FSSA amplifier build thread with review

ClaudeG

Member
2015-01-25 2:12 am
Will post pix anc comments (and questions LOL) of course!

Yep, the VFET has been tweaked as much as I could and it is all about enjoying music. It is a nice little amp that will go into the living room for general purpose where the fun factor is key, paired with efficient speakers (TBD but so far the RP-8000F are a VG match while being excellent VFM).

FFSA will go in my main system. That is in my audio room, with more demanding speakers... and more demanding ears, so expectations are high for that one, hence taking the time to do it very carefully.

Enjoy music very much

Claude
 

Kalamin

Member
2016-11-23 10:07 pm
France
Hi Zman and others,
My FSSA has just been slugging somewhat after I populated the PCBAs; I've now got the heatsinks and need to wire-up 4-off linear ballast supplies, which will be fed by a 800W (so called) switching PSU. I also need to find a suitable and nice housing for it, not convinced by the aluminum ones I found so far, I may try the wooden case route (or mixed wood+alu)..:)
Thing is that life has been so busy this year that even sports had to stay back and wait...for 2022 hopefully! :mad:
 

fab

Member
Paid Member
2004-03-24 1:16 am
Canada
fabaudio.online
Hi Stretchneck
there is no specific ripple value objective, the lower is always the better…. I have used C-R-C of 15kuF- 0.1- 15kuF with 250VA for each mono channel.

For the optional secondary power supply of the front end, a regulated one would be better (more stable front end polarisation during high power demand) of course but the main objective is to get a higher voltage than the one of the output stage. This allows a maximum signal voltage (thus power) to the output stage with same output stage heat dissipation. I have not done it for my USSA this time but already did successfully for my Goldmund Telos clone and other amps.

Fab
 
  • Like
Reactions: 1 user

stretchneck

Member
Paid Member
2018-03-17 6:45 pm
Oxford
Hi Stretchneck
there is no specific ripple value objective, the lower is always the better…. I have used C-R-C of 15kuF- 0.1- 15kuF with 250VA for each mono channel.

For the optional secondary power supply of the front end, a regulated one would be better (more stable front end polarisation during high power demand) of course but the main objective is to get a higher voltage than the one of the output stage. This allows a maximum signal voltage (thus power) to the output stage with same output stage heat dissipation. I have not done it for my USSA this time but already did successfully for my Goldmund Telos clone and other amps.

Fab
Thanks Fab - I ask about ripple because, using something like XRK’s smooth like buttah can achieve just a couple of mV of ripple. However, it is current limited, which is fine for a reasonable bias into class A, but may need additional capacitance to keep class AB happy. I guess adding 15kuF after SMB would meet your design goal?

I’m asking these questions because of the Pass labs designs I have owned (SIT3, XA25 and F7) whilst sounding lovely all had a slight grain to the nature of there presentation… which I presume was caused by higher noise floor associated poor PSSR than, for example a Neurochrome or AHB2 amplifier which I have also owned.

My assumption is that whilst the harmonic nature of FSSA would remain unchanged that the noise floor would drop considerably if using a SLB, or CLC (better than CRC?). But how do you know when the power supply is over designed… is C L C L C C filter totally overkill, or worth trying?

Sorry, lack of direct experience with power supply design, I’m just using past experiences to guide me.

In theory I guess I could use any power supply, even a regulated one, provided that if it was current limited (e.g 5A) I used enough capacitance after it(?).
 

stretchneck

Member
Paid Member
2018-03-17 6:45 pm
Oxford
I think I found the answer in this earlier post. Basically size the caps according to the load on the amp.

FSSA provided with 50vdc rails gives 120w in 8ohms - 120/8 = plan for 15amps of ripple current availability. Worst case would be a continuous 4 ohm load (approx 1.7x.) this would be 204w/4ohms = 51amps of ripple current availability.

Fabs recommended CRC of 15000 uF – 0.15 ohms – 15000 uF, gives a total of 30A of ripple current availability. So given that we're not talking steady state here there seems to be no requirement for capacitance higher than these recommended values. Indeed, with lower supply voltage you could probably decrease the capacitance with no ill effect.

I also like the idea of running a number of smaller caps to achieve this as the cost will be cheaper.

Now just a question of going R or L, or just all C's. What specs do people use? What are the relative merits?
 

ClaudeG

Member
2015-01-25 2:12 am
Quote: I also like the idea of running a number of smaller caps to achieve this as the cost will be cheaper.

That xas the case decades ago when I started this hobby. Paralleling several caps turned out to be both cheaper and to decrease the ESR.

Things have though moved on since. I found that on many occasions it is not chea^per anymore and ESR can be higher, not even taking into account connexions betwen caps!

So better check!

Big lytics have improved on that... downside is they lost though re life expectency... different materials etc.

All IME

Claude
 
  • Like
Reactions: 1 user

Kalamin

Member
2016-11-23 10:07 pm
France
I think I found the answer in this earlier post. Basically size the caps according to the load on the amp.

FSSA provided with 50vdc rails gives 120w in 8ohms - 120/8 = plan for 15amps of ripple current availability. Worst case would be a continuous 4 ohm load (approx 1.7x.) this would be 204w/4ohms = 51amps of ripple current availability.

Fabs recommended CRC of 15000 uF – 0.15 ohms – 15000 uF, gives a total of 30A of ripple current availability. So given that we're not talking steady state here there seems to be no requirement for capacitance higher than these recommended values. Indeed, with lower supply voltage you could probably decrease the capacitance with no ill effect.

I also like the idea of running a number of smaller caps to achieve this as the cost will be cheaper.

Now just a question of going R or L, or just all C's. What specs do people use? What are the relative merits?
Hi Stretchneck, you may perhaps reconsider these maths here...:unsure:
 
So...120W into R=8 Ohms are reached with about 3.9Arms. 4 Ohms load, fed with same rms voltage would deliver: we had P=U*U/R=120W for 8 Ohms; now for r=4 ohms: r=R/2, which ported into previous formula gives newP=U*U*2/R=2P=240W.
240W into 4 Ohms are produced with about 7.75Arms.
That is the easy bit...Computing the current ripple into the caps could be much more complicated according how precise results you expect to get.

To keep things simple, assume current ripple is sine-shaped; so, out of the amp you'd get a pp ripple from each rail of SQRT2*Irms only, as each output signal semi-cycle is coming from a different DC rail (pos/neg)
you could then size your DC rail caps according a targetted max. voltage ripple on the rail, which is not an easy route.... or use as most people would do, a rule of thumb saying: "X thousands microfarads per hundred watts, for my amp, give..." and you're likely to land where you anticipated: 15.000µF for each C of the CRC.
So all in all, you're into the right ballpark...:scratch2:
Hope this would help dissipating some doubts.
Kal.
 
  • Like
Reactions: 1 user

stretchneck

Member
Paid Member
2018-03-17 6:45 pm
Oxford

stretchneck

Member
Paid Member
2018-03-17 6:45 pm
Oxford
So, first time using psud2.

1652892705590.png

Around 60mV of ripple using cheap caps, ok.

1652892779997.png

Still caps only, but now with 2 pairs of Mundorf HC which have 6mOhm ESR each. Ripple down to 12mV due to the ultra low ESR.

Finally a CLC arrangement:

1652892539102.png


Zero ripple predicted after C2 with just 2mH. So am I doing this right? Doesn't seem to make sense otherwise why would people go to more extreme lengths?

How audible is this?
 
Last edited:

stretchneck

Member
Paid Member
2018-03-17 6:45 pm
Oxford
Thanks Fab - silly error on my part. Perhaps I shouldn't do these things when tired!

Remodelled with a 2.5mH choke which has 44mOhms of resistance (Hammond 159ZL) and the simulation is much more representative. Ripple is 1mV and really smooth. Is R1 1kOhm load the most appropriate for representing FSSA?

CLCLC - is a further improvement, adding in two inductors doesn't show any predicted ripple (using this load anyway). But how audible is <1mV of ripple though? Is it really worth having two inductors?

What DC current should the Choke be capable of for FSSA? I'm assuming a 10A choke.

I've noticed that the caps only approach also gives nasty sawtooth, that's really now precludes the use of a cap only solution. The ripple peaks are smoothed slightly with better ESR caps, but still not comparable to introducing a Choke.
 
Last edited:

ClaudeG

Member
2015-01-25 2:12 am
In this amp, as in many, the PS output is modulated by the music signal.

I am not sure you will hear the last mV ripple, but from experience ears can make the difference between so called hard and soft PS. In short, that's related to the output impedance of your PS vs frequency.

I am looking for as low as possible, WHILE retaining a similar value from DC to say 100kHz. It requires a bit of knowledge and time to model it, but the key factor for you is retaining caps that have a low ESR after whatever R or C that you may use to filter (CRC or CLC). bearing in mind Fab has made some provisions for a (additional) reasonable PS cap directly at the output transistors / on the boards

All IMHO, I will leave others giving explanations or their POV... especialy as I haven't even gone beyond gathering parts and chassis for this build, which is intended to be THE quality power amp in my house... the latter being under construction since 2 years, wasting so much of my time :-(
 
  • Like
Reactions: 1 user