| longthrow |
hi guys,
i was wondering which power amp scheme is better:
a power amp with twice a rail voltage or a bridged power amp with supply rails half of the other.
help, suggestions are welcome.... |
|
|
| Gordy |
Consider that for a bridged amp to work correctly the two amps must be exactly matched. Also consider the loop through which the current must flow, including the power supply components (because what flows out must flow back...).
If all other factors are equal, I would choose the higher voltage instead of the bridge configuration. |
|
|
| AndrewT |
Hi,
the answer "better or worse" depends very much on the SOA capability of the output stage and on the power output level.
I tend towards using +-2vs IF the output stage can meet the SOA demands required of it. |
|
|
| richie00boy |
| Bridging loads the PSU more nicely and keeps nasty currents out of ground, but it's more complicated. |
|
|
| longthrow |
| quote: | Originally posted by Gordy
Consider that for a bridged amp to work correctly the two amps must be exactly matched. Also consider the loop through which the current must flow, including the power supply components
|
hi gordy,
yup, all amps are equal, except for the rail supply voltage & amp scheme.. |
|
|
| longthrow |
same power is to be delivered into the load,same impedance..
also,over all system gain is the same.. |
|
|
| janneman |
| quote: | Originally posted by Gordy
Consider that for a bridged amp to work correctly the two amps must be exactly matched. [snip] |
No, that's no requirement. The system will work flawlessly even if the amps are mismatched several dB's. You may not be able to get the max output level, but quality doesn't sufer.
Jan Didden |
|
|
| djk |
"two amps must be exactly matched"
The Yamaha B2-X is a bridge amp. One half runs in class A at ±6V, the other in class AB at ±60V.
Audio (formerly Audio Engineering) published a DIY design to bridge a stereo amp with only one external channel. The 'hot' from the external channel became the 'ground' for the stereo amp. This worked fine at lower frequencies where the bass is common to both channels. |
|
|
| hienrich |
hi longthrow,
musta bai!!!:D
well, for a better speaker excurtion control, I would not go for a bridge amp since damping factor is cut to half when you go bridging....
if you have seen car amps that are usually driven in bridge config, subwoofers are going in and out excessively to the extremes because damping factors are so little to control excursion, this also means that
a bridged amp has a higher output impedance compared
to a single one....
of course power is multiplied by approximately 4.
;)
karon pako bai.......busy kaayo kinabuhi...heheheheheh
macweb |
|
|
| longthrow |
ok ra bay... :)
with the same power level,lets say 500Wrms into 8ohms,
that would be approx 90v per rail while if we use a bridged amp,rail voltage is cut to half (45v per rail)
considering the conditions (post#1 & 6)
well now,which is better in terms of performance, cost, efficiency, safety......etc..
from my point of view, bridged amp costs more since higher current is required; that would be more filter caps & more output devices....
btw,which is more safe for the power supply block; i'm planning use an smps for this amp.. |
|
|
| djk |
"I would not go for a bridge amp since damping factor is cut to half when you go bridging...."
Then how come the Crown amps have DF over 1000, even though they are a bridged design?
"subwoofers are going in and out excessively to the extremes because damping factors are so little to control excursion"
Not related in any way, shape, or form to the DF of those amplifiers.
The excessive cone motion is due to many things, none related to DF. The biggest problem is dynamic off-set due to the cap in the feedback loop of the amplifier becoming mis-charged under large signal conditions. The time constant of this cap and the lower arm feedback resistor is generally below 2hz. When overloaded this cap discharges out-of-phase with the input signal and sends rail-to-rail pulses to the woofer below 2hz (causing the motion you see).
A similar problem can show up with the transformer coupled amps (discussed by you in another thread). The asymetrical nature of audio signals will walk-the-core over to one side and when driven hard will cause the woofer to jerk back when all the energy in the transformer core is discharged (generally a time constant below 10hz).
As regards DF in general, it makes no difference if over about 20. One thing everyone forgets is the DC resistance of the speaker itself (around 6 ohms for an 8 ohm speaker) is in series with the motional resistance of the generator (back EMF of the speaker). So the 'real' DF is this 6 ohms in series with the wire resistsance, connector resistance (on the speaker, box, and amplifier), crossover choke (if used), and lastly, amplifier output impedance.
So does it matter if the amplifier output impedance is 0.001 ohms, or 0.1 ohms when it is in series with 6 ohms?
OF COURSE NOT!
Besides, the instant the amplifier clips, the 0.001 ohms becomes more like 2~3 ohms with the loss of feedback. |
|
|
| richie00boy |
| Well said djk. |
|
|
| Gordy |
| quote: | Originally posted by janneman
No, that's no requirement. The system will work flawlessly even if the amps are mismatched several dB's. You may not be able to get the max output level, but quality doesn't sufer.
Jan Didden |
So if amp 1 has X gain and 0.02% distortion and amp 2 has X + 1.5db gain and 0.2% distortion, then the output is still OK? I am surprised. My experiments with bridged amps produced a (subjectively) worse sound which I thought was down to a measured mismatch. It is interesting that it may be some other factor. Back to the drawing board. |
|
|
| djk |
Sandman, circa 1972 (Wireless World)
Consider two amplifiers, A1 and A2. A1 has 20dB of gain, R1=1K, R2=10K (A2 has the same resistors).
A1 runs off ±52V, A2 runs off ±6V.
A1 is a cheap and dirty class AB+B amplifier, A2 is class A.
If A1 had no distortion the summing input would be at 0V, and the output of A2 would be virtual ground. With 1.414V into A1 the output would be 28.3V (200W/4R).
Assume A1 is driving a ground referred load at 200W/4R with 1% distortion, there will be about 2.828V distortion at its output, and about 0.257V distortion at its summing input. Now hook up A2 and it will output 2.828V that is exactly the same as the distortion out of A1. With the same distortion voltage on both ends of the load the distortion will cancel.
A2 must be able to deliver the same current as A1, but only a fraction of the voltage swing is required. |
|
|
| longthrow |
great example djk...
thanks for the info...:smash: |
|
|
| Gordy |
Very interesting to learn of this. Thanks for the information djk.
However, it does seem excessively complicated that way with dissimilar amplifiers to design, power and drive. My feeling is that the original question of this thread related to using either a single big amp or two similar smaller amps. At least longthrow now has some pointers. My preference would be for the simpler solution of a single amp, and to concentrate on optimising it rather than introduce the complexity of the bridge amp scheme. |
|
|
| djk |
Depends on what you want, and what it's for.
No point in my mind to bridge unless running more than about ±95V, and then I would consider class G.
BSS made asymetrical bridge amps for PA with one half being class A, and the other half either G or H. |
|
|
| hienrich |
hi dgk,
thanks for the insights, specially on time constants.
well, I just based my veiws on actual experience with normal or bridge configurations.
yet so much interested to know more on time constants...
:D |
|
|
| hienrich |
longthrow,
my amp is having plus/minus 90vdc on the rails
yet BULKY and HEAVY enough to break my back!
nice thing your planning on SMPS...
for me bridging eats more space, well , depends on how you arrange and how you design PCB's...
bai, do you have ideas as to where to buy better extruded heatsinks.? |
|
|
| AndrewT |
Hi,
bridging requires twice as many voltage amp stages.
Bridging also requires four times as much smoothing capacitance.
The heatsink and output devices are very similar, although the same number of different devices will be chosen to suit either topology. |
|
|
| longthrow |
| quote: | Originally posted by hienrich
bai, do you have ideas as to where to buy better extruded heatsinks.? |
surplus centers here bay(5 jays;pier3 or side walks along colon area)
bay,for simplicity reasons i usually build single amps/ch..the one that i frequently use is bridgeable(i included a phase splitter),56V per rail..
this new project is for car audio,thats why i'll use smps...:D
| quote: | Originally posted by AndrewT
Hi,
bridging requires twice as many voltage amp stages.
Bridging also requires four times as much smoothing capacitance.
The heatsink and output devices are very similar, although the same number of different devices will be chosen to suit either topology. |
exactly,my option is either to use a higher transformer secondary voltage with less filter caps or a lower voltage secondary at higher current with more filter caps; which costs more...:cannotbe: |
|
|
| djk |
"bridging requires twice as many voltage amp stages."
Sort of, maybe.
(conceptual drawing)
The right-hand amplifier need only have a gain of -1, and it doesn't have to be high quality. IRF type FETs will work here too (otherwise you need a tripple EF for each device shown). Bias details omitted for clarity. See the Pass Zen v5 for ideas. The left-hand amplifier controls the sum, and thus the distortion of the whole amplifier (it must be high quality).
Forgot to label the inputs of A1, the main input is inverting, and the feedback comes to the non-inverting input.
|
|
|
| AndrewT |
Hi Djk,
there are some car audio chip amps that do something similar.
Maybe they could be designed to pass good quality sound, but my little experience with bridged car amps shows that the sound quality is dire. |
|
|
| jcx |
| quote: | Originally posted by djk
"bridging requires twice as many voltage amp stages."
Sort of, maybe...
...The right-hand amplifier need only have a gain of -1, and it doesn't have to be high quality. IRF type FETs will work here too (otherwise you need a tripple EF for each device shown). Bias details omitted for clarity. See the Pass Zen v5 for ideas. The left-hand amplifier controls the sum, and thus the distortion of the whole amplifier (it must be high quality).
Forgot to label the inputs of A1, the main input is inverting, and the feedback comes to the non-inverting input.
|
looks like you forgot a couple of resistors as well, as shown your circuit's A1 has no idea what the sum across the load is (relying on the inaccurate "it doesn't have to be high quality." inverting stage) so A1 doesn't correct for those inaccuracies
the whole fast accurate amp correcting a cheap inaccurate stage approach has fundmental problems - when the "cheap stage" is Class B you have high order crossover distortion that the fast accurate amp corrects by virtue of high loop gain error feedback
but the feedback is limited by the added delay of the cheap stage and the effectiveness of negative feedback at reducing high order distortion is limited - if poor bias thermal compensation in the Class B stage causes transient underbias of the cheap stage the resulting dead zone causes a huge spray of high order distortion products as the fast amp tries to push harder against the deadzone
and yes I am quite aware of "current dumping" and various "error correction" schemes - I beileve effort put into making low distortion output stages is much more quickly repaid than trying to fix up a higher distortion "cheap" output stage |
|
|
| deleveld |
"Bridging also requires four times as much smoothing capacitance."
What for? A bridge is less sensitive to power supply variations because the output is equally affected on both sides thus creating PSRR. This is something a single ended cant do. |
|
|
| djk |
| Crown used to make a bridge model the put out 7KW into 0R5 that only had a single 10,000uF cap. |
|
|
| djk |
People interested in bridge amps should read the two-part article in Wireless World circa 1981. It gives a very good overview and presents two practical designs.
What I was trying to show here (from memory) was the general concept, and that the design of the B amplifier need not be elaborate. While the Zen v5 would work, I probably wouldn’t choose it due to its low PSRR. Even so, the Zen v5 had only about 0.1% distortion mid-band with an AV=-10. With about 20dB of gain for feedback it might be OK (since this design only requires an AV=-1).
The original from the magazine had a single Vas transistor with a bias spreader and a bootstrap current source. It drove an EF output stage. Measurements showed that the A amplifier reduced the distortion in the sum output by an order of magnitude (20dB or better). It should be remembered that a pair of matched outputs will have less than 0.05% distortion (open loop), and don’t forget the feedback from the excess gain of the Vas stage. Current limiting (if desired) only need be done in the B half (unless you expect a short to ground). |
|
|
| AndrewT |
| quote: | Originally posted by AndrewT
bridging requires twice as many voltage amp stages.
Bridging also requires four times as much smoothing capacitance.
The heatsink and output devices are very similar, although the same number of different devices will be chosen to suit either topology. |
| quote: | Originally posted by deleveld
"Bridging also requires four times as much smoothing capacitance."
What for? A bridge is less sensitive to power supply variations because the output is equally affected on both sides thus creating PSRR. This is something a single ended cant do. |
Hi,
a push-pull ClassAB amplifier without bridging requires the PSU to supply half waves of the output signal alternately from either polarity rail.
If you accept that the main high pass filter is located at the amplifier input and that to ensure LF stability, then the two filters associated with this, namely the NFB DC blocking filter and the PSU/Load impedance filter, should be scaled to be some multiple lower in frequency. I have adopted the ratios of square root (2) as suggested by others and it works for me.
I set the input filter to ~90mS and the NFB to ~130mS and the PSU/load to 160mS to 200mS. For a 8ohm load this equates to +-20mF to +-25mF on the supply rails.
If the load impedance is reduced to 4ohms then the smoothing capacitance should be +-40mF to +-50mF if the same ratios of time constants are to be maintained.
Now let's look at the bridged amplifier.
Each amplifier in the pair is loaded by an effective impedance of half the actual load.
The supply rails supply half wave versions of the output signal.
This time the halfwaves are continuous rather than from alternate supply rails. (it looks like a full wave rectified version of the output signal).
The supply rails still supply the current to meet the demand from the effective load on each amplifier. But this time there is no rest period between half waves for the smoothing to recover. The transformer and rectifier work hard to deliver sufficent current during the small recovery period between adjacent halfwaves rather than alternate halfwaves and to compound this both rails must be charged simultaneously.
So I believe the smoothing capacitance needs to be doubled to take account of each amplifier feeding a half impedance effective load and doubled again because there are two amplifiers taking power from the supply. The effective load seen by the supply is one quarter of the actual load, due to two parallel amplifiers working in anti-phase but with half impedance loads on each.
If you separate the PSUs for each amplifier and treat each one as a PSU + amplifier + half load impedance it becomes more clear that each amp needs double the smoothing and we now have double the number of PSUs requiring a doubling of smoothing.
That in my book is four times as much smoothing for a bridged amplifier compared to a single amplifier.
It might even be worse than that because I have not made any allowance for the very much shortened recharge period. |
|
|
| djk |
That makes sense.
I always high-pass filter the inputs at a very high frequency, usually the Fc of the speaker system, and with a Q of at least 1. This does reduce the demands a bit for pole-staggering.
A couple of random notes:
The Crown 10K amp mentioned earlier has 3 phase AC, so it doesn't need much filtering.
The Crown 5K amp (vz5000) has a double integrator on the input and using feedback around the amp to act like an 18dB high pass filter at 40hz (it looks like a 6dB filter on the schematic until you consider the ground reference). They call this filter a 'Loudspeaker Off-set Integrator'. |
|
|
| Steve Eddy |
| quote: | Originally posted by AndrewT
Now let's look at the bridged amplifier.
Each amplifier in the pair is loaded by an effective impedance of half the actual load.
The supply rails supply half wave versions of the output signal.
This time the halfwaves are continuous rather than from alternate supply rails. (it looks like a full wave rectified version of the output signal).
The supply rails still supply the current to meet the demand from the effective load on each amplifier. But this time there is no rest period between half waves for the smoothing to recover. The transformer and rectifier work hard to deliver sufficent current during the small recovery period between adjacent halfwaves rather than alternate halfwaves and to compound this both rails must be charged simultaneously.
So I believe the smoothing capacitance needs to be doubled to take account of each amplifier feeding a half impedance effective load and doubled again because there are two amplifiers taking power from the supply. The effective load seen by the supply is one quarter of the actual load, due to two parallel amplifiers working in anti-phase but with half impedance loads on each. |
Mmmm. If we compare apples to apples (and I think that was longthrow's intent), I'm not sure there's any problem here and it ultimately ends up being a wash.
Let's assume that both the single and bridged amps have the same voltage gain and the same output power. For example, the single amp has a voltage gain of 26dB and each of the two amplifiers being bridged have a voltage gain of 20dB for a total of 26dB.
To deliver the same power, the single amp's rail voltages would have to be twice that of the rail voltages feeding each of the two bridged amplifiers.
So while the single amp is alternately drawing current from one side of the power supply or the other for each half of the signal waveform, the half of the power supply that's being drawn from has to supply ALL of the energy being delivered to the load.
In the bridged amp, although it draws equally from both sides of the power supply on both halves of the signal waveform, each half of the power supply is supplying only HALF of the energy being delivered to the load.
And if you use two transformers, each with the same VA but one whose voltage is half that of the other, and each coupled to the same amount of reservoir capacitance, I don't see that one transformer's working any harder than the other.
se |
|
|
| longthrow |
budget wise..:D :D :D
thanks to all!!! |
|
|
| janneman |
If you really want a single high power amp, you need LOTS of output devices in parallel for reliability, because of Safe Operation limits (I'm talking BJT's here, with Mosfet's it's different) with supplies above 50V or so.
With a bridge, with same power, you can get by with a bit more than half the supply voltage. Each devices sees only half the max Vce and SOA is much greater. You can get by with less than half the number of devices in parallel. So the bridge only needs TOTAL the same number of devices, or less.
Which means no extra cost in transformer, heatsinks, output devices.
Jan Didden |
|
|
| deleveld |
Yes janneman, for power it shouldnt make any major difference. But there are minor differences because the bridge output has voltage drops for two transistors and two emitter reststors instead of one for a single-ended output.
But consider the PSRR angle. A well designed bridge amp should have higher PSRR than a single ended. So any changing power supply lines should be less of a problem for a bridge than a single-ended output. So for a given amount of power supply signal bieng visible on the amp output, a bridge can have a smaller power supply capacity to reach the same quality.
Although, technically it has nothing to do with the bridge vs single-ended issue just PSRR which (should be) better with a well-designed bridge than with single ended. Of course, this is no guarantee that it actually will be...
Doug |
|
|
| janneman |
| quote: | Originally posted by deleveld
Yes janneman, for power it shouldnt make any major difference. But there are minor differences because the bridge output has voltage drops for two transistors and two emitter reststors instead of one for a single-ended output.
But consider the PSRR angle. A well designed bridge amp should have higher PSRR than a single ended. So any changing power supply lines should be less of a problem for a bridge than a single-ended output. So for a given amount of power supply signal bieng visible on the amp output, a bridge can have a smaller power supply capacity to reach the same quality.
Although, technically it has nothing to do with the bridge vs single-ended issue just PSRR which (should be) better with a well-designed bridge than with single ended. Of course, this is no guarantee that it actually will be...
Doug |
Fully agree. Also you are right about the additional Vce-sat and Re drops, that's why I said in my post ...a bit more than half the supply voltage ... ;)
Jan Didden |
|
|
| m2003br |
Friends of Audio,
I don´t resist to give some opinions in this topic.
I have to agree with all points disclosed by Steve and Jan.
About the need of more capacitance in the supply, it´s not a problem, because the elcos may be only half of operating voltage, compared to single ended amps. One elco 10.000u 35V cost less than one of 4700u 70V
But, more important than all, the two biggest advantage of bridged amp is:
a) doubled slew rate for the same single ended circuitry, and,
b) a naturally balanced configuration, with identical impedance in both polarity inputs.
hope my coments are valuable.
Best regards,
Marcos |
|
|
|
