Hi everyone. I am slowly progressing my first amp. This was supposed to be an Aleph 5 clone but I´m thinking to change it as a F5-T (diyaudiostore PCB's). I guess that it would be a better allrounder and I rather not build any boards that have been cloned without permission or so.
I have PSU's ready and measured now (thanks AndrewT). Rails are +/-42.4vdc without load, +/-38vdc under load, 545va per transformer. I don't need very high power amp, I guess that 50w@8ohm is enough, and ability to drive lower impedances also. Don't wanna struggle with temperature and fan will be included. This is going to be monoblock build.
Heatsinks are something like 16" x 14" x 3" (400mm x 350mm x 70mm)
I guess my question is; Is F5-T suitable allrounder for driving loads from 4-8ohm and is my power supply and heatsinks suitable for the job?
Still got some time to think, next up is some cad work and laser cutting parts to chassis, takes some time to get that part ready.
Thanks!
I have PSU's ready and measured now (thanks AndrewT). Rails are +/-42.4vdc without load, +/-38vdc under load, 545va per transformer. I don't need very high power amp, I guess that 50w@8ohm is enough, and ability to drive lower impedances also. Don't wanna struggle with temperature and fan will be included. This is going to be monoblock build.
Heatsinks are something like 16" x 14" x 3" (400mm x 350mm x 70mm)
I guess my question is; Is F5-T suitable allrounder for driving loads from 4-8ohm and is my power supply and heatsinks suitable for the job?
Still got some time to think, next up is some cad work and laser cutting parts to chassis, takes some time to get that part ready.
Thanks!
Attachments
Krell use 2pair of 250W BJT devices for a 50W ClassA amp and 4pair of 250W devices for a 100W ClassA amp.
Those indicate a factor of 20 to be applied. Total Pmax/ClassA power = 20
Take the standard F5 where one pair of 150W FET devices give 25W of ClassA.The factor turns out at 2*150/25 = 12
If we use the factor of 12, then a 75W ClassA amplifier would need 12*75 for Total Pmax
900W 3pair of 150W FET devices meets that requirement. But you could use 4pair if you prefer (this would allow youto run the stage hotter but still maintain an adequate FoS after de-rating for the increased temperature. this in turn allows you to use a slightly lower impedance speaker and stay in ClassA longer.
+-35Vdc in ClassAB hits 50W with the sag at full power.
If you had +-35Vdc after biasing into ClassA you don't have the voltage sag and so you get more ClassA power than does the ClassAB (see my tests of the Krell Klone from ~10years ago).
If you have +-38Vdc when biased into ClassA you will get a lot more. That extra 3V on the supply makes a big difference.
Your transformer is going to give a lot more than 50W into 8ohms so you have to design your build to match the likely output.
Those indicate a factor of 20 to be applied. Total Pmax/ClassA power = 20
Take the standard F5 where one pair of 150W FET devices give 25W of ClassA.The factor turns out at 2*150/25 = 12
If we use the factor of 12, then a 75W ClassA amplifier would need 12*75 for Total Pmax
900W 3pair of 150W FET devices meets that requirement. But you could use 4pair if you prefer (this would allow youto run the stage hotter but still maintain an adequate FoS after de-rating for the increased temperature. this in turn allows you to use a slightly lower impedance speaker and stay in ClassA longer.
+-35Vdc in ClassAB hits 50W with the sag at full power.
If you had +-35Vdc after biasing into ClassA you don't have the voltage sag and so you get more ClassA power than does the ClassAB (see my tests of the Krell Klone from ~10years ago).
If you have +-38Vdc when biased into ClassA you will get a lot more. That extra 3V on the supply makes a big difference.
Your transformer is going to give a lot more than 50W into 8ohms so you have to design your build to match the likely output.
Last edited:
Thanks again Andrew! Power won't hurt, but more than that I need ability to drive 4ohm loads. I guess my concern is about heat produced by these monsters, my heatsinks might barely handle F5-T V3 build with fans assisting airflow inside case?
4 ohms and ClassA won't work, unless you run exceptionally hot.
Your heatsink might manage 6ohms ClassA.
But I would suggest you aim for 8ohms ClassA and feel how hot it gets. Then decide if you can increase the bias to allow for 6ohms. Then after finding how that performs make the decision to adjust for 5ohms.
Your heatsink might manage 6ohms ClassA.
But I would suggest you aim for 8ohms ClassA and feel how hot it gets. Then decide if you can increase the bias to allow for 6ohms. Then after finding how that performs make the decision to adjust for 5ohms.
Juntuin, I would consider getting those huge heatsinks cut in half by a metal shop so that they are 1/2 the height. That way you would have 4 sinks and could build 2 more conventionally shaped F5V3. The sinks will work much more efficiently if cut in half and allow you to mount 2 F5T boards per side of the chassis as in a conventional V3 build. It looks to me that you would end up with 2- 7"h sinks 16"long and 3"deep with a very thick base. That should be good for F5V3 with 38volt supply and a total class A bias of 2amps for over 60watts classA into 8 ohms and 80watts class AB. Your transformers seem big enough to support that level of bias (160w)in a mono block. You would get 30w classA into 4 ohms but classAB rms power output would probably be close too 150watts.
Andrew, I thought the F5-T was happy with a 2 ohm load?
I can understand that there's an inverse relationship between the load presented and the heat given out (the lower the load, the bigger the heatsink required) ... but surely there's also a direct relationship between the amount of power required to be delivered and the heatsink size?
So, surely, a Class A amp which is only required to deliver 1w on average and 10w on peaks (into a 2 ohm load) will not need as big a heatsink as an amp which is delivering 50w on average?
The reason for my Qu is that I bought the F-5T boards and FETs so as to be able drive my 2ohm Maggie ribbons, well.
My CRO measures about 1V across the ribbon terminals of my 3-way active Maggies, when the volume is at my normal listening level:
* V = IR suggests a current of 0.5a
* I^2R produces 0.5w.
So less than a watt on average. So I'm thinking the 80mm high x 360mm long heatsinks which make up the sides of a Modushop "Pesante Dissipante" case should be more than sufficient?
Andy
Yes and so, Andrew? Do the "F5 and variants" produce less heat than pure Class A?
Surely my point still stands that a Class A amp which is running at 1w average ... is going to produce a whole less heat than one that is running at 50w average?
Andy
ClassA into 8ohms requires a particular output bias current.
The ClassA biased amplifier consumes a fixed amount of energy and produces a fixed amount of heat.
When an output is delivered some of that energy is diverted from the heatsink to the load. The power and energy stays the same.
let's look at some numbers.
A 25W into 8ohms ClassA amplifier will run on ~+-25Vdc supply rails and have a quiescent current of ~1.3A
The total dissipation is {25+25}*1.3A = 65W per channel.
If one delivers 1W average to the load, then the total dissipation remains at 65W with 64W going to the heatsink and 1W going to the load.
When the output increases to 20W then the heatsink dissipates 45W and the load dissipates 20W for a total of 65W.
All of the above assumes the amplifier stays in ClassA, i.e. the ClassA maximum current (~2.5Apk) is never exceeded.
The ClassA biased amplifier consumes a fixed amount of energy and produces a fixed amount of heat.
When an output is delivered some of that energy is diverted from the heatsink to the load. The power and energy stays the same.
let's look at some numbers.
A 25W into 8ohms ClassA amplifier will run on ~+-25Vdc supply rails and have a quiescent current of ~1.3A
The total dissipation is {25+25}*1.3A = 65W per channel.
If one delivers 1W average to the load, then the total dissipation remains at 65W with 64W going to the heatsink and 1W going to the load.
When the output increases to 20W then the heatsink dissipates 45W and the load dissipates 20W for a total of 65W.
All of the above assumes the amplifier stays in ClassA, i.e. the ClassA maximum current (~2.5Apk) is never exceeded.
Aaah, OK, Andrew!
For a given output, the less the amp delivers into the load ... the more heat is produced! 
So - given that I can't see the peak power to my 2 ohm ribbons is ever going to exceed 10w - that suggests I should set the DC rails to be, say, +/-10v (so 10w into 8 ohms ... which is 40w into 2 ohms)?
Andy
Last edited:
My original post to Juntuin was intended to give him an option on using his heatsinks more effectively. As far as class A goes an F5 is a class A push pull amp and will go into class AB when the class A current is exceeded producing more power until it runs out of voltage headroom. Given the size and output current capabilities of an F5TV3 output stage the voltage limit is going to limit maximum power for any typical speaker load and supply voltage.
The F5 series of amps is very flexible and scalable in terms of power output. If you want to run class A all the way to the power rails with 4 ohm loads, then design the power supply and heatsinks accordingly.
You also have the option of designing for a given amount of class A output power with a higher class AB maximum power output. It's an equally valid choice for those who want the extra watts into low impedance and don't need or want output to be all classA power.
Andyr, you can certainly design with a +-10 volt supply and high bias current but I'm not sure if the performance is going to be what it could be with higher supply vvoltages. Also, what if you change speakers and need more voltage drive. I'm not sure how you measured the ribbon voltage but my experience is that peak values on tweeters can exceed that of what is on woofers in some source materiel. The average power is not that high but peak voltages are much higher than average.
The class A math Andrew has presented below is correct and is a fact of life you need to deal with when designing with a given set of heatsinks. Small heatsinks will limit the maximum class A power but you can adjust your power supply and bias current to get the most out of them for a given load. What that power will be? You would need to find out through experimentation.
The F5 series of amps is very flexible and scalable in terms of power output. If you want to run class A all the way to the power rails with 4 ohm loads, then design the power supply and heatsinks accordingly.
You also have the option of designing for a given amount of class A output power with a higher class AB maximum power output. It's an equally valid choice for those who want the extra watts into low impedance and don't need or want output to be all classA power.
Andyr, you can certainly design with a +-10 volt supply and high bias current but I'm not sure if the performance is going to be what it could be with higher supply vvoltages. Also, what if you change speakers and need more voltage drive. I'm not sure how you measured the ribbon voltage but my experience is that peak values on tweeters can exceed that of what is on woofers in some source materiel. The average power is not that high but peak voltages are much higher than average.
The class A math Andrew has presented below is correct and is a fact of life you need to deal with when designing with a given set of heatsinks. Small heatsinks will limit the maximum class A power but you can adjust your power supply and bias current to get the most out of them for a given load. What that power will be? You would need to find out through experimentation.
If you want a 40W into 2ohms ClassAB amplifier then you can use any bias you like and any supply rail voltage your heatsinks can cope with.
If you want 40W into 2ohms and stay in ClassA then you need to design the F5 as a ClassA amplifier for that load.
The peak current into a 2ohms load is given by sqrt(2*max power/Impedance) = sqrt(2*40/2)=sqrt(40) = 6.32Apk
The bias should be set to just over 50% of that, i.e. to ~ 3.2A
If you have chosen +-10Vdc for your supply, then the devices will see 10*3.2A = 32W each for a one pair output stage. This is almost the same as a standard F5 (25*1.3=32.5W)
Will +-10Vdc allow the amp to deliver as you expect?
I would be very tempted to look at a two pair output stage with that bias level and probably increase the supply rails to +-15Vdc giving 15V*1.6Abias = 24W per device.
If you want 40W into 2ohms and stay in ClassA then you need to design the F5 as a ClassA amplifier for that load.
The peak current into a 2ohms load is given by sqrt(2*max power/Impedance) = sqrt(2*40/2)=sqrt(40) = 6.32Apk
The bias should be set to just over 50% of that, i.e. to ~ 3.2A
If you have chosen +-10Vdc for your supply, then the devices will see 10*3.2A = 32W each for a one pair output stage. This is almost the same as a standard F5 (25*1.3=32.5W)
Will +-10Vdc allow the amp to deliver as you expect?
I would be very tempted to look at a two pair output stage with that bias level and probably increase the supply rails to +-15Vdc giving 15V*1.6Abias = 24W per device.
Thanks, guys. I'm planning on building an F5T v2 - which I understand has 2 output devices per channel.
It wasn't that I was actually thinking of using only +/- 10v rails ... that was simply an example. But I will have to ponder the value of the DC rails, given this new-to-me concept of the balance of heat generation shifting between heatsinks and load as the average power used by the load changes. But, yes - +/-15v at least!
Andy
the F5 has two output devices per channel.
The F5t usually have more output devices per channel.
increasing the supply rail voltage increases the heat flow to the sink. Increasing the output bias current increases the heat flow to the sink.It wasn't that I was actually thinking of using only +/- 10v rails ... that was simply an example. But I will have to ponder the value of the DC rails, given this new-to-me concept of the balance of heat generation shifting between heatsinks and load as the average power used by the load changes. But, yes - +/-15v at least!
Andy
You need to find a balance between acceptable heatsink temperature and your target voltage & current aspirations.
You can do this by calculation or by trial and error. Developing a model and calculating is cheaper.
I first posted this on "F5 Turbo Illustrated Buil Guide", but I think this thread is a better place to share my problems.
I wrote this:
I have built more than twenty F5V2 steroset, but never run into a problem like this. PCB is cvillers(see photo).
Setting bias give this results on sourceresistors: Q4(NPN) 254 mV but on Q 104(second NPN) 11.4 mV. Q3(PNP) 240mV but on Q 103(second PNP) 28.5 mV. Offset is no promlem to adjust(1-2 mV).
I have other NPN and PNP taken from the same bat. so to find two of each to match will be easy(but not matching NPN and PNP). As long as I can adjust offset close to zero, I find this kind of matching fair.
Any good advices??
I wrote this:
I have built more than twenty F5V2 steroset, but never run into a problem like this. PCB is cvillers(see photo).
Setting bias give this results on sourceresistors: Q4(NPN) 254 mV but on Q 104(second NPN) 11.4 mV. Q3(PNP) 240mV but on Q 103(second PNP) 28.5 mV. Offset is no promlem to adjust(1-2 mV).
I have other NPN and PNP taken from the same bat. so to find two of each to match will be easy(but not matching NPN and PNP). As long as I can adjust offset close to zero, I find this kind of matching fair.
Any good advices??
Attachments
Andrew adviced me to measure sourceresistors and vgs on all four FETs.
This what I have done so far:
Measuring sourceresistance:
On NPN side: 0.4633 and 0.4674. PNP side: 0.4661 and 0.4674. O.K I suppose.
I have used a testrigg with 2.2 kohm(1%) connected to Sorce. Gate and Drain shorted. Drain connected to "earth" and the "free" end of 2.2kohm to 15 V. I have taken the measurement between Source and Drain. I have picked out three NPN and PNP newly bought from Farnell and got this results:
9240 240
0.993V 0.555V
0.995V 0.558V
0.987 0.555V
Results from the F5V2 I have built(and the strange source measurements)
9240 240
0.555V 0.541V
0.715V 0.564V
First: Have I done the right vgs measurement??
It is one measurement that differ from the other: 0.715V, but does this explain the big difference in #4618?
Eivind S
This what I have done so far:
Measuring sourceresistance:
On NPN side: 0.4633 and 0.4674. PNP side: 0.4661 and 0.4674. O.K I suppose.
I have used a testrigg with 2.2 kohm(1%) connected to Sorce. Gate and Drain shorted. Drain connected to "earth" and the "free" end of 2.2kohm to 15 V. I have taken the measurement between Source and Drain. I have picked out three NPN and PNP newly bought from Farnell and got this results:
9240 240
0.993V 0.555V
0.995V 0.558V
0.987 0.555V
Results from the F5V2 I have built(and the strange source measurements)
9240 240
0.555V 0.541V
0.715V 0.564V
First: Have I done the right vgs measurement??
It is one measurement that differ from the other: 0.715V, but does this explain the big difference in #4618?
Eivind S
Measuring vgs
I found this:Nelson Pass' Easy Peasy MOSFET Vgs Measurement ( 2.2 kohm(1%) connected to Sorce. Gate and Drain shorted. Drain connected to "earth" and the "free" end of 2.2kohm to 15 V. I have taken the measurement between Source and Drain).
From the pictures in this thread, the measured values are between 3 and 4 V. In post 4619 my results shows values very low.
Is rigg I use not the right one to measure vgs on FETs??
Eivind S
I found this:Nelson Pass' Easy Peasy MOSFET Vgs Measurement ( 2.2 kohm(1%) connected to Sorce. Gate and Drain shorted. Drain connected to "earth" and the "free" end of 2.2kohm to 15 V. I have taken the measurement between Source and Drain).
From the pictures in this thread, the measured values are between 3 and 4 V. In post 4619 my results shows values very low.
Is rigg I use not the right one to measure vgs on FETs??
Eivind S