Power amp under development

[quasi]

Originally posted by Upupa Epops
Have you any detailed pic of square wave ( 10 kHz ) into eight Ohm ? Or eight Ohm with parallel 100 n cap ? 3 dB bellow clipping ?

Hey Upupa Epops;

Piccy of square wave response at 1khz into 8 ohm resistive load taken at near rail to rail swing.

Cheers

[quasi]

Piccy of square wave response at 10khz into 8 ohm resistive load at near rail to rail swing.



Cheers

[Upupa Epops]

To quasi : Have you connected amp with generator by line with correct characteristic impedance ? I see on input small overshoots .

[quasi]

Originally posted by Upupa Epops
To quasi : Have you connected amp with generator by line with correct characteristic impedance ? I see on input small overshoots .

No I didn't. I just connected the output of my signal generator to the power amp input so the signal is just terminated by the input impedance of around 33kohms.

Is there a better way?

Cheers

[quasi]

1. Yes you can use 1 or 2 pairs of FETS in this design. Without doing any calculations I would say you could use 6 or more pairs.

2. You can use almost any power FET. As a guide FETS with a lower gate capacitance will provide better performance. No schematic changes would be required unless you changed the rail voltages significantly. If the rail voltages were increased then attention may need to be paid to the voltage rating and power dissipation of the input and driver transistors

3. The values of the Zobel network (correct name) are not that critical. The network is intended to prevent RF from entering the amp from the speaker leads. It also provides loading to the output stage for high frequency helping to dampen any potential for oscillation. Many purists don’t like this network, personally I always use it.

4. No coil is needed when driving a subwoofer if you are using a low pass filter before the power amp. Some people also use a coil in conjunction with a resistor and capacitor across the woofer to modify its impedance and response. Whether a crossover (properly designed) is used with a multi-driver speaker or whether a full range is used without a crossover is entirely up to you.

5. I use this symbol (empty triangle) to indicate the “quiet” ground for the signal path from the “noisy” or power supply ground.

6. Yes pretty much. The power delivered is based on the capability of the power supply and the safe operating area of the FETS. The circuit as it stands should deliver near 400 watts into 4 ohms with the right power supply. For driving into 2 ohms I would add 2 more pairs of FETS.

7. Each amp module draws about 20 watts at idle. Other control circuitry that will be added (DC detect protection and soft turn on) will use about 2 watts per channel. Add about 10 watts for the transformer idling and you have about 50 – 60 watts of idle power ( 200mA – 250mA @ 240 volts).

8. Expected amplifier power output from a perfect amp and a perfect power supply is calculated by;
(DC rail voltage * 0.707)² / speaker impedance
E.g. (60 volt rails * 0.707) ² / 8 ohms = 225 watts RMS

The amp cct described in this thread could indeed be used for a multi-channel system and can be used with differing supply rails. You could use lower rails and less output devices for the mid and treble and a higher rails with more output devices for the bass. How much power you allocate to each amp will depend mainly on your crossover frequencies and will be an estimate in the end because it will depend on the signal content you listen to. I would personally run 100 watts per channel for the mid and treble combined and about 300 watts per channel for the bass.

I use DOS (yes DOS) based PCB and schematic design software and there is no easy way of posting these designs. I currently print them then scan them. This is why they look bad. Oh well ……

I hope I have provided some meaningful information. Contact me any which way if I have missed anything.

Cheers

[Upupa Epops]

Look at output impedance of generator and the same impedance give on input pins of amp, but in parallel ( typical output impedance of generators is 600 Ohm or 50 Ohm.

[keypunch]

Quasi,

Thanks for your excellent answers to my questions. I have couple follow-up questions of your answers I need clarification on:

2) With respect to "As a guide FETS with a lower gate capacitance will provide better performance." from the mosfet specs is one of Ciss, Coss, Crss the gate capacitance? If not what is the usual designation in the spec for gate capacitance? I have looked high and low now and in past to find out what in spec states or provides the information from which gate capacitance can be calculated.

6) Would it be fair to say with the current 3 pair of output drivers and ideal power supply that 200W in 8 ohms is possible? If so then basically are we talking 65W into 8 ohms per device (ignoring what is possible with one device at 4 or 2 ohms)? If I get to point of making my system active I likely use one or two output devices, and likely I would use an IFR240N or IRF250N. One device for the mid or tweeter, and likley have midrange no more than 60W in 8 ohms, but for reserve and flux of speaker impedance like to have reserve of 120W in 4 ohms. If I am forced to use a 4 ohm driver I want 240W reserve to 2 ohm for variences in driver impedance. If I could not have the 240W reserve then I will in all likelyhood not use a 4 ohm mid driver. Tweeter would not likely have more than 25W 8 Ohms (50W reserve fo r4 ohms) output acheived via a reduced power supply and exact same drivers. Later if I felt a better choice of output driver suited the tweeter amp I likelyhave many more choices at that power level.

7) Excellent for idle. Any idea how scales up once amps start to actually do something?

It is ok, I still know what DOS is. I date back much farther than PC DOS Does your DOS based software allow you to print to a Postscript printer? My memory is rusty, but I seem to recall CAD based software might also be able to print to HPGL or Laserjet PCL files. I know you likely do nto have a Postscript printer, that is quite ok, all that is important is if the CAD program can print to Postscript, not if you have a Postscript printer. Same holds true for HPGL or LaserJet PCL, no need to have, just if program is able to print to one if had one. If the DOS program can print to a postscript printer, then there are many ways to likely obtain a higher image quality in the .PNG, .GIF or .PDF than you have generated to date from the DOS based CAD program. If cannot print to Postscrip be can to HPGL or Laserjet PCL good possible solutions exist.

Do you scan with DOS or Windows? Do you run the CAD program in a Windows DOS window? If yes to the last, and program can print to Postscript then this will make things much easier than running DOS only PC, though not that much more.


Regards,

John L. Males
Willowdale, Ontario
Canada
12 December 2004 13:49

[Nexus]

Hello JWB:

In reply to: post #57

http://www.diyaudio.com/forums/showt...269#post505269

can you post a schematic of the protection part which is on the pcb in this post?

Thank You in advance.

Nexus

[quasi]

Quasi is unhappy;

Well I have spent (wasted?) a lot of time trying to get this amp to clip symmetrically by using split & higher rails to the driver stages.

The problem I had (maybe common to all quasi-comp MOSFET amps) is that the positive signal swing clipped around 4 to 5 volts before the negative swing.

Well I succeeded in fixing it. I thought by using a higher rail for the driver stages (particularly the positive rail) I now have the negative rail clipping some 10 volts before the positive swing. and the clipping isn't as clean.

What gives??

The amp runs best with common rails all tied together (even with the positive signal swing clipping first).

Do any of you learned gentlemen and ladies have any thoughts?

One thought that had was shifting Q8's collector from the output rail to positive rail.....but this will double the voltage across it.

Hmm....please help......please?

This is the latest cct.

[richie00boy]

keypunch,

IRFP250N has about the same SOA as the older IRFP240. A single pair of these will give you 120 watts into 8 ohms. If you can get IRFP250 (not N version) that will give you even more. I'm talking genuine IR parts only.

Ciss is the capacitance figure you should be concerned about. Also check Qg which is gate charge. You want this to be as low as possible and it's a better thing to use than input capacitance, which is quite non-linear.

quasi,

Your amp is having problems when you use higher driver section rails because the lower half of the output stage has a little gain (yes, I know the feedback results in overall unity gain) thus can swing almost to the rail. So if you move the operating point for the driver of the lower rail to a higher voltage it begins to saturate the MOSFET much earlier as it's referenced to a higher point, as you have found. The only solution with this output topology is to use a higher positive rail only.

If you move the collector of Q8 to the positive rail you ruin the whole feedback around the driver/op and end up with an unstable (both thermally and electrically) amp.