I'm new to the boards and had a question about power supplies, but am not sure if my question should go here or in the Headphones section.
I have a 10+ year old Antique Sound Lab headphone amp that was a prototype of which only about 8 were made: the MG-Head OTL-H. I purchased it off of the Head-fi boards years ago and have had one lingering problem with it: it clips and distorts when the volume is turned up about halfway. There aren't any output transformers and the owner of ASL admitted that it has a weak power supply. Here is some background reading on what I'm dealing with.
A toast to Joe Lau!!!
Help me fix MG Head distortion!
Now, I have measured the voltage at Pin 3 on each 6BQ5 and it's reading about 125VDC. There also seems to be a JFET diode cascode CCS tacked in there, supplying power to the anodes. It's tough to tell what's going on as I can't really get the board out.
How can I improve or rebuild the power supply to provide adequate power to this 6BQ5 OTL design so that there isn't any more clipping?
Thanks in advance,
Dan
I have a 10+ year old Antique Sound Lab headphone amp that was a prototype of which only about 8 were made: the MG-Head OTL-H. I purchased it off of the Head-fi boards years ago and have had one lingering problem with it: it clips and distorts when the volume is turned up about halfway. There aren't any output transformers and the owner of ASL admitted that it has a weak power supply. Here is some background reading on what I'm dealing with.
A toast to Joe Lau!!!
Help me fix MG Head distortion!
Now, I have measured the voltage at Pin 3 on each 6BQ5 and it's reading about 125VDC. There also seems to be a JFET diode cascode CCS tacked in there, supplying power to the anodes. It's tough to tell what's going on as I can't really get the board out.
How can I improve or rebuild the power supply to provide adequate power to this 6BQ5 OTL design so that there isn't any more clipping?
Thanks in advance,
Dan
After more investigating, this amp seems more and more strange.
I have discovered that there are 2 DN2540 MOSFETs inside this amp. There is one DN2540 for each EL84, going to pins 7 and 9. The Gate and Source pins on each unit are strapped together with a 47Ohm resistor, then going out to the plate. I'm not very familiar with this type of architecture. Is this a half-hearted or just a simple CCS?
Would a cascode CCS like the K&K kit work better in this instance?
I have discovered that there are 2 DN2540 MOSFETs inside this amp. There is one DN2540 for each EL84, going to pins 7 and 9. The Gate and Source pins on each unit are strapped together with a 47Ohm resistor, then going out to the plate. I'm not very familiar with this type of architecture. Is this a half-hearted or just a simple CCS?
Would a cascode CCS like the K&K kit work better in this instance?
That's exactly what's going on. There is a slight variation between each CCS that seems to be causing a channel imbalance. If I replaced both with a single cascode CCS, giving me the same B+ for both EL84s, would that give me a more balanced sounding output?
On a different note, what about replacing my whole power supply and keeping a CCS to get a higher B+ on the plate? Right now it's at about 120VDC. Would bringing it up to around 160 help get rid of the distortion?
On a different note, what about replacing my whole power supply and keeping a CCS to get a higher B+ on the plate? Right now it's at about 120VDC. Would bringing it up to around 160 help get rid of the distortion?
Yes- They are all tied together at the same point for each output. Sorry, I'm very new to trying to understand what's going on with tube design. Thanks for helping.
The channel imbalance is probably due to the tubes more than the ccs.
It would help to know more... What is the role of the input tube ? Is there some sort of global feedback ? What's the total gain ? What headphones are you using it with ? What is the voltage of B+ now ?
It would help to know more... What is the role of the input tube ? Is there some sort of global feedback ? What's the total gain ? What headphones are you using it with ? What is the voltage of B+ now ?
Then, the mosfet aren't really used as ccs. It's more an hybrid µ-follower, to get a decently low output impedance.
Last edited:
00940,
Thanks for the advice so far. Here is what I know about this amp and how I'm using it. I have attached the schematic of the headphone amp that my unit is based on. The only apparent difference I can tell is that the output transformers have been replaced by this MOSFET board with the DN2540's.
I know that the channel imbalance is not due to the tubes, as I have swapped out the EL84's and the 12ax7 to multiple variations with the same result. I have 4 or 5 pair of EL84's and 5 or 6 driver tubes that I have tried.
1. The role of the input tube seems to boost gain. I have tried numerous 12ax7 and 5751 tubes in this position.
2. I'm not sure what the total gain is for the design. How would I determine this?
3. I'm using Sennheiser HD-580's with this, so their impedance is 300 ohm.
4. The voltage of the B+ is 164.4 VDC.
Since I'm learning about all of this, how should I proceed? I'm trying to figure out how to simulate this in SPICE, but I'm still climbing up the learning curve.
Thanks in advance for any help.
Thanks for the advice so far. Here is what I know about this amp and how I'm using it. I have attached the schematic of the headphone amp that my unit is based on. The only apparent difference I can tell is that the output transformers have been replaced by this MOSFET board with the DN2540's.
I know that the channel imbalance is not due to the tubes, as I have swapped out the EL84's and the 12ax7 to multiple variations with the same result. I have 4 or 5 pair of EL84's and 5 or 6 driver tubes that I have tried.
1. The role of the input tube seems to boost gain. I have tried numerous 12ax7 and 5751 tubes in this position.
2. I'm not sure what the total gain is for the design. How would I determine this?
3. I'm using Sennheiser HD-580's with this, so their impedance is 300 ohm.
4. The voltage of the B+ is 164.4 VDC.
Since I'm learning about all of this, how should I proceed? I'm trying to figure out how to simulate this in SPICE, but I'm still climbing up the learning curve.
Thanks in advance for any help.
Attachments
Hi,
if I'm not mistaken on some values, going from the picture and the schematic, here's what you've got (trying to respect the names from the schematic):
I calculated a new value for the el84 cathode resistor (r11). This gives about 125V on the tube's plate, as measured.
With a 12ax7 at the input and no-feedback (feedback resistor is missing), gain is crazy high. With a 0.1Vrms input (after the pot), we're already at 5Vrms. I'm surprised you can even turn the pot halfway (the HD580 are very loud at 2Vrms).
Output impedance is also huge (I mean, since the cathode resistor is apparently unbypassed, we're at 3K).
Still... sims show it should work without crazy distortion up to 2Vrms (1% thd point). I assume the 164V B+ is what you get from the external supply ? Is it a measure under load ? Consumption being constant (due to the CCS), weak supply or not, it shouldn't make a difference. If it's unloaded, then we might have a problem.
if I'm not mistaken on some values, going from the picture and the schematic, here's what you've got (trying to respect the names from the schematic):
I calculated a new value for the el84 cathode resistor (r11). This gives about 125V on the tube's plate, as measured.
With a 12ax7 at the input and no-feedback (feedback resistor is missing), gain is crazy high. With a 0.1Vrms input (after the pot), we're already at 5Vrms. I'm surprised you can even turn the pot halfway (the HD580 are very loud at 2Vrms).
Output impedance is also huge (I mean, since the cathode resistor is apparently unbypassed, we're at 3K).
Still... sims show it should work without crazy distortion up to 2Vrms (1% thd point). I assume the 164V B+ is what you get from the external supply ? Is it a measure under load ? Consumption being constant (due to the CCS), weak supply or not, it shouldn't make a difference. If it's unloaded, then we might have a problem.
Attachments
Ben,
I made the B+ measurements from the external supply when it's unloaded. And you're right with there being about 125V on the plate.
Also, you're right about the distortion being when the volume gets to almost unbearable levels.
So, where do I go from here? Why is the 164V B+ a problem unloaded?
I made the B+ measurements from the external supply when it's unloaded. And you're right with there being about 125V on the plate.
Also, you're right about the distortion being when the volume gets to almost unbearable levels.
So, where do I go from here? Why is the 164V B+ a problem unloaded?
126V-125V= 1V for the CCS ? That doesn't seem possible...
edit: about the unloaded thing. If the PS is weak, it will collapse under load. And the CCS won't have enough voltage to operate in between the EL84's plate and B+
edit2: do you confirm having 126V at the red point and 125V at the orange one ? With no music playing ?
edit: about the unloaded thing. If the PS is weak, it will collapse under load. And the CCS won't have enough voltage to operate in between the EL84's plate and B+
edit2: do you confirm having 126V at the red point and 125V at the orange one ? With no music playing ?
Attachments
Last edited:
Yes, that's correct. I just verified that this is the case. Only one volt is dropping with the CCS. I unplugged the input to prevent any shorting issues with my source.
Ben, could you post the SPICe file for this? I'm trying to recreate what I have here and it's taking me forever.
What we want is at least 25 in between B+ and the EL84 anode.
There are a few different ways to proceed.
1. Change the power supply for something that will not sag under load. It means we have to change the transformer and probably the whole PS.
2. Reduce the current through the ccs. B+ will raise a bit with the lighter load. With the same cathode resistor, the anode voltage will also go down a little bit as the biasing voltage goes down. The best way would be to add a pair of 50R trimmers in serie with the two 47r resistors in the CCS. Dial the current down until we have 20V in between B+ and anode.
3. Decrease the cathode resistor to change the el84 operating point and lower the anode voltage a bit. Could be done with a 33r resistor+ 50r trimmer.
Best in my view is option 2. Reducing the current by a few ma should be sufficient and not a problem into high impedance load.
All this being said, I would do option 2 and them mod the amp. Moving the output from the anode to the source of the mosfet (as a µ-follower) and re-establishing feedback would do wonders. See attached schematic and the ltspice file.
I attached the ltspice file, you'll need Koren_Tubes.cir
Attachments
Last edited:
That doesn't seem to be the problem here. I can't find an accurate schemo for this design. You mention using 300R 'phones.
The A Number One problem is that VTs are Hi-Z devices. A 300R load is a nearly vertical loadline. It doesn't matter if the 6BQ5s are being used as cathode followers, or grounded cathode amps, it's the same loadline. Nor does it matter if the ZO of a cathode follower is ~300R. The steep loadline places a severe constraint on the output voltage swing of that heavily loaded output stage. A cathode follower is useful for providing a Lo-Z source, but only to a Hi-Z load. It still runs out of "breathe" if heavily loaded. Nothing you can do about changing the fundamental nature of the beast. It's a high voltage, low current, Hi-Z device. Period.
This is the biggest problem when it comes to hollow state OTLs: getting the amps a relatively Lo-Z load needs. A single 6BQ5 stage isn't gonna produce very much. There aren't all that many types that can source the currents required, and the 6BQ5 ain't one of 'em
It's likely that the final is running out of headroom into a severe load. The only thing that can fix it is a redesign into an OPT coupled stage that presents the 6BQ5s with a friendlier, Hi-Z plate load.
A "stronger" power supply isn't gonna change that, and it looks like the original designer didn't fully understand how VTs actually work.
I posted a very likely schematic earlier on.
I agree with all you wrote but since we have a pair of dn2540 to play with, we can redesign the thing into something useful (see post above) for high impedance headphones.
Ben, is what you're proposing the basic CCS cascode that Morgan Jones describes in his books? This option sounds a lot more promising than adding OPTs.
What about removing that little board with the DN2540's and putting in the K&K Cascode CCS kit? I could then drop the voltage more in a more stable design?
K & K Audio - Lundahl Transformers, audio DIY kits and more
What about removing that little board with the DN2540's and putting in the K&K Cascode CCS kit? I could then drop the voltage more in a more stable design?
K & K Audio - Lundahl Transformers, audio DIY kits and more
The K&K CCS require about the same amount of voltage to properly operate. They're better sure, but you should first fix your amplifier. Currently, with 1V across the CCS, it cannot operate as (badly) designed.
OK. So put a 50R trimpot in series with the 47r resistor and bring the voltage drop up to ~20V? That doesn't sound too hard to start with.
Would I use a single DN2540 per I'd love to really fix this amp up or use what I have to learn and build upon to make something better in the future. Thanks so much for all of your help so far.
-Dan
Would I use a single DN2540 per I'd love to really fix this amp up or use what I have to learn and build upon to make something better in the future. Thanks so much for all of your help so far.
-Dan
Last edited:
Ben,
Looking at CCS designs and calculations, will a simple 50 ohm trimpot in series with the 47r resistor really drop enough voltage? Wouldn't it be better to redesign these 2 DN2540 FETs into an actual cascode CCS and have both anodes share the same power supply?
I'm beginning to think that the channel imbalance is due to the fact that there is no Rset for each DN2540 and that the variation in manufacturing is giving me different anode voltages on each EL84. Is that a possibility?
-Dan
Looking at CCS designs and calculations, will a simple 50 ohm trimpot in series with the 47r resistor really drop enough voltage? Wouldn't it be better to redesign these 2 DN2540 FETs into an actual cascode CCS and have both anodes share the same power supply?
I'm beginning to think that the channel imbalance is due to the fact that there is no Rset for each DN2540 and that the variation in manufacturing is giving me different anode voltages on each EL84. Is that a possibility?
-Dan
- Status
- Not open for further replies.