Lampy cd player

Dedicated to the silver disk spinner
User avatar
Mike H
Amstrad Tower of Power
Posts: 20156
Joined: Sat Oct 04, 2008 5:38 pm
Location: The Fens
Contact:

#106

Post by Mike H »

Oh yes? Image


 
 
"No matter how fast light travels it finds that the darkness has always got there first, and is waiting for it."
Richard
User
Posts: 95
Joined: Fri Mar 18, 2011 10:21 pm
Location: Notts

#107

Post by Richard »

My brain hurts like it's been hit with a cattle prod this week :mrgreen:

The digital stuff must have given it quite a jolt as I found this easy to read and understand,

http://valvewizard2.webs.com/SRPP_Blencowe.pdf

It shows I think that Lukas has read or learned something similar. Half mu and 200R cathode resistors. I had wondered if he'd drawn it incorrectly as he does refer to it as SRPP several times on one page. I can now see the reasoning and how he might have arrived at it (though I think the cons outweigh the pros).

Yesterday afternoon I reworked the stage into Thorsten's SRPP, with 82R at R1 and 22nF at C1 (ok it was a token gesture, should have been 39nF but it was all I had and may still take off some ultrasonics). Didn't fit the de-em network.

Image

After playing with the B+ I got it so I could switch between ECC88 types and 6n2p, getting 180V at the anode with the former and 230V with the latter without blowing my 250V rated supply caps :roll:

The evening was then spent rolling 4 ECC88 types and the 6n2p.

This is rambling so basically,

The scope traces still all look like the ones previously posted.
I can still hear tones at the 15, 18, 20KHz test f but they are very small level.
The sound is much cleaner and expansive at the top in partic with both valve types.
The ECC88 are all quite different in character with one in partic being very good and one pretty dire.
The 6n2p is unbelievably good, huge, powerful, open, expansive, natural in a non hifi way.

All gave level response into 47K but the gain is lower than Thorsten suggests for the valve types, I got 4.4V p-p with ECC88 and 8.8V p-p with 6n2p.

The night (til 4 anyway) was spent reading and trying to digest this against that excellent paper.

Conclusion is that this posted SRPP stage suits 6n2p very well indeed and 6n2p in SRPP suits a 50K upwards load very well indeed. The stage ironically is not optimised for ECC88 and really needs larger Rk for that into 50K load. (So I may not yet have heard ECC88 at their best.)
Max N
Old Hand
Posts: 1451
Joined: Thu Aug 16, 2007 6:10 pm

#108

Post by Max N »

This looks very similar to the circuit in my Audionote DAC kit 1.2
(ecc88 srpp)
Its probably just coincidence but it does look very very similar.:wink:

My kit is pretty old and had an analogue filter after the IV resistors, and then direct-coupled to the ecc88 srpp. The current Audionote kits don't have the analogue filter, but use transformers to couple from the digital to the analogue output. The 3.1 version does away with the srpp and goes for 5687s into 33:1 step down transformers. So if you want to go the whole hog there are some ideas for you.
This hifi collective page has links to the two DAC versions, and from there you can glean quite a lot.
http://www.hificollective.co.uk/kits/audionotekits.html
Richard
User
Posts: 95
Joined: Fri Mar 18, 2011 10:21 pm
Location: Notts

#109

Post by Richard »

Hi Max,

Thanks for that. Can you send me the circuit please or let me know what Rk is?

Ecc88 heaters are the same pinout voltage and current so you might like to try 6n2p in yours depending on those cathode resistors and your load.

You'll know there are advantages to using TX output with a preamp or line drive stage partic in that the low op impedance will allow driving of cables of various lengths and LC and into inputs of various loads perhaps 10k to 1M without big issue. I'm not overly convinced it is a better solution per se.
Interestingly the new AN dac uses the L3 output stage I think which is the same circ as WD Pre3.

Srpp is a good line stage into a known fixed load if optimised for the load but variable if not. Now we see an advantage to the manufacturer and user of using TXs.

Reading Merlin Blencowe's paper I picked up on a few points in partic;

An unbypassed lower triode will greatly reduce the distortion and leave it dominated by warmer 2nd harmonics from the second triode.
(This is as opposed to using a bypass cap when 2nd harmonic dist will largely cancel across the 2 triodes leaving mostly low level odd harmonics.)
I wonder if the bypassed version is where srpp's cool/analytical reputation comes from.

The output impedance of srpp is not a great issue with sensible interconnects as it is more dependant upon matching to load. It delivers max power when load = ra of the valve.
This suggests a "perfect storm" with 6n2p ra 50k optimised for a 50k load, and suddenly this high mu valve starts to look good.

The circ I've posted here errs on the side of lower load (25k) and benefits from a slightly lower Zout of 26k instead of 30k than if set for 50k. It would be interesting to try it again with Rk 1.5k and Ra 3k though the differences may be small as the qualities of this valve shone right through from Lukas' circ I tried previously.

It's still better to have a load a little too high than too low as distortion comes in faster with lower loads.

Thorsten notes that his drawing is presented as a universal circuit and says different valves and loads may be used. Using it as drawn with ECC88 and 1k resistors it would work best with loads around 15k - 20k. For 50k load the theory says the cathode resistors need to be nearer 3k and an Ra of 6k added though I haven't tried it.

Image
User avatar
Nick
Site Admin
Posts: 15694
Joined: Sun May 06, 2007 10:20 am
Location: West Yorkshire

#110

Post by Nick »

I haven't done the numbers, so its just a thought, but remember you are not driving a perfect resistive load, there is L and especially C in there, so the load will fall with incresing frequency, so with the SRPP the distortion will tend to increase with frequency.
Whenever an honest man discovers that he's mistaken, he will either cease to be mistaken or he will cease to be honest.
Richard
User
Posts: 95
Joined: Fri Mar 18, 2011 10:21 pm
Location: Notts

#111

Post by Richard »

Nick wrote:I haven't done the numbers, so its just a thought, but remember you are not driving a perfect resistive load, there is L and especially C in there, so the load will fall with incresing frequency, so with the SRPP the distortion will tend to increase with frequency.
Hi Nick, yes, I've erred on the side of setting it for a lower load to try and counter those unknowns.

At this, even with the 22nF across the input (which I included to try and be sure ultrasonics would be rolled off) there is no roll off evident at 20KHz either into the scope with 47k load, or through the complete system scoped at the speaker posts. Distortion I haven't tried as I think the meter would give up on the 300B se.

I should say the interconnects are 1m of low cap cable and the system input is a 20k pot working shunt against 47k fixed series resistor into Pre3 so a little variable 50k - 70k.

Thorsten's Ecc88 circuit as drawn has a Zout of only 3k6 but it does not drive or sound better than this 6n2p circuit into 50k, quite the opposite, which may be due to it being set for down to 20k as his notes say or some other reason.

I am intrugued by much of that paper and would like to hear what you make of it. Relevant here perhaps is p5 rh column, second para, really saying that a high mu valve with high ra working into a lower load is good in srpp. He says it is unusual, this is a recent paper and I wonder how much is appreciated about its implementation. 6n2p seems better than ecc88 into typical pre loads on that basis which might be counter-intuitive, what do you think?
User avatar
Nick
Site Admin
Posts: 15694
Joined: Sun May 06, 2007 10:20 am
Location: West Yorkshire

#112

Post by Nick »

Yep, interesting paper, I will read more depth later.
Relevant here perhaps is p5 rh column, second para, really saying that a high mu valve with high ra working into a lower load is good in srpp.
I may be missing something but I can't see that point. I can see wher high mu means that the upper Ra can be removed.

I must admit to having a distrust though of any circuit that needs optimizing, but thats just me :-)
Whenever an honest man discovers that he's mistaken, he will either cease to be mistaken or he will cease to be honest.
Richard
User
Posts: 95
Joined: Fri Mar 18, 2011 10:21 pm
Location: Notts

#113

Post by Richard »

I will keep chewing on bits and comment as and when they come out the other end :)

Optimising,

well, Thorsten's as-posted uses Rk 3x less than the "optimisation" suggests (and no balancing Ra which would be greater than the ra of the valve!) when used into 50k (I emphasize that it was me using it, not him saying to do that, but I can imagine many would do that), yet it still sounds very decent until compared with the 6n2p dropped in which is in a different league, maybe for reasons of being better optimised by chance (it is) or maybe just a better valve. Surely every circuit can be optimised to some degree esp when it may be in the order of 3x out of spec if you follow me :)
Richard
User
Posts: 95
Joined: Fri Mar 18, 2011 10:21 pm
Location: Notts

#114

Post by Richard »

Nick wrote:Yep, interesting paper, I will read more depth later.
Relevant here perhaps is p5 rh column, second para, really saying that a high mu valve with high ra working into a lower load is good in srpp.
I may be missing something but I can't see that point. I can see wher high mu means that the upper Ra can be removed.

I must admit to having a distrust though of any circuit that needs optimizing, but thats just me :-)
Ha, I can't even read the number of the pages, should have said p6, sorry,
User avatar
Nick
Site Admin
Posts: 15694
Joined: Sun May 06, 2007 10:20 am
Location: West Yorkshire

#115

Post by Nick »

What is the nature of the difference in sound of the 88 and 6n2p?
Whenever an honest man discovers that he's mistaken, he will either cease to be mistaken or he will cease to be honest.
Max N
Old Hand
Posts: 1451
Joined: Thu Aug 16, 2007 6:10 pm

#116

Post by Max N »

Richard, here is some info on the Audionote implementation:
B+ is shunt regulated by two OB2 in series, so about 216V
Valves are specified as 6dj8/6922
IV resistor is 270R, this was followed by an analogue filter consisting of a 12mH series choke and 10nF shunt cap. This was then followed by a 1K shunt resistor, don't know why. As I said earlier, this analogue filter is not used in the current DACs.
The lower valve in the SRPP had a 470R grid stopper.
RK was 330R for upper and lower triodes. Lower triodes had Rk bypassed with 470uF/16v.
Output coupling cap is 0.47uF but I think this had to be increased if the following amplifier input impedance was not high.
Richard
User
Posts: 95
Joined: Fri Mar 18, 2011 10:21 pm
Location: Notts

#117

Post by Richard »

Max N wrote:Richard, here is some info on the Audionote implementation:
B+ is shunt regulated by two OB2 in series, so about 216V
Valves are specified as 6dj8/6922
IV resistor is 270R, this was followed by an analogue filter consisting of a 12mH series choke and 10nF shunt cap. This was then followed by a 1K shunt resistor, don't know why. As I said earlier, this analogue filter is not used in the current DACs.
The lower valve in the SRPP had a 470R grid stopper.
RK was 330R for upper and lower triodes. Lower triodes had Rk bypassed with 470uF/16v.
Output coupling cap is 0.47uF but I think this had to be increased if the following amplifier input impedance was not high.
Thanks Max, AD1865 dac I gather so you'd leave the input alone and look at the valve stage/load if you fancied a play.

It's this that I find fascinating. Conventional thinking seems to have been applied to the AN srpp as would a common cathode amp in that you use a valve with a low ra and bypass it to keep output impedance low, then hope the following load will be fairly high.

Viewed against this srpp paper it ends up mismatched to the load, with odd harmonic distortion, and passing more current than it needs. The paper seems to say the output impedance is not so important and that it's the matching with the load that counts. Even if my reading/reasoning is wrong the maths are quite different.

How much this differs is that your stage into a 50k load would be,

Rk = 2Rload + ra / mu -1

Rk = 100000 + 5000 / 33 -1

Rk = 3281

If you used the bypass cap you'd also use an Ra equal to Rk and if no bypass an Ra of 2 x Rk = 6560

Now, you could come back quite some way from that and set up for a 25k load as I left for mine but that still makes the resulting cathode resistors many times greater than at present. Output impedance would be higher. Ia would be lower, though still more than 1mA.

Or you could use the circ I posted last (assuming 50k pot or modify it to suit pot) and try 6n2p, same pinout, heaters and HT.

So why is all this so contradictory?

What is the nature of the difference in sound of the 88 and 6n2p?
Hi Nick, the 6n2p has to be heard to be believed. Perhaps it's best to start describing the ECC88 in Thorsten's circuit as drawn. I tried 4 types, Amperex orange prints, Russian or Chinese Colomor brand, E88CC Mullard gold pins and 7308 Sylvania military. The Amperex were ahead by quite a way and were excellent in a hi end hifi way with good tight bass and plenty of air and sparkle well in excess of the standard player on opamps. Putting in the 6n2p was just a different league. Huge expansive sound stage, thunderous bass and drive, natural open mids and shimmering treble that never takes your ears off. Really I've never heard such a change. But the big but is that the sound was similar to that in Lukas' circuit and I agree absolutely with what he posted about it. It's just that it's better in this one. Before it sounded strangely compressed at the top but now it's open and natural and sparkles without losing that great power. The circuit was the same in both these later cases of comparing ECC88 and 6n2p but theoretically suited the latter much better and so I don't know how good the ECC88 would be with changes to the circuit. What's due to the valve and what's due to the circ and which philosophy of setting up is right? Zout doesn't seem a simple answer as that was 3k6 versus 26k and both checked flat 8Hz to 20KHz so what's going on, the valve or its setup, or is a high mu and/or high ra better for this load in srpp?
User avatar
Nick
Site Admin
Posts: 15694
Joined: Sun May 06, 2007 10:20 am
Location: West Yorkshire

#118

Post by Nick »

What would be interesting is pointing the output into a sound card and run some distortion plots of the stages.
Whenever an honest man discovers that he's mistaken, he will either cease to be mistaken or he will cease to be honest.
Richard
User
Posts: 95
Joined: Fri Mar 18, 2011 10:21 pm
Location: Notts

#119

Post by Richard »

Nick wrote:What would be interesting...
Ha ha, so don't you think this is interesting enough :lol: :lol:

Can't do the soundcard but resorted to the distortion meter yesterday.
First checked sig gen into 82R across the meter and it was happy showing 0.016%.

Set signal to 116mV rms to imitate the TDA1541 4mA output into 82R, disconnected the signal lead from the dac chip and fed the amp from the sig gen. (The 82R IV on mine is after this on the valve base so basically the sig gen was used in place of the dac analogue output and at a similar level.)

Distortion meter has 100k input impedance so I put shunt resistors across to give different loads.

Checked both valve types in the Rk=1k Ra=2k circ adapted from Thorsten.

Excellent set of figs for the ECC88 and not so good for the 6n2p.

Then looked to see how they related to Merlin's paper.

Saw they agreed exactly, with a best load 20% less than calc'd for as he'd said. That would be about 12k for ECC88 and 20k for 6n2p a definite notch to see there. Voltage out figs are also interesting and make srpp look very dependant on load.

I prefer the sound of the 6n2p so recalc'd for 65k (20% greater) to go into my 50k preamp load and retested to check. This was the second set of results Rk=1k8 Ra=3k6 and again agrees with the paper.

Had a listen, more later, and checked with test disc in cdp into scope. No loss of hf up to 20KHz at all with 47k loaded usual interconnect and voltage showing 7.7Vpp 0dB so 2.7Vrms agrees with test result.

Image

Image
User avatar
Nick
Site Admin
Posts: 15694
Joined: Sun May 06, 2007 10:20 am
Location: West Yorkshire

#120

Post by Nick »

Good numbers. As you say a clearly defined null in the distortion. As I said before, my only problem with (say) the 50k one, is that a 220pf amp load is 36k at 20kHz, so while the voltage level will remain at that frequency, the distortion specta will be far different to 1kHz. If that matter, is a different case. But its the same reason why I try and avoid driver output valve setups that null distortion, a changing load means that real world distortion varies far more than if it was designed to a initially higher figure.

Its a shame it can't be plugged into a sound card, http://audio.rightmark.org/news.shtml would possibly show some interesting things.

Its nice to know that a little bit of reseach and tweeks can return that much of an improvement in sound.

I just thought, doesn't the 1541 have a DC offset in its output, so that might alter the first valve conditions and possibly indicate that a altered cathode resister may be worth trying to compensate for that?
Whenever an honest man discovers that he's mistaken, he will either cease to be mistaken or he will cease to be honest.
Post Reply