F5 power amplifier

@khaho: since nobody gave you a bit feedback on your board, I jump in, even if it's late.

When layouting a board, usually one routes first the ground, then the rails, (then NFB-path) then the rest. Reason is to have full freedom for the most important parts, like to use space eating star grounding.

Said that your grounding is really a mess; you don't use star grounding and you connect signal ground directly into power ground. That can make problems.

Trace width could be increased a bit as well.

However don't let that stop you from making that board! Go for it and use the tips next time.

Have fun, Hannes

PS: I didn't check your board for errors ;)
 
Actually I did follow your order of making the tracks, maybe the result just is bad.

Thanks for the feedback, I will redo the grounding in next variations, if I will make those. I might even patch the board to have more star like grounding, if needed. Anyway the ground area is very small, because of the circuit design.

I was thinking that 32 mils would be wide enough tracks for the small signal part, maybe it is not. Anyway my layout is much worse than some others posted here, but I think that it will work good enough for me.
 
Hi,

32mils is fine for low current tracks (mA). I can't see what you used for the high current tracks - 100mils? Would be fine as well.

Connect reservoir caps always in a 2 line fashion - one track feeds it, the other goes from the cap to the sink to avoid unnecessary voltage drops.

Have fun, Hannes
 
Is this one better in grounding ? Otherwise it is of course worse than others.
 

Attachments

  • diy-f5-image.png
    diy-f5-image.png
    24.7 KB · Views: 2,309
Hi khaho,
I would suggest that you seperate the "PGND" (for the 2 1,000uF caps) and run a seperate wire to p/supply star earth and this leaves R1,2,10 connected to "GND" only - just a small thing, but good results and easy to do with your posts already there.

Something looks a bit strange about the different connections of the TH1 and TH2 and the ZTX trannies - Not sure, sorry.

A possibility? Add an optional Cmultiplier to replace these caps [IRFP240,9240 - 220uF+47kR on gate - 1uf PP on output] - better result.

Looking good.
 
@khaho: yeah, that's quite nice. However remember that in star ground or star power the connection (wire pad, connector, whatever) is always in the center of the star. That is to avoid, again, voltage drops due to large currents from the output stage.

Signal ground is usually separated from power ground, however a common star ground should work quite well too.

I leave the fine tuning - if any - up to you.

After all it's about building amps and enjoying music and not about making the perfect board-layout ;)

Have fun, Hannes

PS: last comment: maybe you want to use larger pad distances for the resistors? Seems you use 12mm for 10mm resistors, which might lead to problems with broken leads or inner contacts when bent so strongly. But not really an issue since you are not going for mass production :D
 
GNDing

@khaho: V+ and V- are now fine.
But grounding is still a real mess. You need more than one GND stars. One is at the connecting point of main elkos. The other is on the PCB with input circuitry - R10, R1, R2 going to the input RCA. From RCA goes to nearby chassis and from this chassis a thick wire to the main elkos star. Also speakers GND is taken from this elko star - take care speaker currents NOT disturb other grounding currents. Speaker wire from PCB goes twisted with 1000uF elkos grounding and with V+, V- wires as far as main elkos. From main elkos it continues twistwed with speakers GND to output terminals.

This is how I understand grounding. Members with more experience please correct me if I'm wrong. :att'n:
 
There is already thermistor connection to chassis ground from PSU part of F5, so I dont think that GND of elkos (separated from other grounds in PCB) should be connected there again, that would be one ground loop. But if other ground is separated, should that be connected to chassis ground, so it would be tighter in chassis than the PSU ?

I dont understand why speaker ground wire should be twisted with PSU wires ? Maybe because I dont have the experience ?
 
khaho said:
There is already thermistor connection to chassis ground from PSU ....

I dont understand why speaker ground wire should be twisted with PSU wires ? Maybe because I dont have the experience ?

Sorry, I don't understand the need for thermistor. But, sure - avoid GND loops.


The currents are flowing in loops and we must keep these loop areas as small as we can. Speaker GND wire is not twisted with PSU wires, it is twisted with SPKR OUT wire starting from main elkos. SPKR OUT from PCB is twisted with PSU wires & with 1000uF elkos common wire up to main elkos GND star.
 

opc

Member
Joined 2004
Paid Member
F5 Mosfet Questions

Hi Guys,

I have a quick question for Nelson or anyone else who cares to chime in.

I was reading over some mosfet datasheets and I was wondering what the most important characteristics are when trying to match N and P channel fets?

I've been looking at the transfer characteristics (Id vs Vgs), Qg, Crss, Ciss, Rds, Gfs, and the basics like Vdss, Id and Pd.

Nelson uses the FQA12P20 and FQA19N20C, and they don't really seem to be a very good match in terms of gate characteristics, or transfer characteristics.

The FQA36P15 and FQA28N15 seem to have very close transfer characteristics, while the FQA46N15 has very close gate characteristics with the FQA36P15. All three of these fets have much better power dissipation than the 200V parts, at the cost of slightly higher gate charges and input/transfer capacitance.

As of right now, I'm going to try the 36P15 and the 28N15 to see how they perform, and I'll report back with my findings.

Another possibility is using lateral fets like the BUZ901/906 or the Exicon ECX10N16 and 10P16. With the lateral fets you could safely omit the thermal comp and possibly get better performance since they were intended to be used in complimentary pairs.

If anyone has tried any of the above, or if you have some tips for fet selection then please let me know. I can get any of the aformentioned parts, and with only two devices per channel, cost is really not an issue either. I just want to squeeze every bit of performance possible out of this excellent circuit!

Cheers,
Owen
 

opc

Member
Joined 2004
Paid Member
More about Mosfets

Hi Nelson,

Thanks for the quick reply! I kinda figured the only way to tell for sure was to check performance in circuit, so I guess that's what I'll do. I'll order a rail of each of the five Fairchild fets, match them, and give them a test. Luckily I have access to an AP2, so I should be able to get a definitive answer on which fet pairing works best.

When you were developing the F5 did you check the performance of the IRF parts against the Fairchild parts? Does any of this really even make a measurable difference?

Cheers,
Owen
 

opc

Member
Joined 2004
Paid Member
F5 Layout

Hi Guys,

Just a quick posting of my F5 PCB layout before I send it off for fabrication. It uses all SMD components with the exception of a few of the higher power parts.

The layout has a power supply section, with the caps, bridges (mounted beneath the boards) and resistors. Below that are both channels of the F5 with all SMD parts except the JFETS and the output mosfets. The mosfets are mounted below the PCB, and directly against the heatsink. I set it up this way to allow the thermistors to be placed directly above the fets for optimal heat coupling.

On the very bottom is the layout for a very compact B1 buffer again using all SMD parts, including a fancy 1uF film SMD coupling cap, and a very high quality 10uF X7R output cap. Note that the entire circuit is less than 1 square inch. I still have some work to do on the power supply section for the B1, so don't be alarmed by the lack of connections on that part of the circuit.

Red is top layer, Blue is bottom, yellow is silkscreen

If anyone has any comments on the layout, parts, placement, etc... I'd be glad to hear them. I'm sending this out of Friday, so feel free to comment.

As a side note, I have to order 5 of these boards, and I will be keeping one for myself. I also have enough parts for all five boards, so I'll probably be putting the remaining 4 boards, either stuffed or as kits up for sale. Anyone interested?

Cheers,
Owen

P.S. - I'll post the schematics in the next post. This 102 kb limit is going to be the death of me!
 

Attachments

  • first watt f5 layout.zip
    99.3 KB · Views: 444