How do I go about calculating my fuse sizes for the DC rails?

I currently have 5A fast blow fuses thinking that it would protect the TIP41/42C since they can handle 6A max. But they've never blown before the output transistors do.

Obviously there must be a better way of sizing them?
Thanks

 

I need to think about that one. Remember the old axiom.

The transistor usually blows to protect the fuse. You got it wrong.

If a rail fuse blows, you have the other side of the rail across the speaker. Thats why you have to have a fuse in series with the speaker, or do some other heroics.

it's more important to get a fast blow fuse in series with the speaker leads. I Used an AGX fuse rated (I^2)*R
e.g. (3A^2) = 72 W for an amp capable of 100 W.

When I won my so called 100 W speakers in the 70's, the dome tweeters were fused at 1A, I upped that to 1-1/4A, so that's like 13 W. The tweeters have been replaced. They get scratchy after overheating,

 

When they have blown, the collector and emitter have been shorted together when tested with an ohm meter, and the 1A or 1.5A fuses have blown at the same time also. It has always been at least one NPN and one PNP that have blown. The first time they blew I believe three out of the four output transistors had shorted internally. I was under the impression that they should be able to handle more current than an amp or two, so maybe it is more heat related? Is it possible to have a case temperature of say 50 degrees C but the internal junction temperature is actually exceeding 150 degrees?

I haven't look at the circuit yet but I work on guitar and power amps. Most every amp I see has a slow blo fuse in it or the initial surge if not clamped can blow a fast blow fuse quickly.

 

I haven't look at the circuit yet but I work on guitar and power amps. Most every amp I see has a slow blo fuse in it or the initial surge if not clamped can blow a fast blow fuse quickly.

That's going to be the AC line fuse. The filter caps act like a short when the amp is first turned on.

Rail fuses are unusual when I did amplifier repair. Disconnecting the speakers when DC is detected and having a mains fuse is the norm. The consumer can't put a bigger one in.

If you fuse the rails, I'd make it the <= as the speaker fuse except the speaker fuse has to be a faster fuse.
You really can;t use a fuse to protect the transformer. It's the RMS value of the current that heats. So, you can have +-5A p-p which means, it could not possibly blow before the transistor.

So, you have +-24 V rails, so call it 40 V p-p, so about 14 V RMS. If you use 14^2/8, you get about 25W.
Using I^2*8=25, you get about 1-3/4 amps RMS. So, somewhere around there, and likely less.
Use the speaker load you have, not what you designed for. Your power supply figures in too.
For that I need the AC winding current. I'll bet, you did like I did and sized the transformer wrong.
From memory, you have to overrate the transformer current by 1/0.62 or about 1.6x.

Now, the amp should be able to drive at least 6 ohms for an 8 ohm speaker. The lower it can drive, the better, but it drives up the cost.

Amps bay employ VI limiters for protection, but some believe they color the sound.

So, the real idea is to look at all of the limits and ranges, but if you don't know where to look, that's a problem.

 

I'll bet, you did like I did and sized the transformer wrong.

I think part of the initial problem was that I designed the amp with something like a 18-0-18 transformer in mind, but bought a 24-0-24 instead.
Now instead of a safety margin, everything is right on the brink of meltdown

I successfully run if for over two hours with the input shorted @ +-25VDC per rail, set the Vbe multiplier to 2.4v with 3mV across each emitter resistor, resulting in 4.4mA idle current.

Next I applied a 3vp-p 1kHz sinewave with a 100ohm resistor across the output, resulting in 69mA across each emitter resistor. I left it run for just over an hour. TO-220 metal tab only rose from 22 to 23 deg C.

Next I applied a 3vp-p 1kHz sinewave with a 50ohm resistor across the output, resulting in 136mA across each emitter resistor. I left it run for another hour. TO-220 metal tab only rose from 23 to 24 deg C.

Finally, I put an 8ohm resistor across the output and dialed back the volume a bit to run approx 400mA through each emitter resistor.

I'm wondering where to go next? What is a normal bench test?
25% or 50% of max output for X amount time? Full power for X amount of time?

How far do I take it at this point with either a sinewave or a sweep as the input?

Thanks again!

 

I think part of the initial problem was that I designed the amp with something like a 18-0-18 transformer in mind, but bought a 24-0-24 instead.
Now instead of a safety margin, everything is right on the brink of meltdown

I successfully run if for over two hours with the input shorted @ +-25VDC per rail, set the Vbe multiplier to 2.4v with 3mV across each emitter resistor, resulting in 4.4mA idle current.

Next I applied a 3vp-p 1kHz sinewave with a 100ohm resistor across the output, resulting in 69mA across each emitter resistor. I left it run for just over an hour. TO-220 metal tab only rose from 22 to 23 deg C.

Next I applied a 3vp-p 1kHz sinewave with a 50ohm resistor across the output, resulting in 136mA across each emitter resistor. I left it run for another hour. TO-220 metal tab only rose from 23 to 24 deg C.

Finally, I put an 8ohm resistor across the output and dialed back the volume a bit to run approx 400mA through each emitter resistor.

I'm wondering where to go next? What is a normal bench test?
25% or 50% of max output for X amount time? Full power for X amount of time?

How far do I take it at this point with either a sinewave or a sweep as the input?

Thanks again!

Petkan:
The raisl' caps charge to peak value at iddling.
Ready made Buck regulators could be used to regulate the rails to the desired level (they cost around $2-3). The one in the negative rail will have to be re-arranged as Buck-Boost. This means lifting the common to become the negative output and grounding the output. In this mode the Buck will operate from the total Vin+Vout i.e. will have to be sized accordingly (and possibly the input caps replaced). Most Buck cannot exceed 40V. We need at least 60V. (assuming 32V Vin,, -24Vout the total becomes 56V)

 

Petkan:
Here is a 60V Buck example, capable of being set to any desired voltage and current limit.
It will dramatically reduce the risk of damage to the output stage and won't dissipate much heat.
http://www.ebay.ca/itm/DC-DC-LM2596...287513?hash=item2ef3c54599:g:660AAOSwvzRXx~zk

 

I think part of the initial problem was that I designed the amp with something like a 18-0-18 transformer in mind, but bought a 24-0-24 instead.
Now instead of a safety margin, everything is right on the brink of meltdown

I successfully run if for over two hours with the input shorted @ +-25VDC per rail, set the Vbe multiplier to 2.4v with 3mV across each emitter resistor, resulting in 4.4mA idle current.

The 18-0-18 will get you about +-25 V, 24-0-24 gets you about +-34V. So, you used you bench supply?
With a huge heatsink, I's hard to tell if stuff is working. Is your idle current enough? Not sure. It didn't blow up, this time.

What's the limiting factor? Do fuse the speaker output. Barely warm is a good thing. It's not getting warm either because the bias is too low, or the heat sink is too big. I'll bet the latter. No such thing as too big of a heat sink except it take up space.

There's a pre-conditioning period before measurements are made: https://www.ftc.gov/sites/default/f...roducts/001222traderegulationrulerelating.pdf

1hr, 1/8 rated power,1 kHz. Rated -- not max.

I remember stuff, like 200 w RMS/channel continuous, both channels driven, over the bandwidth of the AMP.
Now it's like 1 kHz for normal amps and different for subs.

So, it's probably too low: See http://www.tnt-audio.com/clinica/bias_e.html

My EXPERIENCE w/o a distortion analyzer, was to increase the bias, but keep an eye on the bias. You don't want it running away. 4 mA * 2 might be a bit low. The above link gives you a listening test to set the bias too.

I'd try 25 - 50 mA per paralleled transistor or 50 to 100 mA total. Obviously, if your heatsink gets smaller, the bias has to be smaller.

Look at clipping. Look at square wave rise-time. Look at frequency response.

This https://passlabs.com/press/leaving-class I thinks, sums it up.

 

I used a bench power supply because I can get 30V+- out of it. I assumed that would be close to what I'd get out of the 24-0-24 transformer, minus a few volts.

I'd try 25 - 50 mA per paralleled transistor or 50 to 100 mA total. Obviously, if your heatsink gets smaller, the bias has to be smaller.

Is this for a single NPN/PNP pair or with two pairs? I'm currently just testing with a single pair of NPN/PNP right now...

To set the bias, I monitored the sinewave that is being fed into the base of each driver transistor and increased the Vbe multiplier voltage until there wasn't any "visible" crossover distortion on the oscilloscope.

 

Yep, you can get about 30 V out of the 24 VAC.

The Nelson Pass (well known in the audio world) link effectively said more bias sounds better and essentially puts a $ limit on how much your willing to pay for.

The one thing, I don't know, is "What power output did you design for? and What is the current rating is your transformer? Did you buy 1 or 2 transformers? Is this going to be a stereo Amp?

With no real change in heat sink temperature, you don't know HOW WELL the Vbe multiplier is working.

 

It's a 250VA 120VAC primary dual secondary transformer. Single transformer to power two channels to make a stereo amp.

Still doing testing as we speak....

Swapped a few resistors around the Vbe multiplier and adjusted the idle bias current to 50mA in the output transistors with a single pair of NPN/PNP.

Applied 3vp-p 1kHz sinewave and run at 2.5 watts output power (1/8th of 20 Watts I think I should be able to get with a single output pair).
This was outputting a 4.5Vrms output across an 8 ohm resistive load. After one hour and fifteen minutes, the heatsink temperature rose from 35deg C to 50 deg C (It was already warm from previous testing, and I should have waited).

Next I adjusted the output to 9.05Vrms across an 8 ohm resistive load which should be 10 Watts. It's running right now and is up to 59 deg C on the heatsink after about 15 minutes....

 

Cool.

I'll munch on some of the info you've given.

While your at it, measure the output of the Vbe regulator as a function of temperature. You can have an input signal. Just use a handheld DVM so you can measure differentially.

At some point, you'd also like to have a warm heatsink and no signal and look at the output of the Vbe regulator, temperature and idle current.

Since, you actually trying to "track" Vbe of the output devices, You were setting to aproximately 2.2 V or so because you have 4 junctions in your output transistors. So, you can actually measure Vbe of all for of the transistors while in operation and add them to see how well the regulator tracks.

e.g.
temp Vbe regulator Vbe-Qa Vbe-Qb VbeQc Vbe-Qd Vbe(Qa+Qb+Qc+Qd) (Vbe-Vbe(sum)=K

The Vbe regulator should reduce by about 10 mV/deg C and I'd expect the difference to be close at all temperatures.

So, you designed a 20 W * 2 channels amplifier with a 240 VA power supply? If so, good.

You've come a long way with:
1) The Vbe regulator output depends on temperature.
2) The transistors your compensating and the temperature sensor have to share the same heat sink.
3) The need to Hfe match the output devices in parallel - (Your not quite there yet)
4) The Vbe regulator has to be fast response - your cap was too big.

and you started this project in March of 2015 or sooner.

 

I replaced the TIPs with FJP5200/FJP193 as they seem to look better on paper, and added a "speedup" capacitor on the driver section.

Today and I am currently running all 4 output transistors along with the Vbe multiplier transistor on the same heatsink.
I matched all the hfe values within "1", ex 45 & 46 for NPN, 60, 61 for PNP, etc.

I set the bias current to 50mA through each emitter resistor and I'm monitoring it to make sure it doesn't start to run away.

Just wondering what an acceptable deviation is though each emitter resistor?

With an 8ohm resistive load and 2.5 watt output, on the NPN side I'm sitting at 91mV / 90mV and on the PNP side they're at 91mV / 89mV. They've seemed to track within 2mV of each parallel pair. Is that matched close enough?

I will continue to retest at higher output levels if everything keeps working well.

I run through some tests up to about 65deg C with the TIPs last night and it was pretty cool to see the Vbe track down in voltage as the heatsink temperature rised! It was somewhere around 2mV or 3mV per deg C.

Thanks again!

I've also attached the current LTSpice project with the correct component names to match my other schematic, along with the added transistors.