UCD 25 watts to 1200 watts using 2 mosfets

Input part is like on original schematic, nothing dieferent, but I put 4.7nF parallel with 100pF on input stage. All block capacitors are WMA, MICA and MKT. Output filter 20uH with 1uF capacitor and I put two capacitors 470uF 100V very close to output transistors.

I replaced 330pF on feedback with 100pF ceramic and gate resistors 27 ohms to 20 ohms for higher swiching freq. Output fets are IRFP240.
With +-50V i have little noice but when I connect input to gnd (short) noice gone.
I try this amp with this voltage on 2 ohm load on full power with bass bins and that was amaizing, my house almost distroyed. Little heatsink on amp was cold, output coli little warm, only a little (1 mm Cu wire, 16 turns EE33 ferrite core with 2x1mm gap).

That all of my amp. Now I work on "cliping" indicator (with dual comaparator), speaker DC protection and temp protection with sensor KTY81 for heatsink or output coil. This protection use dual comaparator too (LM319).

When I finished all, amp goes to TEST, big party form dusk till dawn with really powerfull speakers.

Can I ask you something?

What do you think, did I increase swiching freq from original 140KHz?

And what is output power with +-50V 4 ohms?

(50V / 1.41 = 36V 36V x 36V = 1296 1296 / 4ohms = 324W)

right?

Thanks
Regards
 
Your sw freq should be about 220-250 kHz now, thats around optimal. (higher freq would increases switching losses with IRFP240, lower would increase losses in coil)
Changing gate drive resistior wont increase sw freq, but decrease eff. dead time. But if there's no cross-conduction, then its better.
Since high propagation delay decreases modulation limit, your sine output power is lower, about 250W.
Your design is improving very much, now you can easily realise efficiency of class-D. Gapped ferrites are better than iron powder cores (less wire, less losses). You can even improve you inductor by using multiple thin wires (4x0,5) instead of one thick.
I would put some ceramic caps under the 470uF elko, i use SMD ones for that :)
I would use a led vu meter instead of a 1 led clipping indicator. I'm using NTC resistor to control fan speed.
Be aware that SD pin reference is negative supply, when planning protections!
 
Last edited:
Hi !
I have read about this Project and will also try to build D-Class amp.
My Problem ist I.m more Tube -man ,all my Projects are tube designs (amps)
I have some Questions :power suply with 40V -,+ filtered with 120000mikrofarad cond.-must be OK Start?,They are diferent Mosfets used in Project
-what is prefered ? ,Dzony used diferent Coil ? -Problem ist Heat on Coil ,so what is to use ?Amp will be used only for Bass loudspeakers ,what is with protection for speakers -can be used clasic protection via relays or is that kind of protection obsolet?Dzony -PCB ist one sided -its ok ?Thanks for Help
 
Voxac30:
12000uF shuld be well-enough (to kill bus pumping). Mosfets mainly depend on your supplies, and purse. If you can get switching mosfets designed for class-D (for you I reccomend IRFB4228 or FDP2532), its the best. If cannot for low p
The original schematic has some I'd say errors. (it works, but not too good performance)
You should read the reccomendations for coils by Pafi, look at my calculations too. If you do the calculations youn won't have problem with inductor.
You can use the classic relay protection, it's the most reliable, when you can get the right relays.
You can use one-sided PCB, but if you need very great performance (kW and low noise), double sided PCBs with proper ground planes is essential.
I attach my schematic (I USE IRFB4227 insted of IRFP240!), where switching freq is around 220kHz, and there's no input noise.
An externally hosted image should be here but it was not working when we last tested it.


Dzony: i killed noise to a minimal level. I had clues about noise, from your post i got sure. I increased 100p to 1n, and decreased 47k to 22k at input. If all the inputs are open, then there is almost audable hiss (with a horn loaded cab..., with 0,1W output you cannot hear noise :D)
 
Voxac30:
12000uF shuld be well-enough (to kill bus pumping). Mosfets mainly depend on your supplies, and purse. If you can get switching mosfets designed for class-D (for you I reccomend IRFB4228 or FDP2532), its the best. If cannot for low p
The original schematic has some I'd say errors. (it works, but not too good performance)
You should read the reccomendations for coils by Pafi, look at my calculations too. If you do the calculations youn won't have problem with inductor.
You can use the classic relay protection, it's the most reliable, when you can get the right relays.
You can use one-sided PCB, but if you need very great performance (kW and low noise), double sided PCBs with proper ground planes is essential.
I attach my schematic (I USE IRFB4227 insted of IRFP240!), where switching freq is around 220kHz, and there's no input noise.
An externally hosted image should be here but it was not working when we last tested it.


Dzony: i killed noise to a minimal level. I had clues about noise, from your post i got sure. I increased 100p to 1n, and decreased 47k to 22k at input. If all the inputs are open, then there is almost audable hiss (with a horn loaded cab..., with 0,1W output you cannot hear noise :D)


do you have PCB layout ?
 
NMOS:
If the question is, whether my PCB is public, then here is the PCB of the DOT. Its almost the same as the original. I made some modifications in it, changed lots of components to SMD, but not tested, so I show the old, tested one (some component values are not correct, they are on the schematic). If my new PCB will be ok, I'll show it.
An externally hosted image should be here but it was not working when we last tested it.
 
Voxac30:

If you cant find iron core or some ferrite core (tiroid), you can use ETD34 ferrite core. If you cant find etd, you can use ferrite core from pc power supplay. That core is EE33. Simple you remove core from pcb of power supplay, put him in water and cook it :). When water start to boiling, cook it 5 minutes, and while is worm you parted core on two peaces. Remove wire from coil, and use 4 wire 0.5mm and make "one" wire. 16 turns of that wire will be enough. When put together two peaces of ee core, use two small peaces of plastic from CD case (1mm), and put beetwen core, then you have gap on your core. Glue the core and this is your good output coil.Likely to work well up to 600W.
One think, when you wind up your coil, wires must not go over the gap, because will be warm.

Let someone correct if I am wrong about everything
 

Attachments

  • Coil.png
    Coil.png
    10.2 KB · Views: 1,772
Use wire for transformers. No, you not wind diretctly on ferrit core. Use original coil, but remove old wire and yelow tape. When is coil free and empty, start tu wind your wire. You can use a wire of 1,2 mm or 1,5 mm but is defecult to winding and better is multiple wires.
 

Attachments

  • wire.png
    wire.png
    19.2 KB · Views: 1,765
  • default.jpeg
    default.jpeg
    43.6 KB · Views: 1,698
Let me give some issues:
There is no supply decoupling, you must use ceramic caps as close to the fets as possible, and even elkos if you can. Bus pumping can really serius for this design, since power suply capacitors are really far, the trakcs are really thin. Even the RC snubbers has a long path, and if you look at, you can see, that the tracks form a ring, so pcb inductance is not neglectable, so they will be RLC snubbers as Eva mentioned before.
There is a really long way to the inductor (it functions as a radio antenna).
There are design priorities. The easy heatsinking of the FETs should not be the first, since the fets dissipate so little in a good design (for even few hundred watts output poeer), for testing they can even go without sink. Short driver to gate path is good, but there lots of other factors to involve.


please look attachment

have improve PCB, is it ok for Class D...any claims ?

otherwise what can I do better ?
 

Attachments

  • 23.12. 2010Class D PCB.pdf
    94.5 KB · Views: 1,131
  • 23.12 schematic Class D pdf.pdf
    30.2 KB · Views: 994
In the schematic change the followings:
C4 from 330pF to 100-150pF (from 140kHz sw fr to 200-250kHz, otherwise hot inductor, see before)
R5 from 47k to 22k and C3 from 100pF to 1n (mush less noise, see before)
Whats the purpose of your NE555 part?
In the PCB there is a lot of empty space, use the to make planes (ground planes, thicker power paths, V- plane around IR2110...)
The supply decoupling ceramic caps are still missing, you have a lot of space for them. The high freq noise is only filtered by ceramic caps, elkos even with low-esr suffer at high freq.
Your R17 C13 RC snubber is still bad, the R18 C14 is better. Put them right to the legs of FETs.
You dont really using the adventages of two-sided PCB.
Otherwise its getting better.
For +-85V its really reccomended to use aux supplies. You will have much more disspiation in TIP31C, and 5W resistors R22/R23, R21 than the main FETs under normal load conditions. At +-85v i would be a bit scared at bus pumping, speacially if you plan this as a subamp. If you use a linear power supply with bigs caps, no problem. You should calculate bus pumping with dV=V/(8*Pi*f*Rload*Cbus) (where f is the lowest freq possible amplifiing, Cbus is the total capacitance at the power buses)
If + to - supply is close to 200V with pumping be aware (230V is not always 230V, it fluctuates by day), its better to have safe values.
 
In the schematic change the followings:
C4 from 330pF to 100-150pF (from 140kHz sw fr to 200-250kHz, otherwise hot inductor, see before)
R5 from 47k to 22k and C3 from 100pF to 1n (mush less noise, see before)
Whats the purpose of your NE555 part?
In the PCB there is a lot of empty space, use the to make planes (ground planes, thicker power paths, V- plane around IR2110...)
The supply decoupling ceramic caps are still missing, you have a lot of space for them. The high freq noise is only filtered by ceramic caps, elkos even with low-esr suffer at high freq.
Your R17 C13 RC snubber is still bad, the R18 C14 is better. Put them right to the legs of FETs.
You dont really using the adventages of two-sided PCB.
Otherwise its getting better.
For +-85V its really reccomended to use aux supplies. You will have much more disspiation in TIP31C, and 5W resistors R22/R23, R21 than the main FETs under normal load conditions. At +-85v i would be a bit scared at bus pumping, speacially if you plan this as a subamp. If you use a linear power supply with bigs caps, no problem. You should calculate bus pumping with dV=V/(8*Pi*f*Rload*Cbus) (where f is the lowest freq possible amplifiing, Cbus is the total capacitance at the power buses)
If + to - supply is close to 200V with pumping be aware (230V is not always 230V, it fluctuates by day), its better to have safe values.


Thanks for assistance, please check new schematic and PCB layout

1. have changed C3 (before C4) to 100 pf change C2 (before C3) to 1nf and resistor R4 /before R5) to 22k

2. have add supply decoupling ceramic caps, C18, C19 with 0,1 uf parallel to 470uf Elko

you have removed in your schematic cap 1,5 uf from Inductor output to GND (original schematic) , but you have add 1uf cap from Inductor output to negative rail , why ?

any ideas to improve ?
 

Attachments

  • PCB Class D UCD BETA 24.12.pdf
    97.2 KB · Views: 756
  • schematic UCD BETA 24.12.pdf
    30.7 KB · Views: 740
by the way: Merry Christmas to everybody
NMOS:
Getting better. Theres is still a lot fo empty PCB space. Fill it, its more economical (less FeCl3 for making), less noise (solid gorund planes take off audio noise and EMI too), less heat (wider power tracks). And its even more aesthetic.
To V-? Uhh thanks for pointing out an error, the error is only in the schematic noted, on the PCB its okay. (it would even work, but I would need a cap with 2x the voltage rating)