This is probably a stupid question, but why is single output 30V regulated while dual output at +/-15v is unregulated?
Might end up using that toroidal you linked after checking the insane current rating on it, thanks again for that link.
I understand what you were saying about the potential to overload the regulator. However, now I'm not sure how to address that problem and what a solution would like to address it. Would increasing the resistance in the CRC filter prevent voltage from getting high enough to overload the regulator input?
While you're experimenting, you could also take a look at the current boost configuration commonly used with these devices. Basic example here:
LM317 Adjustable Voltage current Boost Power Supply
You said you have fan in your design - and you're going to need it (!) - the 317/337 will get extremely hot you will discover. It's dissipating power (lots of heat) at the current you are drawing in the load (I) x (Vin - Vout).
And as you experiment you'll then realise that you don't want the fan on all the time...so a little thermistor control circuit needed next. Then you'll perhaps want some current limit control. Welcome to experimenting with electronics
Haha exactly!
No, I don't think that would be a reliable solution as the too-high voltage develops on the downstream C in the absence of load.
There are a number of ways to get around this issue. Perhaps your aims could be achieved with a lower transformer voltage? If not then there is the option of bypassing the regulator with a string of high power zener diodes which will clamp the in-out voltage just below the maximum. The clamp only needs to work for a short time, enough to discharge your reservoir capacitor below 40V.
Another solution would be to use an 'HV' variant of the regulator. LM317HV exists and goes up to 60V - to achieve your current requirements you'll need to parallel two or perhaps three - applications for this are shown in the DS. (LM338HV I don't believe exists). The last solution I can think of right now involves using another of the DS applications - the 'tracking pre-regulator'.
I'm interested in implementing the clamp using zener diodes, like you mentioned. From what I've found so far, this method uses TVS diodes to clamp the voltage. Would the following be suitable for this application?
BZW06-33 STMicroelectronics | Circuit Protection | DigiKey
Thanks for you suggestions!
The BZW06-19 might well be suitable. To be sure though you'd want to run a simulation to check it had the capability to absorb the excess charge on the input cap nearest the regulator.
The downside that I can see from employing this solution is the extra in-out capacitance introduced by the TVS diode (of the order of 1nF). This is going to impact the HF rejection of your supply - you might want to improve the input filtering to compensate. Adding more capacitance though can't be a practical solution as this would result in a greater energy to be 'lost' under short-circuit conditions. Looks like a bit of a 'Catch-22'
The downside that I can see from employing this solution is the extra in-out capacitance introduced by the TVS diode (of the order of 1nF). This is going to impact the HF rejection of your supply - you might want to improve the input filtering to compensate. Adding more capacitance though can't be a practical solution as this would result in a greater energy to be 'lost' under short-circuit conditions. Looks like a bit of a 'Catch-22'
The BZW06-19 might well be suitable. To be sure though you'd want to run a simulation to check it had the capability to absorb the excess charge on the input cap nearest the regulator.
The downside that I can see from employing this solution is the extra in-out capacitance introduced by the TVS diode (of the order of 1nF). This is going to impact the HF rejection of your supply - you might want to improve the input filtering to compensate. Adding more capacitance though can't be a practical solution as this would result in a greater energy to be 'lost' under short-circuit conditions. Looks like a bit of a 'Catch-22'
I see what you mean regarding the Catch-22. Trying to make this design work (well) is starting to seem like an exercise in futility. I'm beginning to think that building a linear regulated power supply with a dual output +/-15vdc using a center tapped transformer would be just as effective and far more efficient than trying to make this work.
By simulation, you mean LTspice?
I know this response is going to start a fire storm, what the heck! I have been building electronic projects since the 70's. One of my first audio amps was a Dynakit 70. The Dynakit was what got me hooked on audio. Many of failings of low noise, low THD/TIM topologies was poor power supply designs. In the attached schematic, the AC front end is a 'boiler plate' template I use for all my projects. I am of the old school of thought with linear supplies. Switching supplies have no business in an audio circuit.
The following are, IMHO, MUST's:
* A transorb and EMI/RFI filter. The filter should be at least 2 stage and the transorb at least 1.5KW.
* A series RC RF snubber
Without these parts, you'll couple all that noise through the transformer into the regulation stage.
* Filter/resevoir caps. The standard with many engineers is 10K micro Farads for every 10 volts of
rail. I use that standard and add 10%. Voltage rating: at least 10% above rail voltage. Electrolytics
are crap! They always have been. The closer they get to maximum rating, the faster they wear out.
* A bleeder resistor on the storage caps. A good practice. Really important on larger supplies. Don't
find out why the hard way!
Other 'Best Practices':
* Star, single point grounding. Many noise problems come through the ground plane in the form of
coupling noise and ground currents. Single point grounding eliminates this.
* Twisting rail and ground feeds to the pcb. Twisting the wires doesn't necessarily eliminate noise but
it does make it ratiometric. Using ferrite can often help but can also hinder.
Regarding LateraLiz's Power Supply:
The transformer was changed to a 36 volt secondary with a rating of 56VA. The 36 volts will give you the headroom you need for the LM317. Higher value storage caps will also help with dynamic loads, such as speakers. A 100 Hz bass note, which can send a 8 ohm speaker down to the sub-ohmic region, will demand high current from the supply.
If there's a heat problem with the LM317, you can use a series pass element as shown on the schematic. It will split the load and allow the 317 to run cooler. The MJE2955 is available in a TO-220 case and use a clip on heat sink.
All in all, a good first effort. I hope this helps.
"My advise is free and worth every penny!"
The following are, IMHO, MUST's:
* A transorb and EMI/RFI filter. The filter should be at least 2 stage and the transorb at least 1.5KW.
* A series RC RF snubber
Without these parts, you'll couple all that noise through the transformer into the regulation stage.
* Filter/resevoir caps. The standard with many engineers is 10K micro Farads for every 10 volts of
rail. I use that standard and add 10%. Voltage rating: at least 10% above rail voltage. Electrolytics
are crap! They always have been. The closer they get to maximum rating, the faster they wear out.
* A bleeder resistor on the storage caps. A good practice. Really important on larger supplies. Don't
find out why the hard way!
Other 'Best Practices':
* Star, single point grounding. Many noise problems come through the ground plane in the form of
coupling noise and ground currents. Single point grounding eliminates this.
* Twisting rail and ground feeds to the pcb. Twisting the wires doesn't necessarily eliminate noise but
it does make it ratiometric. Using ferrite can often help but can also hinder.
Regarding LateraLiz's Power Supply:
The transformer was changed to a 36 volt secondary with a rating of 56VA. The 36 volts will give you the headroom you need for the LM317. Higher value storage caps will also help with dynamic loads, such as speakers. A 100 Hz bass note, which can send a 8 ohm speaker down to the sub-ohmic region, will demand high current from the supply.
If there's a heat problem with the LM317, you can use a series pass element as shown on the schematic. It will split the load and allow the 317 to run cooler. The MJE2955 is available in a TO-220 case and use a clip on heat sink.
All in all, a good first effort. I hope this helps.
"My advise is free and worth every penny!"
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
You actually bring up a couple things I have been mulling over!
1. Regarding the inclusion of a 1M bleeder resistor following the large storage capacitors - I understand why it would be wise to include it, however I was under the impression that doing so would introduce a substantial amount of noise?
2. Star ground > Ground plane - if I understood you correctly, the problems associated with using ground planes are effectively mitigated by using a single point star ground?
Thanks for taking the time to share your breadth of knowledge on the subject to a budding DIYer!
> a 1M bleeder resistor following the large storage capacitors - ..., however I was under the impression that doing so would introduce a substantial amount of noise?
You don't have to understand electronics to build stuff. But it DOES help when chosing alternatives.
Resistor hiss power is not according to resistance value. All resistors have the same hiss power.
A high-Z resistor across a low-Z filter cap will have low-low hiss voltage.
You don't have to understand electronics to build stuff. But it DOES help when chosing alternatives.
Resistor hiss power is not according to resistance value. All resistors have the same hiss power.
A high-Z resistor across a low-Z filter cap will have low-low hiss voltage.
In this application looks to me that a 1M bleed resistor would be a waste of time to add - even though it adds no noise.
If you were to choose 10,000uF as the reservoir cap then the time constant in parallel with 1Mohm will be 10,000 seconds. Nearly 3 hours for the voltage to decay by 63%. Chances are the leakage of the 10,000uF cap will decay the voltage faster than that. The quiescent current of the downstream LM317 most certainly will, at least down to low single-digit volts.
If you were to choose 10,000uF as the reservoir cap then the time constant in parallel with 1Mohm will be 10,000 seconds. Nearly 3 hours for the voltage to decay by 63%. Chances are the leakage of the 10,000uF cap will decay the voltage faster than that. The quiescent current of the downstream LM317 most certainly will, at least down to low single-digit volts.
LM338 is pretty straight forward to work with.
just follow the data sheet.
10,000 uf seems excessive
just keep things simple. likewise use a large heatsink
worse comes to worse
you can run 2 in regulators in parallel which allows 10 amp current.
but the idea is to stay with your original design and allow better heat dissipation with 2 devices 2 heatsinks. fan is good idea. might not be needed. otherwise every fan i know of can be turned off and on with a switch.
large heatsink is the easiest
pretty straight forward
headphone amp definitely doesnt need all this voltage and current. but the intention of building a bench supply makes sense.
22 to 25 volt transformer seems feasible. as mentioned smaller transformers tend to have poor regulation. meaning they have higher voltage when not fully loaded
just follow the data sheet.
10,000 uf seems excessive
just keep things simple. likewise use a large heatsink
worse comes to worse
you can run 2 in regulators in parallel which allows 10 amp current.
but the idea is to stay with your original design and allow better heat dissipation with 2 devices 2 heatsinks. fan is good idea. might not be needed. otherwise every fan i know of can be turned off and on with a switch.
large heatsink is the easiest
pretty straight forward
headphone amp definitely doesnt need all this voltage and current. but the intention of building a bench supply makes sense.
22 to 25 volt transformer seems feasible. as mentioned smaller transformers tend to have poor regulation. meaning they have higher voltage when not fully loaded
Adding a fan connected to a sensor to manage heat dissipation is really more about learning how to program and implement sensor based features, and feeling fancy about doing so.
As for the comment about a headphone amp not needing all this voltage and current - "need" is highly subjective. You can power planar magnetic headphones directly from a smartphone. Will they sound good? Not really. Planars like power, and I like options. And who knows, maybe at the end of this I'll come to the same conclusion you reached - headphones don't need quite that much power, after all. There's really only one way to find out!
With regards to transformer regulation, after someone brought that to my attention in the above replies, I've narrowed down my options to transformers rated at 9% regulation, and between 30-36volts to avoid the problems you mention. Thanks for the pointers!
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.