Output Bias current for Zen-Like Amplifier

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I'm working on building up an amplifier loosely based on the Zen series, in this case a common source amplifier (IRF612 (Siliconix)) with complementary output follower (IRFP244 (Harris) and IRFP9240(IR)). The 612 is biased at 50ma, and the output stage at 1A, with +/- 30V rails. The topology I'm using is more complicated than the Zen series, mostly due to devices added for bias stability and a JFET source follower to drive the input MOSFET. The amp simulates nicely in PSpice, with about 0.2% THD, mostly 2nd harmonic, tapering off nicely up through 8th harmonic, at 40Vp-p into 8 ohms resistive load. This aint bad for an open loop design. At any rate, I want to see how close reality is to the simulation. The board at present has places for two pairs of output FETs. I simulated using one pair. Two sets would help to spread out the heat loading a bit, but I don't want to dissipate too much more than 60W/side due to heat sinking limitations. Is 500ma/FET too lean a bias current, or should I stick to just two devices at 1A bias current? For that matter, is 1A too lean for acceptable sound in a real situation with an efficient speaker? I'll probably be using the amp to drive a line array of 16 Parts Express 4" Pioneer drivers plus 4 Peerless dome tweeters per side.
 
Well, I think that even with efficient speakers you're going to want to go with 5 to 10 watts still, at least to have some headroom. That's kind of oppinionated though so it's up to you what you'll need. Your line array from the LS forums will probably end up around 98dB which is absolutely killer. At that efficiency I wonder how loud it would get with just a 50mW opamp driving it :) Or maybe a 50mW opamp per speaker....

Anyway, back on subject, NP says higher bias gives better sound. But we could do some real math to help you out. 40Vpp means +/- 20V into 8 ohms. This is going to be about 2.5 Amps peak current. Your 1-Amp total bias won't support that. On the other hand, you're also talking about 50-Watts of output power. You'll probably want 1-Amp per FET and use both pairs to get closer. Of course that's 30-Watts per FET.

Here's an idea, lower your rail voltages to +/-20Volts and increase your bias to 1.5A total, and split it 750mA per FET so that you're at 30-watts per side and can still do about 10-watts output min. If you feel the need, you still have room to increase the bias and get more power or better sound.
--
Danny
 
Azira has touched most of valuable points to think about and I agree with his suggestions.
However, at 1.5A of biasing level, IRF244 and 9240's opration will be off from optimal input voltage/ output current curve where you can get the best linear charactericstic and so the least distortion. At 20V and 1.5A operation, using IRF530 and 9530 will give you better outcome with cheaper cost. That's exactly what I did for mine and worked like a charm!
 
The 40V p-p figure I mentioned was just because I fed the simulation model with 1V input and didn't bother to change it around. I was more interested in tinkering around for DC bias and output zeroing. Yesterday, I adjusted the input for a little less than 10V p-p output, which would leave the drivers in cllass A. The THD changed to about 0.06%, still with a nice harmonic distribution.
The Pioneer 4" drivers I'm looking at using are good for 5W max apiece, so I wouldn't want to drive them too hard anyway. My boards are laid out for TO-247 output devices, so I will stick with the IRFP244/P9240 I mentioned at the outset. It turns out that both of the heat sinks I will have on hand are about 0.25C/Watt, so I could go to about 120W per heat sink and still have an acceptable heat sink temperature (50-55C). I'll start out with a single pair of output devices biased at 2A apiece, a compromise between the ultimate sound and melting a hole in my carpet...

The actual rail voltage I use for the amps will depend to a large degree on the transformers I have on hand. I'll be doing some tests tonight to determine just exactly what I have. I have several E-bay toroidal no-name transformers and one monster fully enclosed EI-type transformer that is probably good for about 20-25V rails. Since the bias for the input is regferred to the negative rail, I will have to use quiet supplies, so some sort of regulator is in order. I will probably start with a simple capacitor multiplier setup or zener-pass transistor regulator using a MOSFET. This is not the ultimate in output voltage stability, but it will get the ripple down to low levels. If I get desparate, I can always roll my own discrete feedback-type regulator.:hot:
 
I'm also thinking of using a choke input filter for the power supply, as the transformers I have on hand have higher than optimum output voltage. The high class A bias current allows one to do a choke input filter with a relatively small size inductor (5-10mH). I tinkered with a PSpice simulation to get the proper inductor value for continuous conduction. Too low a value of inductance allows the ouput capacitors to peak charge, while too large a value is a waste of iron and copper. I have some steel EI core inductors left over from a failed passive power factor correction project at my former job that would make the nut with a little bit of gap adjustment. This inductance value is also very comfortably in the range for a large powdered iron E core. The powdered iron permeability varies with the AC flux density, providing some "swing" in the inductance.

The advantage of a choke input filter is better transformer utilization, as the peak current through the windings goes way, way down, essentially to the DC load current. The rectifers also recover from a lower peak current, making it a little easier to snub the reverse recovery transients. I'm planning to use 100V Schottky rectifiers anyway, so this shouldn't be too much of a problem. There will also be a lot less output ripple for the same value of output filter cap. The only disadvantage is the size and weight of the filter chokes, though anyone planning to build a class A amp shouldn't be put off by size and weight considerations anyway...
 
Attached are a couple of implementations of a bipolar choke input filter. The circuit on the left is the one I will most likely use, as it has the smallest component count. The circuit on the right requires commutation diodes to provide a return path for the inductor currents.Those looking for more information can consult classic tomes like Terman's Radio Engineering or the Radiotron Designer's Handbook.
 

Attachments

  • choke_input.gif
    choke_input.gif
    6.8 KB · Views: 240
Attached is a photo of an impromptu lashup of a bipolar choke input power supply using a 38-0-38V toroidal transformer bought through Ebay (inexplicably labeled "Barney" by the seller with magic marker, but what the hey). I used a pair of ~20mH powdered iron toroidal chokes that were discards from a failed passive power corrector project at my old workplace. They look to be wound using 17 or 18 AWG wire. A 25A, 200V bridge and a pair of 69kuF, 50V caps completed the setup. The schematic is shown in one of my earlier posts on this thread. A pair of electronic loads was used to load both positive and negative channels to 4A apiece. Output voltage at 115VAC in is 28V each side, indicating that the chokes are in continuous conduction mode and not drying out (not suprising). I wanted to get a rough idea of the temperature rise of various components so I left the setup on for about 1/2 hour. The transformer was barely warm to the touch. The inductors were cold. The hottest component was the bridge rectifier - middling warm ( I estimate about 40C). The next test will be to leave the whole mess on for an afternoon to equilibrate to determine the long-term temperature rise of all the components. I'll also be looking at output ripple and varying the load currents to determine the approximate dryout point of the inductors (not trivial, as the permeability of the powdered iron core material can vary by a factor of as much as 2X as a function of AC flux swing - best to determine dryout current experimentally).
This all looks very promising for a Class A power supply. I might try using a similar setup with a higher voltage transformer for a class AB amplifier with MOSFET outputs. With the choke input filters I can power the output stage with the LC filter and tap off ahead of the inductors to derive a pair of boost supplies to more effectively drive the output stages.
 

Attachments

  • ch_input1a.jpg
    ch_input1a.jpg
    97.9 KB · Views: 199
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.