Amplifier Design - A Documentation

[Bunkey]

It looks like the LM380 is going to be reasonably happy. Looks like about 11-12Vpp with the 15Vcc is ballpark and the output into an 8 or 12ohm load will be 2.25W and 1.5W respectively; which is within device limits as long as I fit a heatsink that brings the 107C/W down to <75C/W (JC=37C/W) and my 70C/W hedgehog sinks should just about do that, assuming this works on a parallel resistance basis.

The Vrms @ 12Vpp is 4.242v, which would make the chip's current draw at full whack:
8ohm = 530mA and 12ohm = 353mA;    *No it wouldn't, see newer posts*

The pre-amp section is drawing <10mA, power indicator LEDs etc. say <20mA

So again this is within the 600mA limit of the transformer...


This is looking good and I certainly don't need much more than 1.5W of output given the size of the cabinet & amp combo. I think more would be too much.


Looks like it's the single LM380 from here on then...

So long bridged 386's, you taught me lots!

[iainpunk]

i''m getting quite curious about the preamp you employ, do you have a schematic?

cheers, Iain

[Bunkey]

If it's peaked your curiosity that probably means I've done something wrong with the calculation 

The pre-amp is a discreet design that I've been working on in some capacity or another for a long frigging time and I'm actually quite proud of it - which is exactly why I haven't posted a full schematic of the amplifier here Maybe when it's finished and I've got a bigger better amp in the works I'll share it. As I said in the OP it's just a simple design using dual BJT's, diode clipping and a LPF/HPF crossover tone stack.

It's really nothing special, I just like the sound at the edge of breakup and spent a while tweaking values to taste. Bit of an Electra style distortion, the tone stack was heavily influenced by the Big Muff Pi. A couple other additions to make it all work.

Of course, when it's slamming against the low ceiling of a massively compromised power amp it sounds a bit.. different.

[PRR]

> Vrms @ 12Vpp is 4.242v, which would make the chip's current draw at full whack: 8ohm = 530mA

Can't be right? 4Vrms in 8 ohms is 2 Watts. 12VDC at 0.5A DC is 6 Watts. Terrible efficiency. And the apparent 6W-2W= 4 Watts of loss would cook the chip.

[Bunkey]

> Vrms @ 12Vpp is 4.242v, which would make the chip's current draw at full whack: 8ohm = 530mA

Can't be right? 4Vrms in 8 ohms is 2 Watts. 12VDC at 0.5A DC is 6 Watts. Terrible efficiency. And the apparent 6W-2W= 4 Watts of loss would cook the chip.

Ah I hadn't thought about it like that.

No of course it's not right because the 8ohm load of the speaker and 4.2Vrms AC signal that I was using to calculate this has no connection with the DC component 

The device dissipation table shows about a max of 1.5W dissipated for the 15v supply with an 8ohm load, so that would suggest 100mA instead?

I'm curious why the dissipation curves the way it does - peaking at around 1.6W output power (extrapolated) - Does this mean it becomes more efficient at higher outputs?

[Bunkey]

100mA seems too low...

The output power must be added to this right? (It has to come from somewhere)

So 2.25w is audio dissipation (represented here as output power) and ~1.5W is heat dissipation (represented here as device dissipation), so it's actually 3.75W total consumption = 250mA or thereabouts @ 15v supply. 


This way everything outputted is accounted for at the input

[PRR]

Efficiency is always zero at zero output. So it tends to rise when it does real work.

Yes, 1/4A is what my thumbs say too.

[Bunkey]

Just a wee update.

It's sounding pretty good into an 8ohm speaker with the LM380; a bit more lively than it is at 12ohm so I'm trying the 5" Jensen in the cab again with a slightly thicker 6mm BB baffle.

I'm gonna need a bigger sink for the chip though, the little 70C/W thing is hot to touch at half volume so the junction temps must be way up there.
Using protoboard as a base isn't ideal when it comes to finding a solution to this - I think the best option for the space might be to fit a TO-220 sink like this:



I've a handful on the way rated at 18C/W. It lines up quite conveniently with the 3mm mounting hole in the board; some sort of spring arrangement to apply pressure directly above the chip would work well.

And this is how I left it all this evening:




The new baffle has indeed sorted the problem of bloom in the mids. I still think the cab is a bit too small for the 5 inch speaker though.. We'll see tomorrow when it gets turned up.

[iainpunk]

do you use thermal paste? in my experience, a tiny bit makes a huge difference in chip temp!!! you don't need the expensive computer stuff, just the cheap grey stuff, or you can make it yourself from petroleum jelly and super fine silica sand.

cheers, Iain

[Bunkey]

do you use thermal paste? in my experience, a tiny bit makes a huge difference in chip temp!!! you don't need the expensive computer stuff, just the cheap grey stuff, or you can make it yourself from petroleum jelly and super fine silica sand.

cheers, Iain

Yeah man of course.

The standard 70C/W CA needs derating to at least 35C/W CA according to the datasheet (@1.67W device dissipation).

Most of the standard heat dissipation is going to happen at the top of the chip; once I stick a heatsink there and cover it, the effective device thermal resistance is going to increase dramatically as the device itself is now only dissipating from its tiny sides..

So instead of being 70 case || 70 sink = 35C/W it's going to be more like >140 case || 70 sink = >46C/W which puts it right on, if not over, the 150degC limit at room temp using the tiny sinks I had for the LM386's and dissipating 1.5W as estimated for the 15v supply.

...That's how I'm looking at it anyway and it makes a bit more sense since the device no longer has that top surface area in addition to the heatsink.


I plugged everything into a calculator which is suggesting 12C/W for the sink in 50deg air to keep the Jt <125degC.
18C/W sink = 104C & 135C @20 & 50C repectively but it's the best I can do here.

...That's using some cheap cpu paste I had lying around, C/W = 0.12 per square inch so I've adjusted that to R = 1.2C/W for the 0.6cm^3 area of the chip.

[Bunkey]

Thought I'd check in with post #50 and a wee game of spot the difference...



Maybe this marks the half way point.


I don't really write much about what I'm doing when I'm actually doing something so yeah it's been a busy few weeks..

The pre-amp has been overhauled somewhat; still the same just different. I went through just about every type of transistor in my possession in every configuration on the board, it was quite remarkable what a difference they make even down to the individual gain of each transistor, regardless of the supporting components.

I won't talk about the cab just yet but obviously there's a lot of work been done, and to be done, there too.

I've since built a DSO138 ocilloscope which is proving really useful - Things started to change for the worse, I think as all the fresh capacitors and bits loosened up, the nice tube-like spongy attack I was getting turned into a hard harshness. From what I gather the LM380 is happy with a few hundred mV of input. My LP outputs peaks upward of 500mV (!), the pre-amp clips to ~300mV then has a make-up stage after the tone stack with about 5x gain, maybe, so I must have been hammering the chip's input at higher volumes and that's exactly what it sounds like. It had a bit of 'give' in the sound at first which was really nice but now it's determined to produce those edges so I need to tame that and maybe try to reproduce the spongy behaviour it was exibiting before, somehow. An LDR compressor might fake it but I don't want to complicate the thing (besides that kinda belongs in a pedal, not an amp) - I think this is more a supply factor with maybe some soft-clip limiting at the power-amp input or something, we'll see.

I had to build a power supply for the scope so I've implemented the same style as the amp to get a feel for the protoboard base and how everything is going to fit together..



I spent an entire day just laying this out on the board. It's quite a head scratcher condensing full-sized components into a tiny footprint but I think the results are going to be pretty cool. I've still gotta fit screw terminals and things to this supply hence the space at the top.
The amp itself will have neater design elements; hopefully a bit of symmetry and pattern to the layout as well as a more full use of the space (board for the amp's supply is half this size!).

[antonis]

Just do not test the amp while vertical drill and washing machine working.. 

[iainpunk]

My LP outputs peaks upward of 500mV (!),

that's not that much, my epiphone explorer has 600mv on its sustain, and over 1.3v on its attack (passive red led clippers are audibly distorting the attack)

just some food for thought:

an input volume control is nice if you have high-ish gain and low headroom.

if you want to soften the clipping of the power amp, you might want to make the power filtering worse, with a higher resistor and smaller cap, it gives the voltage some sag.
you can also decide to have the pre-amp clip and have a passive attenuator so the power-amp doesn't clip when the pre-amp is totally overloaded.

cheers, Iain

[marcelomd]

My LP outputs peaks upward of 500mV (!),

that's not that much, my epiphone explorer has 600mv on its sustain, and over 1.3v on its attack (passive red led clippers are audibly distorting the attack)

Heheh... puny guitar pickups... According to G&L, my L2500 outputs 3V+ in passive mode. They clip the input of almost everything I plug it into.

Also, awesome thread. You motivated me to post the progress on my bass amp. I expect to finish it (as power amp only) in the next couple weeks. Then I'll design an internal preamp.

EDIT: typo

[Bunkey]

if you want to soften the clipping of the power amp, you might want to make the power filtering worse, with a higher resistor and smaller cap, it gives the voltage some sag.

cheers, Iain

Yeah this absolutely does seem to have come about from cleaning up the ground routing and adding a bypass cap to the inverting input of the amp (I was getting sick of listening to all the bullshit on the radio  )

I'd tuned in the sag really nicely with a 100mOhm series resistor on the output of the regulator but I think the caps have broken in a bit and aren't as stiff as they were then so I intend to play around with this some.
Having excess noise taking up the bandwidth does take the edge off really well when it's in the right zone but it's getting fairly inconsistent as the amp settles in and I imagine it'll sound entirely different again in a shielded enclosure, so I need to come up with a more predictable solution to this..
I've a few ideas to try - increasing the supply sag is absolutely one of them. Thanks

My LP outputs peaks upward of 500mV (!),

that's not that much, my epiphone explorer has 600mv on its sustain, and over 1.3v on its attack (passive red led clippers are audibly distorting the attack)

Heheh... puny guitar pickups... According to G&L, my L2500 output 3V+ in passive mode. They clip the input of almost everything I plug it into.

Also, awesome thread. You motivated me to post the progress on my bass amp. I expect to finish it (as power amp only) in the next couple weeks. Then I'll design an internal preamp.

I'd love to see this!
Juicy threads are great - I try my best 

The pre-amp should be happy enough handling 6-7v peaks at its input. I might think about attenuation since your guitars are so brutal 
The 'gain' is supposed to be more of a 'clean breakup' control so this area might need some consideration..

[Bunkey]

[Rob Strand]

hat's not that much, my epiphone explorer has 600mv on its sustain, and over 1.3v on its attack (passive red led clippers are audibly distorting the attack)

It's usually fair safe to allow for 1.2V to 1.5V peak.

[Bunkey]

The pics of the supply were terrible yesterday. Here's a better one..



As a proof of concept for the build style, I think this is looking pretty good.


..More importantly, I got the scope up and running along with a signal generator.
Wow these things are friggin awesome!

So bloody useful - pretty much the difference of being able to build a proper amp instead of just winging it...

So there's nothing wrong with the signal amplitude at the power amp input; about 150mV peak.
I can see where the harshness is though and it's more to do with the tone shaping, which frequencies are being attenuated and how the waveform and amplitude changes with frequency and gain setting.



Rolling the gain right back balances the attack & sustain of the waveform (which sounds really round, warm and nice); high frequencies accentuate that sharp peak where low frequencies do the opposite and make the waveform more rounded; right now this balance point occurs at the start of the gain control so I need to play around with the tone stack to bring this middle point further into the sweep of the control instead. I can see this on the scope and I think it'll make all the difference between having the amp properly dialled in or not.

Also, the clipping that accompanies this harshness is happening at the first gain stage in the pre-amp with amp input signals above 1.5V - it's great being able to see exactly where the ceiling is!



I'm getting this clipping with the guitar volume cranked above 8 so I'd hazard a guess that this LP with its oddball 700-and-something-kOhm pots is putting out similar peaks to those mentioned above.


Very happy that I can see which bits need attending to now 
I really recommend the DSO138 if you don't have an oscilloscope. £25 Amazon

[Bunkey]

With the ESP M-50.

[iainpunk]

cool dude! really like its off-clean sounds! very natural sounding!

do you have a drawn schematic of the full citcuit? i might learn something

cheers
edit: i generally don't have drawn circuits, its all in my head