Questions and rumbles about solid state amplifiers

[FiveseveN]

Sounds like crossover distortion.

[R.G.]

Yep. Crossover distortion.

You could replace that circuit more or less directly with a TDA2030 or LM1875 (considering the +/- 10V power supply) and be fairly certain that you won't have crossover distortion, but will have safe area protection of the power amp and thermal shutdown if it gets too hot.

[tca]

Yep. Crossover distortion.

You could replace that circuit more or less directly with a TDA2030 or LM1875 (considering the +/- 10V power supply) and be fairly certain that you won't have crossover distortion, but will have safe area protection of the power amp and thermal shutdown if it gets too hot.

I was suspecting that! I think I'll go for a LM1875, have to pay a visit to my local store.  I could also increase the power voltage but it is cheaper to buy a new chip!

But it is strange never the less... it seems a come and go phenomenon, that is why I thought about a temperature related effect (Vbe increases with temperature...).

Report soon. Thanks again for your answers.

Cheers.

P.S.
Going to make a sound sample just to future memory, that is a fizz that is worth remembering!

P.P.S
Found the SPICE/TINA-TI model: http://www.ti.com/product/lm1875#technicaldocuments

[R.G.]

I was suspecting that! I think I'll go for a LM1875, have to pay a visit to my local store.  I could also increase the power voltage but it is cheaper to buy a new chip!

That particular circuit has some oddities about it. The nominal "bias" on the power devices isn't usually an issue, as the power transistors don't need bias or do anything at all until the opamp is shoving enough current out to turn them on because of the voltage drop across the resistors in the power supply leads. So the opamp actually supplies the  whole load current for a fraction of a volt away from zero, and when that gets big enough to generate a Vbe voltage across the 100R resistors, the power transistors start pumping current in. Cut-in voltage for a silicon device is about 0.45-0.5, so you're talking a current through the opamp of 0.45V/100R = 4.5ma. Not bad.

The opamp then pulls the base current for the power transistors through its power supply leads. The opamp supplies the base current after the transistor cuts in.

Can the opamp do this OK? Gotta do the math.

A TIP32 is a PNP, and will have lower gain than an NPN. We are hoping it can pull the speaker load up to most of the 10V power supply. A decent objective would be to +9V. So we are hoping the output will put 9V across an 8 ohm resistor, for about 1.125A peak. Notice that I said "resistor". A speaker is a reactive load, and may well want more than twice the current of a resistor, but we'll hope for making a resistor work first.

With a collector current of 1.25 A, what is the base current in the TIP32, and therefore the supply current in the opamp? Easy - it's collector current divided by HFE  - er, and we know HFE varies.    

OK, we can at least see how bad it could be. Time for datasheets. The TIP32 datasheet says the minimum HFE at 1A is 25. Yep, there are those "typical" graphs, but the maker won't guarantee that.  So we'll guess that HFE is as bad as 25, so the current in the base may be as big as 1.125A/25 = 0.05A. The TI datasheet for the NE5532 says that the maximum output short circuit current is 38ma. That's as much as the 5532 can put out, so it's also as much as it can suck in through its power supply lead.

Nope, the 5532 cannot be **guaranteed** to drive the output to as much voltage as the power supply voltage would otherwise let it. Typically?? Maybe. If you have a higher HFE transistor, probably. But if you're an EE, your boss will fire you when he starts getting chewed on by HIS boss for a fraction of the units failing in the field and having to be expensively replaced under warranty.

[tca]

Hi R.G., thanks for your reply. Your description agrees with my idea how the circuits works.

> The nominal "bias" on the power devices isn't usually an issue.
Ok.

> A TIP32 is a PNP, and will have lower gain than an NPN.
Yes, confirmed that experimentally!

>... a higher HFE transistor, probably.
I could try that but the TIP31/32 where just sitting here waiting to be used.

> But if you're an EE...

I'm a completely different animal: a mathematical-physicist! I can do perturbative analysis with Einstein equations but not yet with audio amplifiers.

I've learn a lot so far, time for another configuration?! That chips that you mention earlier deserve some testing.

Thanks again, cheers.

P.S.
One my problems with this circuit is that I could not do a simulation with it, I usually do that with every circuit that I breadboard. PRR confirms my finding that "chip opamp models do NOT simulate dynamic supply current; just a nominal idle current. "

[R.G.]

> But if you're an EE...
I'm a completely different animal: a mathematical-physicist! I can do perturbative analysis with Einstein equations but not yet with audio amplifiers.

Yeah, I knew that. The "EE" stuff is just to show you how us failed mathematician/physicists live. 

[PRR]

> replace that circuit more or less directly with a TDA2030 or LM1875

that is to use a general audio amplifier and use it as a guitar amplifier -tca

Somehow he started on a quest for very *music amp specific* ideas and has come around to a '2030, a boring blob of Watts.

OTOH, looking back over decades of guitar amps come and gone, the only non-general plans which persist are the tube amps. There's a little simple/complex magic between the grids, plates, and transformer.

There's a few oddball transistor amps with fan-clubs; some of them seem to do one thing very well and anything else is ordinary.

This plan: as R.G. shows, it's starved at maximum output. But that's not your complaint. The sand on the cone is: up to 4mA the '5532 drives the speaker directly. '5532 is happy driving 600 ohms, NOT happy driving 8 ohms (even at small level). Gain and linearity are poor. Then up at 5mA-10mA the TIPs kick in. When they get to 5mA they add voltage gain, which goes up and up and up and UP with current. Open-loop gain may be 1/30 low for teeny signal and 30X large at maximum output. Any time you have a change-of-gain with signal level, that's distortion. The negative feedback moderates that, but who can moderate a 900:1 change of gain? I strongly suspect oscillation at high level; at low levels the '5532 gain-bandwidth is far short of spec and NFB essentially fails in the upper audio frequencies. This *can* be nice for music (Bassman Presence control) but is more often just nasty. Especially since it happens only in a narrow signal-level range around 1 milliWatt (decay of bedroom-level amp).

[PRR]

If you can stand the heat, hack-up a John Lindsey Hood:
http://sound.au.com/tcaas/WW4-69.gif
http://sound.au.com/tcaas/jlh1969.pdf
http://sound.au.com/tcaas/

[brett]

Hi
interesting ideas.

There's a little simple/complex magic between the grids, plates, and transformer.

So why don't more people use output transformers in solid state amps? Probably because it's easy to NOT use one. And cheap. And many people have never heard a good SS+output transformer amp. Fortunately, DIY isn't always about easy and cheap and familiar. IMO this is a large field of DIY experimentation that has been lurking in the background for about 10 years. Without the in Brian May's 'Deacy' amp and one or two others (Pignose), the benefits of OTs in solid state amps would be almost unknown.
cheers

[tca]

Hi PRR, thanks for your reply.

>> replace that circuit more or less directly with a TDA2030 or LM1875
> that is to use a general audio amplifier and use it as a guitar amplifier -tca
That can not be a good thing, my statements are haunting me!

> looking back over decades of guitar amps come and gone, the only non-general plans which persist are the tube amps.
> There's a little simple/complex magic between the grids, plates, and transformer.
How far can we take those statements? That feeling thus make a SS guitar amplifier quest unfeasible, almost like a I-told-you -so statement, a highly questionable and waste of time holy grail crusade. My gut feeling is that such a good SS amp can be made with out extraterrestrial technology or other magic.

> There's a few oddball transistor amps with fan-clubs; some of them seem to do one thing very well and anything else is ordinary.
That is also my feeling.

> started on a quest for very *music amp specific* ideas and has come around to a '2030, a boring blob of Watts.
You are correct. I started with some cool discrete designs and converged to a boring blob of watts. Actually what I have on my breadboard is a TDA20* + TL* guitar amplifier! I guess these posts reveal an experimental form of the principle of least constraint (btw, my favorite way of deriving Newton's equations). Going to test some ideas with the 2030... and see how that sounds... But I/you can predict the outcome of that realization.

> If you can stand the heat, hack-up a John Lindsey Hood.

I'm glad you mention that, I did tried a 18V  small version of the JLH, at the time I didn't liked the response of it, but I SHOULD give it another try.

I have a few schematics of high voltage (typical 100V) transistor amplifier with output coupled transformer but I never found a compatible transformer to give it a go.

Cheers.

P.S.
Just for a laugh! Just blew up two LM1875, very sticky (to rail) devices.

[tca]

> If you can stand the heat, hack-up a John Lindsey Hood.
I've been playing with a small 18V version of the JLH... much better now, not to much heat btw, roughly 1W rms . Didn't like the first run, but it seems a bit more knowledge helps. Actually what I'm using is a simpler version, using  the schem of fig. 7: https://www.passdiy.com/project/amplifiers/the-plh-amplifier but with BJT.



I guess I should put back the first device and make something like the original



Cheers.

[R.G.]

Just for a laugh! Just blew up two LM1875, very sticky (to rail) devices.

Many Linn-topology amps (and this is one), including discretes, will exhibit sticking to the rail when the inputs are overdriven. It's an issue of the overdrive and feedback paths activating a positive feedback loop internally.

This is why many amplifiers have back-to-back diodes across the + and - inputs. This pretty much solves the issue of overdrive causing sticking for most cases.

For a guitar amp, it is also a good idea to look at the amount of signal that can go into the amp. If you think about it, trying to drive the input of a solid state amp with lots of feedback further than the supplies is not all that helpful. The knee of clipping is going to be harder than most musicians like, and any ephemeral advantages of "saturating the output stage" like in tube amps is not going to be there.

Thomas Organ took a practical and good-sounding approach in that they put a diode limiter before their power amps. This means that the power amp is never driven outside it's linear region, but exhibits a composite soft(ish) clipping characteristic, and never hits the oddities of the power amp being driven too hard on its input. It's well worth looking at.

[tca]

For a guitar amp, it is also a good idea to look at the amount of signal that can go into the amp. If you think about it, trying to drive the input of a solid state amp with lots of feedback further than the supplies is not all that helpful. The knee of clipping is going to be harder than most musicians like, and any ephemeral advantages of "saturating the output stage" like in tube amps is not going to be there.

I understand that, but it seems very counter-intuitive not being able to saturate the output stage even if using a SS amp. I think the same effect happens, or is similar to, while playing a acoustic instrument, the player rests on the assumption that by picking the strings hard (increasing the input) some type of distortion will occur.

(edit) This could be done by overdriven the preamp of course, but can I call a one BJT booster a preamp?

Thomas Organ took a practical and good-sounding approach in that they put a diode limiter before their power amps. This means that the power amp is never driven outside it's linear region, but exhibits a composite soft(ish) clipping characteristic, and never hits the oddities of the power amp being driven too hard on its input. It's well worth looking at.

I've tested that idea on the previous NE5534 amplifier so that the abs max input would be smaller than 0.6V. I'll have a look on that ref. Thanks.

[tca]

Have you ever looked at the Lab Series L5 amp? It has an awesome crunch tone - think early King's X guitar sound. BB King and Allan Holdsworth were also notable users of that amp. Bob Moog/Norlin design. No output transformer. Granted it's complex, but it was designed specifically for guitar and to sound great, which I think it does.

http://www.diyguitarist.com/Schematics/L5-Schematic.jpg

I forgot to mention this Moog's distortion. I can not see it on that schematic but these resonant filters are easy to find.

Cheers.

[gritz]

I tend to agree that lots of open loop gain coupled with large amounts of global negative feedback is the enemy here. You get a very "stiff" and faithful output, right up until the point where it can no longer follow the input - and then it falls over completely. OK for hifi amps, but not good for guitar.

So, intuition suggests that we either;

a) "fake" our overdriven condition, by tossing it in upstream and make sure that our amp can't rail to rail clip (like R.G. has already suggested). Perhaps we can make our amp's output a little less "stiff" by incorporating the speaker into the feedback loop, as has been done by Marshall and others, or:

b) start from scratch with a topology that isn't from a hifi cookbook, or power IC datasheet. Less loop gain, less (or no) global feedback. It would be nice to avoid the cost and sourcing hassle of an output transformer, but in many ways it makes things a bit easier - single power rail, no worries about DC offset, no need for complimentary output devices. I'm sure it's possible though to build a transformerless design using e.g. complementary output mosfets with a bit of local feedback around the driver / output stages without having to fiddle with trimmers and whatnot every five minutes.

I suppose that if it was easy, we'd all be doing it!

[artifus]

... incorporating the speaker into the feedback loop, as has been done by Marshall and others, ...

do you mean kinda like this? http://www.diystompboxes.com/smfforum/index.php?topic=81956.msg679661#msg679661

http://www.current-drive.info/9

or am i barking up entirely the wrong tree, lost in the forest once more?

[gritz]

... incorporating the speaker into the feedback loop, as has been done by Marshall and others, ...

do you mean kinda like this? http://www.diystompboxes.com/smfforum/index.php?topic=81956.msg679661#msg679661

http://www.current-drive.info/9

or am i barking up entirely the wrong tree, lost in the forest once more?

Yeah, that's the idea! *very basically* the typical hifi style solid state amp (discrete or on a chip) is a bit like a regulated high current power supply and it crowbars the loudspeaker into following it's will. By incorporating the speaker into the feedback loop the amp becomes more responsive to the way that the speaker's impedance changes at different frequencies - particularly at the resonant frequency of the speaker + box it sits in - it allows more of the speaker's "colour" around this point to shine through. An important part of exploiting this characteristic is the resistor and capacitor that are fitted in parallel to the speaker - as the frequency rises above the midrange, the inductance of the loudspeaker coil starts making itself felt, raising the impedance of the speaker. This can make the response very trebly as the amp tries to force more current through the speaker at high frequencies (as the speaker looks like a high resistance at high frequencies). This is tamed by the resistor / capacitor combo. The resistor is around the same value as the speaker's quoted impedance and the cap is tuned by ear to give the required response - bigger = darker, smaller = brighter.

If you have a look at some Marshall Valvestate schematics, you'll see the same principle.

[tca]

^That and most important to protect the output stage from breaking-down and preventing the energy store in the speaker inductance to dissipate in the breakdown region of one of the output devices (zobel network), the values for C and R should be something like:

Rz=Rspeaker
Cz=Lspeaker/Rspeaker^2


Also the schem in http://www.diystompboxes.com/smfforum/index.php?topic=81956.msg679661#msg679661 sounds terrible to my ears (just breadboard that a few days ago).

Cheers.

[tca]

I like your suggestion:

start from scratch with a topology that isn't from a hifi cookbook, or power IC datasheet. Less loop gain, less (or no) global feedback. It would be nice to avoid the cost and sourcing hassle of an output transformer, but in many ways it makes things a bit easier - single power rail, no worries about DC offset, no need for complimentary output devices. I'm sure it's possible though to build a transformerless design using e.g. complementary output mosfets with a bit of local feedback around the driver / output stages without having to fiddle with trimmers and whatnot every five minutes.

I suppose that if it was easy, we'd all be doing it!

I'm still looking for that not-hi-fi-cookbook-or-power-ic-datasheet-topology And yes it is not easy, but it is a lot of fun!

Cheers.

[R.G.]

I understand that, but it seems very counter-intuitive not being able to saturate the output stage even if using a SS amp. I think the same effect happens, or is similar to, while playing a acoustic instrument, the player rests on the assumption that by picking the strings hard (increasing the input) some type of distortion will occur.

I guess that it's not counter-intuitive to me.

The issue is not saturating the output stage, it's what happens to the rest of the circuit when the output stage saturates. Output stages by themselves are not sticky. They largely do what they're told as long as it's no faster than the device physics allows them to turn on and off. And it's very dependent on what's happened at the input and voltage amplifier stage(s). And it depends on whether the input common mode range or the input differential mode range is exceeded.

It's not usually an issue unless the input is overdriven a lot. It's a recognized problem with integrated opamps, and many of them have internal protection diodes for just this reason. A little careful design eliminates it while still letting the thing distort. It just doesn't latch up or stick and go up in smoke.

It's fine if the amplifier limits when driven by too much input, and it will merrily contribute distortion as expected by musicians (although perhaps a different distortion than was expected). But like all active (well, and passive, too, come to think of it) things, it has limits, and its limits are often different from other things.