The heatsink with clip is 666. More expensive, more difficult to mount, worse heat transfer. What is there to like? 🙂
Good question and I should clarify any changes required for operation between 50 and 60V.
For operation above 50V output, RF filter should be omitted and C3 capacitor is not required. Murata RF filter is specified to withstand without failure 125V for a short period. But, max. operating voltage is specified at the 50V and there is no data on long term reliability above 50V. There is no risk if RF filter operates at several V above 50V. I’ve tested it long term at 60V but can’t recommend that high value. Instead of RF filter, two jumper wires should be soldered. If there will be more builders that require 60V, I can add a proper PCB version.
At 60V output, C2 should be rated at 80-100V.
No other parts need any change for the 60V but small heatsinks are required for Q5, Q6, Q7 and Q8 (three heatsinks in total). C10 is high quality Panasonic 63V capacitor and it can certainly operate at the 60V for a lifetime. But, to cover any conditions during first power up and testing, where output voltage could reach 65 -70V at extreme, output capacitor can be replaced with 100V part, like:
https://www2.mouser.com/ProductDetail/Panasonic/EEU-FC2A681?qs=ki9Q8F4jp9A09hu14nNjgg==
For operation above 50V output, RF filter should be omitted and C3 capacitor is not required. Murata RF filter is specified to withstand without failure 125V for a short period. But, max. operating voltage is specified at the 50V and there is no data on long term reliability above 50V. There is no risk if RF filter operates at several V above 50V. I’ve tested it long term at 60V but can’t recommend that high value. Instead of RF filter, two jumper wires should be soldered. If there will be more builders that require 60V, I can add a proper PCB version.
At 60V output, C2 should be rated at 80-100V.
No other parts need any change for the 60V but small heatsinks are required for Q5, Q6, Q7 and Q8 (three heatsinks in total). C10 is high quality Panasonic 63V capacitor and it can certainly operate at the 60V for a lifetime. But, to cover any conditions during first power up and testing, where output voltage could reach 65 -70V at extreme, output capacitor can be replaced with 100V part, like:
https://www2.mouser.com/ProductDetail/Panasonic/EEU-FC2A681?qs=ki9Q8F4jp9A09hu14nNjgg==
Oh silly me. 😆
There is no need to sacrifice all RF filtering at voltages above 50V.
There is a filter in the same form factor rated at 100V (BNX023-01). Sure, you lose several dB of PSRR, compared to the default filter, but not much.
So, 'official' recommendation for 50-60V range is to use filter with this part number:
https://www2.mouser.com/ProductDetail/Murata-Electronics/BNX023-01B?qs=/%2BPMR94VuMYPshF2N9CBoQ==
That was an epic fail from my side, not to check before. 🤣
There is no need to sacrifice all RF filtering at voltages above 50V.
There is a filter in the same form factor rated at 100V (BNX023-01). Sure, you lose several dB of PSRR, compared to the default filter, but not much.
So, 'official' recommendation for 50-60V range is to use filter with this part number:
https://www2.mouser.com/ProductDetail/Murata-Electronics/BNX023-01B?qs=/%2BPMR94VuMYPshF2N9CBoQ==
That was an epic fail from my side, not to check before. 🤣
I was focused on max. filtering performance and reasonably high voltage. Those BNX026 were bought years ago and tested during R21 development and are still better choice. But, some RF filtering may be better than none.
You have used LT4320 and mosfets and that quite expesive solution if you need a power supply with + and - to the ground.
In this case you need 2 chips and 8 mosfets, compared wit ordinare 4 diodes.
With the LT4320 and mosfets voltage loss is lower than even with schottke diodes, and that is important if you need low voltage, let say +-5V what I need.
I simulated if I can with one chip to get +- output bat is not working in that configuration.
In this case you need 2 chips and 8 mosfets, compared wit ordinare 4 diodes.
With the LT4320 and mosfets voltage loss is lower than even with schottke diodes, and that is important if you need low voltage, let say +-5V what I need.
I simulated if I can with one chip to get +- output bat is not working in that configuration.
True, not a cheap solution but design goal was best performance in one compact package.
At high currents, difference in losses compared to the diodes is considerable. This supply can handle 10-20A and it would need large diode bridge with ample cooling. There is also no need to bother with snubbers, as ringing after switching is an order of magnitude lower than with best diodes.
It is a known design limitation that LT4320 can’t be used with center tapped transformer for a dual rail symmetrical supply, so no cheap solution there.
At high currents, difference in losses compared to the diodes is considerable. This supply can handle 10-20A and it would need large diode bridge with ample cooling. There is also no need to bother with snubbers, as ringing after switching is an order of magnitude lower than with best diodes.
It is a known design limitation that LT4320 can’t be used with center tapped transformer for a dual rail symmetrical supply, so no cheap solution there.
The cheap solution is NEVER the best solution. This thing is worth every penny and minute of your time.
Don't know what to say other than if that is the case one would think he would be well aware of the limitations of LT4320.
I will try to keep my enthusiasm for your power supply to myself.
I will try to keep my enthusiasm for your power supply to myself.
I have been reading many @dadod posts with great interest. He knows his stuff, no doubt.
I think his comment was mostly about the expense represented by the R25, particularly in my case, where I won't even draw 1A out of it.
Also, in general, there is a lot of debate about what ends up being audible in terms of supply noise, and it is fair for him to question if the expense is worth the results.
People like me are a bit obsessed and won't leave any stone unturned, and maybe we have more money than sense; although I don't have much money.
I think his comment was mostly about the expense represented by the R25, particularly in my case, where I won't even draw 1A out of it.
Also, in general, there is a lot of debate about what ends up being audible in terms of supply noise, and it is fair for him to question if the expense is worth the results.
People like me are a bit obsessed and won't leave any stone unturned, and maybe we have more money than sense; although I don't have much money.
My $0.02
LT4320 active rectification efficiency when used for Class A amps that draw big current constantly is a major benefit, no heat generated, no trafo snubber calculations and approximately 1V5 “free” bump in output voltage. For me, there was no going back to conventional discrete/bridge rectifiers since the the first time LT4320 was used in a build of mine.
The less heat advantage is lesser when used with Class AB amps and source gear, but the others still hold true.
Cost can be a burden, but this is DIYA! Low cost is not in the fine print 🤣.
Disadvantage: LT4320 can’t be used with center tapped transformers.
The only job I use bridge rectifiers for now is Ground Loop Breakers, hehe.
LT4320 active rectification efficiency when used for Class A amps that draw big current constantly is a major benefit, no heat generated, no trafo snubber calculations and approximately 1V5 “free” bump in output voltage. For me, there was no going back to conventional discrete/bridge rectifiers since the the first time LT4320 was used in a build of mine.
The less heat advantage is lesser when used with Class AB amps and source gear, but the others still hold true.
Cost can be a burden, but this is DIYA! Low cost is not in the fine print 🤣.
Disadvantage: LT4320 can’t be used with center tapped transformers.
The only job I use bridge rectifiers for now is Ground Loop Breakers, hehe.
Vunce - Maybe I missed it, but have you observed any subjective impacts/improvements on the sound of your Class A amps?
I have been spying on @Vunce and other people who know better:
For Q7 and Q8, it appears it is 1 shared heat sink.
Facing the trimmer, the TL431 gets in the way a lot, facing the other way I need to give enough clearance to the resistor.
But, unless I am missing something, it's a shared heat sink no matter what, right?
C3 is not needed above 50V regardless of the RF filter being used or not? Or, is it C3 is not needed when you omit the RF filter.
Yes, one heatsink sandwiched between the BJT’s. Just a dab of goop on the backside of each.
Fitment:
The TL431 was mounted low to the pcb, the heatsinked assembly was mounted high so the lowest “finger” just cleared the top of the TL431.
In order not to stress the leads of Q7/8 and a flat fitment to the heatsink they should be removed. Assemble separately then install back in the board, make sure to double check the orientation of the pins is correct.
If you have fresh BJT’s, use those with uncut leads so you’ll be able to clear the TL431.
Fitment:
The TL431 was mounted low to the pcb, the heatsinked assembly was mounted high so the lowest “finger” just cleared the top of the TL431.
In order not to stress the leads of Q7/8 and a flat fitment to the heatsink they should be removed. Assemble separately then install back in the board, make sure to double check the orientation of the pins is correct.
If you have fresh BJT’s, use those with uncut leads so you’ll be able to clear the TL431.