BigE,
Conrad suggests that doubling the (vertical) length of a heat sink only increases the performance by 40-50%. Conrad Heatsinks - Technical Details. Aavid suggests the correction factor should be .58. - heatsink design tools
Therefore a 10" long profile will probably be only 60-70% better than 3" of the same sink, or somewhere around .4-.5 °C/W. Using AudioSan's temperature criteria (30°C rise), each heat sink is good for ~67 W. At 75W dissipation, your sinks will be roughly 34°C above ambient. If you keep your listening room at 20°C your 10" sink will be OK for one channel.
If you can live with 75W class A peak rather than RMS as AudioSan calculated, you only need 1.5A bias. Using the same 40V rails and the 1.5A RMS, I get dissipation of 120W. You still need two sinks per channel to keep the temperature below 55°C at 25°C.
Class A is a heavy metal project.
Conrad suggests that doubling the (vertical) length of a heat sink only increases the performance by 40-50%. Conrad Heatsinks - Technical Details. Aavid suggests the correction factor should be .58. - heatsink design tools
Therefore a 10" long profile will probably be only 60-70% better than 3" of the same sink, or somewhere around .4-.5 °C/W. Using AudioSan's temperature criteria (30°C rise), each heat sink is good for ~67 W. At 75W dissipation, your sinks will be roughly 34°C above ambient. If you keep your listening room at 20°C your 10" sink will be OK for one channel.
If you can live with 75W class A peak rather than RMS as AudioSan calculated, you only need 1.5A bias. Using the same 40V rails and the 1.5A RMS, I get dissipation of 120W. You still need two sinks per channel to keep the temperature below 55°C at 25°C.
Class A is a heavy metal project.
Let me check my numbers, but I think the 0.4 should be lower. Did you use 20" of heatsink for 4 pairs? Or that there are two pairs on each sink, not 4?
Oh bob, I see you posted.... yes, I got something like that, using a derating factor of the square root of the difference.
Cut in half eh? That's interesting....
Oh bob, I see you posted.... yes, I got something like that, using a derating factor of the square root of the difference.
Cut in half eh? That's interesting....
Last edited:
So without wading through 300+ pages on the F5 turbo thread, I am curious if there is a recommended place to look or phrase to search for to try and find out what size/tap configuration transformer would be recommended when building one of the DIY Universal V3 power supplies. I've spent weeks reading, and am having a hard time sifting through the magnitude and variety of information available.
Jaystone - what is your target power output?
A rule of thumb for CRC class A supplies is your rails will be 1.2 x VAC using a separate winding and bridge for each rail. It doesn't matter whether you use the store boards or some other arrangement.
The formula for RMS power is 1/2 V^2/R where R is the nominal speaker impedance. Solve for V given your speakers and power spec. Add 3-4V for losses in the Source resistors and mosfets to get your target rail voltage. Other amps have different losses but most are in the same ballpark.
A good tool to help understand PSU design and operation is a program called, oddly enough, PSU Designer.
Hope that helps.
A rule of thumb for CRC class A supplies is your rails will be 1.2 x VAC using a separate winding and bridge for each rail. It doesn't matter whether you use the store boards or some other arrangement.
The formula for RMS power is 1/2 V^2/R where R is the nominal speaker impedance. Solve for V given your speakers and power spec. Add 3-4V for losses in the Source resistors and mosfets to get your target rail voltage. Other amps have different losses but most are in the same ballpark.
A good tool to help understand PSU design and operation is a program called, oddly enough, PSU Designer.
Hope that helps.
I recall that the extra 2" ( 12 - 10 ) is a big deal. Still, I am going to look into this cutting of the sinks a bit more.
Thank you.
Do you have an opinion about doubling up the jfets in the input section for a 4-pair mono block?
Really.
You are now saying " I know I am using this wrongly and that it is not recommended by the manufacturers.
But, I will keep doing it the wrong way until it fails !"
you should be okei with one pair of J-fets. around 8mA pairs.
maybe you need to adjust R5/R6 to around 2.2Kohm.
BigE,
Conrad suggests that doubling the (vertical) length of a heat sink only increases the performance by 40-50%. Conrad Heatsinks - Technical Details. Aavid suggests the correction factor should be .58. - heatsink design tools
Therefore a 10" long profile will probably be only 60-70% better than 3" of the same sink, or somewhere around .4-.5 °C/W. Using AudioSan's temperature criteria (30°C rise), each heat sink is good for ~67 W. At 75W dissipation, your sinks will be roughly 34°C above ambient. If you keep your listening room at 20°C your 10" sink will be OK for one channel.
If you can live with 75W class A peak rather than RMS as AudioSan calculated, you only need 1.5A bias. Using the same 40V rails and the 1.5A RMS, I get dissipation of 120W. You still need two sinks per channel to keep the temperature below 55°C at 25°C.
Class A is a heavy metal project.
Hi Bob,
Yes, I was going to use two sinks per channel. N boards on one side, P boards on the other.
I think the question should be phrased, What happens IF you were to put in 2 pr of JFETs in a F5TV3 with say 4pr per channel Mosfets?
Do the extra JFET's allow for lower input signal gain? or perhaps, with higher current output, allow the mosfets to bias lower?
just asking....
Lower Idss Jfets or lower Idss value for the FE requires a higher load resistor and increases the gain of the FE(essentially the load resistor divided by the Jfet source resistor). Pairing up the FE Jfets increasing the Idss and decreases the load resistor, lowering gain of the FE. Either scenario is fine in terms of driving the output stage. The question to ask is how the different variations of gain affect the final CL gain and final distortion spectra. For instance, the BA3 is a single gain stage if you look at the fact that the FE gain is largely degenerated away.
I will let the more educated and knowledgeable explain, as i will surely muck it up, but i do believe your on the right path. It does affect bandwidth, as it is naturally related to Mr. Miller and the input capacitance of the following stage. Increasing gain increases miller, decreasing gain does the opposite. Plenty of drive current helps deal with it all to a degree. Now hopefully someone will say something more intelligent.
Unfortunately, the article only references that doubling the front end fets will enable you to drive 20 pairs of mosfets, but says nothing about driving 4 pairs..... unless you take the schematic as telling you how many jFETs to use to drive 4 pairs..... It would be an interesting experiment for someone with sufficient test equipment to determine if there is a measureable benefit to doubling input pairs with only 4 pairs of MOSFETS per channel.
BigE,
On page 14 NP states
Turn this comment around and increased current increases the bandwidth to a point. Since the V2 is stated to have the same bandwidth as the F5 and two pair is the starting point in the comment about decreasing bandwidth it seems likely that a little more front end current will increase the bandwidth somewhat with 4 pairs. Will it be significant or more importantly, audible? That's up to you to decide. Build it both ways and report back. Those building 4 pair amps seem happy with a single pair, though.
On page 14 NP states
Turn this comment around and increased current increases the bandwidth to a point. Since the V2 is stated to have the same bandwidth as the F5 and two pair is the starting point in the comment about decreasing bandwidth it seems likely that a little more front end current will increase the bandwidth somewhat with 4 pairs. Will it be significant or more importantly, audible? That's up to you to decide. Build it both ways and report back. Those building 4 pair amps seem happy with a single pair, though.