Here, I am only talking about hum as a result of ripple voltage entering the pre-amplifiers.
This ripple is present from the moment you attach *anything* to a simple bridge/filter PSU.
The filter capacitors discharge into the load continuously, but 100 times a second they also charge from the bridge - the result is a sawtooth, very sharp rise and a gentle fall.
Therefore, in my opinion, calling this voltage "DC" is plain wrong, it is modulated DC, it has a base and a variable component. The base component is the smallest voltage as seen on the oscilloscope, the ripple component is at the top of that.
This formula is a very good approximation if you do not have a scope: Ripple voltage peak to peak = I (current drawn) / 2 * f * C, where f is 50Hz for EU.
The more current you draw the worse ripple voltage gets.
This ripple voltage enters your amplifiers and generates hum at the speakers, even with pots turned to 0 sometimes!
This ripple voltage is vile. Assume a typical 12 V DC suppy and a current of say 150 mA to power all your pre-amps, LEDs, tone controls, mixers etc. Your 50VA transformer is man enough for the job and your 4000uF capacitors feel like an overkill for just 150 mA.
However, using the above formula, or a scope, will show a ripple voltage of 380mv peak to peak!!!!
This competes easily with the signal that my guitar puts out, and it completely obliterates other sources, eg a microphone. It will enter the base and the collector paths of my pre-amps and arrive at the power amp amplified dozens of times resulting in a huge hum coming out of the speakers.
There are 4 ways to beat it that come to mind.
1. Include RC filters everywhere you can. Or, if you know from the start how much current you will be drawing then you could have just a single R/C to power all pre-amps in your system. For more flexibility it is better to have local R/C filters at every amplifier / buffer stage, it means you can keep adding/modifying components without affecting the existing ones.
In order to eliminate hum, so that ear or scope cannot see it, I have found you need the single pole R/C to have a frequency of 1 Hz or better. For example if R is 100 Ohm then C needs to be 1600uF. Capacitors like this are bulky and cost quite a lot. You can increase R (which costs nothing) and decrease C accordingly - but increasing R means a greater voltage
drop, and if R exceeds 10% of the load you are driving then you will substantially modify the available voltage and you will have to sit down and recalculate all bias points for every transistor (more or less). Ouch.
2. Take a single, passive R/C, and add a zener diode then buffered by a bulky
transistor. You first calculate how many volts will be dropped on the R (the rest will be on the zener). You then choose an R to allow you to drive the zener so that it does not get hot (5mA is very typical for little, 0.5Watt zeners). Based on that R you choose a C to achieve the single-pole low pass of 1Hz (or better). You then bootstrap a heavy transistor
capable of withstanding shorts (eg a 50W/8A transistor can definitely survive temporary shorts from a smallish, pre-amp PSU). I used the audio transistor MJE15030 which has quite high hFE, and very linear as a bonus, unlike some other rubbish eg TIP31C/BD911 etc. When you connect all your pre-amps to this semi-regulated supply, there will be like 2 dozen 100uF capacitors all eager to be charged, and they will in general blow the likes of BC639/BC337 in an instant. Additionally it is more than likely that you will inadvertently short the supply yourself, and every time you do that it means a blown "lesser" transistor. Hence the need for an 8A transistor capable of withstanding even prolonged shorts. A 100uF capacitor at the emitter of the transistor is also needed to stop oscillations that might arise if you, for example, decide to power a fan from this regulated voltage. Of course, if you can, it is 1000 times better to power fans, relays, LEDs etc from the unregulated voltage and keep the "clean" voltage just for the amps where the audio signal will flow.
3. A combination of the above: a regulated supply and local R/C filters on every stage! Assuming that you can afford a 20-30% voltage drop on the R, then you can use smaller Cs and it should not cost that much. But you will have to design the circuits from the outset - you cannot easily do it retrospectively, eg if you have planned a 12V supply, you cannot feed it 9 volts and expect it to work just as well. Well, unless your circuits are somehow immune to voltage variations, like op-amps.
4. Use monster caps at the PSU. Remember we are talking pre-amps here, and for my typical example of 12V and 150mA, I would need something like 100,000 uF to achieve a 20mV ripple voltage! 20mV is still huge it terms of pre-amps...
I hope the above info will be of some use to someone 🙂
This ripple is present from the moment you attach *anything* to a simple bridge/filter PSU.
The filter capacitors discharge into the load continuously, but 100 times a second they also charge from the bridge - the result is a sawtooth, very sharp rise and a gentle fall.
Therefore, in my opinion, calling this voltage "DC" is plain wrong, it is modulated DC, it has a base and a variable component. The base component is the smallest voltage as seen on the oscilloscope, the ripple component is at the top of that.
This formula is a very good approximation if you do not have a scope: Ripple voltage peak to peak = I (current drawn) / 2 * f * C, where f is 50Hz for EU.
The more current you draw the worse ripple voltage gets.
This ripple voltage enters your amplifiers and generates hum at the speakers, even with pots turned to 0 sometimes!
This ripple voltage is vile. Assume a typical 12 V DC suppy and a current of say 150 mA to power all your pre-amps, LEDs, tone controls, mixers etc. Your 50VA transformer is man enough for the job and your 4000uF capacitors feel like an overkill for just 150 mA.
However, using the above formula, or a scope, will show a ripple voltage of 380mv peak to peak!!!!
This competes easily with the signal that my guitar puts out, and it completely obliterates other sources, eg a microphone. It will enter the base and the collector paths of my pre-amps and arrive at the power amp amplified dozens of times resulting in a huge hum coming out of the speakers.
There are 4 ways to beat it that come to mind.
1. Include RC filters everywhere you can. Or, if you know from the start how much current you will be drawing then you could have just a single R/C to power all pre-amps in your system. For more flexibility it is better to have local R/C filters at every amplifier / buffer stage, it means you can keep adding/modifying components without affecting the existing ones.
In order to eliminate hum, so that ear or scope cannot see it, I have found you need the single pole R/C to have a frequency of 1 Hz or better. For example if R is 100 Ohm then C needs to be 1600uF. Capacitors like this are bulky and cost quite a lot. You can increase R (which costs nothing) and decrease C accordingly - but increasing R means a greater voltage
drop, and if R exceeds 10% of the load you are driving then you will substantially modify the available voltage and you will have to sit down and recalculate all bias points for every transistor (more or less). Ouch.
2. Take a single, passive R/C, and add a zener diode then buffered by a bulky
transistor. You first calculate how many volts will be dropped on the R (the rest will be on the zener). You then choose an R to allow you to drive the zener so that it does not get hot (5mA is very typical for little, 0.5Watt zeners). Based on that R you choose a C to achieve the single-pole low pass of 1Hz (or better). You then bootstrap a heavy transistor
capable of withstanding shorts (eg a 50W/8A transistor can definitely survive temporary shorts from a smallish, pre-amp PSU). I used the audio transistor MJE15030 which has quite high hFE, and very linear as a bonus, unlike some other rubbish eg TIP31C/BD911 etc. When you connect all your pre-amps to this semi-regulated supply, there will be like 2 dozen 100uF capacitors all eager to be charged, and they will in general blow the likes of BC639/BC337 in an instant. Additionally it is more than likely that you will inadvertently short the supply yourself, and every time you do that it means a blown "lesser" transistor. Hence the need for an 8A transistor capable of withstanding even prolonged shorts. A 100uF capacitor at the emitter of the transistor is also needed to stop oscillations that might arise if you, for example, decide to power a fan from this regulated voltage. Of course, if you can, it is 1000 times better to power fans, relays, LEDs etc from the unregulated voltage and keep the "clean" voltage just for the amps where the audio signal will flow.
3. A combination of the above: a regulated supply and local R/C filters on every stage! Assuming that you can afford a 20-30% voltage drop on the R, then you can use smaller Cs and it should not cost that much. But you will have to design the circuits from the outset - you cannot easily do it retrospectively, eg if you have planned a 12V supply, you cannot feed it 9 volts and expect it to work just as well. Well, unless your circuits are somehow immune to voltage variations, like op-amps.
4. Use monster caps at the PSU. Remember we are talking pre-amps here, and for my typical example of 12V and 150mA, I would need something like 100,000 uF to achieve a 20mV ripple voltage! 20mV is still huge it terms of pre-amps...
I hope the above info will be of some use to someone 🙂
I like the idea of the pass transistor best. Try using a CFP instead of Darlington, as it will have lower output impedance (don't know about ripple isolation though, I think it's the same).
Look at the attached schematic for a simple ripple suppression technique that works well.
- keantoken
Look at the attached schematic for a simple ripple suppression technique that works well.
- keantoken
Attachments
I tend to prefer a good, discrete, mosfet shunt regulator. The latest incarnation I got measures around 10uV ripple with a 200mA load, and less than 100uV ripple with a sine load of 40mA p-p over the 200mA DC. That to me is acceptable even for phono stages.
This is the essense of my approach: to design (and understand) as much as possible from first principles. My zener regulator plus R/C filters comprise of 4-5 components in total, and I have experimented and seen on the scope and heard on the speakers the effects of ripple, filters, different capacitors, oscillations, blown transistors due to accidental shorts, current limiting etc.
It would feel like cheating to use someone else's design with two dozen components and techniques I do not (yet) understand - I might as well go buy a ready made guitar amp 🙂
I attach the final circuit I came up with - nothing clever but hum has almost disappeared. I will try (to understand) the shunt regulator suggested here on my next project. Assuming I ever finish this one 🙂
Attachments
akis, it all depends of the circuits you want to feed. It's a big difference between some opamp circuits and a MC preamp built with discrete parts.
To reduce hummm also remember good basic grounding schemes. Earth that metalic case. GND of PSU connect to earth star point too (something I forgot to do once and spent a DAY tracing that damn HUMMMMMMMM). 🙄
Andy
Andy
That circuit, which I drew from memory is slightly wrong, there are 2 * 1N4148 on the back of the zener, to raise the voltage a bit to compensate for the drop across the BE of the transistor. I get 12.60V output from about 14.3V input.
The whole system sits on a wooden plank (piece of plywood). No metal to be seen anywhere. There is a "sort of" star grounding scheme where every ground, from every sub-stage and component, goes back to the PSU in a separate wire, and where signal grounds are separate to power grounds (those used by bypass caps). Power rails also have their own wires all the way back to the PSU. Lots of wires.
So far I have discovered just one ground loop problem, where I had to connect the grounds between the two PSUs : the pre-amp's and the power-amp's. I was connecting the two PSUs together, but that resulted in intermittent and ocassional hum. The proper ground connection now connects the signal ground of the pre-amp output to the amp's PSU and this occasional hum has also immediately disappeared.
I actually use the same circuit for regulating a high voltage smps (200V) vor a stereo tube pre-amp. I used a Mosfet. The capacitance of the mosfet tends to lower some of the glitter and other junk from the smps. Also a Mosfet has a lower Rdson generating less heat.
Are you sure about that? Rdson only relevant when MOSFET fully enhanced (also Rdson also normalized figure dep on heat). Is it the case that when applied in this application (shunt regulator) device dissipates same amount of power regardless of device used (Transistor / MOSFET / whatever)?
I'm probably miles off though, so ignore me!
Andy
This parameter has an importance only if you use the mosfet as a switch. This is the lowest resistance you can get from the mosfet but in a linear regulator is the whole idea not to take advantage of the Rdson. Heat comes from the voltage across drain and source and the current flowing through the mosfet.
The typical Guitar amp will use a multi-stage pi filter. Each stage of the filter knocks the voltage down a bit and has less ripple. The top of the "pi" can be either a resistor or a choke. These multi stage filters are called "progressive filters" sometimes because the power gets cleaner at each tap. The first and less clean tap can be sent to the PP power stage and the last tap to the gain preamp stage nearest the input jack.
Further helping reduce hum is that in a push-pull amp what the speaker "hears" is the difference between the two sides. Power supply ripple is common to both sides and gets nulled out.
Here is a design from the early 1950's doing all of the above:
http://www.schematicheaven.com/fenderamps/bassman_5e6.pdf
Notice in the above that there are three "pi" sections (a choke and two 10K resistors at the tops) the first two, noisier sections feed balanced push/pull stages where ripple is rejected as common mode noise. The input stage gets very clean "triple filtered" power. This circuit was common back when capacitors were very expensive and 47uF was "huge"
The other way to reduce ripple, drastically, is to use a switching mode supply. These "chop" the A/C mains power converting it to about 100KHz and then it rectifies the very high frequency to DC. If there is any ripple remaining in the DC it is well out of the range of human hearing. Almost all consumer electronics uses this method. DIYers don't like it because the engineering is complex and the supply is not easy to build
Further helping reduce hum is that in a push-pull amp what the speaker "hears" is the difference between the two sides. Power supply ripple is common to both sides and gets nulled out.
Here is a design from the early 1950's doing all of the above:
http://www.schematicheaven.com/fenderamps/bassman_5e6.pdf
Notice in the above that there are three "pi" sections (a choke and two 10K resistors at the tops) the first two, noisier sections feed balanced push/pull stages where ripple is rejected as common mode noise. The input stage gets very clean "triple filtered" power. This circuit was common back when capacitors were very expensive and 47uF was "huge"
The other way to reduce ripple, drastically, is to use a switching mode supply. These "chop" the A/C mains power converting it to about 100KHz and then it rectifies the very high frequency to DC. If there is any ripple remaining in the DC it is well out of the range of human hearing. Almost all consumer electronics uses this method. DIYers don't like it because the engineering is complex and the supply is not easy to build
Last edited:
"The other way to reduce ripple, drastically, is to use a switching mode supply. These "chop" the A/C mains power converting it to about 100KHz and then it rectifies the very high frequency to DC. If there is any ripple remaining in the DC it is well out of the range of human hearing. Almost all consumer electronics uses this method. DIYers don't like it because the engineering is complex and the supply is not easy to build
"
Are you sure about this? normally the 100/120Hz ripple is not the problem - standard linear (shunt or series) can copy quite well with this. The problem with switchers - even good ones - is the HF noise and this generally precludes them from very high performance systems, although some companies have managed pretty well (Chord, Linn etc). You are right about the engineering - its a very specialized field.
"
Are you sure about this? normally the 100/120Hz ripple is not the problem - standard linear (shunt or series) can copy quite well with this. The problem with switchers - even good ones - is the HF noise and this generally precludes them from very high performance systems, although some companies have managed pretty well (Chord, Linn etc). You are right about the engineering - its a very specialized field.
- Status
- Not open for further replies.