Basic (switching frequency) ripple is always the easiest noise component to deal with when dealing with the output of an SMPS. Attenuation in the form of caps, inductors works well in a properly designed supply. Heck, I can attain 50mV p-p in a 4:1 input 24v @ 8A 200W supply confined in a 2.5" x 2" x 1/2" DC/DC. No electrolytics at all.
What most people miss it common mode noise. Someone above mentioned common mode noise. That's the key IMO.
I've looked at many SMPS reference designs here since recently joining. One thing that sticks out is the omission of an I/O capacitor. Works wonders for both differential and common mode noise issues. Something in the order of 1nF to 10nF can make a world of difference in terms of both conducted and radiated noise.
What most people miss it common mode noise. Someone above mentioned common mode noise. That's the key IMO.
I've looked at many SMPS reference designs here since recently joining. One thing that sticks out is the omission of an I/O capacitor. Works wonders for both differential and common mode noise issues. Something in the order of 1nF to 10nF can make a world of difference in terms of both conducted and radiated noise.
Nope, but the LM5576 is pretty badass. It's a buck converter, though.
If you look at EPRI (Electrical Power Research Institute) data you'll see most mains overvoltage events don't go more than 20% above normal with a few 25% as outliers. It's statistically unlikely you'll be monitoring the power line with a mulitmeter when an overvoltage happens so just choose the interval of confidence you want. I generally design for 25% over, though I've been known to fudge it a couple percent on occasion.
Conversational level listening is 50 to 60dB SPL and 1mW/channel into a 90dB SPL/W efficient pair of speakers delivers 60dB SPL at a 2m listening position---give or take a few dB that's a pretty typical setup for home audio. Power amps often dissipate 10 to 50W to deliver that ~1mW to the speaker. So it's not uncommon the ripple on an amp's supply rails will be larger than the amp's voltage swing into the speaker.
There are various ways of dealing with this. The most common response is to get fancy with the supply to try to lower the ripple and reduce the PSRR required of the amp circuitry. A much less common response is to size the amp more closely to the power level required and design it for limited quiescent dissipation. However, little class B-ish amps that take advantage of the latter option are really fun as the rail voltages are friendly to a much wider range of regulators.
I know casual listening is typically very low power and I'm quite fond of low power amps, but I was imagining listening to 100mW of music while there was 1V of ripple coming through my 92dB speakers.
Now I get it.I will at some point play around with Class B "ish" and other such amps, but I'm picky about sound and haven't heard much of B and other class solid state amps that I've liked...yet. (I was disappointed by my little Tripath amp, but it was a cheap Chinese knock off so who knows?
So I did a test running my PS with the two 5R/50W resistors in parallel ($0.50 each at a local hamfest, but they look brand new) on a heat sink and they were quite cool running compared to the single/no heatsink.
Unfortunately, the hum was unbearable through the Grados. Quite a lot louder than the SMPS and constant (not tied to the input). Not sure what the source is yet. Of necessity, the grounding has changed a little. Amp circuit is still star gounded, but I've tied that to the PS ground which for now is just a bare wire strung between caps and the rectifier. I've tried it "floating" and tied to the mains ground leg, but no joy.
I'm not giving up, though. I may try to sleuth the hum a bit. Next I'll try the regulated PS, then maybe filter that a bit more. Based on what I've heard so far, though, I'm not inclined to shell out for a $20+ choke unless someone can make a strong case for doing so. (I'm on a VERY tight budget for a while).
Unfortunately, the hum was unbearable through the Grados. Quite a lot louder than the SMPS and constant (not tied to the input). Not sure what the source is yet. Of necessity, the grounding has changed a little. Amp circuit is still star gounded, but I've tied that to the PS ground which for now is just a bare wire strung between caps and the rectifier. I've tried it "floating" and tied to the mains ground leg, but no joy.
I'm not giving up, though. I may try to sleuth the hum a bit. Next I'll try the regulated PS, then maybe filter that a bit more. Based on what I've heard so far, though, I'm not inclined to shell out for a $20+ choke unless someone can make a strong case for doing so. (I'm on a VERY tight budget for a while).
Last edited:
Did you do the hum test with no input wiring and the amplifier input shorted?
Best to rule out any external hum sources first.
Thanks, metalsculptor. I tested with a dvd/cdp connected (but powered down), then pulled the interconnects and tried again (didn't short the inputs). I have a flourescent light overhead that has, on occasion, introduced noise, but never anything like this. If I can find an easy way to short the inputs, I'll try that tomorrow.
PS (off topic), is your username related to your audio work (as in amp casing) or are you a metal sculptor, or something else?
My work is mainly RF and industrial instrumentation, hacking bits of metal to do things the original builder never envisaged is more a hobby.
Many broadband RF amplifiers are class A or AB.
I'd have to sim this design to see how much the FET is operating like a single ended triode but wouldn't surprised if it's kind of in that direction (though normally one gets the feedback from a cathode bias equivalent). If it's second harmonic distortion you're wanting you can also get it by following a suitably configured signal FET with a class AB power stage. Simpler supply that way, particularly as circuits like the triodizer and fetzer valve run happily on 9-15V rather than the 300V the ECC tube family tends to want.
test your grounding properly.
Your report is virtually worthless unless you can show that wiring errors have not led to
AndrewT,
Yep, that's one thing I'm hoping to do once I come up with more ways to revise and/or test the grounding. You could help by suggesting next steps for me...
But in any case, worthless seems a strong word. I've done equally sloppy PS wiring jobs breadboarding in the past and never had hum anything like what I have on this. That makes me suspicious of the combination of this PS and this amp.
Yep, that's one thing I'm hoping to do once I come up with more ways to revise and/or test the grounding. You could help by suggesting next steps for me...
But in any case, worthless seems a strong word. I've done equally sloppy PS wiring jobs breadboarding in the past and never had hum anything like what I have on this. That makes me suspicious of the combination of this PS and this amp.
A few things came to mind re reading the posts.
Is the negative supply rail grounded at the amplifier end? it should always be grounded after the filters and nowhere else. If it is floating it is easy to get capacitive pick-up of mains noise.
Use your CRO to measure the ripple, use a x1 probe with AC coupling, Even the most basic CRO should have no problem displaying ripple down to a few mV (-40dBc with a 24V supply). A multimeter set for AC placed across the power supply rails will show ripple down to 100mV or so.
Is the negative supply rail grounded at the amplifier end? it should always be grounded after the filters and nowhere else. If it is floating it is easy to get capacitive pick-up of mains noise.
Use your CRO to measure the ripple, use a x1 probe with AC coupling, Even the most basic CRO should have no problem displaying ripple down to a few mV (-40dBc with a 24V supply). A multimeter set for AC placed across the power supply rails will show ripple down to 100mV or so.
Right now the amp ground is attached to the PS ground via some solid core, unshielded/un-twisted wires. The B+ is fed to the amp from a similar wire, again not twisted with the ground. These wires jump over the R side of the amp casing to the B+ wires for the amp, which sits further to the L. In other words, the B+ and ground do NOT pass over the transistors or any other amp components (except the SMPS which is not plugged in).
I've tried connecting that ground to the mains ground and keeping it disconnected and floating (I measured about 20mV difference between the mains ground and the amp ground). I have not yet tried to disconnect the amp ground from the PS ground. Not sure whether that's a good idea or not, as there'd be no amp reference for the voltage from the PS, but again, I'm no EE - in fact I have a liberal arts degree!
Thanks again,
Carl
I've tried connecting that ground to the mains ground and keeping it disconnected and floating (I measured about 20mV difference between the mains ground and the amp ground). I have not yet tried to disconnect the amp ground from the PS ground. Not sure whether that's a good idea or not, as there'd be no amp reference for the voltage from the PS, but again, I'm no EE - in fact I have a liberal arts degree!
Thanks again,
Carl
Can you post the complete schematic of your amp and supply?
This will help deciding whether the problem is inherent and structural, or caused by a material/implementation flaw, like poor cabling.
I think it is inherent, which leaves a number of options: the easiest is to correct some part of the amp to make it ripple-resistant.
Otherwise, if you want to avoid brute force and electronic regulation/filtering, you can use resonance in your series inductance to remove the 120Hz fundamental.
It won't be a complete cure, neither for the fundamental because of the low practical Q achievable, nor for the harmonics, but it will bring a worthwhile improvement.
The next step is to actively help the filtering by means of a small audio transformer in series with the last filter cap connected to a ripple cancelling amplifier.
It is as complicated (or uncomplicated) as an electronic filter, but doesn't involve high power components.
This will help deciding whether the problem is inherent and structural, or caused by a material/implementation flaw, like poor cabling.
I think it is inherent, which leaves a number of options: the easiest is to correct some part of the amp to make it ripple-resistant.
Otherwise, if you want to avoid brute force and electronic regulation/filtering, you can use resonance in your series inductance to remove the 120Hz fundamental.
It won't be a complete cure, neither for the fundamental because of the low practical Q achievable, nor for the harmonics, but it will bring a worthwhile improvement.
The next step is to actively help the filtering by means of a small audio transformer in series with the last filter cap connected to a ripple cancelling amplifier.
It is as complicated (or uncomplicated) as an electronic filter, but doesn't involve high power components.
Hi Elvee. Here's the amp itself:
http://diyaudioprojects.com/Solid/ZCA/sch_ZCA.png
As for the PS, it's identical to the PS of this amp above (http://diyaudioprojects.com/Solid/ZCA/ZCAps.png) except for:
The transformer is 24v/3.3 amp.
My first cap is 470uF/50v
2nd cap is 8200uF/50v
Instead of the choke, I've used 2 5ohm/50watt power resistor in parallel
Next cap is 8200uF/50v
I did not use the 0.1uF final cap in the PS. I will probably add that as soon as I can get back to this.
I'm reading about 25v for the B+. Thanks for any ideas you have. I do hope to measure ripple w/my O-scope, both with the current PS and the SMPS. Basically, aside from the headphones (Grado SR-60s), there's not a lot of noise with the SMPS (through my ~92dB Pass BOFU TL speakers - it's a nice combo for the most part). I was simply wanting to experiment with alternative PS b/c I thought SMPS weren't very quiet.
Cheers,
Carl
Last edited:
First thing to do is to move the load to B+: it involves no modification to the circuit itself, yet will yield a valuable improvement. See comparison below.
The sim shows the level of 120Hz leaking into the load for 1V pp on the supply.
There is a ~6 fold improvement. In your case it will be better, because most of the residue is caused by the Cgd capacitance, which is lower in a linear MOS.
The value could be further reduced by lowering the input impedance.
Ultimately, the remainder will be caused by the transistor's output resistance, and there is no simple way to compensate for it (compared with the simplicity of the amplifier itself).
In short, this simple mod should result in a ~20dB improvement
The sim shows the level of 120Hz leaking into the load for 1V pp on the supply.
There is a ~6 fold improvement. In your case it will be better, because most of the residue is caused by the Cgd capacitance, which is lower in a linear MOS.
The value could be further reduced by lowering the input impedance.
Ultimately, the remainder will be caused by the transistor's output resistance, and there is no simple way to compensate for it (compared with the simplicity of the amplifier itself).
In short, this simple mod should result in a ~20dB improvement
Attachments
Hi,
you have a rCCRC supply.
I think the second C could usefully be made larger for a ClassA amplifier. Try doubling it.
The r is the total resistance of the transformer secondary and rectifier circuit up to the first CC. R is your 5r0//5r0. This is bit high for the bias of your single ended amplifier.
Maybe add a third 5r0 and possibly a 4th. But this will require more final capacitance, maybe 20mF to 40mF.
Don't add that 100nF to the PSU. It might make the PSU susceptible to ringing, if a pulse/spike from the mains, or a pulse demand due to interference arrives at the PSU.
you have a rCCRC supply.
I think the second C could usefully be made larger for a ClassA amplifier. Try doubling it.
The r is the total resistance of the transformer secondary and rectifier circuit up to the first CC. R is your 5r0//5r0. This is bit high for the bias of your single ended amplifier.
Maybe add a third 5r0 and possibly a 4th. But this will require more final capacitance, maybe 20mF to 40mF.
Don't add that 100nF to the PSU. It might make the PSU susceptible to ringing, if a pulse/spike from the mains, or a pulse demand due to interference arrives at the PSU.
I'm not sure I understand. My B+ IS tied directly to the load resistors (I'm calling the PS output B+ because I'm mostly a tube builder). It looks like you've got the mosfet output (speaker) tied to ground on one channel and to the 24vdc B+ on the other side. That doesn't sound to appealing to me and my speakers...
Please help me understand what you mean.
Also, I see you show 12k as the input impedance, correct? I need to check to see if that will cause any problems with my sources.
AndrewT, I have a few 22mF caps I could use instead of the 8200uFs, so I'll try them out. Re: the parallel 5Rs, they are what I happen to have on hand (picked 'em up at a hamfest recently for $0.50 along with the big caps, all look to be new). Together they read about 3R, and I will try to find a way to get even lower.
Some questions though. First, what does this resistor have to do with bias (aside from just its effect on the voltage)? The design goal was 12v at the drain and 12v across the load resistor. The lower CRC resistance will push me close to 30v from the PS. Should I run the mosfets at 15v and drop 15v across the load resistors? I'm worried about running them at 15v as I think that's at or close to the max d-s voltage. Also, will I have to rework the rest of the circuit? Alternatively, should I run at 12v at the drain and burn 28v at the load resistor?
Thanks again,
Carl
Last edited:
I am not showing two channels, I show the original situation and the improvement when the load (speaker) is returned to B+. This has to be done on all channels of the amplifier
12K simulates an average worst case value for the 100K input potentiometer.
If it can be lowered, it will improve the residual after the first mod has been done, otherwise it is useless
OK, the schematic looked all tied together so I though you were showing both channels. So if I return the speaker to the B+, the output cap will block the DC from B+ and prevent damage to the speaker? I've never heard of tying the speaker to the B+ so it makes me a little nervous. Also, the voltage swings on the B+ from the signal - wouldn't they act as some kind of feedback to the signal flowing through the speakers?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.