I've built a small headphone amplifier loosely based on the Son of Zen amplifier. I'm having a problem with some sort of distortion that I can't pin down. While the amp sounds ok, it feels like it's always on the verge of distorting, no matter the volume, especially in the lower frequencies and the treble seems a little grainy to me. Also, it seems like the more instruments are playing at once, the worse it gets. Is there a type of distortion that is the result of trying to reproduce multiple frequencies at once? I don't have an oscilloscope so it's very frustrating trying to track it down. All my voltages are pretty close to my calculations. It seems like the bass is really sloppy, like it wants to keep going after the hit (ringing?). My thoughts are either some sort of induced oscillation, bad damping, or perhaps a bad bias point. I'm running the bias at about 360mA per device, regulated by a simple constant current source using IRF510's. It gives me just about 5 volts at the drain of the signal mosfets(also IRF510's). The problem seems especially bad when listening to the radio for some reason, but a cd sounds really pretty good unless i crank it up past about 1 volt input. I've enclosed a schematic of the circuit. One thought was that I'm using wire wound resistors for the 27ohm drain resistor and the 4.7ohm current source resistor (I was out of metal films of that value). All other resistors are metal film. I've enclosed a schematic for reference. Any advice would definitely be appreciated
Attachments
psykod said:
Intermodulation Distortion. As a result of the
nonlinearity the circuit also acts as a sort of 'mixer', where you get not only the frequencies f1 and f2 out, but also sums and differences: f1+f2, f1-f2, maybe even 2*f1-f2 and 2*f2-f1, and so on... depending on how "bent" the transfer function is.
I don't expect a wirewound resistor to be part of the problem. Also, I didn't see the attachment, so I'm not sure what the bias point of the amp would be.
Erik
I can see the pic now.
About 360 mA per half with 27 ohm load resistors would give you 5.28 volts at the drains, sure enough. (What sort of amplitude do you need for your headphones? I dunno what 'typical' would be for 'phones.)
Figuring about -3.7 volts at the Source and you have a VDS of 9 volts, give or take a few dozen mV.
If you take a look at Figure 1 of the IRF510 data sheet you'll see that you would cut through a fairly bent part of the output characteristic, with VDS=9V.
Maybe bump the positive rail up a bit? Or use a smaller Drain resistor? Or a little of both?
Erik
About 360 mA per half with 27 ohm load resistors would give you 5.28 volts at the drains, sure enough. (What sort of amplitude do you need for your headphones? I dunno what 'typical' would be for 'phones.)
Figuring about -3.7 volts at the Source and you have a VDS of 9 volts, give or take a few dozen mV.
If you take a look at Figure 1 of the IRF510 data sheet you'll see that you would cut through a fairly bent part of the output characteristic, with VDS=9V.
Maybe bump the positive rail up a bit? Or use a smaller Drain resistor? Or a little of both?
Erik
I'm sure you're probably right about that. It's been quite a while since my last electronics course. The way I was looking at the data sheet, with a Vgs of 4 volts, ID was so low it's hard to tell what the pinch off range is but it looked like i was well within it's linear range to me. I'm using a hafler pre-amp with a 5 volt output so I don't need a lot of gain, but I was hoping to maybe connect the amp straight to my CD player. My Sennheiser's have a 300ohm impedance and sound pretty good feeding straight off the pre-amp outs (just a bit of weak bass).
Let me see if I got this right:
If I move the drain voltage slightly higher....
The source voltage will always be approx. 3.7 volts below the gate voltage. So if I'm swinging -2.5 to 2.5 volts at the gate (the most my hafler will put out), the source will be about -6.2 to -1.2 volts. I need to maintain a minimum Vds threshold voltage of 2.5 volts (We'll say 3.0 volts). That puts the minimum drain voltage at (-1.2 + 3.0) = 1.8 volts. So if I maintain a minimum of 2 volts and I want an output voltage swing of 10 volts then the drain voltage should be biased at: (2 + (10/2)) = 7.0 volts. So with a 15 volt supply, I need to drop about 8 volts across my drain resistor to acheive 7.0 volts at the drain. If I'm using a 27 ohm drain resistor, then the bias current should be: E=I*R, 8=I*27, I=8/27, I=296.3mA. I think I'm calculating a gain of about 4 (gfs=2 * (Rd || Rs)). So a 5 volt input would give me about 20 volts swing at the output. Since I've biased my drain to 7 volts, the system is going to clip pretty handily. If this sounds good to you, I think I might want to re-think my gain resistors, perhaps replace the 10ohm with a 1k ohm.
Thanks for the help by the way.
FWIW: The transconductance curve on the data sheet for IRF510 looks to be really lousy for Id below about 2 amps. Almost a vertical line. From 0 to about 1.25. From 2 amps on up, it only deviates about .25. Perhaps I shouldn't even be contemplating this set up at such a low bias?
Let me see if I got this right:
If I move the drain voltage slightly higher....
The source voltage will always be approx. 3.7 volts below the gate voltage. So if I'm swinging -2.5 to 2.5 volts at the gate (the most my hafler will put out), the source will be about -6.2 to -1.2 volts. I need to maintain a minimum Vds threshold voltage of 2.5 volts (We'll say 3.0 volts). That puts the minimum drain voltage at (-1.2 + 3.0) = 1.8 volts. So if I maintain a minimum of 2 volts and I want an output voltage swing of 10 volts then the drain voltage should be biased at: (2 + (10/2)) = 7.0 volts. So with a 15 volt supply, I need to drop about 8 volts across my drain resistor to acheive 7.0 volts at the drain. If I'm using a 27 ohm drain resistor, then the bias current should be: E=I*R, 8=I*27, I=8/27, I=296.3mA. I think I'm calculating a gain of about 4 (gfs=2 * (Rd || Rs)). So a 5 volt input would give me about 20 volts swing at the output. Since I've biased my drain to 7 volts, the system is going to clip pretty handily. If this sounds good to you, I think I might want to re-think my gain resistors, perhaps replace the 10ohm with a 1k ohm.
Thanks for the help by the way.
FWIW: The transconductance curve on the data sheet for IRF510 looks to be really lousy for Id below about 2 amps. Almost a vertical line. From 0 to about 1.25. From 2 amps on up, it only deviates about .25. Perhaps I shouldn't even be contemplating this set up at such a low bias?
psykod said:
What's odd is that I've used IRF510s at a lower bias current in my previous Bride of Zen line amp. The VDS was about 30V, though.
Did you try bumping up the voltage at the drain? If so, how did it work?
(You might want to decrease the value of the resistor that ties the Sources together if you reduce your drain resistors, btw)
You mentioned, too, that it sounds reasonably good coming straight out of the preamp, just a little weak at the bottom end. Have you considered a simple source-follower buffer?
Erik
So far I've tried running the drain voltage from as low as 5 volts to as high as 9 volts to no avail. I've also played with the source resistors. I did make a simple mosfet follower based on the Szekeres design and it sounded very good, but of course, I'm a glutton for punishment so I have to have more gain. It's possible that it's a power supply problem or a defective part. This week I'm going to be re-building it from scratch along with a nice brute force power supply (70,000uF + 10mH per rail per channel). That should eliminate the power supply as the culprit. I also found an odd little mosfet at the parts store. A vishay/Siliconix VN46AFD that only has a Vds of 40 volts, and Id of only about 1 amp, and has only 25pF of input capacitance and a 10nS rise time. Seems like that should be just right for my 250-500mA of bias current I'm shooting for.
FWIW, I'm running the IRF610s in my headphone Zen (http://www.diyaudio.com/forums/showthread.php?s=&threadid=12243) at Vds 14V/200mA and they work great. I think in the BSOZ current is even lower.
BTW, have you rewired your cans for balanced operation? I'm not sure the SOZ would make much sense otherwise...
BTW, have you rewired your cans for balanced operation? I'm not sure the SOZ would make much sense otherwise...
Initially I didn't tie the unused input to ground because I thought it was already tied through the 10k resistor. Once I tied it to ground, the sound improved but still sounded "off". I have re-wired my Sennheiser's for dual +/- connections to the amp. I then made a small pigtail that allows me to play them through a regular jack.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.