Of course, the lower value resistor injects less noise into the circuit. 10K ALPS RK27 pot ordered.
I completed the kit. Here is my opinion in mini review
Just one question - 5mV output voltage drift is OK, or I should to play with the trim pots some more ?
In short: I very satisfied with the sound of that amp. It is one of the very best SS amps I heard. Who looking for true hi end headphone amp - this one is strongly recommended!
I completed the kit. Here is my opinion in mini review
Just one question - 5mV output voltage drift is OK, or I should to play with the trim pots some more ?
In short: I very satisfied with the sound of that amp. It is one of the very best SS amps I heard. Who looking for true hi end headphone amp - this one is strongly recommended!
Last edited:
Another build
Have recently finished a Marsh amp using jackinnj's pcb boards. Been working on it for a few months now. Ended up with a power supply based on inexpensive Antek 10VA transformer, bridge, big caps, and amp. Used MPSA06/A56 and KSA1220/KSC2690 transistors. The MPSA06/A56 can be easily matched to within 10%, and at 50ma Id the KSA/KSC even closer. This is for complementary pairs, not same-polarity pairs. I'm not sure why the KSA/KSC parts don't get more press, I've experimented with several different flavors of medium-power parts and KSA/KSC parts can be very closely matched, a fundamental design criteria for the Marsh design.
My initial experiments were with an O2, chip-amps, and then with a Doug Self Blameless Class A design. The O2 was good but not great, chip amps not to my liking, and I could not finesse the Blameless to sound like I thought it should. In retrospect, either a layout issue or player interface issue. Probably the former.
The Marsh is the total opposite: I think I've reached the pinnacle of what a hobbyist can do without an AP or spectrum analyzer. I have a few sets of headphones ranging from 16R Momentum2 to 600R DT880, and I've listened to all of them with this amp. But this build really shines with a non-regulated supply, DT770Rs, and jackinnj's boards with closely matched KSA/KSC pairs. The highs are silky smooth, mids are clear, lows powerful enough to embed shards of ubendium into your eardrums. I started with a regulated supply, but when I changed to simple non-regulated supply... wow!
Anyway, I hope that maybe some of this info will help the next guy find his pot of gold. Thanks to Doug Self for teaching me what little I know, thanks for Richard Marsh for sharing his design, and thanks for jackinnj for making the boards available.
Best regards.
Have recently finished a Marsh amp using jackinnj's pcb boards. Been working on it for a few months now. Ended up with a power supply based on inexpensive Antek 10VA transformer, bridge, big caps, and amp. Used MPSA06/A56 and KSA1220/KSC2690 transistors. The MPSA06/A56 can be easily matched to within 10%, and at 50ma Id the KSA/KSC even closer. This is for complementary pairs, not same-polarity pairs. I'm not sure why the KSA/KSC parts don't get more press, I've experimented with several different flavors of medium-power parts and KSA/KSC parts can be very closely matched, a fundamental design criteria for the Marsh design.
My initial experiments were with an O2, chip-amps, and then with a Doug Self Blameless Class A design. The O2 was good but not great, chip amps not to my liking, and I could not finesse the Blameless to sound like I thought it should. In retrospect, either a layout issue or player interface issue. Probably the former.
The Marsh is the total opposite: I think I've reached the pinnacle of what a hobbyist can do without an AP or spectrum analyzer. I have a few sets of headphones ranging from 16R Momentum2 to 600R DT880, and I've listened to all of them with this amp. But this build really shines with a non-regulated supply, DT770Rs, and jackinnj's boards with closely matched KSA/KSC pairs. The highs are silky smooth, mids are clear, lows powerful enough to embed shards of ubendium into your eardrums. I started with a regulated supply, but when I changed to simple non-regulated supply... wow!
Anyway, I hope that maybe some of this info will help the next guy find his pot of gold. Thanks to Doug Self for teaching me what little I know, thanks for Richard Marsh for sharing his design, and thanks for jackinnj for making the boards available.
Best regards.
I have not built or heard all the headphone amps here and else where. So, it is great to get comparisons from others experiences with specific HP Amps with their headphones. This variety is helpful to determine if the sound quality and design is consistant over a wide range.
Thx-RNMarsh
.
Thx-RNMarsh
.
I have not built or heard all the headphone amps here and else where. So, it is great to get comparisons from others experiences with specific HP Amps with their headphones. This variety is helpful to determine if the sound quality and design is consistant over a wide range.
Thx-RNMarsh
.
Thx-RNMarsh
.
I did not know the schematic. But I tried make simulation and modified it for low THD.
Great testimony! Would you be willing to share your power supply schematic?
Power Supply
Sure. It's so simple it's embarrassing. And then again maybe that's the magic.
If anybody is interested, send PM and I'll send entire KiCad project, including gerbers.
Sure. It's so simple it's embarrassing. And then again maybe that's the magic.
If anybody is interested, send PM and I'll send entire KiCad project, including gerbers.
Just a suggestion -- use C1 and C2 as a snubber in conjuction with the resistor I show (150R) -- will tame ringing from the diode/transformer. The values shown are those Mark Johnson used in a recent article for LinearAudio.
Putting 10R in between the filter caps, turning it into a pi-filter, will add many dB of attenuation at the line frequency.
The bypass cap C9 is unnecessary, even unhelpful. Bypass is useful when placed as close to possible to the active circuitry.
I don't understand the purpose of C7,C8,R2,R3.
I also use 10,000uF filter caps, but as shown.
Putting 10R in between the filter caps, turning it into a pi-filter, will add many dB of attenuation at the line frequency.
The bypass cap C9 is unnecessary, even unhelpful. Bypass is useful when placed as close to possible to the active circuitry.
I don't understand the purpose of C7,C8,R2,R3.
I also use 10,000uF filter caps, but as shown.
Attachments
PS Boards
C&C welcome. Thank you.
I am planning a 2nd (final) build of the PS board. I'd like to clean up the A/C interconnects, plan on running all the AC to one end of the board (mains, xfmer primary/secondary, switch,, LED, etc.) I was also planning on adding a pi filter. D. Self calls them cheap, easy, and "bulletproof". In my mind I had chosen 4R7, but 10R makes better sense. Thanks also for tip using C1, C2, and a 150R resistor. I will definitely add that.
C7 and C8 were added to provide rail to rail decoupling. Added in series to double the voltage capacity. R2 and R3 provide a voltage reference so C7 and C8 are each handling about the same voltage. I chose 4K7 because it provides 5mA, maybe 10K or larger would be better?
Why is the better question. Primarily for 3 reasons. (1) I've read somewhere that adding "hundreds to thousands" of microfarads across the rails helps equalize the voltage, so absolute value of the two do not differ significantly. I can't quickly find the reference. If it's on the internet, it must be true, right? Not sure if this idea has merit or not. (2) D. Self shows an example of rail-to-rail decoupling in the power supply chapter of "Small Signal Audio Design, 2nd Ed." pg. 711, "such decoupling can be useful because it establishes a low AC impedance across the supply rails without coupling supply rail noise into the ground, as C7, C8 are prone to do". Remember I'm a self-taught hobbyist, what little I know comes from the Self, Cordell, Sloane books (primarily Self) and manufacturer app notes. If it's good enough for D. Self, it's good enough for me. And finally, (3) it can't hurt. Maybe it's not absolutely needed, and I could free up some board space, reducing pcb cost and ~ $4 in components. I use OSH Park for the low volume, quality boards, but one square inch costs $5.
I'd be glad to update the schematic, supply some part numbers, doing the "legwork" so to speak and then post again. This might provide a known good starting point as an option for other hobbyist types, not repeating the same mistakes I made. I'm willing to help.
Thanks for taking time to comment.
C&C welcome. Thank you.
I am planning a 2nd (final) build of the PS board. I'd like to clean up the A/C interconnects, plan on running all the AC to one end of the board (mains, xfmer primary/secondary, switch,, LED, etc.) I was also planning on adding a pi filter. D. Self calls them cheap, easy, and "bulletproof". In my mind I had chosen 4R7, but 10R makes better sense. Thanks also for tip using C1, C2, and a 150R resistor. I will definitely add that.
C7 and C8 were added to provide rail to rail decoupling. Added in series to double the voltage capacity. R2 and R3 provide a voltage reference so C7 and C8 are each handling about the same voltage. I chose 4K7 because it provides 5mA, maybe 10K or larger would be better?
Why is the better question. Primarily for 3 reasons. (1) I've read somewhere that adding "hundreds to thousands" of microfarads across the rails helps equalize the voltage, so absolute value of the two do not differ significantly. I can't quickly find the reference. If it's on the internet, it must be true, right? Not sure if this idea has merit or not. (2) D. Self shows an example of rail-to-rail decoupling in the power supply chapter of "Small Signal Audio Design, 2nd Ed." pg. 711, "such decoupling can be useful because it establishes a low AC impedance across the supply rails without coupling supply rail noise into the ground, as C7, C8 are prone to do". Remember I'm a self-taught hobbyist, what little I know comes from the Self, Cordell, Sloane books (primarily Self) and manufacturer app notes. If it's good enough for D. Self, it's good enough for me. And finally, (3) it can't hurt. Maybe it's not absolutely needed, and I could free up some board space, reducing pcb cost and ~ $4 in components. I use OSH Park for the low volume, quality boards, but one square inch costs $5.
I'd be glad to update the schematic, supply some part numbers, doing the "legwork" so to speak and then post again. This might provide a known good starting point as an option for other hobbyist types, not repeating the same mistakes I made. I'm willing to help.
Thanks for taking time to comment.
I ask my friend who sell some components. He have 2SK246 and 2SJ103 JFET. Now, I make simulation use this jfet. I add a capacitance multiplier before output transistors. THD is low, I like the harmonic profile, and it have high slew rate. I will build it.
The PI filter does carry a penalty in terms of I^2*R losses, and Dr. Marsh reports that a higher rail voltage may be preferable. If you detect a 120Hz problem (barring grounding issues) try the PI filter.
In my MOSFET power amplifiers I just use a 1uF polypropylene cap across the rails at the output.
My version of Self's book is 2010 and doesn't seem to have that particular circuit (the book went digital with much better graphics.) I note that in simulation whether one uses the PS you show, or simply adding another 10,000uF seem to have the same effect.
Attachments
Thank you very much for your reply. Based on your comments, I took the PS board, cut some traces and tacked a couple of 10R resistors to the bottom. I also disconnected the 2 series caps across the rails by cutting traces and clipping out the equalizing resistors. A few things happened.
There was a voltage drop, as predicted. The xfmr may be somewhat anemic for this circuit, because I'm trying to keep the footprint small. The DC voltage dropped from just over 16V to 13.8V. The measured AC voltage dropped to about 500uV, a significant improvement. But (and this is entirely subjective) it did not seem to sound as good as the before version. Probably due to preconceived notions, volume level, phase of the moon... I have no way to measure (confirm) this.
I will add another 10,000uF of filtering by tacking some bus wire to the bottom of the board. The two series caps are still there, just disconnected. However, I've been wondering if a second filter would be useful?
The idea of 1uf across the rails is very interesting, new to me. But then again all this is new to me. Perhaps replacing the 2 series caps with another 10,000uF of filtering, maybe another filtering stage (pair of resistors), and 1uF across the output rails is "better"? Better in the sense that it might not be "better" but won't be any worse? What I do not know is the point of diminishing returns, and would like very much to hear the opinions of others.
Thanks again for your thoughts.
There was a voltage drop, as predicted. The xfmr may be somewhat anemic for this circuit, because I'm trying to keep the footprint small. The DC voltage dropped from just over 16V to 13.8V. The measured AC voltage dropped to about 500uV, a significant improvement. But (and this is entirely subjective) it did not seem to sound as good as the before version. Probably due to preconceived notions, volume level, phase of the moon... I have no way to measure (confirm) this.
I will add another 10,000uF of filtering by tacking some bus wire to the bottom of the board. The two series caps are still there, just disconnected. However, I've been wondering if a second filter would be useful?
The idea of 1uf across the rails is very interesting, new to me. But then again all this is new to me. Perhaps replacing the 2 series caps with another 10,000uF of filtering, maybe another filtering stage (pair of resistors), and 1uF across the output rails is "better"? Better in the sense that it might not be "better" but won't be any worse? What I do not know is the point of diminishing returns, and would like very much to hear the opinions of others.
Thanks again for your thoughts.
Finally integrated the Marsh into my Mezmerize preamp chassis! I also picked up a set of Grado GH1's and have been in sonic bliss!
When I first fired it up, I had a bad hum. I had the power supply set to +/-18V which I had run it at before with no problems but with a 50VA toroid. The new toroid is only 10VA, so I dropped the supply down to +/-15V and the hum is gone. I guess the transformer was not up to it! Nonetheless, all is well now and it sounds incredible.
I also removed one of the diodes and now the heatsink barely even gets warm to the touch.
When I first fired it up, I had a bad hum. I had the power supply set to +/-18V which I had run it at before with no problems but with a 50VA toroid. The new toroid is only 10VA, so I dropped the supply down to +/-15V and the hum is gone. I guess the transformer was not up to it! Nonetheless, all is well now and it sounds incredible.
I also removed one of the diodes and now the heatsink barely even gets warm to the touch.
Attachments
I was only able to hit 18V after removing the diodes. Prior to that, at 12V I could fry eggs on the heatsink!
I have some problem with the DC offset at amp output. The problem is due to small - 3-4mV voltage at the output of my source. The question is - how to resolve it better: by decoupling caps like : ????? ?????????? - WIMA - MKS2B046801M00JSSD - CAP, FILM, PET, 6.8UF, 50V, RAD or 1 Pair 2pcs of Mundorf Mcap Supreme Capacitors 2 600 1400V All Values | eBay, or to build a simple DC servo with some fet input opamps, like TL082 ?