How much power for an initial experimentation

I have mostly built tube guitar amps, and one of the draws for solid state hi-fi has been that experimentation can be done a lot cheaper than with tubes, as tubes require a big power transformer no matter what the wattage, and the ever expensive output transformer. Not to mention the tubes themselves.

I would like to build my first working amp. This thing will be a breadboard affair, except for the output transistors which will be on some heatsinks that I scavenged out of an old guitar amp.

So how many watts should a first attempt be? In guitar, the answer is usually a lot less than you might think, but in hi fi, I suspect I want more than I think. I was thinking maybe 20 watts for a start. Using an inexpensive set of Yamaha speakers from a pawn shop (nx-s75's), would 20 watts be loud enough with a enough headroom for me to have something usable? Rock, country, and classical would be the main program material, with occasional metal.

Also, what driver and output transistors should I get? I want to get something that can be scaled up for more higher powered future attempts. The small signal transistors will likely be 3094/3096s since I have lots of 'em

Any thoughts?
 
A 20watt amp, that too on a breadboard will be of no use for future scalability. In fact, there is no upgradation possible in the solid state except may be to increase output drive by paralleling the power devices.

I know you think of 20Watts as you know the tubes but is too less in the SS terms except if you want to go for class A.

If you want to go for breadboard version of the ss, better go for chipamps although I wouldn't recommend them for serious listening.

Upgradation and improvements on one base amp are not feasible and you will have to build and build newer ones till you are satisfied or tired of.

Gajanan Phadte
 
By scaleability I just meant what power transistors should I get that would work ok for a low power amp, but also would work in reasonably higher powered amps. I just don't want to make more than a couple mouser orders here.

If I build this "experiment" on a breadboard, of course I can make upgrades and improvements.
 
@ djk: Couple questions about those transistors. Ok, maybe a few.

What is it about the 5551/5401 that makes it better than 3804/3906? All my books seem to use the 5551/5401 as you suggest, but a quick glance at the datasheet is not showing me what the key difference is.

About the mje340/350 for the VAS. Why such a high voltage/current component here? Or is there something else at play. Also, if I do a multi transistor VAS like an emitter follower driving a cascode, would I want the same part everywhere?

The MJ21195/96 looks like a pretty serious part! Looks like one could build a decently powerful amp with just a pair of those for output. I notice in a lot of pics of builds the output transistors are some sort of to-220 or similar packages in parallel. What would those typically be?

How about current sources and bias spreaders? Is it sufficient that they handle the voltages/currents or are there other criteria.
 

twest820

Member
2009-06-24 10:49 pm
In guitar, the answer is usually a lot less than you might think, but in hi fi, I suspect I want more than I think.
Not that much, really; there's about a 6dB difference in efficiency between pro and hi fi speakers and a 6dB difference between dipole peaks and closed box---call it 10dB between an open back 12" cab and a typical closed box hi fi speaker. Means 1mW nominal for 60dB listening position SPL with a typical home audio speaker build versus a couple hundred microwatts from a guitar amp being used at bedroom volumes.

Either way an LME49600 or LMH6321 provides plenty of crest factor headroom, though having a few watts can be nice at times depending on the neighbors. You can go discrete on the output devices if you want---look at Sanken STD03s for bias tracking---but it's extra cost and complexity for lower performance than a composite chip amp with a good op amp in the control loop---search for nested feedback in the chip amp forum. This approach works up to a few tens of watts, with discrete outputs making more sense above that---look at the LME49811 as a controller.

Of course, if your goal's to muck about with amp innards then ICs aren't so useful.
 
Yeah, I suspect a chip amp isn't what I'm after. I want to "muck with the innards" as you say!

My comment on guitar amp wattage requirements reflects the fact that for a lot of us, headroom is a bad thing with tube guitar amps, so low wattage is used so that the power section of the amp can be distorted at relatively low and non-painful volume levels.

For instance, the amp I currently gig with puts out maybe 9 watts cleanly, but will do 30 with everything up. This would be considered a 30 watt guitar amp. I doubt anyone would describe a hi fi amp as 30 watts if it was clipping as heavily as this guitar amp does at that wattage!

I was wondering if 20 watts into an average efficiency home stereo speaker would be enough headroom to enjoy listening to. I just want to have it loud enough so that I can discern any improvements I make as I go. The 100 watt amp that I will eventually build will be the result of these early lower wattage learning attempts.
 

Ian Finch

Member
Paid Member
2010-04-11 4:22 am
Coffs Harbour, NSW
If you are after scalability, as in increasing the power dramatically, it won't happen by just substituting parts with higher ratings and adjusting other components to suit higher voltages etc. This usually involves a complete redesign, new power supply, caps and all the expensive stuff and experimenting all over again to reestablish similar performance at the higher voltages and current throughout the design.

Within limits, simply tacking on more output devices and heavier drivers allows more current output capability and that does enable more power by driving lower impedance loads such as multi-speaker arrays.

Practically speaking though, a 100W amp. will be very little different in cost to a 25W one. There is almost no point trying to start low and scale that up. However, if you start at 100W and scale up to 400W, there will be quite a cost difference as the major parts and physical size start to be significant. This also has a decided advantage in using the same, hopefully rugged device types of the lower power units and that maintains the same cramped headroom, if that's desirable for guitar use.
 
Last edited:
What is it about the 5551/5401 that makes it better than 3804/3906? All my books seem to use the 5551/5401 as you suggest, but a quick glance at the datasheet is not showing me what the key difference is.
Look at the Vceo breakdown voltages. 2N3904/3906 is only 40V. 2N5551/5401 is much higher.
About the mje340/350 for the VAS. Why such a high voltage/current component here? Or is there something else at play. Also, if I do a multi transistor VAS like an emitter follower driving a cascode, would I want the same part everywhere?
The VAS transistor typically has to endure rail to rail voltages, so this means they must be able to survive without overvoltage breakdown. Also the driver stage(s) must handle rail to rail voltage as well. With hollow state you might get away with pushing the voltage limits of the device a bit but with solid state you will usually end up with the magic smoke being released, and it can't be put back in.:D Make sure you take into account Pd derating curves in the datasheet for the devices. Pd max is not really Pd max.:p
The MJ21195/96 looks like a pretty serious part! Looks like one could build a decently powerful amp with just a pair of those for output. I notice in a lot of pics of builds the output transistors are some sort of to-220 or similar packages in parallel. What would those typically be?

How about current sources and bias spreaders? Is it sufficient that they handle the voltages/currents or are there other criteria.

MJ21195/96 can produce ~100Wrms with class AB bias. Being new to solid state I would suggest you shoot for ~25W if you intend to use class AB bias. Larger amps could be problematic for a first discrete design. If you intend to start screwing with the 'inards' and cause a failure of the output stage (quite easy to do), a larger amp with more expensive output devices will set you back more $. 25W can be done with just one pair of TO-220 transistors. D44H11/D45H11 is a nice pair, cheap and available; I have used them to build an amp of that size before and can be easily breadboarded. If you use a CCS to bias the VAS it will also have to swing rail to rail so it should have the same adequate Vceo. 25Wrms @ 8R has a peak voltage of only 20V, so a +/-22V supply should be adequate taking into account rail voltage sag under load. This opens up many more suitable transistors that are available to use. If you don't already have a power transformer, for a cheap start you might look into the 12.6V -0- 12.6V @ 2.5A transformers available a Radio Shack for $10, at least they use to be $10.........:cheeky: One of these will give you about +/-18V after rectification which should do about 20W @ 8R.
 
Last edited:

twest820

Member
2009-06-24 10:49 pm
Yeah, I suspect a chip amp isn't what I'm after. I want to "muck with the innards" as you say!
Eh, 9 watts into a 100 dB SPL/W is pretty darn loud at home. Loud for a fair number of gigs too. When practicing I run my 22W 6L6 amp with a 45dB attenuator in the effects loop and it's still borderline too loud---driving the cab's internal speaker off the TL072 that buffers the effects send is a little quiet and suffers from tone suck, but that's with a 220 ohm output resistor in series with an 8 ohm speaker. Milliwatt peak powers like that pretty well mean all the distortion's happening in the preamp and the output stage's tonal contributions are from the output transformer and not the pentodes. Generally a desirable arrangement as a push-pull pentode pair has essentially the same clipping behavior as FETs. Same holds at higher output powers; the dirt generally comes from the preamp and output transformer saturation's used to fold the tone back to track the phon curves as the volume goes up.

If you have an oscilloscope just throw it on your home audio speakers and measure the drive levels you use and that'll give you the amp sizing. If not, look at the thread on how much voltage do your speakers need in the multiway forum. Statistically speaking 100W is overkill. You might also reconsider the relation between loudness and audibility; higher output power tends to correlate with degraded performance at typical listening levels. Though whether the degradation's due to a higher noise floor, distortion, or both depends on the amp topology. In particular, the more output device strings that are operated in parallel to mitigate Gm distortion the higher the quiescent current draw of the amp and the higher the PSRR required to keep the resulting ripple off the output. This is where a regulated IC control loop is a win. Even with the tight matching obtained from having devices on the same die the PSRR of the op amp or whatever isn't sufficient to deliver a clean output on its own and needs to be boosted by regulation.

This is why people spend so much time monkeying around with amplifier power supplies trying to get the sound they want. One can always build discrete regulators but discrete layouts are big enough it's quite difficult to hit the same level of performance even if one's willing to spend considerable time and/or money on device matching. At some point you look at an LM7815, LM7915, and an NE5532 and realize they all cost less than a buck and all you have to do is grab 'em and solder them to a board. Not to dissuade you from messing about with amp innards, but spending less time, money, and hassle to get better results does have a way of becoming compelling after a while.
 
@ CBS 240: I actually have a coupe of those radio shack transformers of which you speak, and this was actually the main reason I want to start at such low wattage. The only other transformer I have lying around is 300-0-300 volt beast I have earmarked for tube use. To much voltage and too little current capability for solid state use.
 
@test 820: I realize that there is a cost/simplicity angle to chip amps, but if I'm in this as a hobby, a chip amp circumvents a lot of the reasons the whole business appeals to me.

Keep in mind, I've been gigging for years with guitar amps I designed, so I don't need an easier route to see if the hobby interests me. I think if someone had approached me with the notion of building a chip amp as a first serious electronics project back in the day, I would have been all over it. I am also prepared to concede that I could probably get a better initial result with a chip amp, but I basically want to do discrete so I can have the benefit of the experience and have a better understanding as a result.
 

twest820

Member
2009-06-24 10:49 pm
This usually involves a complete redesign, new power supply, caps and all the expensive stuff and experimenting all over again to reestablish similar performance at the higher voltages and current throughout the design.
Actually, the main way power scaling is implemented in guitar amps is to throw a linear regulator on B+ and knock down the rail voltage by however much is desirable---100-200V is common (another routine solution is to put a 100W or rheostat in series with the driver, often with a line out that's used for direct recording to bypass micing the dimed out cab). A similarly boneheaded approach will work with a sand amp, it's just not very elegant and the need for a high current negative linear regulator makes it a hassle to build.

However, if one selects components for the high end of the voltage range it's often possible to reduce the power just by swapping in a lower voltage trafo. Some retweaking and optimization in a discrete design, yes, complete redesign and power supply changeout, no. With ICs it's not at all hard to build a design with essentially invariant performance across a fairly wide supply range (or at least it has been for the stuff I've built). +-2.5 to +-15V is one range that tends to work well, +-9 to oh +-23V or so is another---the upper end depends mostly on how one chooses to derate for the possiblity of mains overvoltage events blowing 35V caps. At higher voltages it's less attractive as the trafo VA requirements go up, making swapping iron more expensive than changing other things, but from oh, somewhere below the 100 to 200VA range the approach can be attractive.
 

twest820

Member
2009-06-24 10:49 pm
I am also prepared to concede that I could probably get a better initial result with a chip amp, but I basically want to do discrete so I can have the benefit of the experience and have a better understanding as a result.
Be careful with that line of thinking. Chip amps themselves aren't particularly great; it's probably best to think of them as cheap pair of power BJTs that happen to have a few conveniences, such as reduced gain bandwidth variation over current. What governs sand amplifier performance is the quality of the control loop, which mainly comes down to the loop's ability to handle its own deficiencies. The control theory's and the design fundamentals are pretty much the same either way. Much the same is true for tube amps, actually (though for hi fi basically what tubes mean is you get the privilege of spending gobs of money on transformers in order to get the amp's nonlinearity to limit on the tubes rather than the iron whilst simplifying supply design as the loading is class A rather than AB). A different arrangement of nonidealities and parasitics is kind of fun for a change of pace. But probably the main difference between hi fi amps and guitar amps is it's a lot harder to find hi fi amps with onboard DSP. ;)
 
I think what I meant by scaling got lost somewhere. I'm not talking about building a amp with scale-able output power. I simply meant I wanted to know which transistors to get that would work well for a decent range of power applications. I suspect that transistors are different than tubes in that you can get an over spec'd part and still use it effectively at much lower power. With tubes, using a 6l6 when a lower powered 6v6 is called for doesn't work as cut and dry.

In short, I just want to order a pile of parts that will serve me for quite a bit of experimentation. When I decide I want to build something permanent in a proper enclosure, I will order more purpose driven components.
 

Ian Finch

Member
Paid Member
2010-04-11 4:22 am
Coffs Harbour, NSW
Nope, not clear at all but now we get the idea. :idea:
Scaling by starting at the top voltage parts and reducing power supply size is usually OK with solid state devices. The parts suggested by djk are fine for that, though MJL21193/4 might be just as suitable for a lower range power. It's more or less down to cost constraints but >200V parts are probably unnecessary in home hi-fi.
 

djk

R.I.P
2001-02-04 4:23 am
USA
The parts I suggested are all dirt cheap and suitable for building any size amplifier up to ±100V, and as much as 2.5KW bridged.

A single pair of outputs on ±40V will be good for about 60W/8Ω (100W/4Ω), or 20W/4Ω on your 12-0-12 transformer.

The 2N5551/2N5401 from Avnet can run as little as $0.0143 in quantity.
The MJE340/350 from Avnet can run as little as $0.1739 in quantity.
The MJE15032/33 from Avnet can run as little as $0.6571 in quantity.
The MJ21195/96 from Avnet can run as little as $2.16 in quantity.

"though MJL21193/4 might be just as suitable for a lower range power."

The MJ21193/94 cost more, what's the point?

Check Mouser, Digi-Key, and Future Electronics for prices on small quantity. (Digi-Key) About $0.20 each for the 2N5551/5401 in quantity one piece, $0.56 for the MJE340/350, $1.20 for the MJE15032/33, $4.23 for the MJ21195/96, Mouser is a bit cheaper on all. Future Electronics is cheaper than either of the above, but only has the MJE15030/31 (limited to ±75V max).

You could save a few bucks with TO220 plastic outputs if you only wanted a 25W amplifier, but you wouldn't be able to scale it up. I built a great sounding 25W amp years ago with the MJE3055/2955 and MPSA06/A56 as drivers, cost about $2 for a stereo output stage. I used an 18-0-18 80VA transformer from Mouser, and later upgraded it to a 20-0-20 160VA toroid.
 
I mention this with the proviso that I know nothing about guitar amps per se, but in your position I would consider going somewhere close to the Naim route, since in the early days they almost spanned both areas. That means seriously over-rated output devices (current capability rather than sound quality ;) ), perhaps BUV20 or above and lots current through the drivers and VAS. There's also every chance that these heavy old devices will be a bit softer on clipping too.

Then your limitation on power - or voltage at least - will only be the rating of the reservoir caps. If you really want to see what power is needed then you could regulate the power supply and vary the voltage (though I couldn't recommend copying theirs).