RJM Audio Sapphire Desktop Headphone Amplifier

Can you detect any differences in sound quality between closed and open loop?

Well, rather not. I would need to spend more time testing it. With two different headphone sets, the results were different, but it is certainly not "quality" difference, just the different taste. With both options the sound is good - no distortions, ho hum, no noise.

I assembled both option at the start only for test purposes, just to measure things. For the little while I had set them differently I observed the difference of the sound level between the two channels so if was not an option.
When decided to actually listen to sound of the amp, I made both channels the same, and finally stayed with open loop and gain 26db (headphones 300ohm 94db SPL/mW)

Open loop seems to be more colorfull, more pleasant, adding more to the sound signature. Closed loop sounded more sterile, the dynamics was better but overall sound was rather harsh when compared to the open loop.

With high gain I would stay with open loop, but for low gain settings the closed loop option seems to be better.

I put pins in place of resistors and switched between a lot of different options, so I'm quite confused at the moment. I need to listen one fixed configuration little longer to get used to it and then after the change is made it will be easier to tell if the change is for better or not.
 
With high gain I would stay with open loop, but for low gain settings the closed loop option seems to be better.

I'm not doing the high gain thing since I'm using this as a preamp and not as a HA.
The first boards I built were closed loop with 6dB of gain.

For the next boards I assemble, I'm thinking about trying the same 6dB gain and the open loop configuration.

Sounds like you may have confused yourself with not maintaining the same gain/loop configurations for both channels.
 
An externally hosted image should be here but it was not working when we last tested it.

This is my build of sapphire4.
I just started populate, waiting for capacitors delivery.
I also made Power PCB to have transformer and bridge rectifier in one PCB.
following modifications are made on main PCB.
1. One Nichicon KG (20mm dia snap-in) instead of two parallel Nichicon KW 1000uF.
2. Around Zener Regulator I changed 4 Caps from Nichicon KW 100uF/25V (2.5mm ls, 6mm dia) to Nichicon FG 100uF/25V (3.5mm ls, 8mm dia)
3. Around Voltage Amplifier and Buffer stage, I changed 6 Nichicon KW to 6 Nichicon KZ 100uF/25V (5mm ls, 10mm dia)
4. changed 0.1uF ceramic capacitor footprint to 5.5mm/7.5mm FKP Print to use Vishay MKP1837
5. for gain switch, I used a two position DIP switch, instead of two jumper switch.
6. added 3mm dia LED (for each of + and - rail) to identify board is working or not.
 
Last edited:

rjm

Member
Joined 2004
Paid Member
I'm looking forward to see the final assembly.

One small thing: since you've used larger capacitors, it may not be possible to screw together the heatsinks and transistors once the capacitor C10,11 and C2,3,6,7 are in place. Clearance was already tight on the original layout. So you might want to fit the transistors/heatsinks first.
 
The very first modification I would do if I were making new PCB would be to move away C8, C9 so instead of separate heat sinks for transistors I would install one common heat sink vertically across PCB from top to bottom - a cheap bare piece of aluminium that assure that both channels have the same temperature, and you don't need extra place vertically anymore to accommodate typical to-220 heat sinks.

cheap, effective, and useful in term of temperature stabilization of both channels.

On the original PCB I mounted all transistors under the PCB, getting rid of heat sink, using aluminium enclosure case as large heat sink - easy to assemble but difficult to service and dis-assemble.
 

rjm

Member
Joined 2004
Paid Member
The heatsinks are barely needed except when bench-testing the full output power. Each transistor only dissipates 400 mW under normal operation.

I tried various layouts and heatsink arrangements, including oddly enough different iterations of the metal plate idea - my original plan actually. Not saying it couldn't be done, but the layout is fairly cramped and from my point of view the vertical heatsinks were the most space efficient option that was still convenient. Also, I had the heatsinks left over. :D
 
Update :
Power test : works fine.
LEDs are too bright, I need to add some resisters.
Will finish/listening test soon.

Xr51kYk.jpg
 
Last edited:
Had problem with output impedance, just solved.
I ordered takman 4.7ohm but, it was 4.7k ohm :(
Now it works charm.
I temporarily put it in cookie box :)
galtuC0.jpg

I used 10k ladder attenuator from allo,
Mundorf supreme 0.68uF,
Nichicon KG super thru 2200uF,
Nochicon KZ Muse and fine gold.
Mix of takman and PRP metal film resisters.
Mogami OFC cable, BC550C and BC560C without matching, but filtered odds.
BYV27 for rectifier, and RS-component transformer, which is actually Talema's OEM.

Comparing to my previous project - Lehmann clone with discrete OPAMP, Alps RK27, Mundorf AG 4700uF, Silmic II, Clarity Cap ESA for coupling, all takman metal film - Sapphire has even clearer sound and wider stage with same depth of bass. I'm very impressed.

I have tested one transformer configuration vs separate transformer configuration, unfortunately I noticed nothing difference.

I have tested both of open loop and close loop,
Open loop gave me more dynamics, more bass, and much tighter bass. Awesome for pop songs, but ears get easily tired.
Close loop provides wider stage and clearer high tone. It's bass is comfortable. Much more reference sound.
I can't say which one is better, but I still can enjoy both sound color, and which is good thing.

I need to spend more time to define my feeling, but my first impression is very positive and far better than lehmann's diamond buffer.
 

rjm

Member
Joined 2004
Paid Member
Thank you for posting your build notes and listening impressions. I get the feeling we are homing in on the closed/open loop sound difference now with several people reporting fairly consistent opinions.

I imagine other people might be interested in your power supply boards. Would you mind posting the gerber/brd files?

Lastly do you have a link to the allo attenuator? I haven't seen that model before.
 
Attenuator link :
Allo.com - Stepped Attenuator

It came with gold plated Chinese switch, I measured each resister value as below.

L-Ch R-Ch

9.98k 0.5 9.99k 0.4

9.98k 2.4 9.99k 2.3

9.98k 4 4 9.99k 4.3

9.98k 7.4 9.99k 7.3

9.98k 13.4 9.99k 13.3

9.98k 22.4 9.99k 22.3

9.98k 40.5 9.99k 40.4

9.86k 70.8 9.86k 70.7

9.86k 113.1 9.86k 113.2

9.86k 177.9 9.86k 177.8

9.74k 249.0 9.74k 249.0

9.63k 353.0 9.63k 353.0

9.51k 493.0 9.51k 493.0

9.29k 698.0 9.29k 698.0

8.96k 989.0 8.96k 989.0

8.54k 1.403k 8.54k 1.404k

7.94k 1.978k 7.94k 1.978k

7.48k 2.51k 7.49k 2.51k

6.79k 3.14k 6.79k 3.15k

6.03k 3.95k 6.03k 3.95k

4.98k 4.97k 4.98k 4.97k

3.69k 6.31k 3.69k 6.31k

2.04k 7.93k 2.04k 7.93k

0.5 9.96k 0.5 9.96k

You can share gerber of power board, since it was based on your board outline design :)
 
Bridge Rectifier questions

I've been reading many posts in this thread with interest and am keen to go the 4.1m version soon.

I'm trying to understand something in your power circuit: The use of 2 x diode bridges per transformer.

I've attached a diode bridge arrangement from a long running commercially produced headamp which shows one bridge for a dual supply.

Can you explain why the difference?

Also, why such high rating? I recall you mentioned you had spare oversized ones when building the earlier revisions, but is this a requirement on the latest BOM to maintain performance in any way?

Finally, have you considered a snubber RC filter across discrete diodes instead of a potted IC as used?
 

Attachments

  • BCL Power circuit snippet.jpg
    BCL Power circuit snippet.jpg
    60.3 KB · Views: 587
elmura - not sure that putting small value caps across bridge diodes constitutes a snubber. That would be a series R-C series connection. As I see it, the caps alone would just provide a low impedance path for any mains or diode-generated noise into the power supply.

Isn't the purpose of a snubber to damp the LC resonance between the transformer and smoothing caps? If so it should go across the transformer secondary winding not the diodes shouldn't it?

As for the dual bridge question - I think it's because your pic shows the arrangement for a centre tapped secondaries transformer whereas the other is for twin secondaries.

Maybe I'm wrong, anyone care to comment?
 
Last edited:

rjm

Member
Joined 2004
Paid Member
@elmura

1 vs 2 bridges:

The main reason for recommending 2 is it's harder to mess up: for dual secondaries connection the phase of the windings does not impact what wire goes to what. With one bridge, you have to tie the winding together positive to negative for the center tap. If you only have a center tap secondary, of course, you must use only one bridge.

There are other subtle differences: one bridge will give 60 Hz component as a result of any imbalance in the center tap, two bridges only give 120 Hz component. 60 Hz is harder to filter out so...

Downside is double the diode drop, and a small imbalance between the value of the positive and negative voltage rails.

Diodes, and the power supply configuration generally, are completely open. The BOM values are just my long-running suggestion. I like those large bridges as the stock can be used for any project, they are easy to mount to the case, and they are cheap on ebay.
 
@rjm

I'm looking into local supplies of transformers and questioned the reasoning behind dual 12v vs dual 15v output. With various active devices I've read in the datasheets (plus in my own measurement findings) that higher rail volts equate to lower THD. Will the Saphire 4.1m circuit safely handle the higher voltages of 15V?