Op-Amp issues on breadboard

[eschertron]

...and run one LM3886 per speaker. The speakers are effectively acoustically paralleled, so you get 30W per LM3886, up to as many watts as you need...

[tangent] I'm suspecting this scalable SS amp philosophy is behind the recent threads here regarding power soaks/speaker emulators. It will be nice (in the near future) to get the same feel from the same amp at 5, 15 or 150W. [/tangent]

[Zozobra]

Yes, believe it or not the amplifier itself (semiconductors, passives and PCB) is the *cheapest* part.

The bulk $$$ (as well as the physical bulk and the weight bulk) are in power supply, *transformer* , *chassis* , and more or less on cabinet/case, depending on how fancy you get.

And we have not even mentioned the speaker and speaker cabinet.

There is no loss of prestige or bad karma for buying an LM3886 kit: both the typical EBay supplier and yourself will use the same Factory datasheet suggested circuit, and he´ll supposedly use a properly designed and very well made (in China) PCB.

You will have to pre designPCB on a napkin, learn to use a CAD package, do the layout and design and then have it made in China anyway ... buying the kit is exactly the same thing, only you saved a couple steps.

"In principle", the magic is not in the Power amp (we are talking SS here) which just needs to be a robust workhorse, capable of carrying your signal to a speaker, but in the preamp, EQ, etc.
.
A good speaker and cabinet is another full job by itself.

My needs/angle is thus; a friend of mine who is an incredible guitarist but absolutely terrible at not blowing up tube amps has given me a Bugera 1960 JMP clone that had bounced around every tech in town before landing on my desk. The OT was dead and it didn't really seem worth the expense to replace it given the amp could be bought for not much more cash (yes yes, I know, disposable consumerism sucks). I gutted the amp with the intention of using it as a project chassis with a spare 4K OT I had kicking around but the layout of the chassis isn't terribly conducive to fitting any tube gubbins that doesn't have the same foot print as the PCB I pulled, thus I have claimed the PT and choke for another project. In essence what I have is a serviceable chassis for non-tube stuff and a nice wooden box for it to live in. My thought is to try some sort of SS or hybrid topology in the chassis/shell rather than throw it in the bin, which otherwise is it's final resting place as it stands. I'd then give it back to said friend and then hope it lasts longer than a few months before he calls to say it let the magic smoke out.

Preamp - whatever; it can be SS, tube or something in between like the KMG lnd150 stuff. It's be a simple affair no matter what. I was thinking of experimenting with the EQD speaker cranker circuit before the poweramp to simulate a bit of poweramp grit and perhaps soft diode clipping/limiting to protect the output device and to stop it sounding like ass.

Poweramp - whatever I can cobble together reasonably cheaply. I was looking at bridged TDA7294 (populated board and heatsink run around £20) after being inspired by some gut shots of the orange CR120. Motivation here is decent power into higher loads. I don't know what voltage the CR120 runs. The TDA7294 spec sheet states a max power of 150W with +/- 25V rails into 8 ohms (thus I estimate 0.6*150~90W into 16 ohms as a worst case scenario) which I guess will be about the ballpark? Orange claim it is good for 120W into 8 ohms but don't specify exactly what that means. Either way I'd guess I'm looking at a 18-0-18V (of maybe 15-0-15) transformer. I still need to think about the VA rating. Would 160VA be cutting it too fine or should I be punting at 225VA? Is a bit of sag going to be okay here? I'm considering adding soft limiting and the CR120 has a rep for being pretty damn loud anyway so maybe it'll never pull all that's available. Who knows? This all said I can pick up a new toroidal transformer for around £20-30 and much less than that used.


Anyway that's enough thread hijacking, sorry!

[J M Fahey]

In principle it´s doable.

A couple before-morning.coffee comments, and using this datasheet: https://www.st.com/resource/en/datasheet/cd00000017.pdf

1) consistent with "the amp Electronics is the easiest/cheapest part, the problem lies in hardware" thing, I often suggest getting a trashed amp at Goodwill or whatever, which solves cabinet/chassis/transformer.

So your getting a kit PCB and sourcing a proper SS PT is fine.

2) a good part of TDA7294 bad Karma is trusting its stupid specs, printed black on white by ST and others which should know better.

a) title is:

100V - 100W DMOS AUDIO AMPLIFIER

which it is NOT.

* 100V is absolutely maximum NO SIGNAL supply voltage, which must not even be approached, since mains Voltage fluctuates up/down 10% any day of the week.

* 100W is ludicrous.

Exact same size case (and so possible dissipation) is LM3886, rated 50/60W tops which is very reliable, it can NOT dissipate wasted heat coming from a 100W amplifier, NO WAY.
Marshall, Loudbox, SWR and a few others trusted that and shot themselves in the foot.

Reality check: that case is exactly same as "siamese twins" side by side joined 2 x TO220.
We all know 2 x TO220 is NOT reliable for a 100W RMS amplifier, minimum is 2 x TO3 metallic (2N3055 or better) , 2 x TO218/TO247 (TIP142/147) and so on.

2) they contradict themselves, next lines in datasheet are:

VERY HIGH OPERATING VOLTAGE RANGE (±40V)

HIGH OUTPUT POWER (UP TO 100W MUSIC POWER)

Then they get to the real numbers (which should have been in the headline, not buried in the body text):

Test Circuit VS = ±35V, RL = 8 r (ohm)

PO RMS Continuous Output Power d = 0.5%:

VS = ± 35V, RL = 8r 70W

VS = ± 27V, RL = 4r 70W

so presumably 140W into 8 ohm loads and some 80W into 16 ohm which is not bad at all.
You would need a 19+19 VAC transformer, which is not a standard value, but 18+18VAC is very popular thanks to (ahem) LM3886 and is close enough.

I read and trust RG´s dislike for 7294, but I guess it´s far more usable if "downgraded", actually "realistically rated", "as if it were an LM3886".

In fact, if you find a similar ">100W rated" 2 x bridged LM3886 board, by all means choose it instead.

You are not "losing power" by doing so, now you know headline 7294 specs are B.S. , even ST recognizing it once you read the full datasheet body.

As of preamps, feel free to experiment what you like

[tpaairman]

One other question while we’re at it. How do I know what voltage my voltage divider needs to be to bias the MOSFETS? If I have to use a load line, I understand the process (I think....) but I’m not sure what voltage value to shoot for.

[J M Fahey]

With due respect, you won´t be able to design, troubleshoot and build a MosFet amplifier for Christmas ... or any time soon.

For a shock of Reality, *ALL* Mosfet amps in the 80´s, when the "MosFet craze" attacked and *everybody*. from Marshall to Trace Elliot to HH to .... a dozen major manufacturers started using them, everybody cloned the datasheet Toshiba circuit, go figure.

If THEY didn´t design their own, why would you?

To boot, most of them abandoned MosFet outputs, and at least in 2 cases I know, Award Sessions and Hartke issued Bipolar based power modules to REPLACE MosFet powered ones, instead of repairing them.
That should say something

[R.G.]

I once found a GREAT deal on an amp. The local craigslist had a listing for a "Rogue" combo, a big suitcase with two 12's and a 100W solid state amp. The guy was happy to get $40.

The amp and speakers were trash of course. But it came with a solid cabinet, chassis and power supply for about the price of a decent power transformer. The chassis was reasonable for building either a tube or solid state amp into. I hacked a really decent preamp and power amp into it and gave it to my brother in law for Christmas.

One of the slap-you-in-the-face, thanks-Cartwright-I-needed-that things about engineering is that you can't optimize for highest function separate from what it costs, nor separate from how practical it is to build, nor separate from how hard it is to replair. And costs are always waiting in the wings. Experienced designers are the ones who always have tickers running in their heads for the cost, time to to supply, repair consequences, mechanical and thermal issues associated with every single part as they put them into the design.

[Justin Thomas]

One of the slap-you-in-the-face, thanks-Cartwright-I-needed-that things about engineering is that you can't optimize for highest function separate from what it costs, nor separate from how practical it is to build, nor separate from how hard it is to replair. And costs are always waiting in the wings. Experienced designers are the ones who always have tickers running in their heads for the cost, time to to supply, repair consequences, mechanical and thermal issues associated with every single part as they put them into the design.

And this is why I have such respect for you & Juan - you've been down the road & know the REAL "costs" of everything. Even if I do go against your advice & do something conpletely impractical... But I can't (none of us can, actually) say I wasn't warned when I get bogged down in futility!

Justin

[R.G.]

One other question while we’re at it. How do I know what voltage my voltage divider needs to be to bias the MOSFETS? If I have to use a load line, I understand the process (I think....) but I’m not sure what voltage value to shoot for.

I don't mean this snidely, but that's a little like asking "If apples are green, why is Arizona?" You're missing the function of the circuit.

The output MOSFETs form a complementary source follower. Their job is to follow the output voltage of the opamp, but with a huge amount of current gain. The 10K resistors are there just to limply hold the MOSFETs near the middle of the power supplies. The opamp is fully capable (or should be) of dragging both gates up and down as needed.

The real bias is those two LEDs. Enhancement MOSFETs (these are such) don't conduct at all until you pull their gate one "threshold voltage" higher than the source for N-FETs, lower than the source for P-FETs. You want the output to sit at 0V with 0V input, so the sources are sitting at 0V, and the LEDs and 10K resistors are there to pry the two gates apart by a few volts. The two-LED voltage is intended to be approximately one N-FET threshold voltage plus one P-FET threshold voltage.

It is likely that the LEDs are the wrong voltages. Real Class AB push pull output stages (of which, this is one) need a ticklish and precise bias voltage BETWEEN the two gates, set to be just enough to get the FETs to conduct perhaps 25ma or so with no signal. In practice, using two LEDs is not going to be usable. The most professional thing to do is to put a Vbe multiplier in there instead of the LEDs and set the bias by how much current flows through both FETs with zero signal. This is the ever-necessary bias into Class AB that you'd find in reading about solid state Class AB amps. Each MOSFET will likely need between 3 and 5 volts of bias, so you're looking at a total of 6-10V needed where the LEDs are.

Load lines are substantially zero help in understanding this, as you have to design the output stage and bias to fit the actual devices you put in - usually setting it with a trimmer.

[J M Fahey]

Agree and add, now "putting some numbers into it":
* Typical Red Led forward voltage is around 1.9V ;
* a typical "turn on" voltage I measure on the Fairchild, IR or ST IRFP250 Mosfets *I* use, which means that applies to those batches and none else, what I get next Month or tomorrow from a different supplier or, more important, what you will get ftom your supplier, can be any other value, there´s a lot of batch to batch variation so you can´t *assume* what you´ll find next, is often around 3.8V .

So to begin with, "one" Led per Fet is terribly underbiased, and two leds per gate is too much.
I have tried that, and found two things: first it´s uncomfortbly hot bias, 100 to 200mA at idle, may be even more, which is noticeably "warming", you feel it with your finger, but worst is that it does not thermally track, Leds do not care much about temperature, so output stage easily runs away, I have burnt my fingers into serum filled blisters, not kidding (that means way over 100C) trying to use Leds until I gave up.

Soution is, as RG suggested, a Vbe multiplier, carefully setting up bias current between 25 and 50mA, playing some loud music (or a continuous tone into a resistor) until heatsinks are at some realistic "live use" temperature, between warm and hot, and readjust again.
Then you can feel safe

And thermal coupling is paramount: I either clamp bias transistors to heatsinks with a small aluminum strip or drill 5/6mm (1/4") hole in heatsink and bury a TO92 transistor body there, with some thermal grease or at least epoxy them to heatsink.
Yes, thermal coupling needs to be that tight to ensure proper tracking.
A popular solution is to use a TO220 transistor there, not because of dissipation which is nil, but to ensure good thermal coupling. Of course, use thermal grease, mica, etc.

Problem cause is that we are using "switching" MosFets, which are built with a very high threshold (almost 4V) below which they areb basicalloy OFF and above which they jumpinto a lot of current with very little bias increase.
Excellent for a switch, a mess for an Audio device.

Of course, the real Audio ones are Lateral Mosfets, the old 2SK/2SJxxx famously made by Toshiba, etc. but they are hard to find, and expensive.