My understanding is that when a psu is going to be used for audio, then the design goes far beyond nominal rating. I was reading about terms like;
Q factor
Output impedance
Recovery time
Ripple and harmonics rejection
Signal envelope
I can't tell I got a clear picture of all these let alone there may be even more!
So, I thought it would be very educating if you could share the procedure you follow when designing a psu. What do you want from a psu and how do you get it? When are you happy with the result?
Thanks for any reply!
Q factor
Output impedance
Recovery time
Ripple and harmonics rejection
Signal envelope
I can't tell I got a clear picture of all these let alone there may be even more!
So, I thought it would be very educating if you could share the procedure you follow when designing a psu. What do you want from a psu and how do you get it? When are you happy with the result?
Thanks for any reply!
Not an easy one to answer.
Normally you want a voltage source that doesn't sag under load and that has zero ripple and noise present. It should current limit when the maximum designed for limit is reached.
All totally unrealistic goals for anything other than low current small signal circuitry
I'm happy when 'worsening' the power supply by a considerable margin (ripple, sag etc) has no significant impact on the performance of the circuit it feeds. That allows you to sensibly scale parts such as reservoir caps and so on to suit.
There are formulas and design procedures that allow you to calculate the required transformer size, cap size etc and that also predict such things as ripple and ripple currents that the caps might see.
So like everything, its always a compromise.
Normally you want a voltage source that doesn't sag under load and that has zero ripple and noise present. It should current limit when the maximum designed for limit is reached.
All totally unrealistic goals for anything other than low current small signal circuitry
I'm happy when 'worsening' the power supply by a considerable margin (ripple, sag etc) has no significant impact on the performance of the circuit it feeds. That allows you to sensibly scale parts such as reservoir caps and so on to suit.
There are formulas and design procedures that allow you to calculate the required transformer size, cap size etc and that also predict such things as ripple and ripple currents that the caps might see.
So like everything, its always a compromise.
OK! Let's start with this. We need a transformer that can supply the current plus a margin. How much overspecified should it be? If secondary's VA increaces, DCR decreaces and vice versa. So transformer's power rating affects Q factor. If we need to drop some voltage in the filtering then we have to start from higher AC when we use a transformer marginally specified. How do we choose?
I appreciate the whole reply but I would like a little more explanation on this, please!
Sure, that's the easy part. But we need it to sound good so we have to mind for dynamic performance as well.
Well that is more theoretical than practical on a large high current supply. To achieve current limiting requires that you have a series pass element somewhere in the chain, whether that be in the DC rail or even a means of interrupting the incoming AC supply in the event of an overload.
While its all do-able, you would find the power supply was probably more complex than the amplifier (or whatever) that the circuit supplies.
As with anything, you have to draw up your specification first and then see what is possible to achieve that. For example if you want a limit of 3 amps on a rail then you have to ask whether a 3.15 amp fuse would be acceptable or do you want something that trips instantly when say 3.01 amps has been reached. The fuse would work for any voltage you might come across in audio, including valve gear wheras a fast acting electronic version could be quite a challenge to design and get to reliably switch say 450 volts DC.
Draw up your specification and see how far from the ideal you are prepared to compromise.
First the equipment has to work. That is priority. It is no good it blowing up.
What is needed to get the equipment working. Develop the specification.
Once you have it working, then you can experiment with modifications/additions and measure/hear what difference if any they might make.
OK! Using calculations, tests and even a simulator we can reach to a simple and reliable design. From there on, if there is bench equipment and experienced ears, what do we look for?
What do you mean by "Q factor"? Do you mean 'Q' (i.e. the peakiness of a second-order filter) or just general 'quality'?MagicBus said:
That depends on what you mean by "overspecified". Obviously it needs to be able to supply the required current at the required duty cycle without overheating, but it may also need not to drop too mcuh voltage under load. This is another aspect of specification: PSU output impedance.
A PSU should never "sound good", at least, not in a hi-fi amp. It should not sound at all. This may be a matter of getting sufficiently low output impedance at subsonic frequencies - often overlooked. This relates to signal envelope.
Welcome DF96!
By "Q" I mean the damping factor that determines capacitors charging, ringing e.t.c. I've found references both for the Pi filters as well for rectifiers peak current. I didn't manage to understand all this completely.
Yes, psu output impedance is related to and affected by a lot of other parameters. I can't tell if there is a practical way to manipulate all this but for sure is very complex for people like me.
I understand the requirement for an inaudible psu. Please, provide some more info about signal envelope! Perhaps a simple example of how do you work this out in your designs?
So little specifics.
First you start with how much volume you need in the room. then you figure out how wide your dynamic range is. If you're reproducing classical music, you can get away with less transformer and handle the brief peaks with capacitance. If one intends to reproduce house music in a bar or disco, then 90% all the time current needs to flow at the desired volume. More transformer or switcher supply.
Then you figure how much voltage and current the speaker you own or are going to buy, takes to produce that much volume. All speakers have an impedance, which requires higher voltage and less current for 8 ohms than 2. Mostly speakers are not that well specified in current draw versus sound level, but certain PA quality speakers are. What people actually do is figure out how much transformer they can afford or need from their experience with previous systems that did not perform as well as they expected.
Which technology amp you use comes in at this point. Class AB, class A, Class D, class H? AB wastes about 50% of the energy, A more like 80%, D I don't know but it is more efficient and you have to have higher frequency delivery for D than for AB.
for example, I tried to use a 120 W amplifer to fill a 200 seat hall. Volume was okay, but the higher ratio of average to peak than home use meant the thing overheated and set fire to some parts. 8 ohm speakers.
So the next amp I bought had about 4x the watt rating, bigger heatsinks and a fan.
I bought another amp with 11x the 120 watt rating because it was cheap, but would probably need to play a 400 person crowd in the open to need that much wattage. It was cheap in parts because it weighs 55 lb. The transformer is 25 or 30 lb. Newer technology amps have switcher supplies and weigh less.
Some circuit designs have ripple rejection built in. The early stages have constant current sources, a lot of little cheap parts that make the current flow evenly whatever the ripplie is doing. OTOH I hand wire my boards, so 50 more wires of a constant current source is a nuisance. I use designs that require the rail voltage to be regulated, because I can build a regulator with about a dozen wires in an area of the amp where the parts don't need to be RF or hum isolated.
If you're selling, saving every penny counts. If you're building for yourself, volume or complication or parts availability is more important.
For example, in the US, a certain supplier of transformers charges the same price for 400 VA, 500 VA, 600 VA, 700 VA. In that case one would probably buy the 700. If you were selling , you could negotiate a discount for the smaller parts in quantity of 500.
They sell the different voltage levels at the same price, so if you know you're not going to drive 8 ohm speakers, you buy a lower voltage output transformer, and save on heat sink and fan, because less voltage is wasted idling waiting for a peak current demand.
It is all engineering.
I buy old systems and repair them, which gives a very up close & personal view of what is necessary and what is not. If I'm building something, I can use experience from the systems I've owned to decide what I need for the new system. This is not unique to individuals. One of the amps I own, somebody said on here the manufacturer copied the configuration of a boutique amp from a manufacturer who was not commercially successful. A little manufacturing engineering to drive the costs down, voila, a successful commercial product sold by the thousands.
First you start with how much volume you need in the room. then you figure out how wide your dynamic range is. If you're reproducing classical music, you can get away with less transformer and handle the brief peaks with capacitance. If one intends to reproduce house music in a bar or disco, then 90% all the time current needs to flow at the desired volume. More transformer or switcher supply.
Then you figure how much voltage and current the speaker you own or are going to buy, takes to produce that much volume. All speakers have an impedance, which requires higher voltage and less current for 8 ohms than 2. Mostly speakers are not that well specified in current draw versus sound level, but certain PA quality speakers are. What people actually do is figure out how much transformer they can afford or need from their experience with previous systems that did not perform as well as they expected.
Which technology amp you use comes in at this point. Class AB, class A, Class D, class H? AB wastes about 50% of the energy, A more like 80%, D I don't know but it is more efficient and you have to have higher frequency delivery for D than for AB.
for example, I tried to use a 120 W amplifer to fill a 200 seat hall. Volume was okay, but the higher ratio of average to peak than home use meant the thing overheated and set fire to some parts. 8 ohm speakers.
So the next amp I bought had about 4x the watt rating, bigger heatsinks and a fan.
I bought another amp with 11x the 120 watt rating because it was cheap, but would probably need to play a 400 person crowd in the open to need that much wattage. It was cheap in parts because it weighs 55 lb. The transformer is 25 or 30 lb. Newer technology amps have switcher supplies and weigh less.
Some circuit designs have ripple rejection built in. The early stages have constant current sources, a lot of little cheap parts that make the current flow evenly whatever the ripplie is doing. OTOH I hand wire my boards, so 50 more wires of a constant current source is a nuisance. I use designs that require the rail voltage to be regulated, because I can build a regulator with about a dozen wires in an area of the amp where the parts don't need to be RF or hum isolated.
If you're selling, saving every penny counts. If you're building for yourself, volume or complication or parts availability is more important.
For example, in the US, a certain supplier of transformers charges the same price for 400 VA, 500 VA, 600 VA, 700 VA. In that case one would probably buy the 700. If you were selling , you could negotiate a discount for the smaller parts in quantity of 500.
They sell the different voltage levels at the same price, so if you know you're not going to drive 8 ohm speakers, you buy a lower voltage output transformer, and save on heat sink and fan, because less voltage is wasted idling waiting for a peak current demand.
It is all engineering.
I buy old systems and repair them, which gives a very up close & personal view of what is necessary and what is not. If I'm building something, I can use experience from the systems I've owned to decide what I need for the new system. This is not unique to individuals. One of the amps I own, somebody said on here the manufacturer copied the configuration of a boutique amp from a manufacturer who was not commercially successful. A little manufacturing engineering to drive the costs down, voila, a successful commercial product sold by the thousands.
Last edited:
Hi Indianajo,
Thank you for putting this into perspective. A particular PSU will not perform the same on different gain stages. It needs to be designed as a system. Starting with a PSU that will bring tha gain circuit to life and it won't set to fire and then improve it soundwise. At this point I' m confused; the thing has to be fast but then it will sag or trying to get high frequency response we may loose low frequencies.
Thank you for putting this into perspective. A particular PSU will not perform the same on different gain stages. It needs to be designed as a system. Starting with a PSU that will bring tha gain circuit to life and it won't set to fire and then improve it soundwise. At this point I' m confused; the thing has to be fast but then it will sag or trying to get high frequency response we may loose low frequencies.
this is the starting point.....
what load current and what load voltage will the psu see?
what duty factor? is it 24/7? ir just a couple of hours at time?
what ripple voltage is tolerable?
what temperature rise is acceptable?
then finally how much is your budget?
if you can put in numbers, then the rest will be easy...
My 120 W home amp has a 3300 uf 80 v single rail cap for two channels. That is adequate for brief peaks of 30 v out (regulated to 70 v) but the heat sinks are not adequate to do that constantly. The transformer was 6.75 A @ 80 V. They didn't worry about high frequency in 1966, the output transistors wouldn't reproduce it, but I've added .1 uf ceramic caps parallel because the new MJ15003 class output transistors do actually do a nice job on tinkly bells & top octave piano. This transformer never got especially hot even in the 3.5 hour rehearsal where the base stopper resistors caught fire.
The CS800s produces 260 W/ch 8 ohms and 400 w/ch 4 ohms and is entirely adequate to drive the PV1210 speakers I used in 200 seat halls. It gets away with 2000 uf rail cap + and -, apparently because it is a switcher supply and they can count on help from the solid state feed. Transformer they would have put in more rail capacitance I suspect.
The PV-1.3k has 10000 uf + and - rail capacitance, probably a bigger ratio to power rating than the CS800s, because it has a transformer feed. It has local ceramic caps on the driver board, but only .47 uf film caps on the rails of the output transistor board. Doesn't seem to hurt the high frequencies any, but I don't listen to it at the 124 db @ 300 W my SP2-XT speakers would reproduce.
I'm building an amp for my summer camp. I have a portable radio out there, that I can't hear when I'm working on cars or cutting up tree limbs 50' away. So I need at least 9 db more than 1 watt, or at least 8 watts. more would allow volume peaks.
The IC I selected will produce 25 W according to the datasheet. While the IC will take up to 60 V single rail, I noticed reading the SOA tables that wattage out was better at about 40 V. I have three possible transformers in stock. A 24 VCT 1 amp doorbell transformer, a 24-28-30-33-36 @ 2.5 A no center tap transformer, and a transformer advertised as 70 VCT but turned out to be 110 VCT @ 2 amps.
The speaker I have is a monitor speaker rated 8 ohms & 200 W. I've decided the 1 amp I could get from the doorbell transformer is not enough for the volume I want. I've decided using one side of the 55 v transformer single ended wastes too much heat. So I'm using the 2.5 A @ multitap transformer at 24 V. I've bought 3300 uf filter caps because that is enough for the ST120, and they are about $2. The datasheet specifies 2200 uf caps as speaker couplers, but I like a little low bass sometimes, particularly on rock CD's. 3300 is only about 10% more than 2200. I'm putting .1 uf ceramic caps on the rails near the IC, because the datasheet says to - also experience confirms that.
The whole amp has to fit in a bicycle basket, as that is how I motorvate these days, so the 2.5 A multitap transformer comes in a 5" high 19" relay rack chassis that used to be a DC power source.
I also had available a 48 V 20 A linear supply, but that transformer is probably overkill for what I need out there. If it gets too nice, somebody will steal it. Need to look like ****, I always find the front door jimmied open by kids or hunters in spring, but most of what they've stolen I didn't need. Like the jon-boat.
So that is how I approach engineering. What can I buy, how much volume do I need, what has worked in the past?
The above was all class AB. I suspect class D you would need way more high frequency capable capacitance, liked stacked film caps (not wound) or massive banks of parallel ceramic caps, but I've never cut apart a class D amp to know what standard practice is.
The CS800s produces 260 W/ch 8 ohms and 400 w/ch 4 ohms and is entirely adequate to drive the PV1210 speakers I used in 200 seat halls. It gets away with 2000 uf rail cap + and -, apparently because it is a switcher supply and they can count on help from the solid state feed. Transformer they would have put in more rail capacitance I suspect.
The PV-1.3k has 10000 uf + and - rail capacitance, probably a bigger ratio to power rating than the CS800s, because it has a transformer feed. It has local ceramic caps on the driver board, but only .47 uf film caps on the rails of the output transistor board. Doesn't seem to hurt the high frequencies any, but I don't listen to it at the 124 db @ 300 W my SP2-XT speakers would reproduce.
I'm building an amp for my summer camp. I have a portable radio out there, that I can't hear when I'm working on cars or cutting up tree limbs 50' away. So I need at least 9 db more than 1 watt, or at least 8 watts. more would allow volume peaks.
The IC I selected will produce 25 W according to the datasheet. While the IC will take up to 60 V single rail, I noticed reading the SOA tables that wattage out was better at about 40 V. I have three possible transformers in stock. A 24 VCT 1 amp doorbell transformer, a 24-28-30-33-36 @ 2.5 A no center tap transformer, and a transformer advertised as 70 VCT but turned out to be 110 VCT @ 2 amps.
The speaker I have is a monitor speaker rated 8 ohms & 200 W. I've decided the 1 amp I could get from the doorbell transformer is not enough for the volume I want. I've decided using one side of the 55 v transformer single ended wastes too much heat. So I'm using the 2.5 A @ multitap transformer at 24 V. I've bought 3300 uf filter caps because that is enough for the ST120, and they are about $2. The datasheet specifies 2200 uf caps as speaker couplers, but I like a little low bass sometimes, particularly on rock CD's. 3300 is only about 10% more than 2200. I'm putting .1 uf ceramic caps on the rails near the IC, because the datasheet says to - also experience confirms that.
The whole amp has to fit in a bicycle basket, as that is how I motorvate these days, so the 2.5 A multitap transformer comes in a 5" high 19" relay rack chassis that used to be a DC power source.
I also had available a 48 V 20 A linear supply, but that transformer is probably overkill for what I need out there. If it gets too nice, somebody will steal it. Need to look like ****, I always find the front door jimmied open by kids or hunters in spring, but most of what they've stolen I didn't need. Like the jon-boat.
So that is how I approach engineering. What can I buy, how much volume do I need, what has worked in the past?
The above was all class AB. I suspect class D you would need way more high frequency capable capacitance, liked stacked film caps (not wound) or massive banks of parallel ceramic caps, but I've never cut apart a class D amp to know what standard practice is.
Last edited:
This is very true.
My first amp build used 30 volt rails.
It clipped so easily I had to build a new power supply with higher volts.
Its down to the music signal. The music signal is far from a sine wave and has high transient peaks which can easily get clipped.
You really need to be at least 45-50 volts to get a decent sound.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.