But 100mA idle bias gives 200mA peak output current (staying in class A). That is well below the 3500mW at 20 Ohms (600mA peak) that the HPA-1 is rated at.
Terry
Terry
It's 200mA peak per MOSFET, so 400mA peak in total: 0.4*0.4*20 == 3.2W. That's in the ballpark. This can blow up those 20ohm cans 🙂
I'm sure it will knock the headphones right off your head! 😛
I'm just trying to back into the correct bias current based on what I can determine from the specs of the HPA-1. I'll probably reduce the voltage and current to something I can live with.
The specs make it look like it should have 600mA peak current and 11V peak voltage at the output. So 300mA bias per FET and maybe 18 to 20 volt rails looks like a start.
Terry
I'm just trying to back into the correct bias current based on what I can determine from the specs of the HPA-1. I'll probably reduce the voltage and current to something I can live with.
The specs make it look like it should have 600mA peak current and 11V peak voltage at the output. So 300mA bias per FET and maybe 18 to 20 volt rails looks like a start.
Terry
300mA @18V is 5.4W per MOSFET: more than twice as much as 0.1A @24V, thus your heat dissipation requirement will be even bigger. Besides, it will be hard to consider it a HPA-1 clone, I'd say.
I'm trying to understand where the 100mA per FET value came from. It doesn't seem right for the HPA-1 based on the specs for it (unless the power it crosses to class A-B).
Terry
Terry
I went with a 120VA transformer because I'm not skilled enough to figure out what it really needs. (But whatever it is, I'm quite sure it's less than 120VA, so there you go.)
On the bias value, I think I "calculated" that via Newtonian approximation in SPICE. And while SPICE isn't real-world (and Newtonian approximation can perform poorly when presented with non-monotonic curves), I have a hazy memory of Jam confirming that the value I ended up with was pretty close to his....
My case has no ventilation but is somewhat large. It's warm (but not hot) to the touch. It's been on 24/7 for over a year now.
On the bias value, I think I "calculated" that via Newtonian approximation in SPICE. And while SPICE isn't real-world (and Newtonian approximation can perform poorly when presented with non-monotonic curves), I have a hazy memory of Jam confirming that the value I ended up with was pretty close to his....
My case has no ventilation but is somewhat large. It's warm (but not hot) to the touch. It's been on 24/7 for over a year now.
Thanks Jeff,
If the bias is really that low, then it must not stay in class A at full power with low impedance headphones. Based on the specs that are published for it, the bias would need to be about 300mA (to stay in class A) and the rail voltage would need to be sufficient to give 11V peak with low currents (hi-Z phones). Jam mentioned it could drive a tad over 15V peak in pre-amp mode, so I suspect the rail voltage is about 18V to 20V.
I think I'll start there and see where it leads.
18V and 300mA gives 5.4W per device. The heat sink I'm looking at is 3.7C/W so about 20 C rise over ambient. The question will be how high the internal box temperature goes. There will probably be 30 Watts in the box. Vents in the top and bottom may be required.
Terry
If the bias is really that low, then it must not stay in class A at full power with low impedance headphones. Based on the specs that are published for it, the bias would need to be about 300mA (to stay in class A) and the rail voltage would need to be sufficient to give 11V peak with low currents (hi-Z phones). Jam mentioned it could drive a tad over 15V peak in pre-amp mode, so I suspect the rail voltage is about 18V to 20V.
I think I'll start there and see where it leads.
18V and 300mA gives 5.4W per device. The heat sink I'm looking at is 3.7C/W so about 20 C rise over ambient. The question will be how high the internal box temperature goes. There will probably be 30 Watts in the box. Vents in the top and bottom may be required.
Terry
If I had my case, I'd put a power resistor in it and run it at 30 Watts for a few hours to see how much heat rise I get in the case. But it's on a slow boat from China (literally), so I'm proceeding as if it will be workable. What could go wrong? 😕
Rails are 24V.
The bias I'm using is indeed insufficient to drive 3.5W into 20ohms in class A. So either the full power specs are for class A/B, or I need to increase my bias.
But even at the 23ohms of my Ether Flows it needs more than 2V input to get out of class A, which my source can't provide anyway.
The bias I'm using is indeed insufficient to drive 3.5W into 20ohms in class A. So either the full power specs are for class A/B, or I need to increase my bias.
But even at the 23ohms of my Ether Flows it needs more than 2V input to get out of class A, which my source can't provide anyway.
Do you know that the HPA-1 used 24 volts rails? That seems high based on the specs and the statement Jam made about the 30V p-p output level for the pre-amp mode.
Terry
Terry
Spice Runs
I simulated the circuit in PSpice. I don't have a large confidence in the device models. I just went with what I found on the web, with no attempt to validate them against the device curves.
But it shows that running the bias at about 100mA results in a bit lower distortion than running it at 300mA. That kind of surprised me. I assumed the higher bias would be better. I am running it with a 60 Ohm load (since I have AKG K701 headphones).
I also tried changing the power supply rails to +/-18V. This resulted in a bit higher distortion. I didn't spend any time adjusting component values to compensate for the lower supply rails. As expected, the clipping is lower (about +/-15.4 volts at 24V rails and +/- 9.21V at 18V rails). It also lost a little bit of bandwidth with the lower rails.
My headphones max out at 80mA peak, so there isn't any reason for me to run any higher than the 100mA bias. So maybe staying with 24 volt rails and 100mA bias is the way to go after all. I may still have a heat problem in the small enclosure I'm trying to use, but if I do, I could still drop the rail voltage without losing much ( I need less than about 5 volts peak swing for my phones).
Of course I won't know how accurate the Spice sims are until I build some hardware. I'll probably use a 50VA transformer. That still leaves margin.
Terry
I simulated the circuit in PSpice. I don't have a large confidence in the device models. I just went with what I found on the web, with no attempt to validate them against the device curves.
But it shows that running the bias at about 100mA results in a bit lower distortion than running it at 300mA. That kind of surprised me. I assumed the higher bias would be better. I am running it with a 60 Ohm load (since I have AKG K701 headphones).
I also tried changing the power supply rails to +/-18V. This resulted in a bit higher distortion. I didn't spend any time adjusting component values to compensate for the lower supply rails. As expected, the clipping is lower (about +/-15.4 volts at 24V rails and +/- 9.21V at 18V rails). It also lost a little bit of bandwidth with the lower rails.
My headphones max out at 80mA peak, so there isn't any reason for me to run any higher than the 100mA bias. So maybe staying with 24 volt rails and 100mA bias is the way to go after all. I may still have a heat problem in the small enclosure I'm trying to use, but if I do, I could still drop the rail voltage without losing much ( I need less than about 5 volts peak swing for my phones).
Of course I won't know how accurate the Spice sims are until I build some hardware. I'll probably use a 50VA transformer. That still leaves margin.
Terry
@Jeff, in post #57 I can't quite tell but did you deliberately find a 470uF with the same case dimensions as the 220uF to make C61 and C62 look symmetric or is it actually an error on the schematic and C61 is actually supposed to be 220uF as well? Any harm if I make both of them 220uF or make both of them 470?
While I have your attention, does the value for R91 (0R82) matter if I substitute with something close like 1R0?
He he... yes, I have OCD. C61 is a 25V cap and C62 is a 50V cap so that they're the same size.
R91 is there to modify the ESR of C62 (or maybe the time constant?). It's from some app note or paper that I read, but I can't seem to find it now....
In any case, I think the LM7812 is actually pretty stable and none of the values matter much.
R91 is there to modify the ESR of C62 (or maybe the time constant?). It's from some app note or paper that I read, but I can't seem to find it now....
In any case, I think the LM7812 is actually pretty stable and none of the values matter much.
Last edited:
How would you test a fully populated board without the button board? Is it just a momentary connection from the button pad to +12V. The appropriate button pad of course 🙂
Speaking hypothetically of course 😉
Speaking hypothetically of course 😉
