What you think about this schematic?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi all. This is my first post here and my english is not very well and i'm sorry for that :eek:.

I created one schematic using ideas from many others but i don't know is it good , is there something that is wrong or can be made better.

Schematic is working actualy and sound is good.

Any opinions are wellcome.

So good...bad...or....what?
 

Attachments

  • Sch.jpg
    Sch.jpg
    133.8 KB · Views: 1,590
Driver stage is running 40mA which maybe is a bit on the high side, but everybody who has tried adjusting the driver current for IRF (including myself) has learnt that more is better. If you ever lower it, don't go under 13mA.

If it is working and if it sounds good, then you really don't need our opinion, do you? :cheers:
edit: you should throw in a picture of your amp.
 
Looks good to me.
A couple of questions:

I wonder if one pair of IRFP240/9240 is enough?
I would probably go for paralleling 2 pairs. And so can feel safe.
Because if there is a dip in Load impedance so it goes to 4 or 3 Ohm
a teoretical worst case would be like 10 Ampere output current.
This is much for one transistor to handle in a good way.

Bootstrap.
330R and 3.3k is very big difference.
Many times we use 2 with same value: 1.8k + 1.8k
But sometimes can be a smaller difference: 1k5 and 2k2, for example.
 
Looks like a rather conventional SS design. There are a couple of things that seem odd, however. It isn't unusual to bootstrap the VAS, as opposed to active loading, but the series combination of a 3K3 and 330R for the collector definitely is. Most bootstrapped VAS's split the collector resistor into equal halves, with the bootstrapping applied to the junction of the two resistors. Was that a schemo typo?

Also, paralleling the output inductor with 5K6 doesn't look right. The de-Qing resistor for this coil is usually much smaller than that, as a large resistor won't do its job of de-Qing that coil to prevent ringing at its natural resonant frequency.

Otherwise, if it's performing well for you, dunworrybouddit.
 
astankov - the output stage

I have done some SPICE of only the output stage.
As I have those transistors.
If we know the output works well as it should and does not distort too much
then designing the rest of amplifier is easier.

Images.

1. This is the out at idle. 102mA
What I have added is in BLUE.
What values I have changed a bit is in RED.

Nelson Pass in a topic once made me aware to use VBE-multiplier with same characteristic as the output devices.
This means we BJT multiplier for BJT output and HEXFET for HEXFET output.
So I use here IRF510 to set the bias.

I have used 220 Ohm for R3, as 37mA or 41mA does not matter.
But 220 is much much easier to find and buy.

I did some testing for R1/R9. Nelson Pass often use 220 Ohm.
My simulation got the speed ~2.2MHz when using 470 Ohm.
With those 100 Ohm we get speeds upto 10MHz.
Actually speeds higher than 1MHz is not really needed and can cause problems,
when we talk about asmplifiers with 300-500 kHz upper limit.
A good target here is so far I know 1-2MHz.

I have also added a Zobel filet, 100 nF + 10 Ohm.
So far I know, a Zobel filter is more important
than output inductor filter.


Image 2.
You had set 3uH for the inductor.
The usual for this filter is 10 Ohm + 1uH.
THe image shows how filter will effect upper bandwith
with 3 different values for inductor: 3uH, 1uH and 0.5uH
The highest curve is the bandwith BEFORE the filter. With 470 Ohm gatestoppers.
The lowest curve is 3uH.
The thin red is what I have used = 1uH


Image 3.
The output of 50 Watt rms into 8 Ohm.
20 Vrms out is a magic number for +-35 Volt amplifiers.
Means we have +-28.28 Vpeak
... and the margin is less than 7 Volt to supply.
For MOSFET we can take supply voltage and subtract 5-6 Volt
to get maximum voltage swing.
And here we end up at a figure like max 29 Vpeak.
 

Attachments

  • astankov_output1.png
    astankov_output1.png
    11.6 KB · Views: 1,348
  • output_filter.png
    output_filter.png
    8 KB · Views: 1,066
  • astankov_out_50w-8ohm.png
    astankov_out_50w-8ohm.png
    16.4 KB · Views: 1,022
Last edited:
Image 4.
In this setup i use 4 Ohm load.
Then I tested and see where output starts to complain.
Where there is beginning to be an increase in distortion.

This output managed to give out 92 Watt rms, before things get bad.
The voltage is 19.2 Vrms. Means a Vpeak of 27.15 Vp
 

Attachments

  • astankov_out_92w-4ohm.png
    astankov_out_92w-4ohm.png
    16.8 KB · Views: 990
Thank's very much for pictures and simulations

Thank's guys for your opinions. From you I understand that for guy like me where electronics is just a hobby I made this schematic good. But there are some things that need to be improved.

1.To increase gate resistors from 100 to 470R
2.To add Zobel 10R and 100nF
3.To correct the Bootstrap with more equal resistors
4.To use HEXFET for VBE multiplier
5.To use lower value resistor in the output coil

1,2,3 and 4 are easy to done but if I use 10R , 5W resistor for output coil body how much windings and how thick wire to use to make the correct value?

So I will try the changes and will post the result.

Thank's again for your professional opinions - that's what I needed.
 
With a +-35V power supply you might want to use IRFP9140 instead of the IRFP9240 (the 9240 is a 12A unit while the 9140 is a 21A unit, closer to the 19A of the IRFP240).

The 'odd' bootstrap arrangement with 3.3k and 330 ohm resistors will work but you would need to significantly increase the bootstrap capacitor - by a factor of 10. This sort of asymmetry is sometimes used to linearize the bootstrap impedance - the absolute minimum impedance would then equal the 3.3k resistor, which increases the VAS gain at DC in comparison with a regular 50%-50% split, which may help lower DC offset. Also, sometimes this is done for stability if the bootstrap manifests negative impedance. The resistor is in series with this and makes it always positive. In any case you may want to increase the bootstrap cap value - the time constant it makes with the bootstrap resistors must be the largest of all present in the amplifier. It is also worth noting that the quality of the bootstrap cap has quite an influence on the sound of the amp - bypass with a small foil cap (0.1u...) for best results.

The feedback network seems overly complex. I would replace the 470nF cap with a short and remove the 56k resistor, making it a more classical arrangement without the need to correct the frequency compensation.

The Vbe multiplier used to set the bias will work as is. It may well be the best to use as the output is really a composite of BJT and MOSFET devices. Because the MOSFETs are somewhat degenerated by the source resistors, the BJT driver stage may be dominant regarding thermal compensation. I doubt that it would go into thermal runaway with either the BJT Vbe multiplier, or the MOSFET Vgs multiplier - so I would change this last.

Finally, there are two potentially fatal flaws in the design!.

First, the conenction of the bias set pot is dangerous. If the wiper contact fails even very intermittently - and this can happen even on a new pot while the wiper is being adjusted - the idle current becomes 'infinite' which will destroy the output stage. Unfortunately, the actual MOSFETs may even survive blowing well dimensioned fuses but this is not what will kill them - the second flaw will. simply join the wiper on the pot and the junction with the 10k resistor. There is also a marginally better way to do this that requires a bit more rewiring but as a quick fic, this should do just fine.

Second, there is no gate protection for the MOSFETs It seems like this amp began as a BJT design, and the output was then replaced by MOSFETs, A zener in series with a Si diode (like the 1N4148) betweeen the gate of each MOSFET and the output will do the trick. The zener voltage depens on your chosen maximum current - it will be on the order of 7.5-9.1V.
 
I won't touch the feedback network.......not at all.
This type of feedback makes bass stiff , fast and more deep and it is used in other designs so I like it :D. And as you can see on the schematic it is power supply voltage dependent. If someone is interested i will post values for different supply voltages or the way of calculating the values.

About trimpot for bias I agree that is connected dangerously but this is only in prototype design and will be corrected. And I also can add gate protection.

Thank's again for the suggestions.
 
1. In this schematic, what if BD139/140 is removed, and directly connect B to E (emitor resistor 220 ohm is out/removed). The voltage should be enough to drive MOSFET (4 volt by adjusting Vbe), and the current is also enough (10mA).

2. What if Vbe is replaced by simple VR without transistor, assuming using lower current / class AB, or Lateral MOSFET?

Just curious why BD139/140 driver is needed - if current (and thermal dissipation) is still proper for MOSFET final -, and why Vbe multiplier is needed to compensate bias (as this will lower sound quality).

Thx,

Ervin L
 
Are the electrolytics actually installed the way you have them drawn?

Not sure. To be right oval shaped pole on the picture should be positive.

Without driver stage this schematic will also work but 10mA is not enough to drive huge input capacitanse of HEXFET devices (more than 1200pF for IRFP 240/9240). I tested without drivers and sound was awful. I did not tested with LMOS devices but probably will work without driver stage.
HEXFET's really need a thermal compensation and to clarify that read please this article:

Using HEXFETs in High Fidelity Audio
 
1. In this schematic, what if BD139/140 is removed, and directly connect B to E (emitor resistor 220 ohm is out/removed). The voltage should be enough to drive MOSFET (4 volt by adjusting Vbe), and the current is also enough (10mA).

2. What if Vbe is replaced by simple VR without transistor, assuming using lower current / class AB, or Lateral MOSFET?

Just curious why BD139/140 driver is needed - if current (and thermal dissipation) is still proper for MOSFET final -, and why Vbe multiplier is needed to compensate bias (as this will lower sound quality).

Thx,

Ervin L

#1 - Gate capacitance would load the VAS just enough to greatly increase THD. Some designs bias the VAS at over 15ma and run the mosfets directly. With the drivers , you lose another volt on top of the 5v required to forward bias the MOSFET, but isolate the voltage stage which reduces THD. For me , DRIVERS are essential , you would have a junky , cheap design without !!


#2 , with vertical MOSFET's .. they have the same tempco as a typical BJT. So, they would increase their conduction , creating more heat... more conduction , POP .. thermal runaway !! With a very large heatsink this could take a while and how they were biased would be a major factor. Vertical MOSFETS sound the best and simulate with the best THD at @100ma+.

Attached is my MOSFET setup , which I make use of junkpile IRFP240's , grab any ideas or ask ????'s.
OS
 

Attachments

  • quasi.gif
    quasi.gif
    19.4 KB · Views: 421
Last edited:
#1 - Gate capacitance would load the VAS just enough to greatly increase THD. Some designs bias the VAS at over 15ma and run the mosfets directly. With the drivers , you lose another volt on top of the 5v required to forward bias the MOSFET, but isolate the voltage stage which reduces THD. For me , DRIVERS are essential , you would have a junky , cheap design without !!


#2 , with vertical MOSFET's .. they have the same tempco as a typical BJT. So, they would increase their conduction , creating more heat... more conduction , POP .. thermal runaway !! With a very large heatsink this could take a while and how they were biased would be a major factor. Vertical MOSFETS sound the best and simulate with the best THD at @100ma+.

Attached is my MOSFET setup , which I make use of junkpile IRFP240's , grab any ideas or ask ????'s.
OS

I use DoZ (3 transistor, 1 CCS, 1 common source, and 1 for CFB) as LTP and VAS, bias at around 15mA (and already tried to 30mA without problem using BC141-161), drive MOSFET directly at +/- 24 volt. So I wonder if I add another BD139/140 like in your schematics it will improve the THD or add another sound signature improvement.

Currently I use IRFP 240/9240, and bias to 500mA each without any problem. Previously, I bias up to 1.5A but at lower voltage, +/- 13.5V also without any problem (run for hours) - just hot transformers :) . I add 9.2 silence computer fans in chasis (front of heatsink).

I don't use Vbe multiplier as I belive it will degrade sound when hot, while without it, it will have "warmer" sound when being hot (class A bias).

I still use DoZ as it is using CFB, and I believe CFB is still better than VFB/differential amplifier.

Any comment for improvement?

Thx,

Ervin L
 
Last edited:
Disabled Account
Joined 2006
I won't touch the feedback network.......not at all.
This type of feedback makes bass stiff , fast and more deep and it is used in other designs so I like it :D. And as you can see on the schematic it is power supply voltage dependent. If someone is interested i will post values for different supply voltages or the way of calculating the values.

About trimpot for bias I agree that is connected dangerously but this is only in prototype design and will be corrected. And I also can add gate protection.

Thank's again for the suggestions.

Which designs also use this feedback scheme ??? Do you have any literature about it explaining its working ?? I guess the way of calculating the values would help as well.
 
Disabled Account
Joined 2006
My tips for the output stage ..... Use a smaller value source resistor for irfp9240 around 0.15 ohms, this will help match the transconductance compaired to irfp240.

Use a smaller value gate resistor for the P mosfet, around 60-75 percent of the value of the n mosfet, so in this case a value like from 56 to 82 ohm is good if the n mosfet s gate is 100 ohm, that is if its stable with the 100 ohm. For a good explanation see aplication note AN1645 from National.
 
It is used in the design on the picture. On the text file is the formula for calculating the values and on second picture already calculated values for common supply voltages.

Original source: Audio Club Bulgaria - DIY :: Ïðåãëåä íà òåìà - Excela
(Sorry in Bulgarian Only)

Schematic is copyrighted. Free for personal use!
 

Attachments

  • pic.png
    pic.png
    26.3 KB · Views: 421
  • ku.jpg
    ku.jpg
    37.4 KB · Views: 401
  • Ku.txt
    104 bytes · Views: 83
Disabled Account
Joined 2006
Interesting thanks, Ill try get some of it translated, anyway the original schematic is how I use those v mosfets in output stages,Triples, just change the output bjts to Fets, install gate resistors, modify temp compensation, bias fets at 100ma, drivers at 50ma, predrivers at 7-10ma and you have a very good Fet outputstage.

Btw you really shouldnt use mje340 or kse340 for temp compensation, they have the worst caracteristics for this use, they have very low HFE and high voltage.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.