LM317's on positive AND negative rails....? Power supply design query

[Bunkey]

PS. Are you guys telling me in no certain terms that my last idea with the virtual ground and seperate caps isn't going to work;
or are you just trying to avoid the solution of floating L3 on a virtual ground to avoid the possibility of confusing what is what in future?

[iainpunk]

PS. Are you guys telling me in no certain terms that my last idea with the virtual ground and seperate caps isn't going to work;
or are you just trying to avoid the solution of floating L3 on a virtual ground to avoid the possibility of confusing what is what in future?

it would probably work, but its not ''properly done'', i recommend making the full 48v ground referenced, and put in a Vbias, this means taking off the ground on the middle connection of the transformer and adding stacked voltage regulators to get the nice 24 middle voltage. i don't think a 386 can do tubes any justice, you might want to go for a better sounding power amp IC with a bit more headroom, like the TDA2020 or TDA2030 (check their max voltage before getting them, idk if they handle the full 48v)

cheers, Iain

[Unlikekurt]

Not to try to confuse things further, but to toss a completely different idea at this; if the 48v rail is simply for the plate supply, which will draw very little current, why not go with a full wave rectifier, grounding the center tap.  Then use a charge pump/ boost converter to create a high voltage node for the plate; which you’d certainly like.
You will have your 24 from which to make 18 and 12 without an issue as well.  Probably feed the boost converter from the 18v regulator for stability.
The expense wouldn’t be incredible by any means and you’ll also do away with all the funny business with floating grounds.

[iainpunk]

this gives full power for the low voltage parts, and lower power to the plates.
i guess the 18v is for the heaters? on such starved plate situations you really don't need the full heater current, so you can get away with only 12v

as an alternative, you could also just use a 12v transformer and use a voltage multiply circuit for a higher plate voltages:


cheers, Iain

[antonis]

as an alternative, you could also just use a 12v transformer and use a voltage multiply circuit for a higher plate voltages:

Do plates like voltage multipliers high ripple as they like their high voltage..?? 

[PRR]

....Whether the principle is conventional or not isn't a concern.....

It's not about being conventional. If both signal and power are shared across several stages, it is far simpler if one side of power and one side of signal are "common". In the most careless case, you plug it in and smoke comes out. More benignly, you inject VOLTS of unexpected buzz into the audio path.

> Load 1 = 18v / 16mA
> Load 2 = 12v / 150mA
> Load 3 = 48v / <20mA

OK, the 12V is tube heater. Could be 6V at more current. 18V is a high voltage for LM386 but the full-roar current will be far over 16mA (which may be the idle spec, but you won't be idle all the time). Expect over 250mA to the audio power chip. The 48V is the tubes' plates and is only approximate (I feared Phantom Power which sometimes has to be pretty exact).

And you have 18VAC+18VAC (two separate windings?) transformer.

Dogs are crying for supper. Look for post #27.

[Unlikekurt]

The filament supply for one 12A_7 wire in series arrangement will want to see 12.6v 150mA. The 12v regulated supply will do quite well. 
I still say the OP would be happiest using a 12AX7 with traditional plate voltages, which can easily be achieved with a smps boost circuit. A circuit using a 555 timer  ic has been discussed many times on the forum.  One such instance is here:
https://www.diystompboxes.com/smfforum/index.php?topic=124809.msg1185461#msg1185461
Of course, at that point, one would likely toss the valvecaster schematic aside and look to more traditional tube preamp designs to suit one's fancy.

[PRR]

[Bunkey]

Thank you very much Paul, I will give this a try.

The 386's are the Texas Instruments 1W 'N-4' chips - I think the Vmax is 22v. I plan on bridging 2 of them set at low gain so I'll keep that current draw in mind, thanks again.


Once I've got this working I can look at some of the other ideas mentioned here but obviously I still have a lot to learn and I'd like to understand what I'm doing so this low voltage tube / 386 combo is a comfortable starting point. Thanks everybody for the suggestions.

[Rob Strand]

Voltage doublers aren't something I've come across before but I'm guessing it means double the current draw too?

With adding complexity I'm concerned about making the supply noisy but I will look into it for sure, thank you.

The input current is twice the output current.   However it does not *waste* power in the sense of having losses.   The input *power* is more or less equal to the output power.   A 36Vac transformer with a bridge rectifier and a cap will require the same size transformer as a doubler with a 18Vac.

The difference in your case is the doubler would only connect to one winding.

If your transformer is close to its VA rating you should be aware that the full wave rectifier connection with the center-tapped winding and the two diodes has a lower DC output current rating than a bridge rectifier.  Your 18V and 12V rails are current powered this way.

If you can changed the transformer connection from the centre-tapped 18Vac-0-18Vac connection to two parallel windings connected as 18Vac and use a bridge rectifier you will  utilize the power rating of the transformer a bit better.   You have to make sure your transformer can be wired in parallel, also you need to get the phase of the paralleled winding correct.

Doing that loses the 48V because you no longer have 18V-0V-18V.  To get 48V you can use a different form of doubler, which also lets you use a bridge rectifier.  In addition the doubler is powered from both windings.  So this connection fully utilized the transformer power, gives you a common ground, and gives you all the DC rails.   The only thing you need is to make sure the 1000uF cap shown on the schematic is large enough.



(FWIW, I don't have a problem with PRR's solution unless you want to pull max current from the 18V and 12V rails.  PRRs still offers good ratings and is less parts.)

[Bunkey]

Ahh I see - It was my intention to keep the supply bridge rectified and I wasn't quite sure how supplying the 317's from the middle tap was still having the rectifier function as a bridge but that's because it isn't, it's turning it into a full-wave rectifier for the 317's like you say. I guess this would have been the case for the 317's going +24 to 0v ground on my ill-conceived virtual rail schem too (whilst the +24 to -24 L3 48v was fully bridged)?

The current rating is indeed a concern with the higher draw from the 12v & 18v - It's 15VA per winding. These can be parallelled for 30VA so once my brain has cooled down a bit I'll work out what the current demand is actually going to be and see if that's a more appropriate way of handling it.


So how does the doubler in that schem work?
I guess it's something to do with the AC swing pulling on the -ve terminal of Cseries to create potential across it, then pushing on it during the +ve swing and therefore raising the potential of the rail at the +ve terminal when referenced to +17vDC via D1, which then charges Cboost to +34v acting as the rectified dc supply with the addition of D2?

I'm just trying to make sense of it.

[antonis]

So how does the doubler in that schem work?

https://www.electronics-tutorials.ws/blog/voltage-multiplier-circuit.html

[Rob Strand]

So how does the doubler in that schem work?
I guess it's something to do with the AC swing pulling on the -ve terminal of Cseries to create potential across it, then pushing on it during the +ve swing and therefore raising the potential of the rail at the +ve terminal when referenced to +17vDC via D1, which then charges Cboost to +34v acting as the rectified dc supply with the addition of D2?

That's correct.  The ripple on the doubler output cap is at the mains frequency not twice the mains frequency like a full-wave or bridge.   Notice the second double I posted with the bridge steals charge from the lower voltage cap, whereas the previous doubler only gets power from the AC input.


I just took the figures you posted earlier and calculated the DC power from the transformer,
Vreg     Vunreg      Load      PDC
[V]          [V]              [mA]       [W]
18          24              16           0.384
12          24              150         3.6
48          48              20           0.96

Total DC power is about 5W.

Typically the transformer VA needs to be 1.6 to 2.0 times the DC power so a 30VA transformer is going to do that easy.
There's so much margin it doesn't seem worthwhile scraping for a few watts with the doubler + bridge.
Pauls's solution will work fine.

[PRR]

....The 386's are the Texas Instruments 1W 'N-4' chips - I think the Vmax is 22v. I plan on bridging 2 of them.....

Over-volting and bridging would only make sense if you had some odd (high) load impedance. The LM386 (any flavor) will only pass 0.4Amps peak to a load. (Limited by small die and teeny power transistors.) If you jack the supply to 18V and double-bridge, you could hope for 15V peak, at 0.4A peak, is a 37 Ohm optimum load and a mild 3 Watts output at best.

I assume you are locked-in and can't get any more suitable power amp? Even an old car behind the shed? Pull the radio. 12V supply to 4 Ohm load is good for 4 Watts.

[Bunkey]

Man I need to stop and just take a minute to laugh at myself here because when I came into this about a week or two ago I'd just discovered the LM386 and I thought it was incredible;
I never imagined you could get a power amplifier on a chip this size capable of driving an actual 4ohm guitar speaker at 1W. My Blackstar HT1R with 2 tubes and a handful of IC's only puts out 1W. The concept was a game changer to me.
Of course, after scratching my head over it for a bit I came up with the crazy idea of bridging 2 of them together for 2W of power. Mind blown.

It's only now I realise you guys have been having this same conversation with hundreds of people, all thinking and doing exactly the same thing, for the past 20 years or so 

This is hilarious. I am humbled.


I'm honestly just going to play around with this for a bit because I still find the thing fascinating. It's a bit like getting a bigger engine; the more power you have inevitably the more power you end up wanting. I think I'm more of a minimalist in this sense; I like to take the simplest most basic elements and make them as good as I possibly can until I hit the ceiling (if I find it). I think I learn more that way but it keeps the door open to new discovery too. Savour the pages.

Anyway..

If these bridged 386's do draw 250mA each at full whack then that's what.. 16 watts DC total. Plus any indicator LEDs, tone stacks and whatever else I might end up with..

It's probably best to make use of the parallelled secondaries after all.

I've just stumbled across the concept of heat dissipation too and I'm looking at dissipating 3 to 4 watts in that 18v regulator, depending how I set it, so maybe thinking about heat management as well..

This is the most fun I've had in a while 
Cheers

[Rob Strand]

As you go up in power there's more motivation to run the power amplifier from unregulated rails and only run the preamp (low power stuff) from a regulated supply,

Check out, 
https://sound-au.com/project72.htm

If you look up some of the Marshall Valve-state amps they connect the feedback around the amp differently so the amp output impedance has resistance; most solid state amps do this in this day and age.  It tends to sound better for guitar.

[PRR]

....It's only now I realise you guys have been having this same conversation with hundreds of people, all thinking and doing exactly the same thing, for the past 20 years or so  ....

Everybody comes down a different path at a different time, but certain way-stations are super popular stopping points.

I built "that" little speaker cabinet, but as a two-3", about 30 years ago.

Before that I was building discrete power amps. I thought the LM386 was a cheap trick. The LM380 is more like real power. The LM377 was a great stereo mini-system engine; I burned-up a few.

And eventually you think: do I really need Corvette power in my golf-cart? A tenth-Watt from a good pocket radio could annoy the family. More so with that bigger Jensen Special. And these days we can't get far from family, can't go out and annoy different strangers every night.

LM383 is a fairly serious power chip. At 12VDC and 8 Ohms it will deliver a couple Watts with a small hunk of Aluminum heatsink. That's really all we need in 2021.

[Bunkey]

Well I ended up breadboarding one of the LM386's with a 'pre-amp' derived from a mild fuzz pedal I'd been working on (2x BC108a with Ge diode clipping), all running on a 12v wall supply, and it actually works really well as-is - A little smaller of an amp but probably a more suitable config for that chip you would say?



1x Gain and 1x Volume control. A real sparkly vintage edge-of-breakup sound about it. Little Wing vibe.

I also took the Jensen out of the cab and put a pair of 4" paper cone speakers back in (2x6ohm in series) which seem to sound better despite being of lesser quality - The cab wasn't the deepest thing in the world; the 4" speakers are about half the depth to the coil as the Jensen is so I wonder if the extra space between the back and the rear baffle has something to do with it. Along with a bit of tone shaping in the pre-amp it's sounding altogether more balanced.

Just working on an enclosure for this sub-miniature amp to match the cab now..




Nice little side project anyway.


As for the allocation of this power supply - I've ordered a few of the LM380's to run instead, at 18v bridged just the same as I'd planned with the 386's.
I'm gonna up-scale the cabinet design slightly to fit 2 x 5" Jensens, or a single 6" Jensen C6V I've had sat here for a while depending what sounds better.
I think those proportions with a bridged pair of the LM380's will be a bit more deserving of the tube pre-amp, as some of you have stated already.

What is your experience with the LM380, Paul?

I'm planning to use the 8-pin MDIP package with a heatsink pasted to the top but I'm struggling to wrap my head around the requirements for such..
The pins themselves are going to be soldered straight to a 1.2mm dia solid copper wire 'skeleton' circuit that I think will do a decent job of heatsinking alone.
In addition I'm just looking at the little fin sinks about the size of the chip itself (10 x 8 x 6mm) to stick on top.
- Any advice on this?

Cheers

[iainpunk]

In addition I'm just looking at the little fin sinks about the size of the chip itself (10 x 8 x 6mm) to stick on top.
- Any advice on this?

maybe a bit too small, the chip has an efficiency of about 40%, so the other 60% will be heat. on 2.5W that is 4-ish Watt of heat, keep that in mind when chosing the heatsink. most heatsinks of 4 to 5 watt range are around 50mm x 50mm give or take a few.

The pins themselves are going to be soldered straight to a 1.2mm dia solid copper wire 'skeleton' circuit that I think will do a decent job of heatsinking alone

a deadbug style circuit? thats an... ""interesting""... choice, haven't seen that in a long time. last time i build something deadbug style was when i was doing a Fuzz Face circuit, but i totally fudged that, both transistors fried... i was an idiot.
i don't think those wires would be a sufficient heat sink tho, unless you wish your IC to de-solder itself.

cheers, Iain

[Bunkey]

a deadbug style circuit? thats an... ""interesting""... choice, haven't seen that in a long time. last time i build something deadbug style was when i was doing a Fuzz Face circuit, but i totally fudged that, both transistors fried... i was an idiot.
i don't think those wires would be a sufficient heat sink tho, unless you wish your IC to de-solder itself.

Haha those copper rods are mighty efficient at transferring heat aren't they!

I built my first 'proper' pedal in this style and had great fun attaching something to one end whilst something else fell off the other... 
In the end it looked fantastic and sounded terrible but that was another issue entirely.

If I could find PCB's that were attractive I'd consider it. The closest I've come through RS-Components in the UK is the Adafruit 'FeatherWing' boards with their copper eyelets on a black backing but at £4.54 for a 22x50mm modular slice they're not exactly cheap..

Edit: Ignore that, I obviously hadn't googled the plethora of similar boards available on amazon 

I'm open to suggestions for alternative build styles though...