HPS 6.1

Here's for your free non commercial use the last implementation of my phono preamp: HPS 6.1

What's new:

- A new servo in the head amp, allowing the use of an active load in the low noise stage; as a result, the PSRR is improved by an order of magnitude.
- A new power supply, much simpler, using the latest ultra low noise regulators from TI. TPSA7A4700 (positive) and TPSA7A33 (negative).

You will find attached the schematics and the Gerber files (including this time the much requested solder mask). The board size is 124.5 x 81.3mm so the construction is very compact. Any questions, feel free to ask. For the time being, only a few comments:

- It is dual mono, and requires separate raw power supplies (recommended +20V @250 mA -20V @50mA) for each channel (if you want to keep it dual mono all the way). Double the currents if you are using a single power supply for both channels.
- Supply voltages are +/-18V. This being said, you can replace the LME49610 high voltage buffer (now exctinct) with the lower voltage version LME49600
- For the low noise jfets, you can replace the now extinct BF862's with the still in production at OnSemi 2SK3557 or CPH3910 with a less than 5% penalty in noise. For the 2SK3557 choose the -6 rank (lower Idss).
- I have two more versions, one uses the OnSemi CPH5905 integrated cascode (same noise performance as the 2SK3557) and one using standard LM317/LM337 linear regulators (almost the same noise performance, board is slightly larger). The latter was a proof of concept to showcase the very good PSRR of the input stage, but I would still recommend the current version. I may eventually post these versions as well.

And a strong WARNING. This construction is not for the beginners. It is full SMD and while most of the parts are a convenient 1206 size, some are smaller 0402 and the TI ultra low noise regulators are in DFN package, difficult to solder. Skills in SMD soldering and the use of a hot air soldering tool is mandatory for success.

Finally, I have to add the usual disclaimer: I am not into group buys, selling boards, building BOMs for DigiKey or Mouser, etc... life is too short for such, so essentially you are on your own. Information herein is accurate to the best of my knowledge. This will be posted on my web site, as soon as I'll find time to do it (I am hopelessly in behind updating my web site).

Enjoy!

09/17/2019 Schematics corrected here: https://www.diyaudio.com/forums/analogue-source/333000-hps-6-1-a-4.html#post5917513
 

Attachments

  • HPS61.zip
    629.8 KB · Views: 378
  • HPS61-4-gerbers.zip
    106.9 KB · Views: 167
  • HPS61.jpg
    HPS61.jpg
    544.7 KB · Views: 1,381
Last edited:
For those intimidated by the DFN soldering (yes, it needs tools and some practice, YouTube has several videos to start with) here's the version with regular LM317/LM337 regulators. This version also uses the CPH5905 low noise jfet, cascoded with a bipolar device, in the same SOT-23-5 case. As shown, the preamp is fed from +/-22V power supply voltages and requires high voltage opamps, some of which are extinct (LME49610) and some are difficult to source (OPA2604). The prospective builder could easily modify the LM317/337 to fall back to +/-18V and then the same op amps as above could be used.

Due to the excellent PSRR of the low noise stage, using regular LM317/LM337 has absolutely no impact on the measured noise; the CPH5905 JFET (which, according to the data sheet, is a 2SK3557) is onlittle 5-10% noisier than the now extinct BF862. As such, the BF862 HPS version above has 0.33nV/rtHz @1KHz while the CPH5905 version attached here has 0.38nV/rtHz @ 1KHz, but in turn a slightly lower 1/f noise corner frequency. Please use the low Idss grade of the CPH5905.

For any other side effects of using these inexpensive voltage regulators, I was unable to identify any, YMMV.

Attached are the picture, schematics and Gerbers (including the solder mask and silk screen). The comments and disclaimers in the first post of this topic apply here as well.

Coming up (but don't hold your breath, it is in the final design stage) a HPS version that can be powered from 2 x 9V batteries (taking only some 25 mA from the positive supply and 15 mA from the negative supply). Preliminary results show 0.45nV/rtHz with BF862 and 0.5nV/rtHz with 2SK3557. This version uses an extra high Vp and high Idss JFET, to cascode the 8x paralleled low noise JFETs.
 

Attachments

  • HPS_61.jpg
    HPS_61.jpg
    526.1 KB · Views: 1,193
  • HPS61.zip
    568.6 KB · Views: 208
  • HPS61-2-gerbers.zip
    98.9 KB · Views: 112
Disabled Account
Joined 2008
Hi I hadn't seen this thread until now, very impressive.

I like the way you are doing it with one cascode transistor on each jfet instead of one common cascode transistor for all.
Nice Servo BTW.

-Have you used a smt pick and place machine? Every component is perfectly placed.

-What about the temperature of the input devices, is it a difference when you use them in separate packages or both in the same package as in the CPH5905?

Personaly I'm considering buying a manual smt pick and place machine.

Cheers
 
Hi I hadn't seen this thread until now, very impressive.

I like the way you are doing it with one cascode transistor on each jfet instead of one common cascode transistor for all.
Nice Servo BTW.

-Have you used a smt pick and place machine? Every component is perfectly placed.

-What about the temperature of the input devices, is it a difference when you use them in separate packages or both in the same package as in the CPH5905?

Personaly I'm considering buying a manual smt pick and place machine.

Cheers

100% manual soldering. Hot air for DFN, fine point Weller for the rest. Hot air for rework. Cleaned extra flux in isopropylic alcohol.

No major temperature difference between same chip and separate transistors cascode.

Next version is coming up nicely:

- Can be battery powered (bypass the low noise regulators and no gain switching relays, takes 50mA both channels from 2x9V cells). Shown in photo is the full version with +/-18V regulators and gain switching.
- jfet cascode
- 0.45nV/rtHz
- 30% smaller board size, fits in 100x100mm PCB size, $20 for 10pcs, 2 days turn around, including DHL shipping, got them the same week :).
 

Attachments

  • IMG_2029.jpg
    IMG_2029.jpg
    782 KB · Views: 1,098
Last edited:
Superb SMD soldering job - I take my hat off to you. I can place and solder down to 0805 and the small 6 pin dual transistor packages, SOT23, SOD363 etc but they don't look your solder jobs.
:)

I can easily solder manually 1206, 0805 and 0603 and under the microscope 0402. 0201 is tricky (meaning essentially a low yield), can’t do 1005 manually, by any means.
 
Cool syn08, nice job,
yeah I try to make as much as possible in 100x100mm for the special pcb deals these days. I just used jlcpcb, the import tools is great at reading in the data, determines the board outline auto, the gerber viewer is a nice extra check to see that they read it in correctly.
I like the 685 cap too :)
I am using glow core 60/40 lead, no-clean solder, it works great with the default leaded HASL plating, ROHS not, this gov work
I did my DFN by hand but you have to open up the solder mask so that you get a hot tip on the trace so it can wick up to the lead. The first pin is the hardest to do.
Whats with all the 90 degree bends? looks like you run the cheapo orcad auto-router, with no miter function :)
cheers,
 
Last edited:
I am using glow core 60/40 lead, no-clean solder, it works great with the default leaded HASL plating, ROHS not, this gov work
I did my DFN by hand but you have to open up the solder mask so that you get a hot tip on the trace so it can wick up to the lead. The first pin is the hardest to do.
Whats with all the 90 degree bends? looks like you run the cheapo orcad auto-router, with no miter function :)
cheers,

I use 63/37, and solder DFN by hot air, don't need anything specific, the chip self aligns on the pads nicely. There are several videos on YouTube showing the howto.

Orcad indeed but no auto-router. Routed 100% by hand, but I simply don't care to miter anything (unless required). I'm confident electrons won't trip, impede or otherwise hamper at the 90 degs corners.
 
Do you use a stencil/paste for the DFN?

Nope, just soldering the PCB pads with a wide Weller tip before reflowing with hot air. Needs exercise, the whole trick is to work quickly and, as much as possible, keep the solder on all pads level. Air flow has also to be carefully adjusted so that it won’t blow away the part. Temperature is also rather critical, I use to go for 400C and air exhaust tip at 1.5-2cm away from the chip (I measured some 300C at the chip top). Lower temperature and the chip won’t self align, higher temperature (or mucking for too long) doesn’t leave much room for any eventual rework (pads may fly away during the next heating cycle). Always solder first the DFN chips on a fresh board.

To me, using stencils and paste makes sense only for reflowing the whole board in an oven. PCB reflow ovens are not that expensive, but I still doubt the quality of a home made process, and the economics of the whole shebang (including the paste which is anything but cheap) for max 2-3 boards I’m usually assembling of each design. Plus that I find the paste messy to work with.
 
Thx, for your experiences.
I agree do the DFN first so you can work around the part easier. Since yours is a regulator function, it can test before the rest of the circuit needs to be assembled.
Once you can do DFN/QFN, makes the rest look easy :)
Many years ago we/I did a few BGA designs, it was another animal, we even had a extremely expense re-work station in the HP lab, stupid, because we had Celestica down the road doing the production runs ;-) It was all setup, developing profiles fr each assembly. Celestica had 4-pi xray for the BGA, that was cool.
 
100% manual soldering. Hot air for DFN, fine point Weller for the rest. Hot air for rework. Cleaned extra flux in isopropylic alcohol.

No major temperature difference between same chip and separate transistors cascode.

Next version is coming up nicely:

- Can be battery powered (bypass the low noise regulators and no gain switching relays, takes 50mA both channels from 2x9V cells). Shown in photo is the full version with +/-18V regulators and gain switching.
- jfet cascode
- 0.45nV/rtHz
- 30% smaller board size, fits in 100x100mm PCB size, $20 for 10pcs, 2 days turn around, including DHL shipping, got them the same week :).

It is done, works great. Attached are the schematics and the Gerbers.

To run this design on 2x9V batteries:

- Do not install the onboard power supplies.
- Replace D404, D407, D504, D507 with shorts
- Replace the 2Meg R222, R232 with 1Meg resistors
- Do not install the gain relays RL201, RL301
- Replace R231, R331 with shorts
- Adjust the input stage transistors current to 8mA total (about 2.5V across R203, R303)

Feed from 2x9V batteries directly to the terminal block. The amp takes 18-20mA per channel from the positive battery and 8-10mA from the negative battery per channel. Of course, the dynamic range (or overload margin, if you prefer) will be some 6dB lower compared with the +/-18V version, but still over 20dB.
 

Attachments

  • HPS62-schematics.zip
    515.2 KB · Views: 309
  • HPS62-gerbers.zip
    93.3 KB · Views: 134
This is my first post in this forum ... hi everyone !

The TPS7A4700/TPS7A4701 Datasheet describe two modes for setting the output voltage.
TPS7A4700 : "ANY-OUT" option (program pins) only
TPS7A4701 : "ANY-OUT" option (program pins) and "Adjustable" option (FB pin)
http://www.ti.com/lit/gpn/tps7a47

It seems that in the HPS6.1, output voltage is set with FB pin.
So I will be using TPS7A4701 instead of TPS7A4700 (as indicated in circuit diagram)

Am I wrong ?
 
This is my first post in this forum ... hi everyone !

The TPS7A4700/TPS7A4701 Datasheet describe two modes for setting the output voltage.
TPS7A4700 : "ANY-OUT" option (program pins) only
TPS7A4701 : "ANY-OUT" option (program pins) and "Adjustable" option (FB pin)
http://www.ti.com/lit/gpn/tps7a47

It seems that in the HPS6.1, output voltage is set with FB pin.
So I will be using TPS7A4701 instead of TPS7A4700 (as indicated in circuit diagram)

Am I wrong ?

That is correct, use TPS7A4701. I took the symbol from the TI library and it is labelled TPS7A4700 for both models.
 
Is d201 (d1n4735 on circuit) sod-323 ?
I found a zener BZT52C6V2S-TP that is sod-323 and 6.2v 200mw, is it ok ?
Solder pad seems small for 1w (1n4735)

For those who are searching for Wima 6.8u, it may be hard to find. However you can purchase the Wima SMD-PET-KIT2 that contains 3 units of 6.8u in 6054. Or use the 4.7u also in 6054. But I think it will act as a high pass filter in the 0-20 hz range. Don’t know if it will be right ... what do you think ?
 
Is d201 (d1n4735 on circuit) sod-323 ?
I found a zener BZT52C6V2S-TP that is sod-323 and 6.2v 200mw, is it ok ?
Solder pad seems small for 1w (1n4735)

For those who are searching for Wima 6.8u, it may be hard to find. However you can purchase the Wima SMD-PET-KIT2 that contains 3 units of 6.8u in 6054. Or use the 4.7u also in 6054. But I think it will act as a high pass filter in the 0-20 hz range. Don’t know if it will be right ... what do you think ?

That would be fine, 1N4735 is the device I used for simulation, then forgot to change it.

4.7uF or 6.8uF, doesn’t really matter, both will be fine with negligible roll off at 20Hz.
 
I don't seem to find any 47uF/50V or 33uF/50V SMD caps for the regulator sections in case sizes looking like the photos from 6.2 (C401 looks like maybe 1006 or 1206?).

Anyone have suggestions of where I SHOULD BE looking?

It's an error in the schematic, the output cap is 47u/16V (not 50V) ceramic, 1206

C3216X5R1C476M160AB TDK Corporation | Capacitors | DigiKey

The input 33u/25V is also ceramic, 1206

C3216X5R1E336M160AC TDK Corporation | Capacitors | DigiKey

Both are now 0 stock @ DigiKey, replacements:

GRM31CR61C476ME44L Murata Electronics | Mouser Canada

C3216JB1E336M160AC TDK | Mouser Canada
 
Last edited: