My_Ref Fremen Edition - need help on PCB evaluation

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
really strange. :scratch: The schematic is correct.

Perfect

Are you sure that you have actually connected everything according to the schematic? (make sure the 3886 is actually connected before the diodes!).

Yes, I think so

Perhaps the 220u are a bit too small? Don't you have any larger ones?

Not 50V rated...

The problem may also be the new circuit layout.

No, I'm still using TP boards and a breadboard, see attachment.

Where is the cap common ground point connected? consider that with the cell the currents on the 318 closes on that point rather than on the general star ground as they did before (using only a rail-to rail cap after the diodes should restore the previous situation WRT ground currents layout).

In my actual 'temporary' setup there is no much control on it...

Just in case, you may also want to try some different diodes types if you have some at hand.

I have UF4003, SF17 and SBYV27-100

BTW: do you have a scope? can you check the differential voltage across the diodes while driving a real (or real-like) load at normal listening levels?

Sadly no, no funds... :(
 

Attachments

  • IMG135.jpg
    IMG135.jpg
    226.5 KB · Views: 473
No, I'm still using TP boards and a breadboard, see attachment.
:D

ahem... with new layout I meant the changes in the actual current loops produced by the introduction of the D-C cells. ;)

In my actual 'temporary' setup there is no much control on it...
that's a problem. In circuits like this one the actual layout does have a huge impact on the SQ.

I have UF4003, SF17 and SBYV27-100
well, given that with such setup it's very easy, let try them all and "see" if there are any differences.

P.S.: about the C, you may also try connecting in parallel the ones you have (likely a bad idea, but... in this case trying costs nothing).
 
that's a problem. In circuits like this one the actual layout does have a huge impact on the SQ.

I'm sure about that but, since also the 'plain' regulator shares the same bad layout I think it's a fair comparison ;)

well, given that with such setup it's very easy, let try them all and "see" if there are any differences.

There are differences for sure:

1N4007: 'live' sound, materic, precise but a bit hard
UF4003: 'live' sound, materic, precise
SF17: soft, warm sound, maybe a bit dull

between D-C net and without difference is quite small but the D-C (with UF4003) sounds more direct, materic and with a firmer bass.

So far it seems a slight improvement but tomorrow I can do a better comparison since I'll have 50R resistors for the LM317s and BD139-16/140-16 for the shunt transistors.

I'm curious to see if that difference will remain or it will fade...

Oh, BTW: is that setup dual-mono or stereo?

Dual mono: two transformers, two PCBs, two breadboards... ;)
 
Last edited:
I'm curious to see if that difference will remain or it will fade...

Just tried first 50R (23ma) and BC639-16/BC640-16 both with and without the D-C net.

Interesting results...

  • as expected 50R gave back to the sound the same puch it had with LM3x7L (TO92)
  • With 50R the D-C net still improves sound but not much
  • With BC639-16/640-16 are a clear improvement over standard BC639/640, bass much tighter and faster, cleaner sound
  • With BC639-16/640-16 the D-C net seems no more an improvement
It seems also that resitors values needs some fine tuning.

Beta testers will have to work... :D ;)

In the meanwhile I've read again this great W. Jung's article and I'm tempted to try the LM317+MOSFET cascode CCS (image 13c).
 
In the meanwhile I've read again this great W. Jung's article and I'm tempted to try the LM317+MOSFET cascode CCS (image 13c).
I don't like it. There's still the problem of the limited gain/bandwidth of the LM317 internal op-amp, and there's a lot of unneeded complexity. It may work well on paper, but I would not be so sure that it will work as well in practice, under real-world dynamic conditions. IMHO the cascoded MOSFET is better.

(incidentally, it's much the same thing, with the 317 in place of the second MOSFET. Which will have a flatter response WRT frequency. I wonder why he tried a cascode with two 317 and did not try with two MOSFET... I bet that would have won hands down, at least in term of freq. response).
 
Last edited:
I don't like it. There's still the problem of the limited gain/bandwidth of the LM317 internal op-amp, and there's a lot of unneeded complexity. It may work well on paper, but I would not be so sure that it will work as well in practice, under real-world dynamic conditions.

I thought that it could merge the best of both world... :(

BTW I've also made the comparison NatSemi Vs OnSemi LM3x7...

They've a pretty different sound: rounder but a bit harsh the NatSemi, warmer, more refined and tight bass the OnSemi...

I wasn't expecting so much difference...

Tomorrow I'l compare the BC639-16/640-16 with the BD139-16/140-16...
 
Last edited:
This could be going around in circles, unless one is able to generate a criteria in listening tests clearly identifying what is going on, for example:

1. Does the background seem cleaner or more complicated?
2. Is there increase or loss of detail?
3. Are the harmonics of various instruments more separated or more mixed?
4. Is there increased image depth between the front and rear instruments or reduced?
5. Is there more differentiation in recorded room reflections or less among recordings in various types of environments?
6. Is the sound more fatiguing or less?
Repeat the process for various listening levels and consider the same.

Pure listening impressions may vary over time. It would be best if correlation with measurement data could be established. When we think that some configuration is clearly a winner, it is necessary to find out through measurements what really is improved. Good combination of data can pin point to what is going on.
 
This could be going around in circles, unless one is able to generate a criteria in listening tests clearly identifying what is going on, for example:

1. Does the background seem cleaner or more complicated?
2. Is there increase or loss of detail?
3. Are the harmonics of various instruments more separated or more mixed?
4. Is there increased image depth between the front and rear instruments or reduced?
5. Is there more differentiation in recorded room reflections or less among recordings in various types of environments?
6. Is the sound more fatiguing or less?
Repeat the process for various listening levels and consider the same.

Pure listening impressions may vary over time. It would be best if correlation with measurement data could be established. When we think that some configuration is clearly a winner, it is necessary to find out through measurements what really is improved. Good combination of data can pin point to what is going on.
Forgot to add.

7. How does music polarity effect the above results?

Of course it is always good to know how our own ears perform.
 
Last edited:
...
In the meanwhile I've read again this great W. Jung's article and I'm tempted to try the LM317+MOSFET cascode CCS (image 13c).

It might be a major convenience to keep the regulator(s) on a daughtercard, with a standardized pinout - either +IN, GND, +OUT or +IN, -IN, GND, +OUT, -OUT. The spacing on the PCB between +IN/-IN and +OUT/-OUT should preferably be the same as what is currently being used for the 1K/2W resistors (R1/R4). This way, one can assemble the original Rev C BoM with R1/R4, and upgrade to improved regulators later as required. The +/- 12V regulator development can proceed on a separate track at its own pace, and improved regulators can be retrofitted later as required, without holding back the main board layout, which can be frozen and kept the stable for a longer time. It will be more convenient both for you (as board supplier) as well as the DIYer.

My 2c about the shape/pinout of the daughtercard: Rectangular; with +IN, -IN along one short edge; +OUT, -OUT on the other short edge; and GND forming a small star-ground at the centre. It could be a plug-in or soldered (electrically better) module.
 
This could be going around in circles, unless one is able to generate a criteria in listening tests clearly identifying what is going on, for example:
....
Pure listening impressions may vary over time. It would be best if correlation with measurement data could be established.

You're list of criteria seems good (apart the polarity one that I don't know how to identify it) for a complete listening test but I don't know if I will be able to follow it.

My acoustic memory is limited so I listen to short parts of music, make the swap as quick as I can and listen again to the same short part.

I do so with various music until I'm sure the difference is clear.

This method already takes a lot of time, following also your list would take a lot more...

Probably you're a more experienced listener and identify such criteria in few seconds, I'm able to do so with:


  • timbre
  • soundstage
  • bass performance, HF harshness
BTW I'll try to add what I can of your criteria


When we think that some configuration is clearly a winner, it is necessary to find out through measurements what really is improved. Good combination of data can pin point to what is going on.

Sadly, I can't.

I don't have the tools...

It might be a major convenience to keep the regulator(s) on a daughtercard, with a standardized pinout
...
The +/- 12V regulator development can proceed on a separate track at its own pace, and improved regulators can be retrofitted later as required, without holding back the main board layout, which can be frozen and kept the stable for a longer time.

I thought about it, it's smart idea but possibly I prefer to have it on the PCB.

At the end the phase which needs more testing and swapping of different regulators is the development/beta one.

BTW we don't need the best regulator in absolute terms but the most performing which fits the board... ;)

It will be more convenient both for you (as board supplier) as well as the DIYer.

I don't want to be a board supplier... I'll simply organize a GB for the first batch of boards. ;)
 
I thought about it, it's smart idea but possibly I prefer to have it on the PCB.

At the end the phase which needs more testing and swapping of different regulators is the development/beta one.

BTW we don't need the best regulator in absolute terms but the most performing which fits the board... ;)

Maybe it is possible to define the borders of the regulator area on the pcb and the positions of the Ins and Outs. So the separate regulator-development-track can work but has a requirement specification from the main project. This is the usual development approach for complex systems so why wouldn't it work in this case?
If this is possible Sivas idea can take place and your demand of having the regulator on the pcb is met.

Just my 2c.

Best regards
Sven
 
You're list of criteria seems good (apart the polarity one that I don't know how to identify it) for a complete listening test but I don't know if I will be able to follow it.

...
changing polarity is just swapping the the + & - speaker leads in both channels at the same time. Generally this will change the perception of the whole sound field in more complicated performances.

Personally, I only change one channel at a time during listening to shorten the listening session, one is always the baseline. Take turns in the modification until a point where I think could be a good milestone, then make both channels the same. It took some time to learn how to do it efficiently. When the milestone is established, then it gets compared against others to see if I am still on track. So the process is my own perception first, the comparison against existing equipment. Yes, it is time consuming.
 
changing polarity is just swapping the the + & - speaker leads in both channels at the same time. Generally this will change the perception of the whole sound field in more complicated performances.

I'll try

Personally, I only change one channel at a time during listening to shorten the listening session, one is always the baseline.

In my setup there is no balance control, so for me it's not an (easy or fast) option.

BTW a lot of changes in soundstage are audible only if both channels are modified.

Since I always use sockets (and breadboards) my reference are both channels, reverting changes in a very short time. ;)
 
Sound stage change is more audible with both channels modified; however the trick is to listen to two different channels at the same time without changing the balance because we know there is no gain change in the circuit. If under such conditions two channels seem to have a wrong balance, then the modification is creating a significant difference. This is then easier to determine which is better. Then the channel that is worse gets modified to verify whether this design change is truly better.
 
however the trick is to listen to two different channels at the same time without changing the balance because we know there is no gain change in the circuit. If under such conditions two channels seem to have a wrong balance, then the modification is creating a significant difference.

I do agree, I use it as a check.

This is then easier to determine which is better.

Here balance control is needed.
 
I have not used balance control for over 20 years. But if you think you need it, then probably you do. I tend to try and train my ears. As I turn the volume up, the side that gets fatiguing first generally is not the good side, but it's also necessary to mentally focus on the other side to see if some detail penetrates clearly but not so loud as I remember it.
 
Well, you're a more experienced listener.. for sure ;)

BTW

This evening, while swapping transistors and resitsors and I fried one of my monoblocks so no much time for critical listening...

So far BD139/149 seems very similar to BC639/640 so probably, since they can handle more power, they'll be in the production board.

I've also tried to reduce to 200R the 270R resitstors and sound became fuller and rounder loosing a bit of speed.

I've yet to underestand which value is more balanced...

PS

The module is working again thanks to a new LM318
 
Last edited:
So far BD139/149 seems very similar to BC639/640 so probably, since they can handle more power, they'll be in the production board.

I've also tried to reduce to 200R the 270R resitstors and sound became fuller and rounder loosing a bit of speed.

Did you ever consider swapping D1 & D2 with the largest chokes that would be able to fit in there? You could even wind your own chokes on ferrite cores, using either straight rods or toroids. Pre-made common mode chokes are out there, and the LM318 doesn't draw much current compared to the LM3886. Might be worth a try. Maybe one of the readers here with more test equipment could do some measurements.

Just a thought. ;)
 

Attachments

  • Dario Power Supply - 02-07-2012.jpg
    Dario Power Supply - 02-07-2012.jpg
    76 KB · Views: 224
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.