KurtW
500+ Head-Fier
- Joined
- Jun 26, 2001
- Posts
- 971
- Likes
- 13
I have been working for several months now on modifications to the Meier Audio Corda HA-1 headphone amp. This is already an excellent headphone amp for the price, but since I had built mine from a kit and Jan Meier has published articles about the design including schematics, it seemed like a likely candidate for upgrades. My initial changes of substituting the integrated circuits produced a nice improvement but it was not as great as I expected. After extensive modifications and comparisons to a stock Corda was well as a Sugden HeadMaster, I now feel satisfied with the modifications.
I will list the modifications roughly in the order of sonic impact, from most to least.
1. Integrated circuit swap. The National LM6171 has great specs but compared to the higher end HeadRoom amps that use the Burr Brown OPA627, the Corda seemed a little harsher and not as open and airy. It should be noted that this obviously isn’t the only difference between the Corda and HeadRoom amps, but the OPA627, although quite expensive at over $15 each, has a good reputation for very good sound. I swapped out the National chips for the OPA627 in the first stage, taking account that although they are close to pin compatible, the PCB uses some No Connect pins on the 6171 for other routing, so you have to lift up pins 1 and 5 on the 627. This swap did take the sound in the direction that I wanted it to go. The 627 isn’t well suited for driving headphones directly though, but the BB BUF634 is an excellent solution for swapping in the output stage, and its cheaper too. Again, pin 1 needs to be lifted. The BUF634 does have a larger output offset voltage, but on the two units I’ve done so far it was under 60mV which doesn’t seem to cause any problems. So far we have a solution that requires no soldering, but the 627s run a bit hot. The Corda design has a resistor from the output to V+ to put the op amp into Class A operation. This provides 10mA of bias current at idle which is way more than is needed. It’s a bit too much for the 627 and causes the chip to run hotter than I’d like to see, as well as being a fairly big load for a chip rated at 45mA output. Changing the 1.5k resistor to 5k still provides plenty of bias current to keep it in class A and solves the heat problem. Because you loose the gain in the output stage, the overall gain of the 627/637 combo is a bit lower but not a problem for many headphones. The gain can be adjusted up, and is addressed below.
2. Improved Current Source. A better although much more complicated solution to the class A bias in the first stage is to use a current source. With the stock Corda a 2 volt output signal at this stage you give a bias change of 13%. I used a cascode FET current source which regulates the current to well within 1% over this same range. There has been quite a bit posted about these already, so I’ll just comment that I used a 2N5486 and a 2N5484 with a 100 ohm isolation resistor, and selected for 3mA current. The other change is that this is connect to V- instead of V+, taking advantage of the faster N channel transistors in the output stage of the IC. I was quite surprised what a difference this made. The sound seem to open up and become much more lively.
3. Output stage improvements. Particularly with low impedance headphones, the current requred to keep the output stage biased in class A is too high to use the above current sources. However, the BUF634 has a wide band mode that sets the internal bias currents much higher. This can be implemented by connecting pin 1, which is already hanging in the air, to V-. The original design has a 1.5k ohm resistor going to V-, so once this is removed you can use this V- connection for pin 1 of the 634. The penalty you pay for this is heat. I added 16 pin dip heat sinks to the top of the 634s which weren’t needed until the wide band mode was activated. I made a couple of other changes to beef up the bass impact when using low impedance headphones. One was to double up on the 634 buffers, which just entails soldering another buffer on top of the first one. The second change was to change the 47 ohm output resistor to 10 ohms. Each of these changes seem to beef up the bass slam. I was a bit worried about stability but after checking it out with a scope, using a bunch of different headphones connected through a 6 foot extension cable made from Kimber wire, I couldn’t detect any ringing or stability problems.
4. OPA637 vs OPA627. The OPA637 is a higher performance version of the 627, with over double the slew rate and much faster settling time, but according to the datasheet it is not stable at gains less than 5. The Corda has the gain of first stage running around 3, but it can be set to a gain of 5 by changing the 4.7k resistor to a 2.5k for each channel. I used a 2.74k resistor in mine resulting in a gain of 4.6 and it works fine. With this value the gain is similar to the stock Corda, compensating for the gain lost in the output stage when switching to the BUF634. If you look at the graph in the data sheet you can see a gain of 1 is definitely trouble but anywhere near 5 should be okay. The 637’s improvement is like everything that is good about the 627 is just kicked up another notch. They are priced the same, so unless you already have the 627s or don’t want to make the gain change, this is the way to go.
5. Schottky Diodes. The Corda has inductors in the power supply to filter out the switching noise caused by the silicon diode bridge, but I feel the change to schottky diodes caused an improvement in imaging. There are many schottky diode options, but be sure to use one in the DO-41 format as the space is tight and the pcb holes aren’t too big. I used the fairly common NTE585.
6. Capacitor changes. The Corda uses Wima polyester box capacitors in the crossfeed circuit (the red ones). Wima makes polypropylene caps (which are green) with the same pin spacing and are only slightly bigger otherwise. It seems to be well established that polypropylene caps sound better than polyester caps, so I swapped out all of these. Although I didn’t spend a lot of time trying to verify the sonic benefits of this, I’ll have to say they weren’t real obvious. Nevertheless, I feel better about having the polypropylene caps in there. I also added a pair of 1.0uF polyester caps for the buffer decoupling on the back side of the board. I had wanted to use an even larger value but there was not enough room to do so and still keep the leads short. My Corda (an early model) came with 2200uF electrolytic caps in the power supply, but this was later changed to 3300uF. I couldn’t hear or measure any difference in performance. I didn’t try any "designer label" caps anywhere, as there is not a lot of room unless you start spreading things out, and I’m not sold on the idea that they provide a benefit worth the cost.
7. Input RCA jacks. When I was evaluating many changes with my Corda I eventually broke the internal ground connection on the stock RCA jacks. This probably wouldn’t happen under normal use with the cover on, but lots of plugging and unplugging with the cover off put a lot of stress on these and they finally gave. I found that the Cardas GRFA jacks fit nicely in the same holes, you only need to trim the insulating washers slightly. These are very nice jacks and are more fitting to the sound of the modified Corda. If you want to use another jack, be aware that the Cardas were a tight fit, and the jack needs to be insulated from the metal case.
8. The last change I made was to change the red LED to a clear lens blue one, and to adjust the resistor a bit since the blue one was really bright. There were so many changes inside, I felt one small change to the outside would be appropriate. I now have dubbed my unit "Corda Blue".
So, how does it sound? In direct comparisons to a stock Corda, using identical interconnects and power cords and using a Linn Ikemi as a source with various headphones, there is a very noticeable difference. The modified Corda is much more refined sounding, more spacious, more dynamic and better defined from top to bottom. It also seems to have much better rhythm and pace. I also compared it to the Sugden HeadMaster, which is the best solid state headphone amp I’ve heard. The Sugden still has a transparency and three dimensional airiness that even the modified Corda can’t quite match. Of course it’s bigger, much more expensive and doesn’t have crossfeed.
I did make some measurements of the two Cordas, and didn’t find a difference within my modest measurement limits for distortion and noise. Channel matching on either unit was excellent, as was frequency response.
The total parts cost for all of these mods is around $100. I took around 6 hours to implement them. This seems like a lot, but I look my time to be careful and to implement them in a fairly neat and compact way. One issue I had with a factory assembled unit was that the crossfeed switch was held to the front panel by a stator nut. Jan was kind enough to tell me that Farnell sells a stator-nut key (p/n 321-357) for this, but Farnell didn’t answer my email and I don’t know any source in the US to obtain this. If anyone does, please let me know.
Acknowledgements:
I would first like to thank Jan Meier for making the Corda such a solid design at a great price, publishing the design and being very responsive to everyone’s questions. None of these mods are my original ideas, so I’d like to thank everyone who has shared their knowledge about these things on the internet. Special thanks go to Kelly who loaned me his factory built Corda for comparisons, Chris Own whom I exchanged many emails with while we were both trying mods, and Head-Fi-wiser ppl whom I’ve never had any correspondence with but whom I have learned a lot from reading his many posts.
I will list the modifications roughly in the order of sonic impact, from most to least.
1. Integrated circuit swap. The National LM6171 has great specs but compared to the higher end HeadRoom amps that use the Burr Brown OPA627, the Corda seemed a little harsher and not as open and airy. It should be noted that this obviously isn’t the only difference between the Corda and HeadRoom amps, but the OPA627, although quite expensive at over $15 each, has a good reputation for very good sound. I swapped out the National chips for the OPA627 in the first stage, taking account that although they are close to pin compatible, the PCB uses some No Connect pins on the 6171 for other routing, so you have to lift up pins 1 and 5 on the 627. This swap did take the sound in the direction that I wanted it to go. The 627 isn’t well suited for driving headphones directly though, but the BB BUF634 is an excellent solution for swapping in the output stage, and its cheaper too. Again, pin 1 needs to be lifted. The BUF634 does have a larger output offset voltage, but on the two units I’ve done so far it was under 60mV which doesn’t seem to cause any problems. So far we have a solution that requires no soldering, but the 627s run a bit hot. The Corda design has a resistor from the output to V+ to put the op amp into Class A operation. This provides 10mA of bias current at idle which is way more than is needed. It’s a bit too much for the 627 and causes the chip to run hotter than I’d like to see, as well as being a fairly big load for a chip rated at 45mA output. Changing the 1.5k resistor to 5k still provides plenty of bias current to keep it in class A and solves the heat problem. Because you loose the gain in the output stage, the overall gain of the 627/637 combo is a bit lower but not a problem for many headphones. The gain can be adjusted up, and is addressed below.
2. Improved Current Source. A better although much more complicated solution to the class A bias in the first stage is to use a current source. With the stock Corda a 2 volt output signal at this stage you give a bias change of 13%. I used a cascode FET current source which regulates the current to well within 1% over this same range. There has been quite a bit posted about these already, so I’ll just comment that I used a 2N5486 and a 2N5484 with a 100 ohm isolation resistor, and selected for 3mA current. The other change is that this is connect to V- instead of V+, taking advantage of the faster N channel transistors in the output stage of the IC. I was quite surprised what a difference this made. The sound seem to open up and become much more lively.
3. Output stage improvements. Particularly with low impedance headphones, the current requred to keep the output stage biased in class A is too high to use the above current sources. However, the BUF634 has a wide band mode that sets the internal bias currents much higher. This can be implemented by connecting pin 1, which is already hanging in the air, to V-. The original design has a 1.5k ohm resistor going to V-, so once this is removed you can use this V- connection for pin 1 of the 634. The penalty you pay for this is heat. I added 16 pin dip heat sinks to the top of the 634s which weren’t needed until the wide band mode was activated. I made a couple of other changes to beef up the bass impact when using low impedance headphones. One was to double up on the 634 buffers, which just entails soldering another buffer on top of the first one. The second change was to change the 47 ohm output resistor to 10 ohms. Each of these changes seem to beef up the bass slam. I was a bit worried about stability but after checking it out with a scope, using a bunch of different headphones connected through a 6 foot extension cable made from Kimber wire, I couldn’t detect any ringing or stability problems.
4. OPA637 vs OPA627. The OPA637 is a higher performance version of the 627, with over double the slew rate and much faster settling time, but according to the datasheet it is not stable at gains less than 5. The Corda has the gain of first stage running around 3, but it can be set to a gain of 5 by changing the 4.7k resistor to a 2.5k for each channel. I used a 2.74k resistor in mine resulting in a gain of 4.6 and it works fine. With this value the gain is similar to the stock Corda, compensating for the gain lost in the output stage when switching to the BUF634. If you look at the graph in the data sheet you can see a gain of 1 is definitely trouble but anywhere near 5 should be okay. The 637’s improvement is like everything that is good about the 627 is just kicked up another notch. They are priced the same, so unless you already have the 627s or don’t want to make the gain change, this is the way to go.
5. Schottky Diodes. The Corda has inductors in the power supply to filter out the switching noise caused by the silicon diode bridge, but I feel the change to schottky diodes caused an improvement in imaging. There are many schottky diode options, but be sure to use one in the DO-41 format as the space is tight and the pcb holes aren’t too big. I used the fairly common NTE585.
6. Capacitor changes. The Corda uses Wima polyester box capacitors in the crossfeed circuit (the red ones). Wima makes polypropylene caps (which are green) with the same pin spacing and are only slightly bigger otherwise. It seems to be well established that polypropylene caps sound better than polyester caps, so I swapped out all of these. Although I didn’t spend a lot of time trying to verify the sonic benefits of this, I’ll have to say they weren’t real obvious. Nevertheless, I feel better about having the polypropylene caps in there. I also added a pair of 1.0uF polyester caps for the buffer decoupling on the back side of the board. I had wanted to use an even larger value but there was not enough room to do so and still keep the leads short. My Corda (an early model) came with 2200uF electrolytic caps in the power supply, but this was later changed to 3300uF. I couldn’t hear or measure any difference in performance. I didn’t try any "designer label" caps anywhere, as there is not a lot of room unless you start spreading things out, and I’m not sold on the idea that they provide a benefit worth the cost.
7. Input RCA jacks. When I was evaluating many changes with my Corda I eventually broke the internal ground connection on the stock RCA jacks. This probably wouldn’t happen under normal use with the cover on, but lots of plugging and unplugging with the cover off put a lot of stress on these and they finally gave. I found that the Cardas GRFA jacks fit nicely in the same holes, you only need to trim the insulating washers slightly. These are very nice jacks and are more fitting to the sound of the modified Corda. If you want to use another jack, be aware that the Cardas were a tight fit, and the jack needs to be insulated from the metal case.
8. The last change I made was to change the red LED to a clear lens blue one, and to adjust the resistor a bit since the blue one was really bright. There were so many changes inside, I felt one small change to the outside would be appropriate. I now have dubbed my unit "Corda Blue".
So, how does it sound? In direct comparisons to a stock Corda, using identical interconnects and power cords and using a Linn Ikemi as a source with various headphones, there is a very noticeable difference. The modified Corda is much more refined sounding, more spacious, more dynamic and better defined from top to bottom. It also seems to have much better rhythm and pace. I also compared it to the Sugden HeadMaster, which is the best solid state headphone amp I’ve heard. The Sugden still has a transparency and three dimensional airiness that even the modified Corda can’t quite match. Of course it’s bigger, much more expensive and doesn’t have crossfeed.
I did make some measurements of the two Cordas, and didn’t find a difference within my modest measurement limits for distortion and noise. Channel matching on either unit was excellent, as was frequency response.
The total parts cost for all of these mods is around $100. I took around 6 hours to implement them. This seems like a lot, but I look my time to be careful and to implement them in a fairly neat and compact way. One issue I had with a factory assembled unit was that the crossfeed switch was held to the front panel by a stator nut. Jan was kind enough to tell me that Farnell sells a stator-nut key (p/n 321-357) for this, but Farnell didn’t answer my email and I don’t know any source in the US to obtain this. If anyone does, please let me know.
Acknowledgements:
I would first like to thank Jan Meier for making the Corda such a solid design at a great price, publishing the design and being very responsive to everyone’s questions. None of these mods are my original ideas, so I’d like to thank everyone who has shared their knowledge about these things on the internet. Special thanks go to Kelly who loaned me his factory built Corda for comparisons, Chris Own whom I exchanged many emails with while we were both trying mods, and Head-Fi-wiser ppl whom I’ve never had any correspondence with but whom I have learned a lot from reading his many posts.