M³ Project Announcement
Aug 23, 2004 at 1:04 AM Post #76 of 565
Morsel: Yes, there is a DC offset at the opamp output, thus limiting your voltage swing. But the schematic you showed has the same problem. The opamp is connected to the top of the biasing circuitry. Which has to be a few volts above ground to start conduction. To avoid the problem, it would have to be connected to the middle. Of course the bias circuitry would have to be different, since this one doesn't really have a middle. I believe the original Sheldon Stokes article mention this. In a design I am playing with, I use balanced outputs, so even with the higher voltage requirements of Senn's, +-12Vdc rails is more than enough.

AMB: Yeah, single ended is definitely not very efficient. But I thought the project statement indicated that there was no concern for power supply requirements. Also the target enclosure should have more than enough room. PCB area could be reduced by mounting the power mosfets horizontally off the end of the board. I am not sure if project cost would really be impacted that much. Another way to decrease heat dissipation is to change the bias current AND the voltage rails to match the intended load. For example, Grado's need more current and less voltage versus Senn's. But the overall power requirements of both headphones are roughly the same. Just some food for thought...
 
Aug 23, 2004 at 1:11 AM Post #77 of 565
Aos: Yeah, that is an unacceptable sacrifice, given that some of us have 600 Ohm phones, and we are already losing a fair bit to bias voltage. I for one would like to stick with AD8610 opamps, despite their 24V supply limit. PPL suggested a nifty opamp bootstrapping circuit which creates a sliding voltage window to power the opamps controlled by the output voltage, but that adds quite a bit of complexity.

Cameron: I thought our scheme only loses 1/2 of the total bias, or 4V. If we bias in the middle of the gates, I'm not sure how we would bias the opamp into Class A at the same time. We abandoned that idea a while back, but we could reconsider it.

Amb: I don't think we care much about efficiency.
 
Aug 23, 2004 at 1:33 AM Post #78 of 565
The project statement says that we do not make concessions for battery compatibility since this is an AC only design (and indeed it would be impractical to run a class-A amp on battery). However efficiency is still a consideration given that it directly impacts the amount of heat we have to manage. I am comfortable with dissipating one or two watts on those Aavid/Thermalloy heatsinks we picked, but not much more. This is based on my own experience, having built the prototypes as well as a modified SDS Labs MOSFET amp using the same heatsinks. I don't want the sinks to get burning hot
cool.gif
.

On the topic of max output swing, even with the fact that the op amp's output is biased a few volts above 0, we should still be able to achieve 15Vp-p swing with +/-12V rails with the "good" opamps. IMHO that is quite enough to drive most hi-Z phones to a very high volume.

-Ti
 
Aug 23, 2004 at 2:01 AM Post #80 of 565
Are in-line heat sinks really practical for most DIYers? Is there a source that sells them for a reasonable price independent of a case? Also, at relatively high levels of heat dissipation, what about thermal crosstalk?
 
Aug 23, 2004 at 2:04 AM Post #81 of 565
Morsel: The schematic I showed sits at 2.4Vdc above ground at the opamp output with the IRLIZ24N. Or 3.8Vdc for the IRFZ24N. For the push-pull, you could use two variable voltage drop transistors instead of the one. And connect the opamp between them. Even at the max +-13.2Vdc rails for the AD8610s, there is no way you could get 20Vpp with that 4Vdc offset.

AMB: One or two watts per mosfet is all a single ended topology would need. My latest design sits at 1.2W (12Vdc rails at 100mA for HD650s as target load). The IRLIZ24N has a junction to case thermal resistance of 5.8C/W. So, to keep with the Nelson Pass 50C max temp guideline, a heatsink with a thermal resistance of roughly 15C/W should do it. Here is the one I have been using. Even made a sweet Eagle library for this part if anyone is interested.
 
Aug 23, 2004 at 2:47 AM Post #82 of 565
Quote:

Originally Posted by Wodgy
what about thermal crosstalk?


The temperature isn't going to change once it stabilizes in a class A amp, so thermal crosstalk is not going to be an issue at all. Moreover, the only things that are going to heat up are the MOSFETs and the voltage regulators, not the op amps or other items. We are also going to be using single op amps rather than duals, so there is no "dynamic" thermal crosstalk problem there either.

The voltage swing issue is something Morsel and I had already discussed at length offline. It boils down to basically this, if you want lots of voltage swing, then you need either higher supply rails or a balanced output topology. As I said IMO what we have is perfectly adequete for most headphones unless you want to damage your hearing. We have listened to this setup with Morsel's 600 ohm (?) Beyer DT-990s and we had no trouble getting loud, very loud.

Oh, and just to put things in perspective, the difference between 15Vp-p and 20Vp-p is only 2.5dB. Not really night and day...

For those small fraction of people out there who wants even more SPLs, or those that don't mind a much more complex "cost no object" uber deluxe amp, I have a fully discrete MOSFET amp in the works for the future that will surely address all those concerns. However the M³ as it is (power supply, heat sinking, PCB layout, casing and other topics notwithstanding) should still be an excellent performer for most dynamic phones.

-Ti
 
Aug 23, 2004 at 12:27 PM Post #83 of 565
Morsel, isn't it wiser to have a mosfet as a temp sensing element? With a mosfet you will get a more stable bias. If a BJT is used you will get a heavy bias peak until the amp has stabilized. I noticed that myself in my QRV-01 monster. With a BS170 (in my case) I got a very good stable bias with my IRFD120/9120.
 
Aug 23, 2004 at 2:41 PM Post #84 of 565
I think, it is meaningful approach to consider a ground channel as a viewpoint of power supply rather than that of additional amplifer channel.
Here is my idea of virtual ground power supply with two ground outputs.
If we build this kind of power supply, we can enjoy PI-Amps (PPA Inspired headphone Amps) with almost all existing 2-channel headphone amplifer.
orphsmile.gif


http://www.headphoneamp.co.kr/bbs/vi...sijosae&no=181
 
Aug 23, 2004 at 3:37 PM Post #85 of 565
Cameron: I like your heat sink choice. There are many other heat sinks available. I am not firmly committed to the Aavid 531202. Choosing a less obtrusive profile would open up other layout possibilities.

Peranders: I can't imagine us using temperature sensing as it does not fit the KISS philosophy. Once the amp warms up it should be pretty stable.

Sijosae: I think we want to use the same topology for all 3 channels. We will test M³ in 2 and 3 channel mode.

To input the "³" character, hold down the Alt key and type 0179 on the alternate keypad (Windows), or copy it from Character Map (Windows) or the title of this thread.
 
Aug 23, 2004 at 3:49 PM Post #86 of 565
Quote:

Originally Posted by morsel
Peranders: I can't imagine us using temperature sensing as it does not fit the KISS philosophy. Once the amp warms up it should be pretty stable.


Ok, but since you already have the transistor... and isn't it a bit "engineering science" to have constant bias regardless if the amp is in a cold garage or in a hot listening room?
 
Aug 23, 2004 at 3:53 PM Post #87 of 565
Quote:

Originally Posted by morsel
Cameron: I like your heat sink choice. There are many other heat sinks available. I am not firmly committed to the Aavid 531202. Choosing a less obtrusive profile would open up other layout possibilities.


Morsel, how much heat sinking is actually required (or, how much dissipated power do you expect per device)?


/U.
 
Aug 23, 2004 at 5:31 PM Post #89 of 565
I really don't see the bias current drifting slightly do to thermal variations as any kind of a problem. I have not tested the one shown with the push-pull topology. But the one I showed, the bias is set exclusively by the 25 ohm power resistor. If you get one with very low PPM and oversized to take the power dissipation easily, any drift due to ambient temperature should be very small. Also the rate of change is SO slow that there is no chance for interaction with anything in the audio band. Now, if the bias current was super critical for some reason, absolutely, a current sensing resistor would be almost mandatory.
 
Aug 23, 2004 at 7:07 PM Post #90 of 565
peranders, what are you talking about? Your schematics show BD139 used as VBE multiplier to bias the mosfets, not the BS170! Besides BS170 is a TO-92 device, the only way to thermally couple it would be to glue it to the
heatsink which is a hack.

That amp of yours is truly a monster though.
 

Users who are viewing this thread

Back
Top