Hi All. I'm about to make another push-pull power amp (100W) and was wondering if I should have fuses on the supply rails (+-50V) Currently I have an amp that I made and simply have the DC supplies DIRECT from the main power supply. I do however have a LS protection board on it. My thinking is that the fuses (Fast/Normal or Slow?) is that they introduce quite some resistance (non-linear?) to the supply and spoil the very solid 40,000uF I have on each channel. They are, after all, just lengths of very thin fuse wire and I use silver-plated thick connecting wire for minimum resistance. What are the thoughts on these fuses please? I am tempted, as usual, to leave them out. The way I see it is that the only time they will blow is when one or more of the O/P transistors blows (or maybe just a smaller component) and they will in themselves, act as fuses? If any component does blow/burn out it will need replacing anyway so nothing has been saved by the fuses. PS: I will use them for the first powering up session.
Last edited:
The real question is what happens if you do get a short on a rail. Does the primary side fuse blow, does the transformer catch fire... and so on.
Fwiw a Pioneer A80 I once had used no fuses apart the primary side one. The design must have been such that it was 'safe' and it would rely on the safety characteristic of things like the emitter resistors in the power stages.
So everything has to be considered as a whole, you have to look at the big picture. Ask what would happen if you fitted shorting links in place of the output transistors and switched it on and went out for the day. Would you be confident in the outcome.
I would not rely on an output transistor failing open circuit under a heavy current condition, my experience is they fail dead short.
Fwiw a Pioneer A80 I once had used no fuses apart the primary side one. The design must have been such that it was 'safe' and it would rely on the safety characteristic of things like the emitter resistors in the power stages.
So everything has to be considered as a whole, you have to look at the big picture. Ask what would happen if you fitted shorting links in place of the output transistors and switched it on and went out for the day. Would you be confident in the outcome.
I would not rely on an output transistor failing open circuit under a heavy current condition, my experience is they fail dead short.
Thanks for your thoughts Mooly. I understand the risks, but do you think the voltage drop across them compromises the power supplies solidity voltage wise and makes the use of good connecting wire pointless? Are they even a linear resistor? Like a chain is only as good as its weakest link PS: I never leave anything on connected to the mains when I go out, not even a phone charging up, maybe a light bulb but that's it
Thanks Nanofarad. Yes that seems to be a good way. I guess fast acting fuses with large torroidal transformers may cause irritation, but a well-judged low value normal fuse would blow if a big short occured.... Makes me want to check what value fuse I used in my existing amp. PS: What made me ask is that I noticed the Blameless amp uses them on its PCB
I was actually going to say the Blameless used them because I built that one on the official boards back in the day.
In practice I don't think it does compromise the supply to any real extent. Some may disagree but I think a suitable fuse is fine.
Comparing to connecting wire is a bit of a different thing because the fuse has a non linear resistance vs current but the effects of that are very small with a suitably chosen fuse. I would guess for -/+50v rails (lets say 120 watt amp) and a massive capacitor block then a fuse of around T5A or T6.3A would certainly blow if a rail short occurred.
That assumes the mains transformer is also suitably rated (say 500 or 650 VA) and would deliver enough current to blast a fuse if a short occurred.
You also have to think of fusing on the secondaries (not rail fusing) and what happens if the bridge fails or a cap fails. It all has to be safe. Would a primary fuse blow if the bridge went short. It may sounds harsh but if you are unsure then maybe you should try scenarios like that to see what really does happen. Would enough primary current flow to pop the fuse? or would it sit there frying and be a fire hazard.
In practice I don't think it does compromise the supply to any real extent. Some may disagree but I think a suitable fuse is fine.
Comparing to connecting wire is a bit of a different thing because the fuse has a non linear resistance vs current but the effects of that are very small with a suitably chosen fuse. I would guess for -/+50v rails (lets say 120 watt amp) and a massive capacitor block then a fuse of around T5A or T6.3A would certainly blow if a rail short occurred.
That assumes the mains transformer is also suitably rated (say 500 or 650 VA) and would deliver enough current to blast a fuse if a short occurred.
You also have to think of fusing on the secondaries (not rail fusing) and what happens if the bridge fails or a cap fails. It all has to be safe. Would a primary fuse blow if the bridge went short. It may sounds harsh but if you are unsure then maybe you should try scenarios like that to see what really does happen. Would enough primary current flow to pop the fuse? or would it sit there frying and be a fire hazard.
Class AB amps with two rails, each having a fuse, is a stupid idea. If one fuse blows, the output inevitably goes to max DC on the other rail, which would damage your speaker. BTW transistor die leads will always burn to protect a fuse rated for max output current.
A real protection circuit that detects DC on output for longer than 400 ms, and opens a suitable relay to disconnect the speaker, is a much better idea. Unfortunately 99% of protection circuits for sale on aliexp, ebay, amazon, gumtree, have totally unsuitable relays rated to break AC current only. The problem is DC current, not AC. Protection relays inside PA amps from reputable vendors have only an AC rating, but the vendors are reputable because they have tested that relay to break the maximum current their amp will produce in a fault.
Reputation has nothing to do with 5 star ratings on websites, because those ratings are based on visual iinspection of the product, at best maybe a 5 minute function check. Retail shops stock reputable brands because they do not come back to the service department for warranty repairs very often. The lack of retail shops in today's market means a real lack of that sort of valuable information.
A real protection circuit that detects DC on output for longer than 400 ms, and opens a suitable relay to disconnect the speaker, is a much better idea. Unfortunately 99% of protection circuits for sale on aliexp, ebay, amazon, gumtree, have totally unsuitable relays rated to break AC current only. The problem is DC current, not AC. Protection relays inside PA amps from reputable vendors have only an AC rating, but the vendors are reputable because they have tested that relay to break the maximum current their amp will produce in a fault.
Reputation has nothing to do with 5 star ratings on websites, because those ratings are based on visual iinspection of the product, at best maybe a 5 minute function check. Retail shops stock reputable brands because they do not come back to the service department for warranty repairs very often. The lack of retail shops in today's market means a real lack of that sort of valuable information.
I respectfully disagree. In case one side of the amp shorts, the other side tries to compensate by pumping max available current, much more current flow than through the speaker. At the end both fuses will blow. This is my experience with "normal" power amps in the 100W ballpark.
Things are different at real high power. With my former 5kW LatFET-MOSFET amps (long time ago) I burned some latFETs wthout noticing at first, because the bonding wires had evaporated.
And for regular repairs both fuses are a bless. You can replace them by some 100 Ohm power resistor for easy troubleshooting.
The Hafler DH220 uses fuses directly in the power supply lines going to the output FETs, which are lateral devices. A properly sized fuse can indeed protect them but would be ineffectual for bipolar devices. However, they might serve to mitigate a cascading failure if one device blows—it depends on your circuit topology, etc.
As for affecting the performance of your amp, I doubt that the small voltage drop across the fuse would be in any way measurable. Consider that the fuse is effectively within the feedback loop, and in any event, its voltage drop would be small relative to the ripple on the power supply lines. Also, with such large filter capacitors, the charging interval will be small, and the circulating currents in the transformer-diode-capacitor loop will be substantial, likely limited by the impedance of the transformer itself. The effects of this will probably swamp any distortion caused by the fuse.
I would say to put the fuses in. They may not save the output transistors, but they will do no harm and might even provide a convenient way to isolate a portion of your circuit for debugging and troubleshooting.
Just my 2 cents worth!
As for affecting the performance of your amp, I doubt that the small voltage drop across the fuse would be in any way measurable. Consider that the fuse is effectively within the feedback loop, and in any event, its voltage drop would be small relative to the ripple on the power supply lines. Also, with such large filter capacitors, the charging interval will be small, and the circulating currents in the transformer-diode-capacitor loop will be substantial, likely limited by the impedance of the transformer itself. The effects of this will probably swamp any distortion caused by the fuse.
I would say to put the fuses in. They may not save the output transistors, but they will do no harm and might even provide a convenient way to isolate a portion of your circuit for debugging and troubleshooting.
Just my 2 cents worth!
In a well designed differential class AB amp with feedback, voltage at output is not dependent on the power supply voltage variations under non clipping condition.
The only moment that this happens is when voltage at the output is close to the voltage rail and close to clipping condition, so radical distortion will be happening anyways - this is not the hi-fi operating condition.
The prove is that you don't hear the 120Hz ripple from the power supply nor you see the low frequency voltage level variations (5 to 10% in general) in the speaker as the current drain increases.
The only moment that this happens is when voltage at the output is close to the voltage rail and close to clipping condition, so radical distortion will be happening anyways - this is not the hi-fi operating condition.
The prove is that you don't hear the 120Hz ripple from the power supply nor you see the low frequency voltage level variations (5 to 10% in general) in the speaker as the current drain increases.
Thanks LVQ and ron68. I have measured some fuses and the resistance, at most, is less than 100 milliohms and that surprises me. I thought it would be more. However, that is cold so I would imagine it needs to get very much hotter to melt (hundreds of degrees C) at which point the resistance must be very much higher. It would make an interesting experiment to plot the response of current vs resistance up to blowing point. Perhaps the resistance before it blows could be 1 Ohm and maybe more, say when using the amp for high power output? The higher the resistance the greater the voltage drop, so the power rails voltage at the amp would be modulated by the signal frequency more than the LF ripple? I concede that both rails would be modulated by the same amount. If then, I take what you say, regarding no effect on distortion and power, then I might as well use quite thin wires from my power supply and also use a very modest amount of filtering capacitance as the push pull amplifier will reject the shortcomings by its very design?
I'm heavily on the fused side of the fence. I have seen plenty of amplifiers open on supply fuse without the output going to the opposite rail.
Don't worry about the main filter capacitors. There is enough inductance in between already. High current fuses have low resistance, and all are very temperature dependent in resistance. After the fuse and local to the point of use, install bypass capacitors to retain a low HF impedance. And yes, they can really aid in troubleshooting.
Don't worry about the main filter capacitors. There is enough inductance in between already. High current fuses have low resistance, and all are very temperature dependent in resistance. After the fuse and local to the point of use, install bypass capacitors to retain a low HF impedance. And yes, they can really aid in troubleshooting.
Normal rail voltages variations will be much higher in amplitude than the influence of the fuse resistance.
An if the fuse is getting very hot, it means something is already wrong.
You should not design a fuse to work so hot in normal condition.
Non-stabilized power supplies can have variations around 10% or more.
As long as you design your amp correctly, so isolating these variations for the diferential and VAS stages, including good amount of negative feedback, those variations will be only on power BJT collectors and will not be reflected to the BJT emitters, which is the output of your amp. Output stage is composed of emitter followers, so independent of collector voltage as long as collector is higher enough than emitter (some volts).
And even if some voltage variation reaches the differential stage, its nature (differential) will reject the common mode voltage variations.
Again, a good design with current source, all symetrical etc so as to achieve a good common mode rejection.
As you can see, a good amp design have several layers or means of rejecting common mode variation (differential amp, power supply low signal filtration, emitter followers on the output, negative feedback etc).
Thank you ron.... I am becoming happier to include rail fuses now. However, I'm still slightly confused regarding Power Supply filtering. For many years I have always believed that an Un-regulated bridge rectifier type supply needed as much capacitance as you could throw at it. It seems that is not the case, and we are wasting money exceeding quite modest amounts of PS smoothing capacitance? I've always used 40K uF per channel but maybe 4K uF would do just as well as long as the transformers regulation was good by using a large enough value of VA. If this is NOT the case and the objective is to have the lowest supply ripple, then we are back with the fact that the fuses WILL introduce ripple, albeit a very small one?
Wether an amp output will stick to a power rail if the other rail is cut off is topology dependent, FI in a typical blameless
with say a NPN differential driving a PNP VAS the output will stick to the negative rail if the positive rail is cut off,
basicaly the differential is rendered inactive and the VAS CCS will get the OS input to the max negative voltage.
Same for a PNP differential and NPN VAS, in wich case the amp will stick to the positive rail if the negative one
is cut off., other topologies could behave differently, most cautious is to implement a speakers DC protection
and eventualy rely on a fuse at main, although it could well be unable to keep the circuit from a catastrophic
chain reaction failure, at some point when there will be a good short somewhere in the circuit it will blow,
or not by lack of other components to smoke.
with say a NPN differential driving a PNP VAS the output will stick to the negative rail if the positive rail is cut off,
basicaly the differential is rendered inactive and the VAS CCS will get the OS input to the max negative voltage.
Same for a PNP differential and NPN VAS, in wich case the amp will stick to the positive rail if the negative one
is cut off., other topologies could behave differently, most cautious is to implement a speakers DC protection
and eventualy rely on a fuse at main, although it could well be unable to keep the circuit from a catastrophic
chain reaction failure, at some point when there will be a good short somewhere in the circuit it will blow,
or not by lack of other components to smoke.
The ripple on the main filter capacitors is primarily from the current drawn by the output stage.
For class AB, the ripple increases with power level, but is more than the drop on the fuse.
The ripple due to the capacitor only can be estimated from I = C x dV / dT and then dV = I / ( 120 x C )
(where dT is 1/120 seconds, C in Farads, I in Amps, V in Volts).
If at a high power level, I = 10A and C = 10,000uF, this gives dV = 8.3V ripple
Just ran those scenarios in SPICE for the Blameless-type amplifier I'm working on. It wouldn't have been pretty in real life with a speaker connected. VCC-transistor drop on the output. :O
Hi,
The amount of capacitance you put in the power supply has 2 goals: reduce the ripple to a reasonable value and also give some dynamic power to the short transients.
Suppose +/-30V rails, 4ohms and disregard the losses just for an example.
Suppose also that your power supply drops 10% at maximum power output (mostly due to transformer coil resistances).
Peak power (short bursts) will be: ((30*0,7)^20)/4 =110W
Continous power (continuous sine wave 1kHz) will be: ((30*0,7*0.9)^2)/4 = 89W
The more capacitance you put, the longer the burst can be, but after some value, not much magic.
You have to balance how much you spend on capacitors and how much you get in extra dynamic power or ripple reduction.
You can easily simulate this and take your own decision.
I'd say that 10.000uF for 4ohms per channel is good enough.
Others might choose more or even less.
An complementing of what you said, yes, the more VA your transformer has (less coil resistance) less continuous to dynamic power difference you will observe (the difference is based on how much the voltage drops or hit the botton of the ripple wave).
And smal capacitance will not allow you much difference between continuous and dynamic power. A shot burst and the voltage rails will quickly drop to the continuous minimum.
And yes, the more capacitance, the better, but there are cost and size to consider (and also high inrush current at start up if you go crazy with the size of capacitors).
I like the "good enough" way of designing things.
And smal capacitance will not allow you much difference between continuous and dynamic power. A shot burst and the voltage rails will quickly drop to the continuous minimum.
And yes, the more capacitance, the better, but there are cost and size to consider (and also high inrush current at start up if you go crazy with the size of capacitors).
I like the "good enough" way of designing things.
And this is with an amp that was eventualy still working, at least timely, as a too low load and hence too high output current
could well get the fuse opened while the amp is still functional, at this point what was supposed to be preventive end being
a total disaster.