Solid State Project!!

[angelothewolf]

Hi there, i build this project with a pair of 2sc3281 transistors, its working very well, but it has a little backgroung distortion, im wondering what improvments can i do to get rid, or decrease, that distortion... any ideas????

thnxs!!

[loudthud]

My guess would be that you have crossover distortion. What is the voltage across one of the 0.47 ohm resistors with no signal and no load connected? It needs to be atleast 0.005V. Install a 100 ohm trim pot where the 10 ohm resistor is, wired like a variable resistor. This will be your bias adjustment. Set the trim pot to zero ohms. Power the amp up with a light bulb limiter and slowly increase the resistance until you get about 0.005V across one of the 0.47 ohm resistors.


Edit: Please note that the amp has no short circuit protection. Any short across the speaker terminals will likely kill the output transistors. Make sure you have a fuse on the mains side of the power transformer like shown on page 4 of the instructions.

[R.G.]

What he said, +1.

This is a very bare-bones amplifier. Almost all of the niceties of modern amplifiers are missing.

If your distortion is a low-level grinding kind of sound like one of the speakers is rubbing a little, it's either speakers rubbing (and you have to eliminate that by subbing out speakers) or crossover distortion in all likelihood. You have few options with the way that's set up.

If it were mine, I'd replace the two 1N4006's and 10R resistor with Vbe multiplier, like this:

The quasi-complementary output stage needs about three diode drops, but getting exactly the right amount is tricky. Get it wrong and the thing runs away thermally. If you replace the diodes and resistor with a Vbe multiplier as shown, the voltage is Vbe * (R1+R2)/R2, and by adjusting R2, you can set the bias voltage very finely by making R2 a trimmer pot. The transistor is mounted on the output device heat sink for thermal compensation.

The lack of any protection is serious. If you short the output, the output devices will revert to sand by emitting a flash of light and a cloud of smoke in a few milliseconds.

At a quess, short across ONE of the 1N4006 diodes and see if the quality of the distortion remains much the same, but it gets bigger. If so, that's likely to be crossover distortion.

[angelothewolf]

My guess would be that you have crossover distortion. What is the voltage across one of the 0.47 ohm resistors with no signal and no load connected? It needs to be atleast 0.005V. Install a 100 ohm trim pot where the 10 ohm resistor is, wired like a variable resistor. This will be your bias adjustment. Set the trim pot to zero ohms. Power the amp up with a light bulb limiter and slowly increase the resistance until you get about 0.005V across one of the 0.47 ohm resistors.


Edit: Please note that the amp has no short circuit protection. Any short across the speaker terminals will likely kill the output transistors. Make sure you have a fuse on the mains side of the power transformer like shown on page 4 of the instructions.

Ok, thanks for the reply!! i will try that bias adjustment!

[angelothewolf]

ok, thanks! the thing is that this is an mini amp head project, so the enclosure space is very limited, but that Vbe multiplier looks good, and i think its not gonna take too much space! how do i get the right component values??

as i said before, the space is limited so all is very close to the transformer, and to make the things a little worse i build this circuit as preamp. it works very good with that amp, but it has lots of gain, so the noise i think it would be an issue. im using a different transformer (PSU) for the preamp, and im not quite sure how to earth all this stages so, any ideas would be very helpfull!!

i think the short circuit protection its a must, there is some simple way to do it?

thanks a lot!!



http://www.tonepad.com/getFile.asp?id=112

[J M Fahey]

1) RG's idea is excellent.
I'll suggest these values:


A Vbe multiplier emulates "X" times the BE diode of the thermal sensor transistor.
So if you use R1=R2~1K each you have "2 diodes" (same as now, and not enough) and if R1=2 x R2 you have "3 diodes" which is too much.
The beauty is that you cannot add "1/2 diode" or whatever but you can emulate that varying the resistors

So make R1=1K5 and use a 1K preset for R2.

Now, starting with R2 set to max=1K (= 2.5 diodes, cool, huh?) and slowly lowering its value, you'll start "increasing the amount of diodes" at will.
Slowly move it while monitoring current across the 0.47 ohms resistor (any of them), no signal applied and for better results, with no speaker connected.
As soon as you have, say, 5mV across it (follow Loudthud's method), you are fine.

Note: those videorockola guys make a nice effort diffusing auto construction among hobbyists, but they strip their amps "too much".

2) no matter what, you need a "Zobel network", a 0.1uF capacitor in series with a 10 ohms 1/2W resistor, from output to ground.

3) you *may* use an amp without short circuit protection, but only in a combo or powered cabinet, where it's hardwired straight to the speaker, no connectors or long cables.

Being a Music Instrument amp, live stage use means amps are transported often, and connected/disconnected.

A bad plug or cable (something very common) or some idiot who steps on the cable or some roadie who only plugs halfway ... and your amp dies.
Worse: it will die when you are about to play at a Club or something, so .... hardwire it as I said or choose another amp project.

[angelothewolf]

1) RG's idea is excellent.
I'll suggest these values:


A Vbe multiplier emulates "X" times the BE diode of the thermal sensor transistor.
So if you use R1=R2~1K each you have "2 diodes" (same as now, and not enough) and if R1=2 x R2 you have "3 diodes" which is too much.
The beauty is that you cannot add "1/2 diode" or whatever but you can emulate that varying the resistors

So make R1=1K5 and use a 1K preset for R2.

Now, starting with R2 set to max=1K (= 2.5 diodes, cool, huh?) and slowly lowering its value, you'll start "increasing the amount of diodes" at will.
Slowly move it while monitoring current across the 0.47 ohms resistor (any of them), no signal applied and for better results, with no speaker connected.
As soon as you have, say, 5mV across it (follow Loudthud's method), you are fine.

Note: those videorockola guys make a nice effort diffusing auto construction among hobbyists, but they strip their amps "too much".

2) no matter what, you need a "Zobel network", a 0.1uF capacitor in series with a 10 ohms 1/2W resistor, from output to ground.

3) you *may* use an amp without short circuit protection, but only in a combo or powered cabinet, where it's hardwired straight to the speaker, no connectors or long cables.

Being a Music Instrument amp, live stage use means amps are transported often, and connected/disconnected.

A bad plug or cable (something very common) or some idiot who steps on the cable or some roadie who only plugs halfway ... and your amp dies.
Worse: it will die when you are about to play at a Club or something, so .... hardwire it as I said or choose another amp project.

Brilliant!! would make this mods tomorrow and let you know, and maybe upload some pics,, thanks for the help!!

cheers

[R.G.]

If you're going to add some of these steps, try these:

1. Make R2 be a 1K, and the trimmer be a 500 ohm. That way if you sneeze when adjusting, the amp doesn't overheat so quickly. Also, hook up R2 (i.e. series 1K and 500R trimmer) with the body of the trimmer connected series with the 1K, then connect the wiper to one end of the trimmer. That way, if the wiper goes open (it happens!), the amp reverts to underbiased with crossover distortion, not Light Emitting Amplifier.
2. See this 100W guitar amplifier project, figure 2 for how you might add current limiting. The idea is to sense the current in the output device emitter resistors, and have that turn on a smaller transistor that steals the base current for that device.

[angelothewolf]

If you're going to add some of these steps, try these:

1. Make R2 be a 1K, and the trimmer be a 500 ohm. That way if you sneeze when adjusting, the amp doesn't overheat so quickly. Also, hook up R2 (i.e. series 1K and 500R trimmer) with the body of the trimmer connected series with the 1K, then connect the wiper to one end of the trimmer. That way, if the wiper goes open (it happens!), the amp reverts to underbiased with crossover distortion, not Light Emitting Amplifier.
2. See this 100W guitar amplifier project, figure 2 for how you might add current limiting. The idea is to sense the current in the output device emitter resistors, and have that turn on a smaller transistor that steals the base current for that device.

and what transistor can i use for the Vbe multiplier,, BC549???

[J M Fahey]

Any small TO92 case NPN , such as a BC547 or its family and a ton others.
That transistor must be epoxied or clamped to the heatsink so it's at the same temperature.

1. Make R2 be a 1K, and the trimmer be a 500 ohm.

A small correction: if you want to split R2 into a fixed and an adjustable part (fine with me), the total must still be 1K so use a 500 ohms pot and a 470 ohms resistor.

Otherwise the adjustment range won't be enough.

[angelothewolf]

Any small TO92 case NPN , such as a BC547 or its family and a ton others.
That transistor must be epoxied or clamped to the heatsink so it's at the same temperature.

A small correction: if you want to split R2 into a fixed and an adjustable part (fine with me), the total must still be 1K so use a 500 ohms pot and a 470 ohms resistor.

Otherwise the adjustment range won't be enough.

ok, i think i will use an 1k5 for R1 and 1k trim for R2 and see what happens,....
the zobel network, works as short circuit protection_?

[R.G.]

Any small TO92 case NPN , such as a BC547 or its family and a ton others.
That transistor must be epoxied or clamped to the heatsink so it's at the same temperature.

I always liked high-gain (100 or more) TO-220 or TO-126 devices because they have a mounting hole intended for clamping them to a heat sink. The leads are also more robust and less subject to breaking off. It's a smoke-and-flame class of failure if the wires to your Vbe multiplier break. Ask me how I know this.

I consider one part of good engineering to be visualizing the predictable failures and making them be as non-exciting as possible.
In the case of bias compensation for AB output stages, I like to put a string of diodes out on the heat sink itself that is maybe one Schottky more than the correct bias. That way, if the wires to the Vbe multiplier open, it goes overbiased and overheating, but it does not go fully into cross-conduction. It's a much slower failure.

A small correction: if you want to split R2 into a fixed and an adjustable part (fine with me), the total must still be 1K so use a 500 ohms pot and a 470 ohms resistor. Otherwise the adjustment range won't be enough.

I heavily recommend making the adjustments have just enough range to cover all the possible issues. One of the predictable failures is human misadjustment of the controls.

I just eyeballed the thing when I recommended values. Let's do the math.

The Vbe of the TIPs and 3280s are about 0.5 to 0.7V at low to moderate currents. The nominal target voltage is then two TIP Vbes, one 3280 Vbe, and some voltage across that 0.47R emitter resistor. Let's say it's nominally 3x0.6 +0.05V or 1.85V. Max might be 0.7V for each Vbe, 0.1V on the emitter resistor, for a total of 2.2V, and minimum might be 1.55V. So your Vbe multiplier needs to cover at least the range of 1.55V to 2.2V, perhaps just a touch more in case you get a weirdly low or high transistor in the bag.

The voltage across it is Vbe for the multiplier transistor times 1+R1/R2. We pick an R1 or an R2, then compute the result for the other. We probably ought to pick the R2 pot first then pick everything else from that. 500R is a good value for this kind of thing. However, *pots* have some tolerance too. Many small pots are 20% tolerance. So a nominally 500R pot can be 400-600 ohms.

We have Vbias = Vbe* (1+ R1/R2), so R2 = R1/ [(Vbias/Vbe)-1] which is suitably complex to scare the math-averse. I would have killed for a spreadsheet when I was an engineerling, and almost *did* kill for a four-function calculator - my first three undergrad years I did math on a slide rule.

Worse yet, we have to sub into this thing the min and max values of each part; the pot can be anywhere between 400 and 600 ohms, the bias voltage between 1.55 and 2.2V, the R1 value being +/-5% around 1K or 1500, and so on.

I did the spreadsheet, and it comes up with some interesting results. For a Vbe of 0.5V (which varies, too!) on the Vbe multiplier, if you make R1 be 1K and pick R2 as a 270R and the pot as a 500R nominal, you get a range of 1.074V to 2.33V of adjustment, and all the variations of the parts let you cover the range of 1.55 to 2.2V. If you pick R1 = 1500R, a fixed resistor of 390R lets you cover the range of 1.362 to 3.28V. Either of these works, and limits the possible values of bias voltage on the output stage. It's been a while since I did a worst-case matrix like this, so it's possible I missed something. But then very few setups will have a combination of all worst case tolerances in the worst-case directions, either.

I personally like the 1K/270 combination because the maximum voltage is not as much above the maximum 2.2V. But it works either way.

[J M Fahey]

I would have killed for a spreadsheet when I was an engineerling, and almost *did* kill for a four-function calculator - my first three undergrad years I did math on a slide rule.

Looks like we studied Engineering at about the same date (me, from 1969 to 1978), also used slide rules (still do, to amaze the younger generation), then also almost killed for a 4 function one and finaly had a friend buy a black Bowmar 901 in USA for around 150U$ (some 500U$ today).

Yet (surprise surprise) I had access to some kind of spreadsheet.

The Faculty computer was an ancient 4 cubic meter (40 cu ft) IBM1620

and if I wanted I could (and did) write my own Basic or Fortran program to solve the equations I wanted.

FWIW Isaac Asimov's Year 3000 spaceship pilots pull slide rules from their belts when they want to solve some astrogation problem

[R.G.]

That brings back some memories, OK. I learned to program in FORTRAN on the same model, a 1620.

The 1620 was nicknamed the CADET. That's an acronym, stands for Can't Add, Don't Even Try.

There is (was!) no arithmetic calculation unit inside the 1620. It did arithmetic by doing table lookup, much the way kids learn (used to learn??) arithmetic, by using two digits as indexes into a lookup table, returning with a digit and a carry/borrow. I was always mildly astonished by this.

The weight of punchcard decks in the computer room far outweighed the entire computer.

[angelothewolf]

I always liked high-gain (100 or more) TO-220 or TO-126 devices because they have a mounting hole intended for clamping them to a heat sink. The leads are also more robust and less subject to breaking off. It's a smoke-and-flame class of failure if the wires to your Vbe multiplier break. Ask me how I know this.

I consider one part of good engineering to be visualizing the predictable failures and making them be as non-exciting as possible.
In the case of bias compensation for AB output stages, I like to put a string of diodes out on the heat sink itself that is maybe one Schottky more than the correct bias. That way, if the wires to the Vbe multiplier open, it goes overbiased and overheating, but it does not go fully into cross-conduction. It's a much slower failure.



I heavily recommend making the adjustments have just enough range to cover all the possible issues. One of the predictable failures is human misadjustment of the controls.

I just eyeballed the thing when I recommended values. Let's do the math.

The Vbe of the TIPs and 3280s are about 0.5 to 0.7V at low to moderate currents. The nominal target voltage is then two TIP Vbes, one 3280 Vbe, and some voltage across that 0.47R emitter resistor. Let's say it's nominally 3x0.6 +0.05V or 1.85V. Max might be 0.7V for each Vbe, 0.1V on the emitter resistor, for a total of 2.2V, and minimum might be 1.55V. So your Vbe multiplier needs to cover at least the range of 1.55V to 2.2V, perhaps just a touch more in case you get a weirdly low or high transistor in the bag.

The voltage across it is Vbe for the multiplier transistor times 1+R1/R2. We pick an R1 or an R2, then compute the result for the other. We probably ought to pick the R2 pot first then pick everything else from that. 500R is a good value for this kind of thing. However, *pots* have some tolerance too. Many small pots are 20% tolerance. So a nominally 500R pot can be 400-600 ohms.

We have Vbias = Vbe* (1+ R1/R2), so R2 = R1/ [(Vbias/Vbe)-1] which is suitably complex to scare the math-averse. I would have killed for a spreadsheet when I was an engineerling, and almost *did* kill for a four-function calculator - my first three undergrad years I did math on a slide rule.

Worse yet, we have to sub into this thing the min and max values of each part; the pot can be anywhere between 400 and 600 ohms, the bias voltage between 1.55 and 2.2V, the R1 value being +/-5% around 1K or 1500, and so on.

I did the spreadsheet, and it comes up with some interesting results. For a Vbe of 0.5V (which varies, too!) on the Vbe multiplier, if you make R1 be 1K and pick R2 as a 270R and the pot as a 500R nominal, you get a range of 1.074V to 2.33V of adjustment, and all the variations of the parts let you cover the range of 1.55 to 2.2V. If you pick R1 = 1500R, a fixed resistor of 390R lets you cover the range of 1.362 to 3.28V. Either of these works, and limits the possible values of bias voltage on the output stage. It's been a while since I did a worst-case matrix like this, so it's possible I missed something. But then very few setups will have a combination of all worst case tolerances in the worst-case directions, either.

I personally like the 1K/270 combination because the maximum voltage is not as much above the maximum 2.2V. But it works either way.

Ok,, thanks for the knowledge, its helping a lot... the amp is now almost mounted, and suffering some tests. i uploaded some pics for you to see how its going.. its my first loud amp project. its everything very tight but for now is working very well. i thought the heat will be a problem but driving a 8ohms 4X12 speaker cabinet at full power it gets a little warm, will have a fan too.
I still have some doubts about the zobel network and the output protection. but for now exceeds expectations!