GC power supply performance

[SheldonD]

I built Brian GC premium kit, love the performance, quiet and clean clear sound.

But I have some demanding records like Bach Organ pieces and I thought I noticed compression on loud passages. I checked the voltages on the +/_ rails and found considerable sag on the DC voltages.

I have Plitron 2x22 volt 400 VA which gives about 34 volts DC per rail.
On heavy passages with lots of bass I was getting up to 10%. voltage sag. This is on the rectified DC only. Monitoring the AC inputs to the rectifier circuit there was absolutely no drop. So the Xfrmer has excellent regulation.

So this leaves diodes and filter caps, the standard BG4.7 ufd and the 1500 ufd at the chips.

PHASE I

I then added in parallel rectifier board and added 2x5600 ufd Nichicon per rail.
The voltage sag now is under 3% and I feel incresed clarity and more open sound though not tested A/B fashion.

PHASE II
I added 2 LM338 complete with protection diodes and extra tantalum 1 ufd at reg inpiut and output, follwed the specs in the LM338 datasheet. 120 ohm and 5k adjpot to get the3 volt drop.

There did not seem to be any better regulation and the voltages were now down by 3 volts, the sound was not as good but I did not change the chip caps, and also left the 2x5600 on before the regs.

I have since removed the reg circuits and am running as of PHASE1.

Has anyone else checked the rail voltages dynamically under heavy load? With or without regs?

SheldonD


I built Brian GC premium kit, love the performance, quiet and clean clear sound.

But I have some demanding records like Bach Organ pieces and I thought I noticed compression on loud passages. I checked the voltages on the +/_ rails and found considerable sag on the DC voltages.

I have Plitron 2x22 volt 400 VA which gives about 34 volts DC per rail.
On heavy passages with lots of bass I was getting up to 10%. voltage sag. This is on the rectified DC only. Monitoring the AC inputs to the rectifier circuit there was absolutely no drop. So the Xfrmer has excellent regulation.

So this leaves diodes and filter caps, the standard BG4.7 ufd and the 1500 ufd at the chips.

PHASE I

I then added in parallel rectiier board and added 2x5600 ufd Nichicon per rail.
The voltage sag now is under 3% and I feel incresed clarity and more open sound though not tested A/B fashion.

PHASE II
I added 2 LM338 complete with protection diodes and extra tantalum 1 ufd at reg inpiut and output, follwed the specs in the LM338 datasheet. 120 ohm and 5k adjpot to get the3 volt drop.

There did not seem to be any better regulation and the voltages were now down by 3 volts, the sound was not as good but I did not change the chip caps, and also left the 2x5600 on before the regs.

I have since removed the reg circuits and am running as of PHASE1.

Has anyone else checked the rail voltages dynamically under heavy load? With or without regs?

SheldonD

[ble0t]

Interesting...I have a basic kit from Brian running on an Avel 330VA with 25V on the secondaries which gives me about 35V-36V DC. I've done a little testing and a lot of listening with stuff like Chemical Brothers and Daft Punk and haven't noticed anything too serious...my voltages did drop by a couple volts, but I wouldn't say my sound was affected adversely.

Even so, I like the results you and others seem to have gotten from regulating with those LM338's...I think I am definitely going to look into doing that with my next GC.

[NealG]

Hi Sheldon,

Yes, found exactly the same. I was using a 300VA 2x18 toroidal unregulated and noted a rail drop of 10%, adding more capacitance (10,000uF) per rail did not alter this (naturally ) but it did lower the ripple to 10% of the supply vs 34% with just 2,000uF / rail.

With the LM338 I changed the trafo to a 300VA 2x25V one and now measure no sag. Currently I am using 10,000uF before the reg, .22uF polyprop film and foil immediately after it and 100uF at the 3875 it's self, these are also bypassed with .1uF box Wima. This configuration is way better to my ears than the unregulated supply.

As Carlosfm indicated the cap values and types play a major role on the sound quality and I have found this to be true, I am still experimenting. I think Carlos uses 33uF at the 3875 and .047uF at the 338, could be wrong though.

Neal

[carlosfm]

Originally posted by NealG
As Carlosfm indicated the cap values and types play a major role on the sound quality and I have found this to be true, I am still experimenting. I think Carlos uses 33uF at the 3875 and .047uF at the 338, could be wrong though.
Neal

Me?
I use the LM3886, the rest is correct.

[carlosfm]

Originally posted by SheldonD

PHASE II
I added 2 LM338 complete with protection diodes and extra tantalum 1 ufd at reg inpiut and output, follwed the specs in the LM338 datasheet. 120 ohm and 5k adjpot to get the3 volt drop.

Remove that tantalum cap and put a small polyester cap (47~100nf).
Change the 1500uf caps on the amp for 33 to 100uf.
You're almost there.
Listen and tell us what you think.

[SheldonD]

Thanks for replies, folks:

Carlos: You're suggesting remove only the one tanatalum after reg? And replace with polyester, I can leave the first one in.

BTW have you measured voltage drop/regulation under load? I feel this standard PS as in Brian's kit to be inadequate, unless you do not play at high volume and have efficient speakers. It is a good start.

Even Peter has commented on batteries improving the sound,
this almost certainly has to be the constant voltage supply. At least the major reason.

Regards,

SheldonD

[carlosfm]

Sheldon,

Read this and enjoy:
High-End Regulated Buffered Inverted GC

On the first post there's a link for another thread, you must read it first.
So take a break, drink a cold beer, read this.
It's everything here, isn't it?

[Peter Daniel]

Hi Sheldon,

Are you using those 5,600 brown Nichicons from Sayal?
I tried them on LM4780 and they completely messed up the sound, single 1,500 Panasonics are much better.

[carlosfm]

With unregulated PSU you shouldn't pass 1500uf, and 2x5600uf kills the sound.
With regulated PSU you can use what you want before the regs.

[SheldonD]

Yes Peter they are Sayal brown caps.

They may be far from the best but:

I listened for some time a coiuple of weeks and liked the standard GC from Brian but I noticed a sort of strain(?) to high volume particularly with much bass content. Bach pipe organ pieces.

So I checked the DC supply rails under power and was dismayed to see 10% sag on the rails.

As others have pointed out this is just the overall voltage it does not measure the greatly increased ripple under these conditions.

I do not have scope just DDM so can't measure this.

The addition of caps after the PS PCB reduced this to less than 3 % and to me there was a more open sound under the above conditions. To my ears a definite improvement.

The voltage drop must lead to compression of the sound on peaks and increased ripple. Either of which cause distortion.

I believe the battery approach would be ideal if somewhat impractical.

I do have fuses on the rails immediately after the PS PCB and this forms a small resistance, which may affect the sound.

I will be following up on Carlos' suggestions, and will post, after getting some new caps.

Regards,

SheldonD