Hi.
I have been toying with the subject line as an idea I woukd explore when I have the time/space.
I have been reading much of the Super regulator threads and the ESP pages. I have been considering a discrete design, as well as a error amp scheme.
The main goals are:
Stable, but responsive regulation up to at least 30kHz
Low ripple and noise.
Good Current capability, 3-8 amps peak.
Output voltage between 30 & 40 Volts
I also have more of the TI chip amp ICS than I'm every likely to use.
Couldn't such a regulator be made using say LM3875? Or even LM675/1875.
Anyway I plan to try and do just that, use the surplus to experiment making regulators.
I'm certain that at some point the magic smoke will escape the strange multi-legged oblong thingummies...
Either way I'm going to have a shot, and I'll update here with any failures and with my usual comical self deprecating way.
Has anyone tried this?
I have been toying with the subject line as an idea I woukd explore when I have the time/space.
I have been reading much of the Super regulator threads and the ESP pages. I have been considering a discrete design, as well as a error amp scheme.
The main goals are:
Stable, but responsive regulation up to at least 30kHz
Low ripple and noise.
Good Current capability, 3-8 amps peak.
Output voltage between 30 & 40 Volts
I also have more of the TI chip amp ICS than I'm every likely to use.
Couldn't such a regulator be made using say LM3875? Or even LM675/1875.
Anyway I plan to try and do just that, use the surplus to experiment making regulators.
I'm certain that at some point the magic smoke will escape the strange multi-legged oblong thingummies...
Either way I'm going to have a shot, and I'll update here with any failures and with my usual comical self deprecating way.
Has anyone tried this?
I have a portable sized DAC/AMP where a TDA2030A has been used as a rail splitter.
I've tried using smaller amp ICs as regulators - AD815 worked but that's unity-gain stable. I'd guess not having unity gain stability would be a fly in the ointment, you'll probably need compensation networks to get around that.
You'll still need output capacitance due to the fact that chipamps have the usual gain roll-off vs frequency. Then you'll need an output inductor to keep it all stable.
I've tried using smaller amp ICs as regulators - AD815 worked but that's unity-gain stable. I'd guess not having unity gain stability would be a fly in the ointment, you'll probably need compensation networks to get around that.
You'll still need output capacitance due to the fact that chipamps have the usual gain roll-off vs frequency. Then you'll need an output inductor to keep it all stable.
The LM3875's datasheet says that it is stable when operated at a closed loop gain of 10 or greater (top of p.19: "Layout, Ground Loops, and Stability"). You can easily achieve this by setting your VREF to (1/20th) of the desired VOUT. Voila, gain is 20X and honor is satisfied.
For VOUT = 30 volts you'd want VREF = 1.5V. I would consider an LM329 voltage reference IC whose output is 6.9V, followed by a resistive divider, {Rtop= 36K, Rbot= 10K} , giving 1.5V VREF. As a handy side-benefit, the thevenin equivalent output resistance of the divider gives you the R part of an RC lowpass filter, for free. This filter attenuates noise and ripple on VREF.
Or you could use Walt Jung's GLED431 voltage reference (output = 2.5V) instead of the LM329, with appropriate jiggery pokery upon the resistive divider.
Although the LM3875's datasheet contains a misprint for input offset voltage, its sister device LM3886 says 1mV typical, 10mV max. Multiplied by a gain of 20X that's an output error of 200mV. But since you're going to be adjusting the output anyway, to get whatever value you like between 30V and 40V, the potentiometer is already there and so the extra cost is zero.
On the other hand the max output current is only guaranteed to be 4 amperes (6 amps if you're lucky enough to get a batch of "typical" units) and max power dissipation is only guaranteed to be 40 watts (56 watts if you're lucky). If your input voltage is 48VDC then when you've set the regulator to 30V output you will only be able to source (40/18) = 2.2 amperes before hitting the power dissipation limit. If you need an 8 ampere supply you will need to figure out how to connect 4 of your LM3875 regulator boards in parallel without current hogging. That might not be easy.
For VOUT = 30 volts you'd want VREF = 1.5V. I would consider an LM329 voltage reference IC whose output is 6.9V, followed by a resistive divider, {Rtop= 36K, Rbot= 10K} , giving 1.5V VREF. As a handy side-benefit, the thevenin equivalent output resistance of the divider gives you the R part of an RC lowpass filter, for free. This filter attenuates noise and ripple on VREF.
Or you could use Walt Jung's GLED431 voltage reference (output = 2.5V) instead of the LM329, with appropriate jiggery pokery upon the resistive divider.
Although the LM3875's datasheet contains a misprint for input offset voltage, its sister device LM3886 says 1mV typical, 10mV max. Multiplied by a gain of 20X that's an output error of 200mV. But since you're going to be adjusting the output anyway, to get whatever value you like between 30V and 40V, the potentiometer is already there and so the extra cost is zero.
On the other hand the max output current is only guaranteed to be 4 amperes (6 amps if you're lucky enough to get a batch of "typical" units) and max power dissipation is only guaranteed to be 40 watts (56 watts if you're lucky). If your input voltage is 48VDC then when you've set the regulator to 30V output you will only be able to source (40/18) = 2.2 amperes before hitting the power dissipation limit. If you need an 8 ampere supply you will need to figure out how to connect 4 of your LM3875 regulator boards in parallel without current hogging. That might not be easy.
Thanks Mark,
your comments match my thinking on the subject exactly!!! A gain of 10V/V or greater for stability, in the case of the LM675.
It's could be as simple as Vref on the IC non inverting input itself, or expanded to use an error amp feeding the LM.
Thanks Abraxalito,
Your comments make perfect sense to me - output inductor and reservoir will certainly be required (although my train of thought hadn't got that far!)
One thought I have had is that it's a little wasteful of a chip amp IC.... using in this way, aren't I just using one half of the IC transistors?
your comments match my thinking on the subject exactly!!! A gain of 10V/V or greater for stability, in the case of the LM675.
It's could be as simple as Vref on the IC non inverting input itself, or expanded to use an error amp feeding the LM.
Thanks Abraxalito,
Your comments make perfect sense to me - output inductor and reservoir will certainly be required (although my train of thought hadn't got that far!)
One thought I have had is that it's a little wasteful of a chip amp IC.... using in this way, aren't I just using one half of the IC transistors?
Last edited:
Why not just use LM317? (vin-vout < 40V) Audio op amps are not designed for voltage regulator application.
LM317/337
I certainly haven't missed the existence of these devices (in fact I regularly use the cheaper 78**/79** parts.
The supply itself don't require facility to adjust to voltage beyond trimming the voltage, and thus can be accomplished without an adjustable regulator IC.
I have looked at these devices with a view to boosting the current output with a discrete output stage. However, doing so renders the IC current limiting/protection useless.
Hence my thought to repurpose a chip amp for the task (some level of protection will exist)
Of course I could also design and add current protection, device failure protection circuits, but I am stupid so I like to K.I.S.S.
FWIW the inspiration for this came from the lowly TDA2004, which can supply a large current relatively. The downside is the supply voltage, which does not meet my requirements. This lead me to consider the LM3875/6, 3886 or 4780 in parallel to meet the requirements.
I even have some LM12CLK that I can try (though I've been warned these are a beggar to get stable )
I certainly haven't missed the existence of these devices (in fact I regularly use the cheaper 78**/79** parts.
The supply itself don't require facility to adjust to voltage beyond trimming the voltage, and thus can be accomplished without an adjustable regulator IC.
I have looked at these devices with a view to boosting the current output with a discrete output stage. However, doing so renders the IC current limiting/protection useless.
Hence my thought to repurpose a chip amp for the task (some level of protection will exist)
Of course I could also design and add current protection, device failure protection circuits, but I am stupid so I like to K.I.S.S.
FWIW the inspiration for this came from the lowly TDA2004, which can supply a large current relatively. The downside is the supply voltage, which does not meet my requirements. This lead me to consider the LM3875/6, 3886 or 4780 in parallel to meet the requirements.
I even have some LM12CLK that I can try (though I've been warned these are a beggar to get stable )
- Status
- Not open for further replies.