Hienrich,
My apologies for not answering earlier!
Diodes are fine where there is little voltage to move; if you want the CCS to rise to within about 1V of the rail, two 1N4148s are acceptable.
However, there's a problem generally. As temperatures change, you want Vbe to match the diode forward voltage drop. This is fine if there's one of each, but you need at least two diodes to make the CCS function! With two junctions, as temperature rises the voltage across the two diodes will reduce faster than the voltage across base/emitter junction; so current output will reduce with rising temperature, and since the 'window' voltage (Vb - Vbe) is small, this reduction in current will be marked.
Of course, there are situations where changing current with temperature won't matter. A bit of give and take on a VAS is really no problem, but it's not good if the CCS is used to set up a voltage regulator.......
Generally, you want to use a larger 'window' voltage if you can since this reduces noise as the emitter (degeneration) resistance is higher for the same current. LEDs are good in this regard, as they are visible indications of circuit function, they are cheap, and they give you good temperature control with very low noise.
As you can see, the CCS, like most circuit blocks, is the artful management of compromise..... bit like life itself!
Cheers,
Hugh
My apologies for not answering earlier!
Diodes are fine where there is little voltage to move; if you want the CCS to rise to within about 1V of the rail, two 1N4148s are acceptable.
However, there's a problem generally. As temperatures change, you want Vbe to match the diode forward voltage drop. This is fine if there's one of each, but you need at least two diodes to make the CCS function! With two junctions, as temperature rises the voltage across the two diodes will reduce faster than the voltage across base/emitter junction; so current output will reduce with rising temperature, and since the 'window' voltage (Vb - Vbe) is small, this reduction in current will be marked.
Of course, there are situations where changing current with temperature won't matter. A bit of give and take on a VAS is really no problem, but it's not good if the CCS is used to set up a voltage regulator.......
Generally, you want to use a larger 'window' voltage if you can since this reduces noise as the emitter (degeneration) resistance is higher for the same current. LEDs are good in this regard, as they are visible indications of circuit function, they are cheap, and they give you good temperature control with very low noise.
As you can see, the CCS, like most circuit blocks, is the artful management of compromise..... bit like life itself!
Cheers,
Hugh
Hi Lumanauw,
A resistive divider as voltage reference on a CCS has the advantages of simplicity and varying current with applied voltage. There are situations where this could be useful; voltage regulator circuits, for example, where the regulated voltage must always be less than the input, raw voltage.
However, most audio circuits need a constant current regardless of rail voltage, so a voltage reference like a LED or Zener is normally used.
Hienrich,
Green LEDs come in three common sizes; 3mm, 5mm and 10mm. I like the 5mm, as it's the cheapest!
Cheers,
Hugh
A resistive divider as voltage reference on a CCS has the advantages of simplicity and varying current with applied voltage. There are situations where this could be useful; voltage regulator circuits, for example, where the regulated voltage must always be less than the input, raw voltage.
However, most audio circuits need a constant current regardless of rail voltage, so a voltage reference like a LED or Zener is normally used.
Hienrich,
Green LEDs come in three common sizes; 3mm, 5mm and 10mm. I like the 5mm, as it's the cheapest!
Cheers,
Hugh
What about the two transistor version?
Ok, I've been reading through all the posts now but I never saw someone considering a two transistor CCS, why is that? I think it is the best of all, maybe a little more complex to understand, but still more precise and not really more expensive, well, just the cost of one transistor instead of a diode.
Is this kind of design too precise, too stiff, or is there another reason?
I wrote an article on it on my site which you can find at my constant current source page.
Kind regards, Jef
Ok, I've been reading through all the posts now but I never saw someone considering a two transistor CCS, why is that? I think it is the best of all, maybe a little more complex to understand, but still more precise and not really more expensive, well, just the cost of one transistor instead of a diode.
Is this kind of design too precise, too stiff, or is there another reason?
I wrote an article on it on my site which you can find at my constant current source page.
Kind regards, Jef
Unless you feed a two transistor CSS with another CSS to establish base current of the output transistor, or unless you establish it via a resistor from a regulated voltage, you actually get quite a lot of variation of output current vs input current.
I have actually used a trim pot as the Rb for the output transistor to vary the bias current of an amplifier, so I know it works (and the current sensing transistor was used as a thermal sensor on the heatsink).
If it is desired to have a CCS swing as close as possible towards a power rail, then one good solution is to use a current mirror with one side fed by a CSS referenced from a convenient potential. The current mirror can swing very close to it's power rail. This approach is very often used inside OPamps.
I have actually used a trim pot as the Rb for the output transistor to vary the bias current of an amplifier, so I know it works (and the current sensing transistor was used as a thermal sensor on the heatsink).
If it is desired to have a CCS swing as close as possible towards a power rail, then one good solution is to use a current mirror with one side fed by a CSS referenced from a convenient potential. The current mirror can swing very close to it's power rail. This approach is very often used inside OPamps.
Just thought I'd drop in a quick note pertaining to resistors used as current sources. The basic function of a current source under a differential is to provide a reasonable approximation of an infinite resistance. Low resistances--on the order of, say, a few hundred ohms--will cause the differential to give poor results as a phase splitter or as a negative feedback entry point; the two sides of the differential simply don't communicate very well because the signal falls into the resistor instead of getting to the other side of the differential where it can do some good. As an example, a phase splitter will commonly have 10dB or more difference between the two phases. That's useless in the practical sense.
But...
Many people would be surprised at how small a resistor will do a good job. So, how much resistance do you need? As little as 5 to 10k can yield results virtually indistinguishable from a current source.
I remember a thread where someone posted the schematics to one of the early Krell designs. Almost immediately, someone made fun of the fact that the front end differentials were biased by resistors and didn't D'Agostino know any better and gee the thing must have been designed back in the Jurassic before they had such modern conveniences as current sources and...
Sigh.
I don't remember the value of resistors that Dan used, but I seem to recall that they were reasonable. The amps sounded pretty good. Good enough, in fact, that they are regarded as classics, whereas many designs that used "perfect" current sources have fallen by the wayside.
Advantages: Assuming a good quality part, resistors do not exhibit frequency-dependent compliance problems. Resistors have consistent and predictable noise performance. Resistors are reliable. And last but certainly not least, resistors are simple.
If you choose to use a resistor, there are two things to remember. Don't forget to calculate your heat dissipation. On larger amps (hence higher voltage rails) they may need to be more than just a puny 1/8W. The other is that resistors will give a different bias current if the rails collapse. Whether this is a bug or a feature depends on your point of view and the rest of the circuit. I tend to regard regulated rails as mandatory on low level stuff like preamps, so it's never an issue there. On amplifiers, where regulated rails are a bit more cumbersome, it may become an issue.
It's not the end of the world, either way. It's just another option.
Grey
But...
Many people would be surprised at how small a resistor will do a good job. So, how much resistance do you need? As little as 5 to 10k can yield results virtually indistinguishable from a current source.
I remember a thread where someone posted the schematics to one of the early Krell designs. Almost immediately, someone made fun of the fact that the front end differentials were biased by resistors and didn't D'Agostino know any better and gee the thing must have been designed back in the Jurassic before they had such modern conveniences as current sources and...
Sigh.
I don't remember the value of resistors that Dan used, but I seem to recall that they were reasonable. The amps sounded pretty good. Good enough, in fact, that they are regarded as classics, whereas many designs that used "perfect" current sources have fallen by the wayside.
Advantages: Assuming a good quality part, resistors do not exhibit frequency-dependent compliance problems. Resistors have consistent and predictable noise performance. Resistors are reliable. And last but certainly not least, resistors are simple.
If you choose to use a resistor, there are two things to remember. Don't forget to calculate your heat dissipation. On larger amps (hence higher voltage rails) they may need to be more than just a puny 1/8W. The other is that resistors will give a different bias current if the rails collapse. Whether this is a bug or a feature depends on your point of view and the rest of the circuit. I tend to regard regulated rails as mandatory on low level stuff like preamps, so it's never an issue there. On amplifiers, where regulated rails are a bit more cumbersome, it may become an issue.
It's not the end of the world, either way. It's just another option.
Grey
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.