A regulated SMPS for noiseUnit

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I've been working on my noiseUnit speakers - small high quality 2-way speakers, with associated power amps, active crossover, and preamp for a little while.

Some details of the speakers at noiseUnit - transmission line computer speakers.

The preamp is featured at Discrete preamp / headamp based on cascaded diamond amp

With the first iteration of the preamp, I tried using a SMPS. I did this for the following reasons:

  • Lower 50Hz ripple compared to linear.
  • Much more compact, enabling the whole thing to fit in a 35mm thick Hammond 1455 enclosure.
  • Much lower quiescent power than linear, due to lack of magnetising current.

The preamp output stage is hung off the +/-15V rail pretty-much directly from the SMPS modules (Traco TMG15115 15V, 15W units). I have some regulation for the input stage on the preamp board.

So I've been using the preamp for a while and there are a couple of things that I think can be improved on.

  • Firstly 50Hz is sill present when I view output spectra. Indeed it's bigger than all other distortion harmonics This is despite high PSRR in the preamp amplifiers.
  • There's other stuff in the audio spectrum too, audible with headphones with no input on the preamp and the volume turned all the way up. It sounds like a quiet murmur. I'm thinking it's the SMPS modules dropping into a low-power mode when there's no need for large power output, and pulsing every few milliseconds.
  • The +/-15V rails are annoying for driving relays. I'd prefer a 24V rail, and to drive my preamp output at +/-20V rather than +/-15V (another 2.5 dB!). This is the highest I can realistically push the preamp design without having to swap from the 2N3904/3906 transistors that it's built from.

I had a go at filtering the output of the SMPS modules, but ran into the foldback current limiting on the modules when simply adding capacitance. Turns out they're specced for 470µF max output capacitance, and that's simply because higher capacitance trips the foldback current limiting, and leaves them in a state where they burp, unable to start up.

Also I would like a moderate current, low noise, very clean +/-15 to +/-18V supply for running test gear, so I'm motivated to make some changes.

So this is the result. I'm sticking with the Traco SMPS modules, simply because there's still no way to get anywhere near the power density otherwise. I've added a post regulator, which is a fairly conventional series pass setup, using D45H11 and D44H11 pass transistors. I've used BC547/557 transistors in the diffamps for reasonably low noise, and added rather a lot of filtering to the reference, also for low noise.

Q11-14, plus associated resistors, are an output enable circuit, allowing preamp power to be controlled form the micro on the front panel. I was using a relay for that previously. With the enable line floating power consumption is reduced to a few mA. Grounding the enable line turns the output on.

Q9 & Q10, in conjunction with R14-17 provide current limiting for each supply, at around 600mA. This cuts in before the foldback in the SMPS modules, meaning that I can add larger output caps should that be useful without tripping the foldback.

The Traco SMPS modules are happy with a DC input, so I've made the most of that and added an external line rectifier, with 450V 100µF cap at the SMPS inputs. That should improve 50Hz rejection straight off the bat.

I couldn't make my mind up whether I wanted a low dropout design, so there's a couple of links that can be added to swap from a conventional output to a (relatively) low dropout one.

Anyway the whole regulator gives me around 70dB ripple rejection, and should be reasonably low noise to boot.

I'll post up layouts when I've made some.
 

Attachments

  • Screen Shot 2019-09-17 at 6.57.35 am.png
    Screen Shot 2019-09-17 at 6.57.35 am.png
    872.1 KB · Views: 240
Interesting Design, discrete designs are always fascinating.
Im rather new to learning differential inputs, current mirrors/sources et etc

Likewise was interested in learning and seeing what common low noise devices are used
for such designs.

Not sure if i have much advice for improving noise reduction.

But im guessing the BC556 BC546 are closely related to what your already using just maybe more voltage capability's. unless by chance they have better ratings. I didnt cross reference specs on your current

I was shifting through some old Fairchild Data sheets
and the BC556 is quoted for high voltage design
and for lower noise specs suggested the BC559 / BC560

I guess more of a question than a suggestion, instead of using general purpose 2n2904/06 devices for mirrors and sources. Would using low noise transistors offer any benefit

likewise recently i was looking at some musical instrument designs. which yes usually offer nothing amazing as far as low noise.

A Italian Design i was reviewing was using a switching supply but also included another inductor and capacitor after the switching supply and then a final linear regulator .

ill assume the inductor offers extra filtering but would improve rejection since as you mentioned its more ideal to keep capacitance on the low side.

likewise unless circuits are located a long distance from the regulators, large capacitors usually are not recommended after regulators. but again wondering if say a small 68uf and 100n capacitors were placed after your linear regulator would be of any benefit
 
A brave attempt, undoubtedly.

Some improvement ideas:
-Add a resistor of a few tens of ohms in series with the bases of Q9, 10: in case of a hard short, the power section is slow to react, and this can damage the BE of the transistors in question; no need to rework the PCB to do it.

-If another version of the PCB is made, use a single reference, U3 for example, a single tail resistor, R11, connect the base of Q2 to the GND and R19 to the reference (or to the positive output, with a value equal to R21): this will spare some components and ensure a good +/- tracking
 
So I've been using the preamp for a while and there are a couple of things that I think can be improved on.

  • Firstly 50Hz is sill present when I view output spectra. Indeed it's bigger than all other distortion harmonics This is despite high PSRR in the preamp amplifiers.
  • There's other stuff in the audio spectrum too, audible with headphones with no input on the preamp and the volume turned all the way up. It sounds like a quiet murmur. I'm thinking it's the SMPS modules dropping into a low-power mode when there's no need for large power output, and pulsing every few milliseconds.
And it didn't occur to you even once that you might have a ground loop problem?

These Traco modules have all but 4 connections: L, N, out+, out-. Where do you think do their mains filter leakage currents are going? Clearly dissolving into thin air isn't an option. You bet you have some capacitive coupling between out- and (L+N)/2.

You can try something like 100 nF || 1k between power ground and PE and see whether that helps. It still won't be 100% ideal, you'll probably want balanced inputs on top of that. Your preamp's high PSRR is all for naught if what you need is CMRR.

SMPS are all fine and dandy if you can afford to go IEC Class I with your device. Otherwise they can be a real pain in the rear end, especially if (like here) you don't even have a PE connection on them that would allow you to acces the mains filter. If you did have that plus something like Elvee's Cancellator, even Class II would work pretty well.
 
Thanks for the note on the current limit base resistors, Elvee.

Here's the PCB: 120mm x 68mm. The TO-220 output transistors will need to be on a bit of a heatsink - a little bit of T-shaped extrusion will do, as the power dissipation is only a couple of watts. It's designed to go into a Hammond 1455P case, plus also has some holes for mounting in larger cases.

I reorganised the schematic a little to put the enable stuff on the upstream side of the current limit circuitry.
 

Attachments

  • Screen Shot 2019-09-28 at 11.03.05 pm.png
    Screen Shot 2019-09-28 at 11.03.05 pm.png
    881.3 KB · Views: 138
  • Screen Shot 2019-09-28 at 10.52.20 pm.png
    Screen Shot 2019-09-28 at 10.52.20 pm.png
    703.6 KB · Views: 131
  • Screen Shot 2019-09-28 at 10.53.44 pm.png
    Screen Shot 2019-09-28 at 10.53.44 pm.png
    866.3 KB · Views: 124
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.