Understanding what's going on with my PS

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
The background:

I've been running the amp on a SMPS capable of 5-6 amps with no problems, but wanted to play around with a standard, well filtered PS. I'm just trying to learn all I can.

So, I built a Pi filtered PS not unlike the PS in the amp design (but CRC instead of CLC) with a 25amp bridge into a 470uF/50v cap parallel to a 8200uF/50v cap. That goes into a 5 ohm/50w resistor then another 8200uF cap. PS output goes to the V+ input for the ZCA amp. After that it goes past some more caps, through a 15ohm load resistor and then to the mosfet/output. Simple enough. I'm trying to get about 24v for B+ at the "top" of the load resistor and drop that to 12v at the mosfet.

Right now it's running at 22.5v at the end of the PS/top of the load resistor. At the output of the rectifier it's 30v. That means the 5ohm resistor is dropping 7.5v, and therefore passing 1.5A. What concerns me is the 5ohm/50W resistor is getting quite HOT. Within a few minutes it's too hot to touch.

1.5A means it's dissipating less than 12W, and that doesn't seem to me to be a big deal for a 50W resistor. So why is it getting so hot? Is there danger of frying it? Should I heatsink it? Something else?

Also, at 1.5A, that 15ohm load resistor SHOULD be dropping 22.5v, correct? But I'm reading 11.6v or so at the bottom end of the load resistor, so dropping about 11v through the load resistor and running 0.73A. How is that possible? If the PS is running 1.5A, I thought that had to go through the load resistor, and according to ohm's law, that would mean that resistor should burn 22.5v, right?

I'm so confused!

Also, I'm hoping to play around with a LM317 regulated supply using virtually the same filtering and a tip3055 to up the current rating (something like this: Build an Adjustable 0-34 volt power supply with the LM317, with the filtering preceding the regulator). Is there anything I should know about doing this?

Thanks so much!
 
I'm guessing you're measuring with a multimeter and that the multimeter is getting confused by ripple---if so, check whether the meter's true RMS and what its bandwidth is. Measuring with a higher bandwidth tool (like an RTA or oscilloscope) or simulating in Spice will give you more insight into what's going on.

Check the thermal derating on your resistor's datasheet. 50W parts often derate to zero at 250C so, assuming the usual derating starting 25C, the expected Trise for 12W would be (250C - 25C) * 12W / 50W = 53C. One usually wants to design for a Trise of 40C or less.
 
Also, that 50W is usually on a specified heatsink if the resistor is one of the common gold coloured ally cased ones, it will not do 50W without a heatsink.....

Actually, too hot to touch is not automatically all that much of a problem, that usually corresponds to 60C or so, no big deal for a modern power component (If the solder melts in use you might have a problem!)....

Regards, Dan.
 
the R in the CRC is far too high.
Expect the R to drop <<10% of the total available voltage. A target value for reasonable filtering may be around 1V of Vdrop for a 35Vdc single polarity supply to a ClassA single ended amplifier.

If you are passing 1.5A then 0r68 would be about right.
Pd would then be about 1.5W and a 5W resistor would be adequate.
 
Thanks, everyone. I'm using a Fluke 12B so I doubt it's getting confused by ripple. I don't think the meter is true RMS, but since we're talking basically sine wave until filtering, then DC, does it really matter? It's bandwidth is up to 600v but I have it set to "auto" range. No problem there.

Dan, yep, it's one of those gold cased power resistors. So otherwise, I shouldn't really worry too much about the resistor. It may be too high R (I'd rather not drop quite so much v anyway...) but heat sinking should help. I may do a longer test with the resistor bolted to a heatsink.

I used the 5 ohm b/c that's what I had on hand - and was basically guessing anyway. Hmmm, maybe try 2 5Rs parallel since I have them on hand? Right now I want to see if there's any sonic advantage to this type of PS over the SMPS so I don't want to spend anything on it yet (all the PS parts are from my parts bin). Note that I'm also trying this amp as a head amp with output pulled straight from the speaker outputs and the SMPS just puts out tons of noise through my 32 ohm Grado SR60s. That was part of my motivation to try a filtered linear PS.

So AndrewT, are you saying I should aim to drop the PS voltage from 30v (what's coming out of the bridge despite the roughly 35v unloaded transformer voltage) to 29v? If so, I'd probably aim for 15v at the drain. I think the 2sk1058 can handle those conditions (the gate to source v might be close to the max) but I'm not sure and I'm new to transistors. I'd be burning 15w on the 15R load resistor, but it's 50w and heat sinked, so shouldn't be a problem.

Hmmm. So much to learn. Thanks a ton for all the great responses.
 
So one other question. I'd like to try a regulated PS and was thinking of trying this: Build an Adjustable 0-34 volt power supply with the LM317. I know the LM317 is supposed to be a good regulator and with the TIP3055 the current shouldn't be a problem. Would this schematic on it's own provide a pretty ripple-free output, and how would it compare to a brute-force CRC PS like I'm working on right now (sonically)? Would there be any advantage to adding CRC filtration to the regulated supply?
 
Why anyone would use a resistor as opposed to an inductor in a staged filter arrangement is beyond reason.

Why burn power inside a power source!?

Makes no sense.

Also, Fluke multimeters are notoriously bad in noisy measurement environments.
Especially hand helds.
 
Would there be any advantage to adding CRC filtration to the regulated supply?
Depends on the PSRR of the amp's control loop and your desired SNR on the output(s). This is an individual assessment, both on a per design and per designer basis ;), but in most cases the answer is that the regulator provides sufficient ripple rejection that additional filtering's not particularly useful if the control loop is well implemented and the bypass caps in the supply are reasonably sized. The corner to pay the most attention to in the analysis the low power limit as that's where the ratio of output power to total power the lowest.

The other thing to check is the peak input voltage to the LM317 if the mains run high. Back of the envelope I'm guessing the odds of blowing the part's 40V are acceptably low. The LM338 and LM350 may also be interesting to look at.

Why burn power inside a power source!?
It's class A so the dissipation's invariant of whether the power's burned off in a resistor, inductor, regulator, output device, or driver. Yeah, it costs some output swing to burn it off a CRC but the OP stated this is a deliberate design choice to knock 30V down to 24V. Plus since the amp pulls 1.5A it'd take a 16 ohm speaker for a 24V swing to be required. Not as efficient as aligning the trafo with the desired output voltage but, well, this is class A... ;)

(Full disclosure: I have 20V 100VA trafos for sale in the swap meet forum for cheap. :))
 
Have you thought of using an LC filter on your noisy SMPS?, inductors which effectively block SMPS frequencies are small and common these days. I have used filtered SMPS supplies with class A amplifiers and got good results. The high frequency noise that an SMPS produces does require proper layout and attention to signal routing if it is to be effectively supressed
 
Yooper:

1. doing a rough search in Digi-Key for >1mH and >6A gives us the cheapest at 1mH / 10A and this would give us about 0.75 ohms reactive at 120Hz for $19 US.

the next one is 2.8mH / 10A for $28 US (2.1 ohms)

a 1 or 2 ohm 20W resistor (even at Digi-Key prices) is between $3 and $6

so price would be a factor

2. "...but wanted to play around..."... so Carlp is experimenting and learning.

Makes a lot of sense to me.

Keep playing Carlp
 
Last edited:
OK, Yooper, please remember I'm a newbie to PS (so far I've mostly just built what designers spec). But I'm trying to learn. I'd consider a CLC (from what I've read, it seems like a good route in many cases) but first I wanted to see if there's any noticeable benefit to going this PS route using parts on hand. And, as twest says, it's Class A so the burning doesn't seem to matter from an energy standpoint... (Oh, and I didn't know that about Fluke meters. I know the 12B is designed for electricity not electronics per se).

I did at one point try further filtration of the SMPS (again, just a CRC) but I only had a 100R resistor at the time and it burned WAY too much voltage. I'll look into sizing the inductor for a CLC but frankly I don't know anything about the quality of the noise on the SMPS. I've got a boat anchor scope I'm trying to learn how to use but I'm slow to figure out how to measure such things.

Depends on the PSRR of the amp's control loop and your desired SNR on the output(s).

Twest, I don't know a control loop from a SMPS noise signal (see above, but I'll read up on it) so don't know how to make the tradeoff you mention. Also, I'm not sure just now how to decide on appropriate SNR as it seems to depend on so many factors and I'm someone who LOVES to swap gear to try different combos. Where I have a choice, I gravitate toward less noise is better. But then I'm a card-carrying cheapskate and try to get by with scavenged parts whenever possible.

BTW, my goal of 24v at the PS output is only b/c that's what the designer built. I'm reading up on if/how to push that to closer to 30v (with ~15v at the drain) and may go that route eventually. I'll look at your iron, twest.
 
Would this schematic on it's own provide a pretty ripple-free output...

Not too bad... red is output. Adding C2 and D6 recommended. :)
 

Attachments

  • LM317 + 2N3055 ckt.png
    LM317 + 2N3055 ckt.png
    39.4 KB · Views: 124
  • LM317 + 2N3055 plot.png
    LM317 + 2N3055 plot.png
    19.2 KB · Views: 121
Also, I'm not sure just now how to decide on appropriate SNR as it seems to depend on so many factors.
The limiting factor in the linearity of audio systems that's hardest to get away from is the drivers' phase modulation distortion. Usually that's around -65dB. If you want the rest of the system to reveal that floor then the accumulated distortion from the source to the power amp should be around -80dB or lower. From there it's just a matter of counting up all your stages, adding up their distortions, and seasonng to taste. A good rule of thumb is THD+N on any individual stage should be -100dB or lower.

For example, say you have a power amp outputting a fairly typical home audio level of 100mV RMS that has 1V of ripple. For the control loop to hit -100dB on the signal that means 100dB + 1V/100mV = 120dB of PSRR is needed. Typically the long tailed pair and VAS stages in a discrete amp manage around 70dB PSRR, meaning you need another 120dB - 70dB = 50dB PSRR from somewhere. Most regulators provide around 60dB. So you get 70dB + 60dB = >120dB PSRR and life is good.

Looking back at the topology you're using here I see it'll have rather less than 70dB PSRR, though. The easiest way to address that is to change the amp topology, but that's probably a different project and a topic for the solid state forum. You could throw more engineering at the supply here but past the first regulator you toss in it'll be subject to diminishing returns. If you like tweaking that's probably fine though; class A has the worst ratio of quiescent power (ripple) to output power (signal) but it's kinda fun as all sorts of supply lowpassing that doesn't work so hot with class B/D/G/H operation is admissable for class A.

As an aside, be careful with the structure of the ground star as the 0dB CMRR of the unbalanced input isn't something that improvements in the supply's circuit topology can mitigate---it's a layout thing, both in managing where the currents go supply and in feeding the FET's gate.
 
I did at one point try further filtration of the SMPS (again, just a CRC) but I only had a 100R resistor at the time and it burned WAY too much voltage. I'll look into sizing the inductor for a CLC but frankly I don't know anything about the quality of the noise on the SMPS. I've got a boat anchor scope I'm trying to learn how to use but I'm slow to figure out how to measure such things.

In an example application, I had 28V with 200 mV p-p ripple, filtered this with a 1mH common mode choke followed by 2000uF of SMPS RB caps folowed by a 38uH toroidal choke then 8000uF RB caps. Twisted pair cable used thoughout single point earthing at the filtered output etc.
Ripple was unmeasurable with a CRO, a spectrum analyser showed a 30dB drop in the class A amplifier noise floor. This type of filter is not so effective on noise less than ikHz so the SMPS needs to have good line rejection. For this application I planned on using a linear reg to remove the sub 1 kHz noise but the customer was happy with filtered result.
 
So AndrewT, are you saying I should aim to drop the PS voltage from 30v (what's coming out of the bridge despite the roughly 35v unloaded transformer voltage) to 29v? If so, I'd probably aim for 15v at the drain. I think the 2sk1058 can handle those conditions (the gate to source v might be close to the max) but I'm not sure and I'm new to transistors. I'd be burning 15w on the 15R load resistor, but it's 50w and heat sinked, so shouldn't be a problem.
where did that come from?
I never mentioned 30Vdc nor 29Vdc.
I said
Expect the R to drop <<10% of the total available voltage. A target value for reasonable filtering may be around 1V of Vdrop for a 35Vdc single polarity supply
Vdrop across the R of the CRC should be a lot less than 3.5V for a 35Vdc supply.
I showed an example where 0r68 drops ~1V
 
where did that come from?
I never mentioned 30Vdc nor 29Vdc.

My mistake, AndrewT. I read your post as saying I should aim for a 1v drop, and since my loaded iron was putting out about 30v rectified, not 35v, I figured you meant I should aim for 29v out of the PS. And that's why I asked you - I wasn't sure I got it right. Thanks for clarifying.

Also, can you elaborate on why not to use a regulator on a Class A amp?
 
Not too bad... red is output. Adding C2 and D6 recommended. :)

Thanks, and thanks for the LM317HV. I didn't know about that item. I'm also going to look at the LM338. Ever seen a LM338HV? Mouser doesn't show one. Anyway, our power is pretty stable - I've never seen it go over 122v, and after rectifier losses and resistor drops, I don't see the LM317 ever seeing over about 32-35v. But that is fairly close to the 40v max.

For example, say you have a power amp outputting a fairly typical home audio level of 100mV RMS that has 1V of ripple. For the control loop to hit -100dB on the signal that means 100dB + 1V/100mV = 120dB of PSRR is needed. Typically the long tailed pair and VAS stages in a discrete amp manage around 70dB PSRR, meaning you need another 120dB - 70dB = 50dB PSRR from somewhere. Most regulators provide around 60dB. So you get 70dB + 60dB = >120dB PSRR and life is good.

Twest, thanks for fleshing this out with numbers (though I'm confused about a 100mV signal output with 1V ripple). The numbers help me visualize what's going on. And yes, a different topology is for another project (I'm mostly a tube builder, but interested in learning about transistors). Also grounding (I do use star grounding) is certainly a big issue for me. I'll be reading more on that to learn where problems arise.

Carl
 
don't use a regulator to feed a ClassA power amplifier.
Is that personal experience? If so some circuit details would be of interest. My experience class A amplifiers and regulated supplies has been good, it is pretty tough to get power supply noise below 60 dBc with passive filters. Attention to the noise specification and the transient response
of the regulator circuit is a must.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.