This one is for the sake of curiosity. Coming from tubes, I want to start experimenting with a bit more power. Solid state being the obvious solution. One thing I cant get my head around is why certain amps and designs are spec'd with a minimum load of 4 ohms. Is there a clear reason for doing so, or could most amps be used into 2 ohms with and adequate PSU and the proper cooling? Or is there more to it than just the dissipation of the power stage and PSU strain?
Assuming the PSU is able to handle the increased power requirements:
With linear output stages like AB, lowering the load impedance might cause exceeding the SOA rating. Althought linear mosfets have a slightly different failure mode than bipolars, both can fail due to high instant power dissipation. Think secondary breakdown.
With class D, there is no linear mode by definition, the mosfets could fail thermally due to high on state currents if not protected.
With linear output stages like AB, lowering the load impedance might cause exceeding the SOA rating. Althought linear mosfets have a slightly different failure mode than bipolars, both can fail due to high instant power dissipation. Think secondary breakdown.
With class D, there is no linear mode by definition, the mosfets could fail thermally due to high on state currents if not protected.
With bipolar power transistors there is a thing called "second breakdown". The transistor will self-destruct because the current gets concentrated onto a small part of the die.
With MOS-FET power transistors it is the dissipation. And, relatively speaking, the voltage loss over the devices is relatively tolerable to 8 ohms and 4 ohms, but grows in significance quickly.
Also, most parts are obtainable up to +/-100 V rails. Most parts can take 10 to 25 A of current. Everything more than that gets quickly more expensive.
Nominal impedances of 1 or 0.5 ohms are a big thing in autosound, because at least in Finland the categories are (or at least were) by "amplifier power into nominal 8 ohms", so with 1 ohm speakers you gain 8-fold and 0.5 ohm speakers 16-fold.
With MOS-FET power transistors it is the dissipation. And, relatively speaking, the voltage loss over the devices is relatively tolerable to 8 ohms and 4 ohms, but grows in significance quickly.
Also, most parts are obtainable up to +/-100 V rails. Most parts can take 10 to 25 A of current. Everything more than that gets quickly more expensive.
Nominal impedances of 1 or 0.5 ohms are a big thing in autosound, because at least in Finland the categories are (or at least were) by "amplifier power into nominal 8 ohms", so with 1 ohm speakers you gain 8-fold and 0.5 ohm speakers 16-fold.
If you're starting from scratch, you can design an amp to drive any power into any load impedance. 4 ohm minimum load just happens to be an industry standard.
If you want to modify a 4 ohm minimum design to drive even lower load impedances, you probably need to add more pairs of output transistors, and maybe modify the circuitry that drives them so it can supply more current. It's not trivial.
If you want to modify a 4 ohm minimum design to drive even lower load impedances, you probably need to add more pairs of output transistors, and maybe modify the circuitry that drives them so it can supply more current. It's not trivial.
There is also the issue of efficiency.
An amplifier designed to drive 1R and lower loads needs extremely low circuit impedances and resistances (read: expensive) in the power supply and output stage components, PCB tracks, connections and cables. This calls for some special construction techniques at medium power levels because just adding fat speaker cables doesn't do it, whilst efficiency drops from say 50% at 4R to 40% at 3R to 30% etc. as circuit losses become a significant, if not the dominant load themselves.
Unless the traditional constraints of 12V supplies and low rail voltages in, e.g.car audio applies, the performance of higher impedance loaded amps is almost always better in every way. 4R is about as low as I would like to cater for in a domestic environment for cost, cable appearance and performance reasons. There are indeed some crazy-low impedance speakers and drivers but as long as I don't use them, I won't have to worry 🙂
An amplifier designed to drive 1R and lower loads needs extremely low circuit impedances and resistances (read: expensive) in the power supply and output stage components, PCB tracks, connections and cables. This calls for some special construction techniques at medium power levels because just adding fat speaker cables doesn't do it, whilst efficiency drops from say 50% at 4R to 40% at 3R to 30% etc. as circuit losses become a significant, if not the dominant load themselves.
Unless the traditional constraints of 12V supplies and low rail voltages in, e.g.car audio applies, the performance of higher impedance loaded amps is almost always better in every way. 4R is about as low as I would like to cater for in a domestic environment for cost, cable appearance and performance reasons. There are indeed some crazy-low impedance speakers and drivers but as long as I don't use them, I won't have to worry 🙂
Things are getting clearer, thanks. See if I get this right. Let's say a PSU has got a low enough output impedance and doesn't buckle under the current demand, the only real limitation into a 2 ohm load is the dissipation/SOA of the output BJT's. Below 2 ohms, other factors come into play.
I wasn't aware of secondary breakdown and it makes a lot of sense. Am I correct to assume a rule of thumb to design for a maximum of half the dissipation stated in the data sheet? Proper cooling provided off course.
As an example, lets take Rod Elliot's project 68 (link). Simulations show a peak dissipation of 170W into two ohms with four pairs of output BJT's. Adding another four pairs lowers this to a 'safe' 87W. It seems adaptable that way to accommodate a heavier load. But I have to ask myself if it's worth the extra hassle. The 1000VA xformer and huge capacitor bank... 2 ohm loads will only be used a fraction of the time.
I wasn't aware of secondary breakdown and it makes a lot of sense. Am I correct to assume a rule of thumb to design for a maximum of half the dissipation stated in the data sheet? Proper cooling provided off course.
As an example, lets take Rod Elliot's project 68 (link). Simulations show a peak dissipation of 170W into two ohms with four pairs of output BJT's. Adding another four pairs lowers this to a 'safe' 87W. It seems adaptable that way to accommodate a heavier load. But I have to ask myself if it's worth the extra hassle. The 1000VA xformer and huge capacitor bank... 2 ohm loads will only be used a fraction of the time.
Hi,
It turns out 8 ohm is about the ideal load for a pair of power output transistors,
of any size and power rating, just because of the way they are built.
You can go to 2 pairs and 4 ohms, and 4 pairs and 2 ohms.
If you want a particular pair to work well into 4 ohms, the rails need dropping,
If you want a pair to work into 2 ohms, the rails need to drop so much its
pointless. As said the sweet spot for output from a pair is about 8 ohms.
FWIW my ancient Peavey MkIII bass amp has 4 output pairs and does 3 ohm.
The rail voltages are a little higher than is best for a pair into 8ohms.
FWIW I suggest you peruse the service manuals of classic bass amplifiers
for the schematics and layouts, you will find, well I did in the case of
Peavey's, the power amplifiers make no sense in any hifi sense.
Its a long time since I looked at it but my MkIII has a pretty dreadful
power amplifier in any hifi terms. However its sounds great for bass.
rgds, sreten.
It turns out 8 ohm is about the ideal load for a pair of power output transistors,
of any size and power rating, just because of the way they are built.
You can go to 2 pairs and 4 ohms, and 4 pairs and 2 ohms.
If you want a particular pair to work well into 4 ohms, the rails need dropping,
If you want a pair to work into 2 ohms, the rails need to drop so much its
pointless. As said the sweet spot for output from a pair is about 8 ohms.
FWIW my ancient Peavey MkIII bass amp has 4 output pairs and does 3 ohm.
The rail voltages are a little higher than is best for a pair into 8ohms.
FWIW I suggest you peruse the service manuals of classic bass amplifiers
for the schematics and layouts, you will find, well I did in the case of
Peavey's, the power amplifiers make no sense in any hifi sense.
Its a long time since I looked at it but my MkIII has a pretty dreadful
power amplifier in any hifi terms. However its sounds great for bass.
rgds, sreten.
Last edited:
No limitations ... a standard "slewmaster" output stage with 5 pairs BJT
will run 2-4R subwoofers .... provided the supplies are 50V or less.
For the common man , he has 8 ohm speakers ... and needs the headroom.
(higher rails). The OEM's will use 1 pair outputs and 60v+ rails with
a under-current "frail" power supply. This supply will collapse with
the load of paralleled speakers.
The "more to it" you refer to is the SOA (safe operating area -below)
of the semi's vs. power supply (SOA at X voltage for X time)
vs. cooling (the thermal de-rating of the SOA).
PS - any amp rated >4R would need a higher current / lower voltage supply
to run a 2R load. And most likely , greater cooling capacity. Do not
"obey" SOA , you get magic smoke or premature failure. 😱
OS
Attachments
Alpine's chief amplifier engineer once explained to me over beer and sushi that amplifiers don't really deliver power, they deliver current, and the power is a byproduct. The cost is really related to the output current more than the output power, in part because the losses inside tend as I^2*R (proportional to current squared) --> if the load impedance is halved, the losses increase 4X. I'm sure that's not all exact and it depends and exceptions blah blah blah, but it was a useful concept to put in my head.
Ian Finch is thus correct, let me put words in his mouth to say that we'd be better off with higher impedance speakers driven by high voltage amplifiers, and less current all through the chain…except that
- Efficiency of speakers has an inverse relation to impedance (not direct, it's not quite that simple, but there is an effect).
- That doesn't get maximum power out of transistors. To mimic sreten, why are Ford speakers 3.2 ohms DC resistance? Because someone looked at a spec sheet and found that from a 12V supply, whatever common transistors they were using in their radios gave maximum VxI at 3.2 ohms.
Nowadays, for maximum efficiency in electric cars, they should probably run 16 or 32 ohms or something, if cost was no object (a big if!).
Ian Finch is thus correct, let me put words in his mouth to say that we'd be better off with higher impedance speakers driven by high voltage amplifiers, and less current all through the chain…except that
- Efficiency of speakers has an inverse relation to impedance (not direct, it's not quite that simple, but there is an effect).
- That doesn't get maximum power out of transistors. To mimic sreten, why are Ford speakers 3.2 ohms DC resistance? Because someone looked at a spec sheet and found that from a 12V supply, whatever common transistors they were using in their radios gave maximum VxI at 3.2 ohms.
Nowadays, for maximum efficiency in electric cars, they should probably run 16 or 32 ohms or something, if cost was no object (a big if!).
I make amplifiers and speakers, so can answer knowing the limitations on both sides of the fence.
1) Amplifiers.
Earliest (for many decades) amplifiers were tube powered, and having output transformers which you could wind with any turns/impedance ratio, in theory they could accomodate any speaker impedance, so the limitation does not come from that side.
2) So the problem must lie on the speaker itself.
The nominal ohms parameter in a magnetodynamic speaker (the standard kind) is based on the voice coil DC resistance, which in due time depends on diameter and length of the coil wire.
There's 2 extra constraints:
a) the coil is moving back and forth at audio frequencies which means it must have a very low mass ... or inertia will kill its movement at mid and high frequencies.
b) air has incredibly high resistance to magnetic flux, so the gap must be as short as possible, typically 1.5 to 2 mm .
So the standard voice coil is just 2 layers, and that only because you must go "back and forth" with the wire to have continuity.
There's only 2 "exceptions" and not very far from standard: single layer in some tweeters or edgewound ribbon wire to have lowest possible mass and even narrower gaps , and 4 layer in some woofers where you only care about the lowest frequencies and can accept the weight penalty.
On the other side there's other limiting parameter , wire diameter.
Too thick, the coil is heavy and does not fit in the gap with enough safety margin to move without danger of scratching ; too thin and it becomes difficult to wind.
Remember scatter winding is not possible, each wire must lie perfectly parallel and in contact with the one besides it to occupy minimum space.
So all this gives you, in practice, a very limited range of wire diameters, which in due time means only a narrow range of VC impedance is practical .
Mid sized coils (from 3/4" to 2") are reasonably easy and fast to wind for 8 ohms , and no big deal to halve or double that.
Any higher or lower you are in trouble, so it's easy to see why 8 ohms became an unwritten standard.
Pure Darwinism 😉
3) SS amplifiers.
Well, to begin with, 8 ohms had been the standard for decades, and they normally don't have output transformers, so they had to be designed around that impedance ... or multiples/submultiples based on using more speakers to share the load.
FWIW I started designing and making my first SS amps in the early 70's, and way back then the standard most used transistor was 2N3055 in any of its variations.
They stand 70V and can pass nominal 15A , in practice no more than 8A ; with 10A being the absolute limit.
It's easy to get 20V RMS and adding more in parallel you can get higher current, so the unwritten standard gave you around 50W into 8 ohms and 100W into 4 ohms (2 x 8 ohms speakers in parallel).
Some MI/PA amps (Acoustic - Peavey - etc.) gave you 200W into 2 ohms and I made Bass amps : 400W into 1 ohm ... with 2 cabinets of 4 x 12" 8 ohms speakers in parallel each for 2 ohms per cabinet .... anyway we were still using individual 8 ohms speakers to achieve that.
Today's unwritten standard for such amps is around +/- 70V rails which gives you around 200W/8r or 300W/4r ... similar to the old standard only today 140/200V transistors are cheap and plentiful.
So in a nutshell: apparently arbitrary, 8 ohms speaker standard comes from a very practical reason: reasonably easy to produce voice coils.
1) Amplifiers.
Earliest (for many decades) amplifiers were tube powered, and having output transformers which you could wind with any turns/impedance ratio, in theory they could accomodate any speaker impedance, so the limitation does not come from that side.
2) So the problem must lie on the speaker itself.
The nominal ohms parameter in a magnetodynamic speaker (the standard kind) is based on the voice coil DC resistance, which in due time depends on diameter and length of the coil wire.
There's 2 extra constraints:
a) the coil is moving back and forth at audio frequencies which means it must have a very low mass ... or inertia will kill its movement at mid and high frequencies.
b) air has incredibly high resistance to magnetic flux, so the gap must be as short as possible, typically 1.5 to 2 mm .
So the standard voice coil is just 2 layers, and that only because you must go "back and forth" with the wire to have continuity.
There's only 2 "exceptions" and not very far from standard: single layer in some tweeters or edgewound ribbon wire to have lowest possible mass and even narrower gaps , and 4 layer in some woofers where you only care about the lowest frequencies and can accept the weight penalty.
On the other side there's other limiting parameter , wire diameter.
Too thick, the coil is heavy and does not fit in the gap with enough safety margin to move without danger of scratching ; too thin and it becomes difficult to wind.
Remember scatter winding is not possible, each wire must lie perfectly parallel and in contact with the one besides it to occupy minimum space.
So all this gives you, in practice, a very limited range of wire diameters, which in due time means only a narrow range of VC impedance is practical .
Mid sized coils (from 3/4" to 2") are reasonably easy and fast to wind for 8 ohms , and no big deal to halve or double that.
Any higher or lower you are in trouble, so it's easy to see why 8 ohms became an unwritten standard.
Pure Darwinism 😉
3) SS amplifiers.
Well, to begin with, 8 ohms had been the standard for decades, and they normally don't have output transformers, so they had to be designed around that impedance ... or multiples/submultiples based on using more speakers to share the load.
FWIW I started designing and making my first SS amps in the early 70's, and way back then the standard most used transistor was 2N3055 in any of its variations.
They stand 70V and can pass nominal 15A , in practice no more than 8A ; with 10A being the absolute limit.
It's easy to get 20V RMS and adding more in parallel you can get higher current, so the unwritten standard gave you around 50W into 8 ohms and 100W into 4 ohms (2 x 8 ohms speakers in parallel).
Some MI/PA amps (Acoustic - Peavey - etc.) gave you 200W into 2 ohms and I made Bass amps : 400W into 1 ohm ... with 2 cabinets of 4 x 12" 8 ohms speakers in parallel each for 2 ohms per cabinet .... anyway we were still using individual 8 ohms speakers to achieve that.
Today's unwritten standard for such amps is around +/- 70V rails which gives you around 200W/8r or 300W/4r ... similar to the old standard only today 140/200V transistors are cheap and plentiful.
So in a nutshell: apparently arbitrary, 8 ohms speaker standard comes from a very practical reason: reasonably easy to produce voice coils.
It is entirely possible for an amp to produce 15-20Wrms at 1R from 1 pair of TO-220 devices if the correct ones are used with the correct driving stage(s), and the correct power supply to match the output Z to 1R. The problem is when an amplifier is designed to drive 8R and is connected to 2R it will fail for the reasons explained above. Generally amplifiers are designed to drive typical speaker impedance that is available in market.
The MJ15015/16 pair are rated at 15A, but the gain drops above 4A.
To drive 8Ω at 200W needs 56.6V peak, about 7A (or 28A for 2Ω).
So, do you want to use 14 outputs per channel, 7 NPN, 7 PNP (plus rail switches)?
Various companies actually used only one pair of these in their 200W/8Ω amplifiers, and they would blow up their driver transistors when trying to drive 4Ω.
To drive 8Ω at 200W needs 56.6V peak, about 7A (or 28A for 2Ω).
So, do you want to use 14 outputs per channel, 7 NPN, 7 PNP (plus rail switches)?
Various companies actually used only one pair of these in their 200W/8Ω amplifiers, and they would blow up their driver transistors when trying to drive 4Ω.
Thanks to all! Very informative stuff.
Good stuff. But I don't understand the 300W into 4 ohm @ +/-70V rails. 500W (63,2Vp) into 4 ohm should be obtainable(?). What am I missing.
Yes, and since I was pretty much set on a 500W into 4 ohm class A/B topology for my first build, I now realize it's not worth the hassle to account for 2 ohm use.
For a solid state noob like me, the schematic of the MKIII is pretty intimidating. I still fail to see why one is better than the other...
Xformer for +/-70V already in house, so I'll keep to 4 ohm minimum.
Point taken
Hence the floating question marks around my head since I'm a tube guy 😀
Good stuff. But I don't understand the 300W into 4 ohm @ +/-70V rails. 500W (63,2Vp) into 4 ohm should be obtainable(?). What am I missing.
Different designs.
Instrument amplifiers usually have very light PSUs for cost, safety and performance reasons, so you get a lot less power output than the supply voltages might suggest. As load impedance drops say by half, the power increase is nowhere near the doubling you might be expecting if a very stiff power supply were used with a generous output stage.
Use a low current, E-I transformer power supply as in a guitar amplifier and you'll soon see JMF's rule-of-thumb figures play out.
Instrument amplifiers usually have very light PSUs for cost, safety and performance reasons, so you get a lot less power output than the supply voltages might suggest. As load impedance drops say by half, the power increase is nowhere near the doubling you might be expecting if a very stiff power supply were used with a generous output stage.
Use a low current, E-I transformer power supply as in a guitar amplifier and you'll soon see JMF's rule-of-thumb figures play out.
"What am I missing. "
Power supply regulation.
The Adcom GFA555 has a 700VA toroid and runs at about ±77V, yet is only rated at 325W/4Ω, and 450W2Ω.
The Leach LSR&D101 has a 1KVA E&I type transformer and runs at about ±63V, yet delivers 300W/4Ω, and 450W/2Ω.
Power supply regulation.
The Adcom GFA555 has a 700VA toroid and runs at about ±77V, yet is only rated at 325W/4Ω, and 450W2Ω.
The Leach LSR&D101 has a 1KVA E&I type transformer and runs at about ±63V, yet delivers 300W/4Ω, and 450W/2Ω.
Most Tube/Valve Power Amplifiers have taps for driving 4ohms and 8ohms speakers.
Is there any reason for not connecting a 2ohms speaker to these taps?
Sometimes the 4 Ohm spec may be to meet temperature specs for UL, CSA etc. It can do the power but might get too hot so it isn't part of the official specs.
for what it's worth the Crown Macro-Tech 10000 monoblok was spec'd as 'designed to drive very low impedance loads (<1.0 ohm). 8,395 watts @ 1.0%THD 700hz into a .5 ohm load.
You could strap multiples together internally to drive even lower impedances.
The plug is 3-phase 208v.
That's what paralled outputs gets you I guess. Use 0 guage flexible welding wire to speaker binding blocks.
So that's another item to consider: needing heavy speaker wires LOL
You could strap multiples together internally to drive even lower impedances.
The plug is 3-phase 208v.
That's what paralled outputs gets you I guess. Use 0 guage flexible welding wire to speaker binding blocks.
So that's another item to consider: needing heavy speaker wires LOL
- Status
- Not open for further replies.