Peak output current is
sqrt(2*Pmax/Rload)
Bias current is set to slightly above 50% of peak output current for your highest ClassA power.
If you target 50W into 4ohms as your highest ClassA power then Peak output current is sqrt(2*50W/4r) = 5Apk
Set bias to 2.6A
Supply rails assumed to be +-29Vdc
therefore dissipation is (29+29V)* 2.6A = 151W
That is pretty high and will need a very large heatsink, or a fanned heatsink.
If you stay with that ClassA target of 50W then the smallest transformer will be 300VA per channel and your biggest will be around 500VA per channel.
All of the above is due to selecting 4ohms as your ClassA target.
Try putting in the numbers for 3ohms and for 6ohms and see what happens.
sqrt(2*Pmax/Rload)
Bias current is set to slightly above 50% of peak output current for your highest ClassA power.
If you target 50W into 4ohms as your highest ClassA power then Peak output current is sqrt(2*50W/4r) = 5Apk
Set bias to 2.6A
Supply rails assumed to be +-29Vdc
therefore dissipation is (29+29V)* 2.6A = 151W
That is pretty high and will need a very large heatsink, or a fanned heatsink.
If you stay with that ClassA target of 50W then the smallest transformer will be 300VA per channel and your biggest will be around 500VA per channel.
All of the above is due to selecting 4ohms as your ClassA target.
Try putting in the numbers for 3ohms and for 6ohms and see what happens.
Last edited:
Keep in mind that as you have observed, your average power requirements are quite modest. It's all about heat and getting it away from the mosfets and heatsinks. The great part about rolling your own amp is that you can adjust bias and listen. You may not need to run all that class A power for your application.
When you are running really hot the long term reliability will suffer and that is a trade off you have to evaluate when setting your bias. There is no magic number for all circumstances or all ears. Adjust to suit your own taste while respecting the limitations of your heatsinks, mosfets and power supply.
When you are running really hot the long term reliability will suffer and that is a trade off you have to evaluate when setting your bias. There is no magic number for all circumstances or all ears. Adjust to suit your own taste while respecting the limitations of your heatsinks, mosfets and power supply.
Thanks, very much, Andrew.
Sorry to belabour the point but do I understand right ... that the amount of bias current sets the amount of power produced in Class A, at the selected load (before it moves into Class AB)?
You did the calcs at 2.6a bias ... so setting the bias current to 2a will yield 30w Class A power (into 4 ohms)?
Re. heatsinks - I am planning to use a Modushop 3RU "Pesante Dissipante" case for each monoblock. So the PCB for the P-MOSFETS will have 360cm^2 of finned heatsink on one side of the case and the PCB for the N-MOSFETS will have 360cm^2 of finned heatsink on the other side of the case.
Sure - that I understand!
So if I size the heatsinks for 50w into 4ohms ... I can start off with a bias current of 2a (or even less - to reduce heat output) and then see how it sounds ... and whether I can hear it ever going into Class AB? (In which case, I have to increase the bias.)
Thanks,
Andy
Last edited:
Thanks, very much, Andrew.
Sorry to belabour the point but do I understand right ... that the amount of bias current sets the amount of power produced in Class A, at the selected load (before it moves into Class AB)?
You did the calcs at 2.6a bias ... so setting the bias current to 2a will yield 30w Class A power (into 4 ohms)?
Re. heatsinks - I am planning to use a Modushop 3RU "Pesante Dissipante" case for each monoblock. So the PCB for the P-MOSFETS will have 360cm^2 of finned heatsink on one side of the case and the PCB for the N-MOSFETS will have 360cm^2 of finned heatsink on the other side of the case.
Sure - that I understand!
So if I size the heatsinks for 50w into 4ohms ... I can start off with a bias current of 2a (or even less - to reduce heat output) and then see how it sounds ... and whether I can hear it ever going into Class AB? (In which case, I have to increase the bias.)
Thanks,
Andy
Sizing for 50w 4ohms is not that easy on a one off build. I would buy as much heatsink as you can afford. That would allow some headroom if calculated values don't match up with real world experience. If you are going 3u high I would go 400mm deep at a minimum and consider going to 4u if you can afford it.
My personal opinion is that I wouldn't be able to hear the transisition to class ab at 30w. I'm running an Aleph J which is 25watts. With 95db 8 ohm speakers it will play really loud and at that volume I don't think I would notice any transition to class ab. I'll get to test that theory when I build my own mono F5t amps. I won't have much time to work on them until mid May due to other commitments but I have all the parts and am looking forward to getting them built. Of course YMMV and that's why it's great to be able to adjust bias, listen and adjust to suit yourself.
Ref. to post 4618, 4619,4620 and 4621.
I have built two vgs test-rigs, one for NPN and one for PNP. Every IRFP MOSFETs I have have been tested. The resitanse I used in both cases was 2.2 kohm. This gave me a result between 3V and 3.7V. FETs taken from the same batch was, not suprisingly, very close. But I have expected the difference between NPN and PNP to be not so close.
My be I should have measured with 56 ohm and the MOSFETs mounted to a heatsink. This would be closer the working condition in a F5 amplifier. The reason why I ask myself this is that in spite of a close matching(using the results from 2.2 kohm), again I measure very different bias over the source resistance: 270mV (on Q104) and
146 mV (on Q 103). Approx. the same results on Q4 and Q3. See my earlier post with a picture of the cviller PCB I use. I have also brought the results of the measurement I have done on the source resistors. They are very, very close matched.
Again, any good advices??
I have built two vgs test-rigs, one for NPN and one for PNP. Every IRFP MOSFETs I have have been tested. The resitanse I used in both cases was 2.2 kohm. This gave me a result between 3V and 3.7V. FETs taken from the same batch was, not suprisingly, very close. But I have expected the difference between NPN and PNP to be not so close.
My be I should have measured with 56 ohm and the MOSFETs mounted to a heatsink. This would be closer the working condition in a F5 amplifier. The reason why I ask myself this is that in spite of a close matching(using the results from 2.2 kohm), again I measure very different bias over the source resistance: 270mV (on Q104) and
146 mV (on Q 103). Approx. the same results on Q4 and Q3. See my earlier post with a picture of the cviller PCB I use. I have also brought the results of the measurement I have done on the source resistors. They are very, very close matched.
Again, any good advices??
Last edited:
I match power fets at high currents. Perhaps you should try using a higher current. Even something like 100ma, a current that would be more manageable for heat than 1 amp. Just take 2 that matched up using the low current method, mount them on a heatsink and test at higher currents. This should tell you if your original method is at fault.
One other thing comes to mind. Is it possible that you are getting HF oscillations in the amp? If so, this can cause strange DC readings. Something to check with a scope if you can.
I red the article about the F5 turbo v1 they just connected another pair of Power Mosfets on the Output. Is this all I have to do to cascode ?
Bfpca
I will follow your advice and do a test with higher current. About oscillation; 1000pF mounted acrosss the two 220 ohm in parallell(gain). NP gave also other advice to avoid this problems increasing R29/30 475 to 680, if I remember well. What is the "rule" here: Will a higher vgs increase the current trough the source resistors?
Eivind S
I will follow your advice and do a test with higher current. About oscillation; 1000pF mounted acrosss the two 220 ohm in parallell(gain). NP gave also other advice to avoid this problems increasing R29/30 475 to 680, if I remember well. What is the "rule" here: Will a higher vgs increase the current trough the source resistors?
Eivind S
As an experiment, I tried to change the source resitor to the FET I measured only 140mV. A resistanse of 0,4646 ohm was changed to 0,5855 ohm . This increased the bias from 140 mV to 190 mV. I am temted to try 0.68 ohm to come closer to 270 mV. Is this a recommended(and easy) way to get better bias balance on the FETSs?
Eivind S
Eivind S
advise asked: I currently own a nad c510 dac/pre-amp and an exposure 3010S2 integrated amp. I am considering building a DIY power amp to connect directly to my NAD C510 DAC-pre-amp. I am considering building a F5T V2 at 50 watts. I wonder whether people here know or can advice me on whether this amp will fit with my Nad C510 dac-pre-amp and my PMC twenty 24 loudspeakers (imp: 6,5-8 ohms, power: 30-200 watt, sensitivity 85-90 dB -->from different magazines, speakers have a transmission line of 3m).
Or are there other DIY amps that do fit my PMC 24 loudspeakers better, and why? the reason for a DIY power amp is that I would like to try a a class A design that is relatively simple to build as this is my first power amp. The 100 W version seems a bit too much from heat dissipation perspective......
Thanks in advance for your help!
Or are there other DIY amps that do fit my PMC 24 loudspeakers better, and why? the reason for a DIY power amp is that I would like to try a a class A design that is relatively simple to build as this is my first power amp. The 100 W version seems a bit too much from heat dissipation perspective......
Thanks in advance for your help!
Last edited:
One more question... I built the cascode now and I could do the measurement on the IRFP Power Transistors. My DMM shows around 0,625V on them which I think is okay. But the cascode gets hot anyway ( I can put my finger for 6-7sec on it) Is this normal ? Or do i have to change the resistors in the cascode ? I built it like described in the F5T V3 Manual from Nelson Pass.
Kind regards
Kind regards
Last edited:
If BC550/560 are the cascode you can expect them to always run warm, but they are usually safe depending on the temperature inside your case and the rail voltage/resistor combination.
The JFETs enter their lowest noise region below 12V drain-source voltage, hence the cascode needs to drop about 20V (from reference 32V rail). At 8mA drain current, this works out to 160mW - enough to make a TO-92 run really hot but still around 100C internally, and far from its destruction temperature.
Ideally you would like a TO-126 or TO-220 device in this position for better long-term reliability, but a tiny bit of metal will solve this issue once you tack it on with a cable tie.
The JFETs enter their lowest noise region below 12V drain-source voltage, hence the cascode needs to drop about 20V (from reference 32V rail). At 8mA drain current, this works out to 160mW - enough to make a TO-92 run really hot but still around 100C internally, and far from its destruction temperature.
Ideally you would like a TO-126 or TO-220 device in this position for better long-term reliability, but a tiny bit of metal will solve this issue once you tack it on with a cable tie.
Okay thanks, does anyone know btw how I connect a music source on the F5 ? because when I solder INGND and IN on a Chinch input Terminal and connect a music source on it, I just get this noise hum sound. But when I connect a function generator on IN and INGND and "play" a sinus sound everything sounds perfect.
ps :I found the problem, the RCA connectors are faulty... when i connect the wires directly it works
Last edited: