advertisement


LM338K equivalent?

martin clark said:
Thats superb! Well done.
Whats it like under a decent load resistor. Half an amp, say..?
Click to expand...
It's powering my Le Monstre. Bias on each of the output devices is 0.68A, so 1.36A on each rail.

I've set the DC offset and rail voltages into a pair of 8R resistors, although I've not run it with a 2kHz input etc.
 
Last edited:
I spent most of yesterday fettling this PSU and I'm inclined to agree with @Arkless Electronics that it's "a bodge". The reason is that I just can't seem to get it any better than the 400uV that I showed earlier.

It's quite straightforward to get the +ve rail to a noise level that can't be measured by my Fluke scope meter. The noise is all on the negative rail and all (I think) you can do is throw more capacitance at it.

I actually tried to see at what point I could get the +ve rail to show some noise. I've a 1000uF Elna as the first 'C', 0R5 and a 10,000 Epcos before the reg. A 47uF MLCC and a 470uF lytic on the reg output and still nothing! The negative has 2x 15,000uF as the first 'C' and 33,000uF as the second and still gives 400 to 500uV. I guess it could probably be brought down to 200uF with more capacitance, but it's getting a bit silly - also if I go low ESR on the 1st 'C' the rectifier rings like a tuning fork and the noise jumps to 3mV!

I'm wondering / seriously considering an LT1033 for the neg rail. Might give it a go.
 
And another question.. Given the above setup, why am I getting 13.6V across the final caps (before the reg) on the +ve rail and 14.8V on the - ve?

Something not right, or just a product of the increased capacitance?
 
Dan K said:
I spent most of yesterday fettling this PSU and I'm inclined to agree with @Arkless Electronics that it's "a bodge". The reason is that I just can't seem to get it any better than the 400uV that I showed earlier.
Click to expand...

ON the output of any 3pin reg that's really, rally good; and you are unlikely to do better; even if the input were a 'perfect silent DC', these things have internal noise sources from the way they generate the internal voltage reference, which is then ..multiplied-up by the way you set the output voltage.

400uV is as good as it gets, at the output voltage you want, from such regs (any of them)

AND

Voltage 'noise' output has v little to do with ..what you hear, providing that noise is de-correlated with the output! The thermal noise component inside the reg is irrelevant.
IMO - when very much more significant is the load-transient-response of the regulator - which reflects as low output impedance, and can be seen by measuring in how little of the output signal from the device it powers, is seen in the regulated voltage rails. That is the measure of a 'good' regulator. Static 'noise' alone, is not.

ETA: and layout, and the exact positions of connections of bypass caps etc... will dominate at this level. Don't let the idea of 'better' be the enemy of 'Good'. You already have 'better' - but in one static measure, pursuing more of which I ..don't actually think is useful.

Go enjoy listening to the thing awhile : D
 
Oh, another thing - from a philosophical PoV:

What is upstream, supplying the input signal to the power amp?

Since the SNR of that stream is then multiplied by the gain of the power amp - advances upstream are everything. They utterly dominate the outcome at the speaker, to a degree that worrying about low 100s of mV ripple on the power amp rails is daft. Probably even for a thing with as dismal a PSRR as the Monstre.

Any good power amp is usefully-silent with its inputs-shorted (if not - it's no use as a power amp; and, if DIY - there is likely something rather awry in the exact way it was wired-up. )
 
martin clark said:
What is upstream, supplying the input signal to the power amp?

Since the SNR of that stream is then multiplied by the gain of the power amp - advances upstream are everything. They utterly dominate the outcome at the speaker, to a degree that worrying about low 100s of mV ripple on the power amp rails is daft. Probably even for a thing with as dismal a PSRR as the Monstre.
Click to expand...

Fair point. Pre is a Croft 25R with the line stage replaced with a Broskie Aikido (an augmented cathode follower). It's extremely good/ quiet, but I doubt it's 400uV quiet.

The thing is that I managed to get the SLB CapMx down to +/-200uV at 12V (maybe that's +/-249uV) and this reg'd supply was supposed to 'better' it for both noise and dynamics.

Subjectively the amp may now be 'faster', bass certainly is darker (but that's the added black gates). Not a great deal in it so far
 
Last edited:
Dowser said:
How does it sound? :)
Click to expand...
Initial subjective impression is that the amp is faster/ more 'Naimy', it seems to be more 'caffeinated' than with the SLB.

I'm now wondering about the noise level from the SLB. It has occurred to me that the 200 to 249uV noise that I was getting on each SLB rail (at 12V) is self cancelling. This reg'd supply is not. Noise seems to be 400uV on the negative rail and that's it.

I did discover that I can make the speaker earth returns a lot quieter by routing via the amp boards, rather than returning directly to the PSU, so that's a win.

I was hoping to get better seperation and soundstaging, but I'm not entirely convinced that has really happened.

With this supply I've noticed that the DC offset is wandering up/down all the time, tiny amounts - the SLB was rock solid.

I've also got some high freq hash coming from somewhere.. Not the traffo ringing (have installed snubbers). Could it be the Mosfets in the rectifiers, or maybe the regs themselves??
 
Last edited:
This is the speaker terminals into 8R

Yes, thats almost no noise, but what are those spikes about? Something somewhere is ringing... Or there's a cheap switch mode PSU on the mains
 
Last edited:
Spikes are 10ms apart, so they are rectification spikes.
I also sometimes see bursts of low level, high frequency stuff getting into circuits. Not oscillation as not present at different times of the day, so must be cr4p on the mains.
The slight variations in dc offset could be due to low frequency, minor variations in the regs that are not correlated as these will change the quiescent conditions in the amp. I had a similar problem with a phono amp...where the very highest gain is at low frequencies.
 
a.palfreyman said:
Spikes are 10ms apart, so they are rectification spikes.
I also sometimes see bursts of low level, high frequency stuff getting into circuits. Not oscillation as not present at different times of the day, so must be cr4p on the mains.
The slight variations in dc offset could be due to low frequency, minor variations in the regs that are not correlated as these will change the quiescent conditions in the amp. I had a similar problem with a phono amp...where the very highest gain is at low frequencies.
Click to expand...

The spikes might just be picked-up direct into the loop area of the meter's probe wires - try twisting those loosely together, and trying again.
 
You could also replace the V ref resistors (860R?) with zener diodes. Did you have 33uF low ESR across these? Leave them in and get a bag of 11V zeners. Typically 5% tolerance so you can put 12mA through them on the bench (from you 100R set resistance) and pick a pair at 10.7V. That should stop the low frequency 'waffle' from the regs.
 
a.palfreyman said:
You could also replace the V ref resistors (860R?) with zener diodes. Did you have 33uF low ESR across these? Leave them in and get a bag of 11V zeners. Typically 5% tolerance so you can put 12mA through them on the bench (from you 100R set resistance) and pick a pair at 10.7V. That should stop the low frequency 'waffle' from the regs.
Click to expand...
Huge thanks for this tip. Any chance of a link to a picture, or an example?
 
I've started working on the amp's voicing..

Tried with & without the 470uF Black gate F's and 47uF ceramics on the reg outputs, appologies if you think BGs are BS, to these ears the 470uF BGs are best without the smaller cap across them.

Next step will be to try the 47uF MLCCs in place of the (quite large physically) 33uF K73-16 bypass caps

Sounding very good now, might well have the measure of the SLB. Not looking pretty though
 
Dan K said:
Next step will be to try the 47uF MLCCs in place of the (quite large physically) 33uF K73-16 bypass caps
Click to expand...

This didn't sound good either, strange that I convinced myself I could hear a difference. My preference was for the larger K73-16 caps. More musical/ flowing presentation. The larger value / physically tiny MLCC caps seemed to subjectively slow everything down
 
Last edited:
You must log in or register to reply here.


advertisement


Back
Top