F5 Turbo Builders Thread

It sounds like your thermistor has too much resistance to work as it should in that circuit. It should have a hot resistance low enough to charge the capacitors to 90% or more of the final voltage in a few seconds. I have over 300Kuf in my F5V3 monoblocks and they have the thermistors in the circuit for less than 10 seconds and do not blow fuses when the thermistors are shorted out by a relay.

I modelled this on LTspice years ago.

If I remember rightly, the resistance was chosen to limit the spike in the inrush current on startup, given the size of the cap banks. It is higher than usual.

The delay was chosen to limit the spike during switchover.

Initially, the caps charge *very* quickly. For the switchover spike to be under 15 amps, the voltage level at the cap bank needed to be, as you say, near 10% of the rail voltage. If it was say 11 or 12% , the spike would be well over 20 amps, and could take out a fast blow fuse.

It takes a lot more time to fill the cap bank near the end that it does at the start. All that waiting is to fill the last couple of percent. And still, a very close rated fast blow fuse will open after being stressed a few times. I've switched them to slow blow to avoid that nuisance.
 
Now I have disassembled the amplifier to check the build quality and errors.
J-fets are cascoded. Source resistors have right values.
Output fets are poorly mounted, especially the problematic Q4. There is no tight contact with heatsink.
Some screws spin into the gap when I tighten them :(.
I don’t know how long the amplifier operated under such conditions. Would make sense to replace the output transistors or can i test them somehave?
 

Attachments

  • IMG_20210113_181953.jpg
    IMG_20210113_181953.jpg
    230.4 KB · Views: 264
  • IMG_20210113_181958.jpg
    IMG_20210113_181958.jpg
    248.3 KB · Views: 253
  • IMG_20210113_182027.jpg
    IMG_20210113_182027.jpg
    317 KB · Views: 253
  • IMG_20210113_182102.jpg
    IMG_20210113_182102.jpg
    264.5 KB · Views: 246
Nice discovery! That is the most probable explanation. In my wrecked inherited build, the holes were equally screwed, no pun intended.

Not nescessary to replace fets, unless you have easy acess to matched ones. As long as they draw similar amounts of current / similar voltage drops, they should be OK. As Zen Mod says: a fet works, or it does not work. There is not much in between.

I would, as is, adjust bias waaay down, and see if they draw similar ampunts of current then. 8-10 percent pocket should be fine. If you compare now with high bias, q4 will heat more than the paired one and hence possibly draw more current.

I think you are good. The important part now is to unfu** the previous owners mistakes.

If it is not possible to use other holes, i would make a new bigger hole and use a little bit bigger bolts. Another option is new screws: sometimes that will do it. But ensure you clean the bolt hole properly first. Conpressed air and tightening a brand new bolt in and out a few times, will help cleaning it all the way down.

Sounds like they have soldered fets first, then tightened them. A road to certain disaster.

You probably know this, but I would recommend the following, once all holes actually work:

1: desolder fets once you know q4 works properly (90% certain)
2: clean holes, and / or make bigger onces
3: mount the fets to sink again, tighten not too much. Ps: ensure thermal pads are still good!
4: put in the PCB, ensuring all fet legs line up.
5: align fets and tighten goood. 1-1,2nm on all will ensure even dissipation, but requires a special tool.
6: resolder, with a BIG tip.

And of course remove all flux, and ensure all other joints are good.

I think you found your culprit! Nice!
 
Last edited:
Now I have disassembled the amplifier to check the build quality and errors.
J-fets are cascoded. Source resistors have right values.
Output fets are poorly mounted, especially the problematic Q4. There is no tight contact with heatsink.
Some screws spin into the gap when I tighten them :(.
I don’t know how long the amplifier operated under such conditions. Would make sense to replace the output transistors or can i test them somehave?

If you need to use the same hole and the thread is gone you can use Helicoils. Bigger bolt may not go through the fet.
 
Can anyone tell me what size fuse to use for my F5 turbo monoblocks with 41V rails? I have 88000uf per rail in the power supply, if that makes any difference.

I was thinking 4amp slo blow?

Zen Mod recommends this:

if proper soft start is in place, calc is simple

VA of xformer/ mains voltage = Fuse rating

go for first bigger standard value, slow blow

And, NP suggested using a 6 AMP Slow Blow in the same thread.

My F5 turbo monoblocks are also around 40v rails and 80k uf per. They have a 800VA xformer / 120 line voltage = 6.66 amp slow blow. I went down to a 6.3 amp slow blow ceramic and have no issues.
 
I modelled this on LTspice years ago.

If I remember rightly, the resistance was chosen to limit the spike in the inrush current on startup, given the size of the cap banks. It is higher than usual.

The delay was chosen to limit the spike during switchover.

Initially, the caps charge *very* quickly. For the switchover spike to be under 15 amps , the voltage level at the cap bank needed to be, as you say, near 10% of the rail voltage. If it was say 11 or 12% , the spike would be well over 20 amps, and could take out a fast blow fuse.

It takes a lot more time to fill the cap bank near the end that it does at the start. All that waiting is to fill the last couple of percent. And still, a very close rated fast blow fuse will open after being stressed a few times. I've switched them to slow blow to avoid that nuisance.

With a class A amplifier you need to account for the high bias current that is also flowing through the series thermistor resistance. This lowers the maximum cap bank voltage during the initial thermistor limited charge period and makes the current draw much higher when the thermistor is switched out. So you need to use a lower value thermistor in order to get the charge up. The reason that the current is higher than expected when the thermistor is bypassed is that the energy stored goes up with the square of the capacitor voltage. So double the voltage=4 times the energy stored. 90 % voltage charge is storing 81% of the energy, leaving 19% to be topped up. Slow blow fuses seem to be very reliable when properly sized for this application.
 
Last edited:
F5Tv2 Build

Hi hi. After initially investigating the ACA, I think I would be better served by an F5. My speakers are a HiVi 3.1A kit with the perfectionist crossover mod at 4ohm, around 86db sensitivity, rated up to 150W. I use them in nearfield, so I'm thinking an F5 Turbo V2 should be more than enough. I just want to make sure I'm properly understanding everything, given the variants. I'll need the 5U case, one F5T PCB kit, one F5 stuffing kit, one LSK170/LSJ74 Matched Quad kit, one Universal Power Supply PCB kit, everything on the PSU BoM, an Antek AN-6224 transformer, input and output terminals, 10-way binding posts, hookup wire, and some Keratherm. Am I missing anything? Thanks!
 
Hi hi. After initially investigating the ACA, I think I would be better served by an F5. My speakers are a HiVi 3.1A kit with the perfectionist crossover mod at 4ohm, around 86db sensitivity, rated up to 150W. I use them in nearfield, so I'm thinking an F5 Turbo V2 should be more than enough. I just want to make sure I'm properly understanding everything, given the variants. I'll need the 5U case, one F5T PCB kit, one F5 stuffing kit, one LSK170/LSJ74 Matched Quad kit, one Universal Power Supply PCB kit, everything on the PSU BoM, an Antek AN-6224 transformer, input and output terminals, 10-way binding posts, hookup wire, and some Keratherm. Am I missing anything? Thanks!

Or you can go with the BA-3 complimentary input and output =) A dual mono F5T is a complex build, but if you have the means and experience, go for it. Not sure the F5 stuffing kit will do it though. And remember value of the gate resistors and optional C’s to avoid oscillation.

And remember, Nelsons 16 parallell caps in the PSU can’t nescessarily be replaced with 2x8 in dual PSU with the same ripple reduction, I would think. The low R in the PSU is to a certain degree compensated by many caps, as well as providing extra capacitance of course.
 
Last edited:
Member
Joined 2019
Paid Member
I just learned something new. I did not know PSU ripple was affected by load until just now. Still learning PSUD, but it showed a reduction with 1/2 the load with my hacked simulated PSU. Neat, now I need to get my brain wrapped around why. Is it as simple as the caps not draining/charging proportionally as much => thus lower ripple? Thanks! :D
 
Quite new at it too. But quite interesting. I do not get the physics. But I think of it like this: there is always an AC component, it can be very very small, and large, depending on number of variables. This component increases in some kind of proportion to the DC current draw. Hence, as you increase the load (DC current draw), you increase the AC component too. For example, when I have tried little filtering and very high bias, ripple seems to affect bass (120hz) as the amp nears clipping. As I reduce the bias a bit and hence the load, and apply more filtering, the bass is in fact firmer and less volumious. I might be wrong here, but that’s my subjective listening experience wrt ripple and filtering. Dunno if taking the bias down or increasing filtering, or both, were the causes though.

Wrt excactly how and why this happens, I am still learning.

If you look at some of Krells A/B amps, they have very little filtering. Ripple is seldom an issue with such amps at quiescent current levels, but might affect sound at high outputs even though maybe not so audible.

NB: if ripple levels are adequately low to begin with, you can increase the load accordingly with more headroom before ripple becomes an audible issue.
 
Last edited:
Member
Joined 2019
Paid Member
Thank you! :) I went back to R/C time constants along with the storage capacity along with the discharge/charge rate of the caps and my brain melted (again). :D I'll get my head wrapped around it at some point, but I think I've got the general concept. It all makes perfect sense now, but I had never even thought about it.
 
Thank you! :) I went back to R/C time constants along with the storage capacity along with the discharge/charge rate of the caps and my brain melted (again). :D I'll get my head wrapped around it at some point, but I think I've got the general concept. It all makes perfect sense now, but I had never even thought about it.

Nelsons F5T PSU is interesting in several ways in this regard. I haven’t gotten around to using PSUD yet, but am planning to construct a PSU there soon.

If you try to double the amount of caps in PSUD, do you see the program assessing only halv the ripple amount? You also might see the total ESR/impedance decrease accordingly. Though some say the opposite wrt impedance. I have to install that program soon =)
 
Last edited: