors charge. This has the side benefit of lowering intial current through the series DC blocking cap also.
Hoep this helps ...
mlloyd1
A valid question, to be sure ...
I guess I should have posted a safety warning with that circuit since it installs on the primary side of the transformer.
I'm not the original designer - I found this post back on netnews in the rec.audio group MANY years ago (before the web) and thought it presented a clever solution to the problem. So far, I have never been able to trace the origination. So, I have no insight into the designer's thoughts about parts selection, etc.
Some respected designers on this board who have/had commercial products in the field have commented that one should choose a capacitor for this circuit that has adequate ripple current handling capacity. I seem to recall perhaps even some manufacturer recommendations were made as well. Having experienced a cap or two exploding in my day, I think that is great advice.
I know of at least one manufacturer (who offers VERY long product warranty periods) having a similar circuit (also posted earlier in the thread) installed in products that are famous for reliability. Unfortunately, I no longer have my notes about the particular components and specs for the parts that manufacturer chose to use. Maybe someone who has one or has access to such products could comment or upload some pictures.
And finally, most of us using amplifiers with "monster"power supplies also use some type of "soft start" circuit on the primary side of the transformer (usually series resistance of some sort or maybe even a triac) to limit currents as the rail capacitors charge. This has the side benefit of lowering intial current through the series DC blocking cap also.
Hoep this helps ...
mlloyd1
AndrewT said:
wuffwaff said:
A fair question, to be sure. I've figured that with amps pulling rated power, and other stuff connected, it might load the transformer to 1.2kVA. The transformer was oversized somewhat to provide margin and less voltage droop when loaded.
However, ALL of the loads have solid state bridge rectifiers in them, which means the actual current draw will come as brief spikes, with a peak value much higher than average. Some people in the power industry refer to this as crest factor, BTW.
AndrewT said:
Yep, that's where I began to question my assumptions, as I was equating RMS current draw with ripple current, not even including a crest factor, which might be 3 or so.
I was also looking at 6.3Vdc caps -- much smaller.
AndrewT said:
I understand your train of thought, but am somewhat confused by a 100% peak occuring 23 times in an hour. Rather than a random event, I'm inclined to think of this as a circuit which can support the rated loads of the eletrical devices connected, yes or no.
Maybe the real fundamental question I have is, is the ripple current rating of a capacitor based on RMS, or peak current? That would reduce this done to two variables -- current to pass and voltage drop.