A high-voltage, but low-current amplifier for midranges and tweeters?
It isn't news to some, but I've discovered that the voltage swing required of a midrange or tweeter amplifier in an active system can actually be larger than that of the amp for the bass or subwoofer, despite the low RMS power requirement.
We discussed it here...
And I blogged about it here...
So, I've made the common mistake of putting low-powered amplifiers on my tweeters and mids, and larger ones on the bass and subwoofer. (GFA-535s with 48V rails on tweet and mid, GFA-555 with 85V rail on bass, and Crown K2 with 95V? rails on sub.)
Problem is, while these smaller amps have many times more power than needed for midrange and tweeter duty, they don't have enough voltage swing! Indeed, if I crank the system to eleven, the midrange amp is the first to clip.
The simple answer is to put GFA-555's on the midranges and tweeters, so they can swing the voltage needed.
But I don't need so much power! I just need the voltage. Also, many people say the GFA-535 sounds smoother than the big 555. Perhaps this is due to the 535 circuit fitting all on one circuit board, instead of a lot of spaghetti wiring to the output modules. Or maybe the use of fewer and smaller transistors has something to do with it. Maybe it's nothing, but I know I like the sound of the 535 on my tweeters and mids.
So, tell me if this is crazy... how about modifying a GFA-535, by installing a higher-voltage power supply, about 85V rails, but with not so much current? 200VA should be plenty. It would have a big filter bank to sail through those transient peaks, perhaps 60mF. The circuit in the 555 is basically the same, except for higher-voltage transistors and perhaps a few other details. I would modify it to match the circuit to the 555.
One concern is the increased idle current. Maybe it's not that much heat. But how about going down to just one pair of output transistors instead of two? Average power is extremely low. Those outputs would have more current being asked of them, and would thus experience some beta droop, causing more load on the drivers and upstream stages, but is this any concern when it's just transients? Am I missing anything? Is a single pair better for linearity in some way in this low-average-current situation?
It isn't news to some, but I've discovered that the voltage swing required of a midrange or tweeter amplifier in an active system can actually be larger than that of the amp for the bass or subwoofer, despite the low RMS power requirement.
We discussed it here...
And I blogged about it here...
So, I've made the common mistake of putting low-powered amplifiers on my tweeters and mids, and larger ones on the bass and subwoofer. (GFA-535s with 48V rails on tweet and mid, GFA-555 with 85V rail on bass, and Crown K2 with 95V? rails on sub.)
Problem is, while these smaller amps have many times more power than needed for midrange and tweeter duty, they don't have enough voltage swing! Indeed, if I crank the system to eleven, the midrange amp is the first to clip.
The simple answer is to put GFA-555's on the midranges and tweeters, so they can swing the voltage needed.
But I don't need so much power! I just need the voltage. Also, many people say the GFA-535 sounds smoother than the big 555. Perhaps this is due to the 535 circuit fitting all on one circuit board, instead of a lot of spaghetti wiring to the output modules. Or maybe the use of fewer and smaller transistors has something to do with it. Maybe it's nothing, but I know I like the sound of the 535 on my tweeters and mids.
So, tell me if this is crazy... how about modifying a GFA-535, by installing a higher-voltage power supply, about 85V rails, but with not so much current? 200VA should be plenty. It would have a big filter bank to sail through those transient peaks, perhaps 60mF. The circuit in the 555 is basically the same, except for higher-voltage transistors and perhaps a few other details. I would modify it to match the circuit to the 555.
One concern is the increased idle current. Maybe it's not that much heat. But how about going down to just one pair of output transistors instead of two? Average power is extremely low. Those outputs would have more current being asked of them, and would thus experience some beta droop, causing more load on the drivers and upstream stages, but is this any concern when it's just transients? Am I missing anything? Is a single pair better for linearity in some way in this low-average-current situation?
Last edited:
Simple Ohm's law tell us that a high-voltage peak pulls equally high current peak from a amplifier, so tweeter amplifier must supply the same peak current as a mid or bass amp. But, power supply and heatsink for a tweeter amp can be much smaller in size, because high frequency peaks are with short duration and average power is small.
Edit: one pair of output transistors may be enough.
Edit: one pair of output transistors may be enough.
Last edited:
I thought that a large filter-bank should do the trick, yes?
This is kind of how the NAD and Proton class-H amps worked AFAIK. They have huge capacitors on the high-voltage section of the amp, backed by a small-ish transformer.
This is kind of how the NAD and Proton class-H amps worked AFAIK. They have huge capacitors on the high-voltage section of the amp, backed by a small-ish transformer.
The large capacitor bank is to provide the current peaks so the transformer can be smaller...current still has to be available for short duration (as Sonce posted).
Not even necessarily a large filter bank, actually, as the big capacitance is mainly needed to extend power bandwidth to lower frequencies, particularly below rectifier recharge rate. For a tweeter amp, something normal-sized should be plenty, in fact you might consider a bunch of smaller caps in parallel (some of those close to the actual amplifier).
Having huge dynamic headroom on a tweeter amp is somewhat problematic as you want as little crossover distortion on that as possible, meaning that a bit more standing current won't hurt. Not that good for the power bill if your amp is plain AB. Seems like a good case for a Class G or H job, or maybe some other advanced topology.
This is an interesting topic. I think for a back-of-the-envelope calculation of power demands, regarding music as pink noise (equal power per log division) doesn't make too bad an approximation.
For your choices of 70, 300 Hz and 3 kHz, we get:
Sub: 1 octave and change, definitely <2 in practice
Bass: 2 octaves
Mids: 2.5-ish octaves
Highs: 2.5-ish octaves
With that in mind, there are two factors to balance out:
1. Speaker (+room) efficiency
2. Amplifier power
The standard recs do kinda make sense if you consider that high-frequency drivers like dome tweeters often exhibit high sensitivity, as in >=100 dB/W/m. Noise levels tend to become a more important concern compared to power then. By contrast, bass and sub drivers tend to be at 85 dB/W/m or even below, depending on size. Now 15 dB is 2^5 times = 32 times the power, which is a much bigger factor than half an octave more or less.
If sensitivities in your system are different, then of course your amplifier sizing will be different as well.
The surprise factor comes in when you consider that modern music tends to contain all kinds of noises that do not follow pink noise characteristics, like a lot of those emanating from a close-mic'd drum kit. It should never be an overly large surprise though, as the mix engineer would always be shooting for a sonic result that ultimately does sound natural, but crest factor may nonetheless remain higher.
BTW, the time constants for tape recording were chosen based on statistical considerations very similar to the above, back in the day. Which is why you can fairly easily tell a tape recording on more modern music (like 1980s and later) that peak-wise doesn't look like pink noise at all, especially when recording levels tended to be in the red quite often.
Perhaps the best way to put this whole discussion is that the crest factor of most music increases with frequency, so that plus respective sensitivities of drivers should be accounted for in amp selection.
So you still need the peak power to drive your entire loudspeaker to the same peak dB across the spectrum, but can "get away" with an amp that isn't massively overbuilt.
So you still need the peak power to drive your entire loudspeaker to the same peak dB across the spectrum, but can "get away" with an amp that isn't massively overbuilt.
Thanks. My drivers are of fairly similar efficiency.
Tweet: Hiquophon OW1 - 89db
Mid: Focal 5K4211 - 89db
Woof: Focal 10K6411 - 92db
Sub: Have not measured, but relatively high due to 14cuFt enclosure.
Empirical testing with the scope, by playing music and by watching the clip lights on the Adcoms, I get clipping in the midrange and treble well before the bass or sub amps.
Tweet: Hiquophon OW1 - 89db
Mid: Focal 5K4211 - 89db
Woof: Focal 10K6411 - 92db
Sub: Have not measured, but relatively high due to 14cuFt enclosure.
Empirical testing with the scope, by playing music and by watching the clip lights on the Adcoms, I get clipping in the midrange and treble well before the bass or sub amps.
I'd say, it's all in the relative time constants. How fast does your particular peak pull down the supply... i.e. how much energy required over time vs what R/C time constant you have.
It's been a while since I spent time observing this, but yeah, I remember a strong tendency for big main power supply rails to not droop much unless its a long, slow, low frequency transient. HF events, (>1khz) not so much.
If you're only have to handle the very fast ones, yes, you wouldn't need as large supply caps, and moving at least some of them closer to the amp/output stage would make them more effective. (less R with the C) I'd think doing everything possible to make the amp "higher current" would help.
So, you would still need a high peak power capability, but your average power requirements would be lower. You could save on filter caps and heat sink, but it's probably a deeper conversation on what other aspects would make a "good tweeter amp". (low R layout, parallel transistors, smaller emitter resistors?)
Agreed...
Good points about the power on the peaks.
My guess would be to stick with the original two pairs of output transistors. I don't suppose they'll droop much, and should have no trouble delivering the goods. The amp will idle a bit hotter than a stock gfa-535, due to the higher rails, but it will also not get much hotter than that due to low rms power.
I think I'm going to go ahead with this! For the power supply, Hammond makes a torroidal 300VA 60-0-60 transformer. For filters, maybe 4x1000uF, mounted right near the rail fuses. There are 47uF rail caps on the board too.
My guess would be to stick with the original two pairs of output transistors. I don't suppose they'll droop much, and should have no trouble delivering the goods. The amp will idle a bit hotter than a stock gfa-535, due to the higher rails, but it will also not get much hotter than that due to low rms power.
I think I'm going to go ahead with this! For the power supply, Hammond makes a torroidal 300VA 60-0-60 transformer. For filters, maybe 4x1000uF, mounted right near the rail fuses. There are 47uF rail caps on the board too.
But when the peaks are delivered to the tweeters the OS cant borrow in the low average currents periods to compensate for the peak currents...
I mean when those peak occur the OS will require the same number of power transistors as the bass channel to produce this current at a same linearity, moreover highs mandate a better linearity than at low freqencies.
So high peak power is not enough....
Wahab, isn't that what the power supply capacitors do? Charging up between peaks when demands are low?
AFAIK, the large numbers of power outputs in a 200WPC amplifier has more to do with SOA than actual current delivery. They need yo be able to put out that 200WPC for sustained periods. I just need the peaks.
Is there any reason to think that two pairs of 150W transistors would be insufficient to deliver the current peaks without a lot of beta sagging?
AFAIK, the large numbers of power outputs in a 200WPC amplifier has more to do with SOA than actual current delivery. They need yo be able to put out that 200WPC for sustained periods. I just need the peaks.
Is there any reason to think that two pairs of 150W transistors would be insufficient to deliver the current peaks without a lot of beta sagging?
Yeah, that's pretty much what I'm saying, that you have to have a high peak power capability, power meaning voltage x current of course. So you need the higher rails, and whatever amp factors that provide high current.
The big difference in the "tweeter amp" case would be the inherent lower average power that music tends to have in the higher octaves. So, you can probably get away with a smaller heat sink, and probably scale down some with the power supply caps, but you would still have to have whatever else you need to make that peak power when it's needed.
This may mean an ideal "tweeter amp" might not be very different from a regular high voltage and current amp.
But, I wouldn't conclude that "ideal" vs. "practical" wouldn't be different as well. I think it would be worth running some more numbers, measuring some more real world music, even trying a few amp examples... (hi current/voltage amp vs amp with very small filter caps or resistance inserted in rails/output?) to see where these kinds of things become performance issues.
I'd say it boils down to how much can you save in an amp design to deal with lower peak to average tweeter duty. Will it be a little, or a lot? Would be fun finding out... and maybe the design community will learn something new in the process.
Last edited:
^ You just start having to look at it from a systems approach. E.g. a tweeter or mid that's much, much more sensitive than the rest of the system is going to need far less power and far less gain. And that's before worrying about whether it's appropriately sized for the crest factor we expect for it.
Sure... I agree with you there. If you've got a 96dB tweeter, paired with an 84dB woofer in a 2way, you'd need -12dB of tweeter padding to even things out.
That's one of the benefits with active crossovers, you don't have to put a loss in there, you just lower the gain on your tweeter amp. And, the implication is you could get by with an amp that (theoretically) needs 12dB less output.
I was more speaking to the OP's idea that maybe you could/should build a specialized "tweeter amp" for the purpose. The more we talk about it, the more it looks like there might not be as many differences, but I at least think its worth looking into a bit more.
Phloodpants did observe that while the average power went down in the music he looked at, the peaks remained pretty high. So the crest factor in the upper octaves may go up more in relation to the lower ones. I remember looking at the intro to "Hell Freezes Over" version of "Hotel California" on a scope while playing my system, to see where/what was providing the largest peaks, and IIRC, it wasn't that huge drum, but surprisingly the "tick" of that Latin instrument I can't remember, the two sticks of wood that get hit together as a percussion instrument. Cymbal hits and such can have tremendous peak power...
So far, the likely key features seem to be: a smaller amp, higher rails than you might find on a smaller amp, (for higher peak to average capability) higher current capability than you might build in a smaller amp, and some lower needs on heatsink and bulk filter cap size.
That's one of the benefits with active crossovers, you don't have to put a loss in there, you just lower the gain on your tweeter amp. And, the implication is you could get by with an amp that (theoretically) needs 12dB less output.
I was more speaking to the OP's idea that maybe you could/should build a specialized "tweeter amp" for the purpose. The more we talk about it, the more it looks like there might not be as many differences, but I at least think its worth looking into a bit more.
Phloodpants did observe that while the average power went down in the music he looked at, the peaks remained pretty high. So the crest factor in the upper octaves may go up more in relation to the lower ones. I remember looking at the intro to "Hell Freezes Over" version of "Hotel California" on a scope while playing my system, to see where/what was providing the largest peaks, and IIRC, it wasn't that huge drum, but surprisingly the "tick" of that Latin instrument I can't remember, the two sticks of wood that get hit together as a percussion instrument. Cymbal hits and such can have tremendous peak power...
So far, the likely key features seem to be: a smaller amp, higher rails than you might find on a smaller amp, (for higher peak to average capability) higher current capability than you might build in a smaller amp, and some lower needs on heatsink and bulk filter cap size.
Last edited:
Jon--that's my takeaway, too. My point was more that in DIYaudio land, rather than necessarily buying a pre-packaged component set, we can build our system up from an end-goal perspective. So our amplifier choices can get tied into driver choices more carefully.
You are both right.
You start with your speaker and it's drivers.
What do they need?
Then work back towards an amplifier specification.
I'm still unclear on one point: My understanding is that the reason that big amplifiers have many output devices in parallel has more to do with SOA than actual current delivery. Is there any reason to think that having only two pairs of outputs would insufficient in some way? Will they experience beta droop when the voltage/current spikes on these transients?
have a look at the temperature de-rated SOA and see what current the device can pass when Vce is approximately half of one supply rail.
That combination of current and voltage is very common when driving a slightly reactive load.
A worse case for SOA is when Vce equals the supply rail. This requires high output into a severely reactive speaker load and is not so frequent, but can destroy badly designed amplifiers.
Now go back to the hFE vs Ic and see what typical base current will be needed to pass worst case transient current. This is where 2pairs, or even 3pairs, gives a big improvement in reducing base current demand. The reduced base current is effectively an increased load impedance that the driver transistor needs to drive. Very high base current when the output is feeding a severely reactive load can destroy the driver transistors and this then blows up the output stage.
That combination of current and voltage is very common when driving a slightly reactive load.
A worse case for SOA is when Vce equals the supply rail. This requires high output into a severely reactive speaker load and is not so frequent, but can destroy badly designed amplifiers.
Now go back to the hFE vs Ic and see what typical base current will be needed to pass worst case transient current. This is where 2pairs, or even 3pairs, gives a big improvement in reducing base current demand. The reduced base current is effectively an increased load impedance that the driver transistor needs to drive. Very high base current when the output is feeding a severely reactive load can destroy the driver transistors and this then blows up the output stage.
- Status
- Not open for further replies.