Hello, fellow DIYers and Experts!
A while ago I read the TNT online magazine as well as the Zero Tolerance online magazine and followed the articles for solid state power supply. They recommend using 2 bridge rectifiers for a polarised PSU instead of one bridge rectifier.
The authors in these two sites claim that using 2 bridge rectifiers instead of the conventional one bridge has the following benefits:
TNT Schematic
(1) Significantly improved rectification.
(2) increased power handling capabilities
(3) much better channel separation
(4) produces more ground planes
Zero Tolerance Schematic
(1) isolates the transformer from the audio electronics
(2) it makes the plus and minus sides independent of each other
(3) it reduces power supply output impedance
(4) it doubles the current capability of the power supply, allowing for better transient reproduction and faster charging of specified and bigger electrolytic capacitors.
These are in comparison to the Conventional topology
Now when I am about to put my PSU together, I re-read these articles in details. I can't work out how each of those benefits can be REAL.
Let me walk through these claims:
TNT:
(1) Significantly improved rectification
- how? The difference is only that a diode has half the voltage across it in comparison. Would diodes rectify better with lower voltages? Does a diode only conduct or not conduct?
(2) increased power handling capabilities
- I can't see how. The bridges will have sufficient rating (they are dirt cheap) and power handling is practically determined by the transformer, not the bridges.
(3) much better channel separation
- Since each winding of the transformer is independent in both cases I can't see the reason why.
(4) produces more ground planes
- Ground plane is subject to implementation.
Zero Tolerance:
(1) isolates the transformer from the audio electronics
- I just don't see how it differs from the conventional topology and what isolation we are talking in here.
(2) it makes the plus and minus sides independent of each other
- In the conventional topology the plus and minus sides are independent so what we are talking about here?
(3) it reduces power supply output impedance
- I don't see how. In fact, if you observe the current return from the ground, you will find that the current must pass 2 diodes instead of one. There needs to be some voltage applied before diodes conduct so I think using 2 diodes will possibly increase the impedence not reducing it.
(4) it doubles the current capability of the power supply, allowing for better transient reproduction and faster charging of specified and bigger electrolytic capacitors.
- How? the current capabilithy is mainly determined by the transformer and the diodes don't have much to do with it. It only doubles the current capability of the bridges because we use two but it can't double the current capability of the power supply. How fast the caps are charged will be affected by the impedance but as said in above I can't find that the impedance is reduced but likely the opposite is true.
I have no intenstion to argue with the authors or anyone. I simple don't think those particular articles have much truth in them. I hope I can get some answers so that I can learn from them. My questions may well be due to my very basic understanding of electronics (I am a DIYer not an EE). I am thinking about abandoning this topology I chose sometime ago when now I am about to hook my PSU up.
Regards,
Bill
A while ago I read the TNT online magazine as well as the Zero Tolerance online magazine and followed the articles for solid state power supply. They recommend using 2 bridge rectifiers for a polarised PSU instead of one bridge rectifier.
The authors in these two sites claim that using 2 bridge rectifiers instead of the conventional one bridge has the following benefits:
TNT Schematic
(1) Significantly improved rectification.
(2) increased power handling capabilities
(3) much better channel separation
(4) produces more ground planes
Zero Tolerance Schematic
(1) isolates the transformer from the audio electronics
(2) it makes the plus and minus sides independent of each other
(3) it reduces power supply output impedance
(4) it doubles the current capability of the power supply, allowing for better transient reproduction and faster charging of specified and bigger electrolytic capacitors.
These are in comparison to the Conventional topology
Now when I am about to put my PSU together, I re-read these articles in details. I can't work out how each of those benefits can be REAL.
Let me walk through these claims:
TNT:
(1) Significantly improved rectification
- how? The difference is only that a diode has half the voltage across it in comparison. Would diodes rectify better with lower voltages? Does a diode only conduct or not conduct?
(2) increased power handling capabilities
- I can't see how. The bridges will have sufficient rating (they are dirt cheap) and power handling is practically determined by the transformer, not the bridges.
(3) much better channel separation
- Since each winding of the transformer is independent in both cases I can't see the reason why.
(4) produces more ground planes
- Ground plane is subject to implementation.
Zero Tolerance:
(1) isolates the transformer from the audio electronics
- I just don't see how it differs from the conventional topology and what isolation we are talking in here.
(2) it makes the plus and minus sides independent of each other
- In the conventional topology the plus and minus sides are independent so what we are talking about here?
(3) it reduces power supply output impedance
- I don't see how. In fact, if you observe the current return from the ground, you will find that the current must pass 2 diodes instead of one. There needs to be some voltage applied before diodes conduct so I think using 2 diodes will possibly increase the impedence not reducing it.
(4) it doubles the current capability of the power supply, allowing for better transient reproduction and faster charging of specified and bigger electrolytic capacitors.
- How? the current capabilithy is mainly determined by the transformer and the diodes don't have much to do with it. It only doubles the current capability of the bridges because we use two but it can't double the current capability of the power supply. How fast the caps are charged will be affected by the impedance but as said in above I can't find that the impedance is reduced but likely the opposite is true.
I have no intenstion to argue with the authors or anyone. I simple don't think those particular articles have much truth in them. I hope I can get some answers so that I can learn from them. My questions may well be due to my very basic understanding of electronics (I am a DIYer not an EE). I am thinking about abandoning this topology I chose sometime ago when now I am about to hook my PSU up.
Regards,
Bill
Well, I toyed around with Duncan's PSU designer using with 600VA 30V transformer, a single bridge, 20,000uF capacitor and a 4 amp load. Peak current on the bridge was 29 amps. The bigger the transformer you use and more capacitance you have, the larger and sharper these peaks will be.
With two bridges you have double the current capacity and independant rails, but, there will be more I2R losses. If it really matters and is tenable, use Schottky diodes.
ESP Audio also seems to favor dual supplies for SS class AB use.
With two bridges you have double the current capacity and independant rails, but, there will be more I2R losses. If it really matters and is tenable, use Schottky diodes.
ESP Audio also seems to favor dual supplies for SS class AB use.
HiFiNutNut,
I think you should not underestimate yourself, nor your sensitivity to audio-baloney (which unfortunately there is a lot of)
as all of your questions are more than reasonable.
Having at one time experimented with various topologies of unregulated power supplies, I can tell you that there is indeed a small advantage of having separate rectification for your power rails, but only if all other things are properly and equally done in both cases.
Two things to remeber:
1) Always think about the loops rectified current has to make. Loop area has to be minimized (preferably it should be zero) or the resulting magnetic field will induce an 'imprint' of the rectifier current into other circuitry. With a single rectifier, ytou have the advantage of only one (but larger) diode voltage drop from transformer windng to the rail, but you have a common return path for both rails, right to the transformer itself. Depending on transformer and rectifier construction, it may be more difficult to reduce the area of the current loop - you need to twist or braid the three wires together, which never gives you as perfect magnetic cancelation as a simple two-wire twist. Additionally, although we always think of the point where the two filter caps of the two rails connect, as a point, it actually has a non-zero impedance. What happens is, that when connecting this point as the center of a star ground, you may still get voltage variances between the 'points' of the star, and these variances may contain components of the rectifier current.
With dual rectification, you trade one disadvantage, namely, a higher voltage drop (two diode drops per rail) for a much more easyer controlled rectification current loop problem. In essence, you only need to twist the two wires, from transformer to rectifier, from rectifier to filter, to reduce the loops to a minimum. Further, when you connect the two terminals of the filter caps to form a ground, there is virtually no rectification current flowing through that connecton (perhaps only a mA due to capacitive winding coupling in the transformer). This is a rather stark contrast to tens, and not uncommonly over a hundred A peak currents in the single rectifier model. All of these consideratins become more pronounced the more power rails you have in a design.
Most power amplifiers do not use DC current (as oposed to voltage). The current from the rails is half-wave rectified audio output current. In the dual rectifier model, the DC ground stays a point where these half-waves cancel out, without the possibility of rectification current polluting the cancelation - this is very important since the DC ground is a reference point for input, output and NFB.
2) In a stereo amp with a common supply, things are twice as difficult since you get twice as many loops to consider from the amps and one more ground line to contend with. In the dual rectifier design, at worst, what you get if you do not create a perfect star gound, is lower channel separation (the lowering may be very slight). With teh single rectifier model, you can get a difference between the ground references of the two channels, which is a scaled-down representation of the rectification current. The symptoms are usually that the amp does not have a hum 100Hz/120Hz residual as long as the left and right ground is not connected together (perhaps on the other end of the input cable). Now immagine this same scenario in a multichannel amp...
Fortunately, there are ways to reduce the effect for singe-rectifier designs, however, the reduction in multiple rectification is usually in orders of magnitude, while the methods used for single rectifier designs reduce problems by 2-4 times at best.
I think you should not underestimate yourself, nor your sensitivity to audio-baloney (which unfortunately there is a lot of)
as all of your questions are more than reasonable.Having at one time experimented with various topologies of unregulated power supplies, I can tell you that there is indeed a small advantage of having separate rectification for your power rails, but only if all other things are properly and equally done in both cases.
Two things to remeber:
1) Always think about the loops rectified current has to make. Loop area has to be minimized (preferably it should be zero) or the resulting magnetic field will induce an 'imprint' of the rectifier current into other circuitry. With a single rectifier, ytou have the advantage of only one (but larger) diode voltage drop from transformer windng to the rail, but you have a common return path for both rails, right to the transformer itself. Depending on transformer and rectifier construction, it may be more difficult to reduce the area of the current loop - you need to twist or braid the three wires together, which never gives you as perfect magnetic cancelation as a simple two-wire twist. Additionally, although we always think of the point where the two filter caps of the two rails connect, as a point, it actually has a non-zero impedance. What happens is, that when connecting this point as the center of a star ground, you may still get voltage variances between the 'points' of the star, and these variances may contain components of the rectifier current.
With dual rectification, you trade one disadvantage, namely, a higher voltage drop (two diode drops per rail) for a much more easyer controlled rectification current loop problem. In essence, you only need to twist the two wires, from transformer to rectifier, from rectifier to filter, to reduce the loops to a minimum. Further, when you connect the two terminals of the filter caps to form a ground, there is virtually no rectification current flowing through that connecton (perhaps only a mA due to capacitive winding coupling in the transformer). This is a rather stark contrast to tens, and not uncommonly over a hundred A peak currents in the single rectifier model. All of these consideratins become more pronounced the more power rails you have in a design.
Most power amplifiers do not use DC current (as oposed to voltage). The current from the rails is half-wave rectified audio output current. In the dual rectifier model, the DC ground stays a point where these half-waves cancel out, without the possibility of rectification current polluting the cancelation - this is very important since the DC ground is a reference point for input, output and NFB.
2) In a stereo amp with a common supply, things are twice as difficult since you get twice as many loops to consider from the amps and one more ground line to contend with. In the dual rectifier design, at worst, what you get if you do not create a perfect star gound, is lower channel separation (the lowering may be very slight). With teh single rectifier model, you can get a difference between the ground references of the two channels, which is a scaled-down representation of the rectification current. The symptoms are usually that the amp does not have a hum 100Hz/120Hz residual as long as the left and right ground is not connected together (perhaps on the other end of the input cable). Now immagine this same scenario in a multichannel amp...
Fortunately, there are ways to reduce the effect for singe-rectifier designs, however, the reduction in multiple rectification is usually in orders of magnitude, while the methods used for single rectifier designs reduce problems by 2-4 times at best.
HiFiNutNut said:
I agree with your analysis. My bench set-up is actually like the one proposed by TNT, but for a different reason: to have a flexible PS that can be configured either as +/- rails or two indepedent PSs.
ilimzn said:I think you should not underestimate yourself, nor your sensitivity to audio-baloney (which unfortunately there is a lot of)
baloney seems to be in greater supply in the audio land,
ilimzn said:
You can always split the ground and use two wires instead of one, and braid +/ground, and -/ground together.
Tweeker,
Thanks for pointing me to re-read Rod's article and it makes a lot of sense. I read it about a year ago and at that time I barely understood half of it. However, Rod uses a dual bridge design for 2 separate channels whereas the TNT/Zero Tolerance use a dual bridge for 2 rails. They are completely different ball games.
tlf9999,
Thanks for your input. Please read further and contribute.
ilimzn,
Thank you for taking your time sharing your valuable knowledge.
I agree with tlf9999. I think some of the issues you addressed to be the disadvantage of a single bridge may be solved by layout, component positioning and wiring. I still don't understand the advantage of having dual bridges if wiring is very well thought out. e.g. the "cancellation" you mentioned should apply to both a single bridge topology AND dual bridges.
Some time ago I was thinking about having separate ground wiring and using a single star ground. I posted my question here: Dual mono amp star grounding questions - http://www.diyaudio.com/forums/showthread.php?s=&threadid=61961
In that, I was planning to use this layout:
After some excellent advice from forum gurus AndrewT and ChocoHolic, I revised the layout:
And this is what it looks like now:
It is still under design, revision and construction.
(This is getting too long so let me make a new post below).
Regards,
Bill
Thanks for pointing me to re-read Rod's article and it makes a lot of sense. I read it about a year ago and at that time I barely understood half of it. However, Rod uses a dual bridge design for 2 separate channels whereas the TNT/Zero Tolerance use a dual bridge for 2 rails. They are completely different ball games.
tlf9999,
Thanks for your input. Please read further and contribute.
ilimzn,
Thank you for taking your time sharing your valuable knowledge.
I agree with tlf9999. I think some of the issues you addressed to be the disadvantage of a single bridge may be solved by layout, component positioning and wiring. I still don't understand the advantage of having dual bridges if wiring is very well thought out. e.g. the "cancellation" you mentioned should apply to both a single bridge topology AND dual bridges.
Some time ago I was thinking about having separate ground wiring and using a single star ground. I posted my question here: Dual mono amp star grounding questions - http://www.diyaudio.com/forums/showthread.php?s=&threadid=61961
In that, I was planning to use this layout:
An externally hosted image should be here but it was not working when we last tested it.
After some excellent advice from forum gurus AndrewT and ChocoHolic, I revised the layout:
An externally hosted image should be here but it was not working when we last tested it.
And this is what it looks like now:
An externally hosted image should be here but it was not working when we last tested it.
It is still under design, revision and construction.
(This is getting too long so let me make a new post below).
Regards,
Bill
(Continue...)
After staring at the schematics and the layout for hours and days and weeks, I sort of figured out that there is possibly no advantage (and possibly disadvantage, not only cost wise but also sound wise) of using dual bridges.
As in the layout diagrams above, I initially tried to separate all the channels and rails to an extreme.
What I believe now is that channel or even rail separations are really not as important. Given the very high PSRR of modern SS amps there is really little gain having such extreme separation. I think monoblocks sound better mainly because they have twice the power capability, not because they are separate.
I think what is important in the PSU design and layout to achieve minimum noise and hum (which should enhance sonic because there will be less modulations) is to consider current returns, current returns and current returns. When designing, the first step is to identify current loops, then make the length of any current loops as short as possible and make current loops as independent as possible with no or little overlapping.
So what was the problem with the above diagram 1? Because wires are not resistance free therefore when high current is passing through there will be voltage developed. When high current is passing the star ground the relative voltages seen from other grounds get changed (this is highly abstract because the concept of the star ground is the zero reference point). That is why some describe that it makes the star ground "dirty".
Further more is what was suggested by ilimzn about the "cancellation" effect. If we look at the conventional PSU schematic we should find that current flows out from the bridge to the + rail when currents are consumed and the caps are being filled, and at the same time on the negative rail, current flows in an opposite direction from the caps to the bridge. The current return path is via the PSU ground shared by both rails but the positive and the negative currents that are usually having equal amount cancel each other. So the loop actually appears going around from the positive rail to the negative rail while at the same time the ground is relatively quiet.
That is why I change it to diagram 2. In that case, high current loops no longer go through the star ground. I think it is much better to have the PSU common ground that groups all PSU component ground together and have a single wire going to the star ground.
Even so, I was puzzled by one problem: in diagram 2, obviously the bridge rectifiers' current return path of one positive rail and one negative rail (the loops are quite long) will go through the grounds of other relatively quieter negative and positive rail grounds within the PSU ground. This is definitely not the "ultimate". It is something that violates the above point of my own and what ilimzn said must be avoided.
But in practice, I am restricted by the physical size of the enclosures and must make some compromises. I can't think of any better layout with my brain of limited intelligence. But I guess that I should have little problem because I have used the wide aluminium bar in my design. This is indeed a very large ground plane. The resistance within the aluminium flat is at the minimum therefore there should not be much voltage developed within it. The voltage variations caused to the quieter caps, when reaching to the PCB rails should make no consequence due to the PSRR of the amp. Given this is totally within the area of the PSU and no signal ground is attached to it within the bar, I am now hoping what I believe will turn out to be true.
Please let me know if you think any of my belief or assumption is wrong so that I can correct it.
Regards,
Bill
After staring at the schematics and the layout for hours and days and weeks, I sort of figured out that there is possibly no advantage (and possibly disadvantage, not only cost wise but also sound wise) of using dual bridges.
As in the layout diagrams above, I initially tried to separate all the channels and rails to an extreme.
What I believe now is that channel or even rail separations are really not as important. Given the very high PSRR of modern SS amps there is really little gain having such extreme separation. I think monoblocks sound better mainly because they have twice the power capability, not because they are separate.
I think what is important in the PSU design and layout to achieve minimum noise and hum (which should enhance sonic because there will be less modulations) is to consider current returns, current returns and current returns. When designing, the first step is to identify current loops, then make the length of any current loops as short as possible and make current loops as independent as possible with no or little overlapping.
So what was the problem with the above diagram 1? Because wires are not resistance free therefore when high current is passing through there will be voltage developed. When high current is passing the star ground the relative voltages seen from other grounds get changed (this is highly abstract because the concept of the star ground is the zero reference point). That is why some describe that it makes the star ground "dirty".
Further more is what was suggested by ilimzn about the "cancellation" effect. If we look at the conventional PSU schematic we should find that current flows out from the bridge to the + rail when currents are consumed and the caps are being filled, and at the same time on the negative rail, current flows in an opposite direction from the caps to the bridge. The current return path is via the PSU ground shared by both rails but the positive and the negative currents that are usually having equal amount cancel each other. So the loop actually appears going around from the positive rail to the negative rail while at the same time the ground is relatively quiet.
That is why I change it to diagram 2. In that case, high current loops no longer go through the star ground. I think it is much better to have the PSU common ground that groups all PSU component ground together and have a single wire going to the star ground.
Even so, I was puzzled by one problem: in diagram 2, obviously the bridge rectifiers' current return path of one positive rail and one negative rail (the loops are quite long) will go through the grounds of other relatively quieter negative and positive rail grounds within the PSU ground. This is definitely not the "ultimate". It is something that violates the above point of my own and what ilimzn said must be avoided.
But in practice, I am restricted by the physical size of the enclosures and must make some compromises. I can't think of any better layout with my brain of limited intelligence. But I guess that I should have little problem because I have used the wide aluminium bar in my design. This is indeed a very large ground plane. The resistance within the aluminium flat is at the minimum therefore there should not be much voltage developed within it. The voltage variations caused to the quieter caps, when reaching to the PCB rails should make no consequence due to the PSRR of the amp. Given this is totally within the area of the PSU and no signal ground is attached to it within the bar, I am now hoping what I believe will turn out to be true.
Please let me know if you think any of my belief or assumption is wrong so that I can correct it.
Regards,
Bill
I'm not sure about dual bridges per supply; I don't believe it much enhances the sonics.
However, I found during development of the AKSA that a huge contribution came from using two completely independent power supplies. A couple of smaller transformers rather than one big one; two separate bipolar supplies with shared common earth.
The benefits were chiefly in the area of imaging peformance.
Cheers,
Hugh
However, I found during development of the AKSA that a huge contribution came from using two completely independent power supplies. A couple of smaller transformers rather than one big one; two separate bipolar supplies with shared common earth.
The benefits were chiefly in the area of imaging peformance.
Cheers,
Hugh
Re: Re: TNT SS PSU - Is it real or illusive?
Actually, you can only do that if you already have a transformer with separate secondaries (not a center tapped one) and only up to the joining point of the + and - rail capacitors. This however is a two edged sword - on one hand, you get twisted pairs of wires from the transformer, which insures rectification current cancelation for each winding in turn - a positive effect. On the other hand, you get two attachments for the returns of the rectification current onto a star ground, which is always more of a problem than one (in general the less 'points' the star has the easyer it is to make it beave like a true star ground). Rectification current from one point goes alternatingly to one or the other filter cap, ditto for the other point but with oposite phase. This means that if these two points are even slightly apart, you get a voltage drop between them, and, their 'distance' adds to the common resistance path for the rectification and amp supply current - and this is in series with capacitor ESR, not a good thing. This sort of thing can be almost completely avoided with dual rectification, especially if multiple parallel filter caps are used - they can be configured to minimize this common impedance by layout to levels that at least an order of magnitude better than what you can achieve using similar layout tricks with single rectification and center tapped windings. In other words, if you lay everything out very carefully, with two reurn wires to the transformer, the improvement will only be marginal, but of course, that may be welcome in some instances.
Actually, you can only do that if you already have a transformer with separate secondaries (not a center tapped one) and only up to the joining point of the + and - rail capacitors. This however is a two edged sword - on one hand, you get twisted pairs of wires from the transformer, which insures rectification current cancelation for each winding in turn - a positive effect. On the other hand, you get two attachments for the returns of the rectification current onto a star ground, which is always more of a problem than one (in general the less 'points' the star has the easyer it is to make it beave like a true star ground). Rectification current from one point goes alternatingly to one or the other filter cap, ditto for the other point but with oposite phase. This means that if these two points are even slightly apart, you get a voltage drop between them, and, their 'distance' adds to the common resistance path for the rectification and amp supply current - and this is in series with capacitor ESR, not a good thing. This sort of thing can be almost completely avoided with dual rectification, especially if multiple parallel filter caps are used - they can be configured to minimize this common impedance by layout to levels that at least an order of magnitude better than what you can achieve using similar layout tricks with single rectification and center tapped windings. In other words, if you lay everything out very carefully, with two reurn wires to the transformer, the improvement will only be marginal, but of course, that may be welcome in some instances.
HiFiNutNut said:
I actually think that one large transformer is advantagous to multiple small transformers in a multi-channel amp: because it is unlikely that all channels will see high current demand at the same time. you will actually see less fluctuation in rail voltages with one large transformer.
this is pretty much dictated by the law of large numbers - fluctuations in the sum of current in each channel is smaller than fluctions in the current of each channel.
HiFiNutNut said:
I think the flaw in your reasoning is to assume that the same current draw takes place on both rails. In a typical class B amp, you will not have that: current draw will either be on the positive rail or the negative rail at any given time, not from both rails at the same time.
In that case, you will see current returns on either rail (let's not consider the idle current for now - as it is DC). So in ilimzn's example, you will have current going from positive rail to ground during positive output cycles, and then from ground to the negative rail during negative output cycles.
That inicates
a) in a three braided design (+, -, and ground), the cancellation exists at any given time between the positive rail and the ground, or the negative rail and the ground. Either case, it is a two-wire cancellation - the third wire carries no current at all.
b) if you split the ground into two wires, and braid + and the ground, and then braid - and the ground, you will also get perfect two-wire cancellation.
So it seems to me either way, you will not get the problems ilimzn mentioned.
OK, let me try to explain a bit more what I had in mind in my original post. I'm going to start this 'from the top down', i.e. fully independant power supplies for all rails and all channels, that would be on the top of the list as far as PSUs go. Next from the top are separated transformers for each channel, and then next down is a common transformer but with 4 separate windings. The general schematic that follows, still applies, with the exception that interchannel coupling is possible through a common core (but this really occurs only at the lowest of frequencies).
On the picture you can see that all current loops are quite isolated from each other. Ground separation is very simple, and if power rail caps local to the amp board are needed, they go between the power rails and the power ground point, as close to the point where the load and quiet ground separates. The local star ground is very simple, with only 4 points, of which one is very low current (the 'clean' ground). Since power rail return grounds only connect in one point, the only way for a ground current loop to happen is through capacitive coupling inside the transformer windings. Although the capacitances are on the order of nF, the ground line impedance is low that this should not be an issue at all. Still, making the area of this loop very large, enclosing ferromagnetic components or unequal lengths of the ground wires from transformer to clean ground may result in some spurious signal injection, but this is likely to be a problem only if you intend to have your input level directly from a MC cartridge
On the picture you can see that all current loops are quite isolated from each other. Ground separation is very simple, and if power rail caps local to the amp board are needed, they go between the power rails and the power ground point, as close to the point where the load and quiet ground separates. The local star ground is very simple, with only 4 points, of which one is very low current (the 'clean' ground). Since power rail return grounds only connect in one point, the only way for a ground current loop to happen is through capacitive coupling inside the transformer windings. Although the capacitances are on the order of nF, the ground line impedance is low that this should not be an issue at all. Still, making the area of this loop very large, enclosing ferromagnetic components or unequal lengths of the ground wires from transformer to clean ground may result in some spurious signal injection, but this is likely to be a problem only if you intend to have your input level directly from a MC cartridge
Attachments
(continued)
A little trick that works very well with multiple parallel caps used instead of one large filter cap... Also, PS for the previous post, normally the load is located remotely from the load attachment terminals. If the amplifier uses a Zobel network or Boucherot cell at the output, it is important that it be attached as close as possible to the actual amp, in the place of the idealized load on the above schematic.
A little trick that works very well with multiple parallel caps used instead of one large filter cap... Also, PS for the previous post, normally the load is located remotely from the load attachment terminals. If the amplifier uses a Zobel network or Boucherot cell at the output, it is important that it be attached as close as possible to the actual amp, in the place of the idealized load on the above schematic.
Attachments
Lets consider for a minute a simplified situation where two independant windings and rectifiers feed a filter bank ending with a common ground. Not a problem until you try to use a single rectifier.
Since one winding alternately charges one and then teh other rail, you get a comon charging current path and discharging current path. It may not be a major problem as long as you can truly guarantee that the connectioon point of R4 and R4' is truly a point. The only remaining problem is that if the return wires to the transformer center point are still separate, due to different charging paths and discharging paths, the cancellation of currents in the twisted wire loops is not complete. Still, good layout will minimize the consequences. This also holds if you only have C3 and C3', i.e. one single pair of filter caps.
Since one winding alternately charges one and then teh other rail, you get a comon charging current path and discharging current path. It may not be a major problem as long as you can truly guarantee that the connectioon point of R4 and R4' is truly a point. The only remaining problem is that if the return wires to the transformer center point are still separate, due to different charging paths and discharging paths, the cancellation of currents in the twisted wire loops is not complete. Still, good layout will minimize the consequences. This also holds if you only have C3 and C3', i.e. one single pair of filter caps.
Attachments
An interesting variation that also works slightly better at eparating charging and discharging loops when there is only a single cap pair, C2 and C2' in the picture. Pondering a bit shows that a common path for charging and discharging closes around more resistors of a higher value. In a typical implementation, this is done by making a slit inside the metal piece or PCB copper that joins the caapcitor terminals. Still, analysis shows that you only get a couple od dB attenuation on the charging current spikes.
Attachments
Final thoughts:
As can be seen, when rectification and filtering is completely independant, there is considerable freedom in wiring without compromising the ground reference. Ths is because it is simple to have all the various supplies connect only into on ground point.
This is particulairly useful if there are local caps on thea mplifier board, creating a local power ground. With completely independant rectification, this only means that you end up joining (or not, as you wish) the quiet grounds, where very little current flows. The only consideration left is inductive coupling, and this is theone area where your only good weapon is mechanical layout and shielding.
With a single rectifier, or more properly, common rectifier, things get difficult if it supplies multiple amplifiers. In theory, it is enough to keep the power ground point, where load and quiet ground split, common between both amps. In practise, this can ge VERY difficult, and in fact you may not be able to get a satisfactory solution if you have two or more amplifier boards with power ground points relatively far from each other. At first, one may not think this is a big issue - just use a common star ground. The problem arises when the distance is such that wiring to the star becomes long, in which case, power rail capacitors local to the power amp board may become ineffective due to increased effective inductance of the wire to the star ground. While in theory the wiring is faultless, you may get an oscillating amp, or one with severely increased load current induced hash on the power lines.
The usual solution lets the common point of the main filter caps become the star ground, from where two power ground lines go to each power ground point of the power amp boards. here is where the compromises begin. In separating the power ground, we now have separated clean grounds, when most sources have a common ground. Since the separated clean grounds galvanically conenct in the same point (star ground), and have separate paths, if they conenct at the oposite end, a low impedance voltage divider and a ground loop susceptible to inductive coupling, is created. This is always a problem, and the solution is usually a matter of compromise... usually, it will include a low value (10 ohms) resistor from te point where the local power ground and the local clean ground split, as well as careful consideration on where to put the feedback reference point of the amp. Some solutions I have seen even deliberately couple input lines with certain parts of the star to power ground connection to cancel out ground current or voltage differences. I will try to draw up some pictures early next week...
As can be seen, when rectification and filtering is completely independant, there is considerable freedom in wiring without compromising the ground reference. Ths is because it is simple to have all the various supplies connect only into on ground point.
This is particulairly useful if there are local caps on thea mplifier board, creating a local power ground. With completely independant rectification, this only means that you end up joining (or not, as you wish) the quiet grounds, where very little current flows. The only consideration left is inductive coupling, and this is theone area where your only good weapon is mechanical layout and shielding.
With a single rectifier, or more properly, common rectifier, things get difficult if it supplies multiple amplifiers. In theory, it is enough to keep the power ground point, where load and quiet ground split, common between both amps. In practise, this can ge VERY difficult, and in fact you may not be able to get a satisfactory solution if you have two or more amplifier boards with power ground points relatively far from each other. At first, one may not think this is a big issue - just use a common star ground. The problem arises when the distance is such that wiring to the star becomes long, in which case, power rail capacitors local to the power amp board may become ineffective due to increased effective inductance of the wire to the star ground. While in theory the wiring is faultless, you may get an oscillating amp, or one with severely increased load current induced hash on the power lines.
The usual solution lets the common point of the main filter caps become the star ground, from where two power ground lines go to each power ground point of the power amp boards. here is where the compromises begin. In separating the power ground, we now have separated clean grounds, when most sources have a common ground. Since the separated clean grounds galvanically conenct in the same point (star ground), and have separate paths, if they conenct at the oposite end, a low impedance voltage divider and a ground loop susceptible to inductive coupling, is created. This is always a problem, and the solution is usually a matter of compromise... usually, it will include a low value (10 ohms) resistor from te point where the local power ground and the local clean ground split, as well as careful consideration on where to put the feedback reference point of the amp. Some solutions I have seen even deliberately couple input lines with certain parts of the star to power ground connection to cancel out ground current or voltage differences. I will try to draw up some pictures early next week...
Tweeker said:
This should not be a problem, in either case they have to be dimensioned to withstand it, you just use more of the same. In any case, for large trafos you really need a soft start circuit if nothing else to be able to protect the amp properly with a fuse. One that survives 'bare' inrush will probably also survive everything catching fire... one of those cases where, somewhere, in the burnt-down ruins of your house, you hear a 'flink' of the fuse going open, when the foul deed is already done...
Please correct me if I am wrong. From the amplifying circuits' point of view, what they see from the PSU and the ground is the voltage, not current. We want to reduce the current loops and have individual current return paths of the PSU only because we want to reduce the voltage variations seen from the amplifying circuits because large current creates voltage variations due to the finite resistence of wires or conductors.
I believe what ilimzn said is very true. But I have been thinking about ways of reducing the problems ilimzn pointed out of which he has put in considerable efforts eliminating them. Refering to my 2nd and 3rd images, the way I layout the amp is to have an ample distance between the bridges and the amplifying circuits. With that sort of distance I don't think I need to worry about the radiation from the wires, twisted (good idea) or not. The radiation may influence the filter caps but that has far less consequence comparing to influence to the circuit board.
As for current returns, I use the big aluminium flat for the PSU common ground. Aluminium conducts very well and such large area of conduction will ensure that the resistence is at the minimum. Let's say a peak current of 10A is being returned to the bridge via the PSU ground. If we have a wire resistence of 0.001 ohm it will create a voltage of 10mV. But if the aluminium flat has a resistence of 0.0001 ohm the the voltage developed due to the ground returning current is only 1mV. This voltage is entirely within the PSU common. Since the signal common is far from the PSU common I really can't see I have to go through all the troubles of separating the grounds, due to the use of a large aluminium flat. Conceptually, if the resistence of the aluminium flat is so low in comparison to other wirings, electrically the entire aluminium flat can be regarded as a single point. In that case there is no separate current return paths to talk about.
So perhaps topology is one thing, while implementation makes it work.
Again, please correct me if I am wrong.
tlf9999,
Thanks for correcting me on the "cancellation" assumption. I fully agree with your points.
I believe what ilimzn said is very true. But I have been thinking about ways of reducing the problems ilimzn pointed out of which he has put in considerable efforts eliminating them. Refering to my 2nd and 3rd images, the way I layout the amp is to have an ample distance between the bridges and the amplifying circuits. With that sort of distance I don't think I need to worry about the radiation from the wires, twisted (good idea) or not. The radiation may influence the filter caps but that has far less consequence comparing to influence to the circuit board.
As for current returns, I use the big aluminium flat for the PSU common ground. Aluminium conducts very well and such large area of conduction will ensure that the resistence is at the minimum. Let's say a peak current of 10A is being returned to the bridge via the PSU ground. If we have a wire resistence of 0.001 ohm it will create a voltage of 10mV. But if the aluminium flat has a resistence of 0.0001 ohm the the voltage developed due to the ground returning current is only 1mV. This voltage is entirely within the PSU common. Since the signal common is far from the PSU common I really can't see I have to go through all the troubles of separating the grounds, due to the use of a large aluminium flat. Conceptually, if the resistence of the aluminium flat is so low in comparison to other wirings, electrically the entire aluminium flat can be regarded as a single point. In that case there is no separate current return paths to talk about.
So perhaps topology is one thing, while implementation makes it work.
Again, please correct me if I am wrong.
tlf9999,
Thanks for correcting me on the "cancellation" assumption. I fully agree with your points.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.