P= Ipk^2 * R / 2 for a sinewave.
If you increase the number of output devices you can increase the total output bias.
Ipk can be the ClassA current = 2*Ib, or it can be the ClassAB current limit, which will bring in voltage clipping first (P= Vpk^2 /R / 2).
The standard F5 has for each device Pq ~ 25V * 1.3A = 33W.
If you take a parallel pair and run each device to the same Pq, then total output Pq is ~120W instead of ~65W.
You can use the extra output dissipation capability to increase the rail voltage, or increase the output bias, or to combine both, with an infinite range of Vrail & Ib, to approach your 120W of Pq
If you increase the number of output devices you can increase the total output bias.
Ipk can be the ClassA current = 2*Ib, or it can be the ClassAB current limit, which will bring in voltage clipping first (P= Vpk^2 /R / 2).
The standard F5 has for each device Pq ~ 25V * 1.3A = 33W.
If you take a parallel pair and run each device to the same Pq, then total output Pq is ~120W instead of ~65W.
You can use the extra output dissipation capability to increase the rail voltage, or increase the output bias, or to combine both, with an infinite range of Vrail & Ib, to approach your 120W of Pq
To chose the rail voltage, you have to consider:
Which load you intend to use.
How much class A and class AB power you want.
How many output pairs you intend to use.
Then, for a given rail voltage, the maximum bias is limited by dissipation into output devices.
For instance, i have chosen 33v rails for 50w classA and classAB into 8 ohms load, two pairs.
For higher classAB power, you can increase the rail voltage and lower the bias which means less classA power.
For higher classA power , decrease the rail voltage, this allowing higher bias.
Which load you intend to use.
How much class A and class AB power you want.
How many output pairs you intend to use.
Then, for a given rail voltage, the maximum bias is limited by dissipation into output devices.
For instance, i have chosen 33v rails for 50w classA and classAB into 8 ohms load, two pairs.
For higher classAB power, you can increase the rail voltage and lower the bias which means less classA power.
For higher classA power , decrease the rail voltage, this allowing higher bias.
Last edited:
Thanks. Finally clarity.
Now a more foolish question:
Where do most folks find the best outcome with a modified F5, more class A or more class AB?
The other funny observation is how incredibly well designed the stock F5 is, every time I push in a new direction, I see why NP made the F5 the way he did - he's a genius!
Now a more foolish question:
Where do most folks find the best outcome with a modified F5, more class A or more class AB?
The other funny observation is how incredibly well designed the stock F5 is, every time I push in a new direction, I see why NP made the F5 the way he did - he's a genius!
I have made my own personal choise
making it a 40watt, with maybe "just" 20watt classA
but its "experimental", and I may still end up with 40watt classA
I still have to look at the cost of electricity
how will I use it
for how many hours each day, etc
that will have influence on how I do it
2. Yes and Yes and No. First off, Ciss drops further with higher voltage (ie Vds) and becomes almost flat which leads to lower distortion. From memory I think they (IR devices) perform their best around 40V to 50V Vds. However, the answer to your question will be different for everyone. It will depend on how loud you listen to music, the efficiency of your speakers, and the nominal and reactive impedance of your speakers.
If you have low impedance speakers then maybe 30V and 1.8A is a good compromise. It is hard to give a definitive answer.
Let your ear be the judge. If you have two transformers you can try high voltage low bias vs low voltage high bias.
3. Low Noise devices are nice however since the amp is low gain the signal coming into the amp is a lot higher than most other amps so you may not notice a big difference.
Last edited:
Also, P = U x U /R and current flows through load only and only as a consequence of voltage, meaning that if you want higher current through the same load you have to supply higher voltage first.
I tried multiple outputs tonight. I could bias to 2.3A at 36V. The heat-sink was a little too hot (30C over room temp).
I can see how a Cascoded F5 would be interesting but I may head back to the Balanced version.
Juma,
Can you show an example? I'm not sure how one calculates U.
I can see how a Cascoded F5 would be interesting but I may head back to the Balanced version.
Juma,
Can you show an example? I'm not sure how one calculates U.
That does not seem possible using your heatsink.
Your dissipating 165W.
There is no way you would only be at 30 degrees above ambient unless you have very poor thermal contact between your mosfets and the heatsink.
I know because I have four of those heatsinks and did quite extensive tests for Patrick. I found that at 130W the heatsink is bloody hot (just acceptable).
Your dissipating 165W.
There is no way you would only be at 30 degrees above ambient unless you have very poor thermal contact between your mosfets and the heatsink.
I know because I have four of those heatsinks and did quite extensive tests for Patrick. I found that at 130W the heatsink is bloody hot (just acceptable).
Last edited:
V=IR
IF R = 8 Ohms and you want the maximum Class A Power using a bias of 2A (=4A class A for push pull)
Then V=4x8 =32V
IF R=4
Then V=4x4 =16V
That does not mean you have to use those values. Some people bias higher knowing that their speakers drop to 3Ohms even though the nominal impedance might be 8 Ohms (espescially important for single ended amps). Other people might bias higher because it sounds better or they might bias lower because it sounds better.
There are always exceptions to the rules.
How long after turn on, are you taking these temperature measurements?
What are you using to measure?
It does not seem right at all. Your only 21 degrees above ambient.
Nelson designed the F5 for 25w/ 8 ohms using 1 pair.
I t seems reasonable to me, when using two pairs, to limit the classA power to 50w.
In your example, you are reaching 64W
Max voltage swing is about 36v-4v=32v (4v = losses)
32*32/2*8=64W classAB
4*4*8=128W peak =64w ClassA
ClassA power= classAB one which is fine but a bit high.
It's possible but allows no error margin.
Personally, i prefer 33v rails.
Edit: crossed posts.
I t seems reasonable to me, when using two pairs, to limit the classA power to 50w.
In your example, you are reaching 64W
Max voltage swing is about 36v-4v=32v (4v = losses)
32*32/2*8=64W classAB
4*4*8=128W peak =64w ClassA
ClassA power= classAB one which is fine but a bit high.
It's possible but allows no error margin.
Personally, i prefer 33v rails.
Edit: crossed posts.
Last edited:
I'm measuring continuously. I have a digital thermometer, I put the tip in one of the thread holes in the heat sink. I touch the tip to the Mosfet cases for that measurement.
When I thought I was 30C over room temp, I may have mis-measured the room (temp changed as time went on).
I'll try again tomorrow and log all the temps vs time.
When I thought I was 30C over room temp, I may have mis-measured the room (temp changed as time went on).
I'll try again tomorrow and log all the temps vs time.
Maybe best to use an example.
Amp a) Push Pull biased at 2A
Amp b) Single ended biased at 4A
Lest assume we connect both up to an 8 Ohm speaker.
Amp a) Will produce (4x4)x8 =128W pk =64W avg
Amp b) Will produce (4x4)x8 =128W pk =64W avg
So far so good, except this speaker has impedance dips down to 2Ohms.
Lets assume both Amps have 32V rails with no losses.
At 2 Ohms Amp a) simply goes into class B operation and produces 16A (32/2=16) =16^2 x 2 = 512W pk =256W avg
Amp b) does not go into class B operation and will only output a maximum of 4A =4^2 x 2 = 32W pk = 16W avg
So Amp b) experiences severe current clipping while Amp a) just goes into class b operation and is still happy.
So the point here is when considering the bias point for a single ended amp don't just think about the nominal impedance but also consider what is the minimum impedance. The Aleph current source gets us out of trouble a bit here because it is capable of exceeding the bias current by a certain percentage depending on how it is setup. However Nelson has said it does not sound that great while operating in this mode. So best not to rely on it.
I should also point out that this was an extreme example for illustration purposes.
For my loudspeakers, I know I don't exceed 16V music peaks and I know the impedance doesn't fall below 3Ohms. So if I were to build a single ended Amp then I would bias around 5A. Even 4A would probably be enough knowing that it would only be on rare occasions that there would be a Voltage peak at the impedance dip, and this rare occasion could be tackled by the Aleph current source without significant audible effects.
For my loudspeakers, I know I don't exceed 16V music peaks and I know the impedance doesn't fall below 3Ohms. So if I were to build a single ended Amp then I would bias around 5A. Even 4A would probably be enough knowing that it would only be on rare occasions that there would be a Voltage peak at the impedance dip, and this rare occasion could be tackled by the Aleph current source without significant audible effects.
hey Melon Head, are you forgetting that the P-P(balance) amp experience a 4ohm speaker as being a 2ohm load ?
meaning that the "worst case" situation actually happens to the balanced amp with any "normal" 4ohm speaker
A singled ended balanced output amplifier has the same loading effect.
The difference is that someone designing a balanced amplifier does so for the intended loading.
Assembling a pair of unbalanced amplifiers in bridged topology is where mistakes could be made and often are.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.