Desktop Diamond Buffer (DDB) Headphone Amp

I will admit, I'm only comparing it to other headphone amps that can be used as preamps, but still...

I have the DDB HPA, an ACP+ HPA, and a Whammy HPA. I find that the music I play through the DDB sounds more dynamic and has more punch. Especially when playing classic and hard rock which are two of my favorite categories.
 
  • Like
Reactions: xrk971
Hi X,
first let me thank you for the DDB module, it literally saves my otherwise failing project. I will come back when finished with my implementation and measurements.
But I need your help too. While the module powers up and works fine it still does not bias as expected to ~40 mA. I barely get 20 mA, even less actually as it warms up. Both channels are equal. I measured all resistor values on the board, they are ok and as per schematic. The culprit appears to be the reference LEDs. They only develop 1.87 V on them while obviously there had to be somewhere around 2.7 V as follows from your simulation schematic in the PDB thread using LED NSCW100 type. 2.7 V would indeed provide enough reference for output biasing to 40 mA or so. Power rails are ±15 V.
Can you advise please how do I best go about it to bring up the bias. Replace LEDs? Reduce resistor values in CSs or emitter resistors of output BJTs?
Thanks in beforehand!
 
I am using the Lite On LED with specs showing voltage drop as circa 2v.

https://optoelectronics.liteon.com/upload/download/DS22-2000-228/LTST-C191KGKT.PDF

It might be that your transistors HFe’s are different.

you can cut the value of the large emitter resistors R212 and 213 to half the value (2.2R) to double the current. Or change the resistors on the front end to increase the bias current (preferred way).

Perhaps @jhofland can chime in as to how to fix this.

It might be that R204, 205, 206, 211 could be cut in half to double the current in the front end?

1731439471221.jpeg
 
Thank you @xrk971 for the quick response.

Ok, I see, the LED datasheet says that Vf = 2V is measured at 20mA. Since the LED in the DDB is fed by only ~1.2mA, my measurement of Vf = 1.87V makes sense. I can of course increase the forward current to maybe 5 mA but that would not change much the output bias current.

What did you mean exactly by "change the resistors on the front end"? The 100 Ω resistors R201 and R202? The simulator prompts that these resistors would need to be increased to at least 3 kΩ to get the output bias closer to 40 mA. Would that be right?
 
Hi X,

another happy user of DDB here!

I have found the reason for the low bias (had to change two resistors) and once at it I also took the liberty to adjust few other components to my liking. Those SMDs are so small for hand soldering but the learning curve alone was worth the effort.

The DDB board found its place inside an inferior project from a kit I picked up at Ali/Bay. I was about to take out all reusable components and ditch the rest when I came across the DDB thread. The DDB board dimensions were fitting the space for the amplifier section so perfectly that it looked like being designed just for this case. Miracle! I kept the PSU (modified to CRCRC), the protection, DAC and self-designed input switcher in their places and added the DDB board instead of the former amplifier section. Works now perfectly both as headphone amp and pre-amp.
DDB.png

DDB_amp_front_back.png


The photo also shows some odd point-to-point additions on top of the DDB board. Those were necessary for removal of DC offset on the output. Since my DDB board came pre-populated with NE5532, the volume pot could not be connected to the board in the same way as in case of the FET input OPA1642. Rewiring the pot wiper connection to before the input DC blocking capacitor expectedly resulted in higher DC offset. I had negative ca 60 mV, which was in agreement with what simulator had told me. The added positive voltage injection network now keeps the offset within 2 mV.

I also had to retain zobel on the output even though simulator indicated no need for it. I tried the board without the zobel but got small scale local HF oscillation (~80 mVpp) somewhere in the circuit with nothing connected to the output. I did not dig further to find out.

Step response is perfect in my view and the waveform remained undisturbed with any capacitive loads (I tested up to 220 nF). Note however that I am also using 2.7 Ω output resistor which does improve load immunity as well.
10k_sq_3Vpp.png


The THD measurement was in full alignment with your results - the 2nd harmonic dominates, there is a little bit of the 3rd but the rest stays in "grass". Below is 1 VRMS into 33 Ω:
1kHz_33R_1Vrms.png


Since my gear cannot reliably measure HD for higher frequencies, I resort to CCIF IMD. Same 1 VRMS into 33 Ω. Looks good to me:
CCIF_33R_1Vrms.png


Overall, very happy result! Lots of new learning and experience. DDB created the best thing possible out of otherwise miserable project.
Thank you very much for this board!
 
This amazing little amp has been discussed as part of the DB thread but I think it deserves its own thread given how it has quite different objectives from the Pocket Diamond Buffer (PDB) HPA.

This desktop variant uses TO-126 output BJTs with dedicated heatsinks and does not have the separate MOSFET rail switching needed for the pocket one. The setpoints have been optimized for a moderate 40mA nominal bias current that will let you run in full Class A operation for most headphone uses. It will leave Class A around 250mW into 32 ohms. Of course, you can adjust the emitter resistors to run higher bias currents. I have used it with +/-9v, +/-12v, +/-15v and +/-18v supplies and they all work great. It was designed for 15v rails but I don’t think performance suffers much from using lower voltage. For your particular use case you may want more voltage. High impedance 300ohm cans like more voltage, for example. There is lots of room for input cap rolling and I have used huge 4.7uF 400v MKPs and also smaller 0.47uF film to bypass 10uF Elna Silmic electrolytic caps.

Here’s a photo of a built up amp board in operation:
View attachment 1260866
With larger MKP caps only:
View attachment 1260877

I initially struggled with an oscillation but chased it down to a feedback compensation cap that was too large. That’s all fixed now rock solid. A stability analysis on LtSpice pointed me in the right direction. So the schematic shows a 470pF cap needing to be replaced with 47pF. It’s an easy swap and I’ll do this on the SMT pre populated boards that I will be offering in my shop. I will also have bare PCBs for those wanting to assemble it themselves.

Here is schematic for the overall amp, very simple. Power supply in, audio in, audio out and volume pot with MicroMatch IDC cable to an RK09 pot helper PCB (included).

View attachment 1260868
Here is schematic of the Diamond output stage:
View attachment 1260869

Here is a SMT prepopulated PCB so all you need to do is to install all the through hole parts:
View attachment 1260870
Both sides:
View attachment 1260874
Here is the board assembled with big 4.7uF MKP caps:
View attachment 1260871

Standard BOM calls for ECB pinout TO-126 BJTs (Toshiba TTA004 and TTC004) but if you have some classic Toshiba 2SA1837 and 2SC4795 (BCE pinout) you can mount them in the bottom like I did here - leaves a cleaner look on top giving you lots of access to the parts:
View attachment 1260879View attachment 1260878

Testing on the bench:View attachment 1260872
View attachment 1260873
Nice low distortion but second harmonic dominant distortion profile for 1Vrms into 33ohms:
View attachment 1260876
You will find that this headphone amp sounds very natural and clean, but has an incredible power reserve to deliver bass slam while maintaining great control authority of the driver transducer cone. It’s a wonderful headphone amp to listen to for hours. For normal headphone use (under 500mW power) basic AC/DC modules designed for delivering circa 350mA into 15v can be used with a 7812/7913 voltage regulator and it will work fine. You can also make really fancy PSU with linear trafos, cap multipliers, CLC, and low noise LDO’s etc. with a proper PSU capable of 750mA and +/-15v you can drive up to 1.5W into 32 ohms (for those fans of the HiFiMan HE-6).

I’ll be offering the SMT preassembled PCBs for $59. Bare PCBs for $22. Boards are all 1oz copper and ENIG finish.

https://xrkaudio.etsy.com/listing/1659118057

BOM is here. Note that BOM calls for OPA1642 but the pre-populated PCB has NE5532 installed. This was due to availability issues but I can assure you that the NE5532 sounds fantastic and measures well too. If you want to swap it out later, you are welcome to do so.
QuantityReferencesValueManufacturerMPNloadtypevoltageFootprint
9C107, C108, C109, C201, C202, C203, C301, C302, C3031uFSamsungCL21B105KAFNNNEX7R25VC_0805_2012Metric_Pad1.18x1.45mm_HandSolder
4C205, C206, C305, C306330uFPanasonicEEU-FM1V331LB35VCP_Radial_D8.0mm_P5.00mm
3C110, C113, C114100nFWIMAMKS2D031001A00MSSDPET100VC_Rect_L7.2mm_W2.5mm_P5.00mm_FKS2_FKP2_MKS2_MKP2
2C101, C1023u3WIMAMKS2B043301H00KSSDPET50VC_Rect_L7.2mm_W5.5mm_P5.00mm_FKS2_FKP2_MKS2_MKP2
2C103, C10433pFSamsungCL10C330JB8NNNCC0G50VC_0603_1608Metric_Pad1.08x0.95mm_HandSolder
2C105, C10647pFKyoceraKGM15ACG2A470FTC0G100VC_0603_1608Metric_Pad1.08x0.95mm_HandSolder
2C111, C1124u7Audyn027-114DNPpolyprop400VCAP_Jantzen_17mmDx35mm_axial
2C204, C304180pFSamsungCL10C181JB81PNDC0G50VC_0603_1608Metric
8R203, R208, R209, R210, R303, R308, R309, R31010RYAGEORC0603FR-0710RLR_0603_1608Metric_Pad0.98x0.95mm_HandSolder
6R204, R205, R214, R304, R305, R31422R1YAGEORC0603FR-0722R1LR_0603_1608Metric_Pad0.98x0.95mm_HandSolder
4R103, R104, R107, R1081K21YAGEORC0603FR-071K21LR_0603_1608Metric_Pad0.98x0.95mm_HandSolder
4R109, R110, R207, R30710KYAGEORC0603FR-0710KLR_0603_1608Metric_Pad0.98x0.95mm_HandSolder
4R201, R202, R301, R302100RYAGEORC0603FR-10100RLR_0603_1608Metric_Pad0.98x0.95mm_HandSolder
4R206, R211, R306, R311221RYAGEORC0603FR-07221RLR_0603_1608Metric_Pad0.98x0.95mm_HandSolder
4R212, R213, R312, R3134R75YAGEORC1206FR-074R75LR_1206_3216Metric_Pad1.30x1.75mm_HandSolder
2R101, R10215RYAGEORC1206FR-0715RLR_1206_3216Metric_Pad1.30x1.75mm_HandSolder
2R105, R1064K75YAGEORC0603FR-074K75LR_0603_1608Metric_Pad0.98x0.95mm_HandSolder
1R11110RYAGEORC1206FR-0710RLR_1206_3216Metric_Pad1.30x1.75mm_HandSolder
2D201, D301greenLite_OnLTST-C191KGKTLED_0603_1608Metric
1U101OPA1642Texas InstrumentsOPA1642AIDRSOIC-8_3.9x4.9mm_P1.27mm
4Q201, Q204, Q301, Q304MMBT5401Diodes IncMMBT5401-7-FSOT-23-3
2Q202, Q302MMBT5551Diodes IncMMBT5551-7-FSOT-23-3
2Q203, Q303DMMT5401Diodes IncDMMT5401-7-FSOT-23-6
2Q205, Q305DMMT5551Diodes IncDMMT5551-7-FSOT-23-6
2Q206, Q306TTC004ToshibaTTC004B,QTO-126-3_Vertical
2Q207, Q307TTA004ToshibaTTA004B,QTO-126-3_Vertical
1RV10110K audioAlps AlpineRK09L12D0A1TRK09L_dual
2HS201, HS301HeatsinkAavid513002B02500Gheatsink_AAVID_53100202500G_TO220
2J102, J104WE_WR_MM_8Wurth Electronik690367180872CONN_6910367180872_WE_WR_MM_8
2J105, J108XH_3pinJSTB3B-XH-A(LF)(SN)JST_XH_B3B-XH-A_1x03_P2.50mm_Vertical
1J103XH_4pinJSTB4B-XH-A(LF)(SN)JST_XH_B4B-XH-A_1x04_P2.50mm_Vertical
1J107spadeTE Connectivity62409-1FASTON_TE Connectivity_62409

If you want to boost the output BJT bias current (80mA) for operation to a higher power under Class A, use this 2.2ohm metal thin film emitter resistor (R213/213/312/313):
https://www.mouser.com/ProductDetail/Vishay-Dale/TNPW12062R20DEEA?qs=vHuUswq2%2BswIE18TmZVkaQ==

Note that the heatsinks will get significantly hotter and the PSU also needs to provide more quiescent power.

For the 8pin Wurth IDC cable with matching connector for the potentiometer, get this cable (or its equivalent).

Dumb question time from a newbie.

Aren’t there better performing BJT than those older Toshiba ones in the BOM?
Isn’t their transition freq lower than let’s say another older BJT like the BD140?
Or their SMD equivalent? ( I think they are called BCP56?)

Thanks !
 
  • Like
Reactions: lineup
Dumb question time from a newbie.

Aren’t there better performing BJT than those older Toshiba ones in the BOM?
Isn’t their transition freq lower than let’s say another older BJT like the BD140?
Or their SMD equivalent? ( I think they are called BCP56?)

Thanks !
If you can find some new old stock of 2SA1837 and 2SC4795, they are better. But the TTA004 and TTC004 are modern BJTs that will perform better than you need. As can be seen in the recent measurements by nonills, it doesn’t get much better.

1740401336705.png
 
  • Like
Reactions: Nightjar
thanks so much for the replies!
Appreciate it.

One more question.
When increasing the Bias voltage
This amazing little amp has been discussed as part of the DB thread but I think it deserves its own thread given how it has quite different objectives from the Pocket Diamond Buffer (PDB) HPA.

This desktop variant uses TO-126 output BJTs with dedicated heatsinks and does not have the separate MOSFET rail switching needed for the pocket one. The setpoints have been optimized for a moderate 40mA nominal bias current that will let you run in full Class A operation for most headphone uses. It will leave Class A around 250mW into 32 ohms. Of course, you can adjust the emitter resistors to run higher bias currents. I have used it with +/-9v, +/-12v, +/-15v and +/-18v supplies and they all work great. It was designed for 15v rails but I don’t think performance suffers much from using lower voltage. For your particular use case you may want more voltage. High impedance 300ohm cans like more voltage, for example. There is lots of room for input cap rolling and I have used huge 4.7uF 400v MKPs and also smaller 0.47uF film to bypass 10uF Elna Silmic electrolytic caps.

Here’s a photo of a built up amp board in operation:
View attachment 1260866
With larger MKP caps only:
View attachment 1260877

I initially struggled with an oscillation but chased it down to a feedback compensation cap that was too large. That’s all fixed now rock solid. A stability analysis on LtSpice pointed me in the right direction. So the schematic shows a 470pF cap needing to be replaced with 47pF. It’s an easy swap and I’ll do this on the SMT pre populated boards that I will be offering in my shop. I will also have bare PCBs for those wanting to assemble it themselves.

Here is schematic for the overall amp, very simple. Power supply in, audio in, audio out and volume pot with MicroMatch IDC cable to an RK09 pot helper PCB (included).

View attachment 1260868
Here is schematic of the Diamond output stage:
View attachment 1260869

Here is a SMT prepopulated PCB so all you need to do is to install all the through hole parts:
View attachment 1260870
Both sides:
View attachment 1260874
Here is the board assembled with big 4.7uF MKP caps:
View attachment 1260871

Standard BOM calls for ECB pinout TO-126 BJTs (Toshiba TTA004 and TTC004) but if you have some classic Toshiba 2SA1837 and 2SC4795 (BCE pinout) you can mount them in the bottom like I did here - leaves a cleaner look on top giving you lots of access to the parts:
View attachment 1260879View attachment 1260878

Testing on the bench:View attachment 1260872
View attachment 1260873
Nice low distortion but second harmonic dominant distortion profile for 1Vrms into 33ohms:
View attachment 1260876
You will find that this headphone amp sounds very natural and clean, but has an incredible power reserve to deliver bass slam while maintaining great control authority of the driver transducer cone. It’s a wonderful headphone amp to listen to for hours. For normal headphone use (under 500mW power) basic AC/DC modules designed for delivering circa 350mA into 15v can be used with a 7812/7913 voltage regulator and it will work fine. You can also make really fancy PSU with linear trafos, cap multipliers, CLC, and low noise LDO’s etc. with a proper PSU capable of 750mA and +/-15v you can drive up to 1.5W into 32 ohms (for those fans of the HiFiMan HE-6).

I’ll be offering the SMT preassembled PCBs for $59. Bare PCBs for $22. Boards are all 1oz copper and ENIG finish.

https://xrkaudio.etsy.com/listing/1659118057

BOM is here. Note that BOM calls for OPA1642 but the pre-populated PCB has NE5532 installed. This was due to availability issues but I can assure you that the NE5532 sounds fantastic and measures well too. If you want to swap it out later, you are welcome to do so.
QuantityReferencesValueManufacturerMPNloadtypevoltageFootprint
9C107, C108, C109, C201, C202, C203, C301, C302, C3031uFSamsungCL21B105KAFNNNEX7R25VC_0805_2012Metric_Pad1.18x1.45mm_HandSolder
4C205, C206, C305, C306330uFPanasonicEEU-FM1V331LB35VCP_Radial_D8.0mm_P5.00mm
3C110, C113, C114100nFWIMAMKS2D031001A00MSSDPET100VC_Rect_L7.2mm_W2.5mm_P5.00mm_FKS2_FKP2_MKS2_MKP2
2C101, C1023u3WIMAMKS2B043301H00KSSDPET50VC_Rect_L7.2mm_W5.5mm_P5.00mm_FKS2_FKP2_MKS2_MKP2
2C103, C10433pFSamsungCL10C330JB8NNNCC0G50VC_0603_1608Metric_Pad1.08x0.95mm_HandSolder
2C105, C10647pFKyoceraKGM15ACG2A470FTC0G100VC_0603_1608Metric_Pad1.08x0.95mm_HandSolder
2C111, C1124u7Audyn027-114DNPpolyprop400VCAP_Jantzen_17mmDx35mm_axial
2C204, C304180pFSamsungCL10C181JB81PNDC0G50VC_0603_1608Metric
8R203, R208, R209, R210, R303, R308, R309, R31010RYAGEORC0603FR-0710RLR_0603_1608Metric_Pad0.98x0.95mm_HandSolder
6R204, R205, R214, R304, R305, R31422R1YAGEORC0603FR-0722R1LR_0603_1608Metric_Pad0.98x0.95mm_HandSolder
4R103, R104, R107, R1081K21YAGEORC0603FR-071K21LR_0603_1608Metric_Pad0.98x0.95mm_HandSolder
4R109, R110, R207, R30710KYAGEORC0603FR-0710KLR_0603_1608Metric_Pad0.98x0.95mm_HandSolder
4R201, R202, R301, R302100RYAGEORC0603FR-10100RLR_0603_1608Metric_Pad0.98x0.95mm_HandSolder
4R206, R211, R306, R311221RYAGEORC0603FR-07221RLR_0603_1608Metric_Pad0.98x0.95mm_HandSolder
4R212, R213, R312, R3134R75YAGEORC1206FR-074R75LR_1206_3216Metric_Pad1.30x1.75mm_HandSolder
2R101, R10215RYAGEORC1206FR-0715RLR_1206_3216Metric_Pad1.30x1.75mm_HandSolder
2R105, R1064K75YAGEORC0603FR-074K75LR_0603_1608Metric_Pad0.98x0.95mm_HandSolder
1R11110RYAGEORC1206FR-0710RLR_1206_3216Metric_Pad1.30x1.75mm_HandSolder
2D201, D301greenLite_OnLTST-C191KGKTLED_0603_1608Metric
1U101OPA1642Texas InstrumentsOPA1642AIDRSOIC-8_3.9x4.9mm_P1.27mm
4Q201, Q204, Q301, Q304MMBT5401Diodes IncMMBT5401-7-FSOT-23-3
2Q202, Q302MMBT5551Diodes IncMMBT5551-7-FSOT-23-3
2Q203, Q303DMMT5401Diodes IncDMMT5401-7-FSOT-23-6
2Q205, Q305DMMT5551Diodes IncDMMT5551-7-FSOT-23-6
2Q206, Q306TTC004ToshibaTTC004B,QTO-126-3_Vertical
2Q207, Q307TTA004ToshibaTTA004B,QTO-126-3_Vertical
1RV10110K audioAlps AlpineRK09L12D0A1TRK09L_dual
2HS201, HS301HeatsinkAavid513002B02500Gheatsink_AAVID_53100202500G_TO220
2J102, J104WE_WR_MM_8Wurth Electronik690367180872CONN_6910367180872_WE_WR_MM_8
2J105, J108XH_3pinJSTB3B-XH-A(LF)(SN)JST_XH_B3B-XH-A_1x03_P2.50mm_Vertical
1J103XH_4pinJSTB4B-XH-A(LF)(SN)JST_XH_B4B-XH-A_1x04_P2.50mm_Vertical
1J107spadeTE Connectivity62409-1FASTON_TE Connectivity_62409

If you want to boost the output BJT bias current (80mA) for operation to a higher power under Class A, use this 2.2ohm metal thin film emitter resistor (R213/213/312/313):
https://www.mouser.com/ProductDetail/Vishay-Dale/TNPW12062R20DEEA?qs=vHuUswq2%2BswIE18TmZVkaQ==

Note that the heatsinks will get significantly hotter and the PSU also needs to provide more quiescent power.

For the 8pin Wurth IDC cable with matching connector for the potentiometer, get this cable (or its equivalent).

thanks so much for the replies!
Appreciate it.

One more question.
When increasing the Bias current to 80mA, will this increase the THD a bit, but raise the H2 and lower the higher order harmonics ?
 
If you can find some new old stock of 2SA1837 and 2SC4795, they are better. But the TTA004 and TTC004 are modern BJTs that will perform better than you need. As can be seen in the recent measurements by nonills, it doesn’t get much better.

View attachment 1426774

Pardon but another newbie question.

Yeah, that’s a great performance. And nice succession of harmonics.

The THD doesn’t matter too much to me, the above is perfect, but could adding a supercapacitor bank lower the noise floor more and get rid of the small ‘ humps’ below 1khz in the power section ?

Or just adding more capacitance after the voltage regulators would do that ?

I know I know it’s not a DAC to need such a low noise floor. But I do believe in a super low noise floor.

Thanks and forgive me if my questions are tiring.
 
It’s a combination of things to get noise floor down. Adding caps only helps so much. Physical arrangement of components and where wires are routed make a big difference. Look up “Loop area” for EMI pickup. Basically an amp with wire paths that make a large loop will pickup more noise than ones with smaller loops or loops that are balanced with a bifilar wire of opposite phase.

Using cap multipliers is a big help. About -50dB to be gained there.

Good EMI filters with CMC chokes.
 
  • Like
Reactions: Nightjar
Thanks so much for all the answers.

Ok hopefully this isn’t tiring you guys.

But do you hear a difference in the input Caps C111 and C112, if using let’s say wima or a higher end Film Foil like the Audyn or let’s say Mundorf?

Or is it such a small diff you wouldn’t tell in a blind test