CCS mode for K5881/5-20 etc.

[NealG]

Been messing about with K5881 again, here’s a simple mod for anybody with a PP amp that uses a long tailed pair phase splitter, I’m thinking of the Mullard 5-20, K5881, KiT88 etc.

The mod adds a constant current sink/source to the tail of the splitter replacing the usual tail resistor.

Why? A good introductory write up can be found on Gary Pimms site where the basic idea/cct was pulled. See: http://www.paclink.com/~gpimm/Active...t_control.html CCS’s have been used by others on the BB for different applications, I think Nick and James use them in their phono stages.

The cct uses depletion mode mosfets, DN2540N5, which unfortunately are near enough impossible to get over here because they are not compliant with the EU RoHS directive. However, Kevin at K and K Audio (USA) can still supply them in low quantities at $3 each. If anybody can find a similar device in Europe I would be interested to know.

The 1K resistors are there to keep the cct stable and the set resistor is found by simple ohms law.

I fitted this to my K5581 and found it worked with no issues. I suggest building the cct up on a small piece of copper strip board, see diagram.

You need two DMOSFETS per channel, 4 total. As these are silicon devices they can ‘break’ very easily, they are not like valves! You might want to have a few spares. Double check all soldering and cct connections, any mistake could kill the DN2540. Also earth yourself if possible to minimise any static discharge, as this will also kill the DN2540.

Determine the current flow through the tail resistor of the existing splitter and divide 2V (the set voltage) by the current to give the required value in Ohms. For K5881 the standard draw is 0.82mA, so 2V / 0.00082 gives an awkward value of 2K44 ohms, upping the draw to 1mA gives a nice and easy 2K value. Or you can use a variable resistor to set the value instead of a fixed item, a small 4K7 pot would do fine. If you decide to use 1mA for K5881 then ideally the Anode Load resistors R9/10 need to be reduced from 150K to 120K to keep the voltage drop more or less the same as before. Also the load resistors now need to be as close to each other in value as possible for best balance, the closer the better!

As the current draw for K5881 is so low no heatsinks are needed but at higher levels it might be wise to use one.

I’ve had success with this cct in my 2A3 PP and it seems to be good in K5881 as well, if anybody else tries it I’d be interested to hear what they think.

..........

[Max N]

Hi Neal
I've been using DN2540 as a ccs tail for the 6sn7 diff amp stages in the PP amp that I'm breadboarding. I'm actually using the little pcb's from Kevin Carter, they are very neat. FWIW, I'm getting 0.04% THD at 100 volts output from the input/driver circuit.
Have you done any tests to determine how much voltage 'headroom' they require? I have found that a single one (not cascode) set to 18ma required at least 12 volts across it to work really well, and 20v for a cascode. This is a bit inconvenient, requiring a negative supply, so I'm thinking of building one of Morgan Jones cascode ccs circuits - he gets away with 10v for a cascode using bc549. Has anyone tried these?
Cheers
Max

[NealG]

Hi Max, no, not check the minimum voltage. I would say your figure are about right, I've only used the CCS in the tail of the splitter where voltage in not a problem!

[Richard]

Hi Neal

Thanks for posting this!

I know it's hard to do unless you have 2 amps side by side but did you notice a difference with CCS?

Regards
Rich

[NealG]

Not yet Rich, I'll see if I get time this weekend, I was going to make it switchable but I've already corrected RL2 to get a closer balance and I need to lower RL to 120K anyhow so I don't think that idea will fly.

After reading the Mullard book the other day I'm not sure how much benefit this is going to be with high mu valve like the '83 but we shall see!

[NealG]

I’ve now had time to evaluate the CCS a little more.

First I checked the splitter performance with just the resistors, at 1Khz 1W o/p there was a little, visible difference between the two outputs, at 1Khz full o/p there was a small difference visible on the ‘scope. I then checked at 10Khz 1W, very similar to the 1Khz but at 10Khz full output there was about 5v difference.

Putting the CCS in place I could see no difference at 1Khz 1W and full o/p or at 10Khz 1W o/p, however at 10Khz full o/p the amplitude of the outputs was the same but one of them was badly distorted.

I found that not only do RL1 and RL2 need to be as close as possible but both halves of the ’83 do as well! I swapped in a number of spare 7025 and 5751’s and found two that gave an undistorted output.

At first I was not sure about the sound and I don’t really know if what I’m hearing is down to the CCS or the fact I changed the valves. It’s hard to put into words and I can’t easily go back to standard but I think overall it is better. The amp seems to hold my attention for longer, it seems less ‘bland’. I can detect a little more space in the sound and a clearer brighter treble, however, as I said it could just be down to the valve change.

I think it’s worth mucking about with, the DN2540’s are not expensive, be interested to hear from anybody who has given this a shot.

[NealG]

Quick update. The sound is starting to grow on me, it seems the amp has more bounce and seems more lively than before. It's not as grey sounding and I'm less aware of it being an amp with feedback. All subjective I know!

[Richard]

Originally Posted by NealG

Quick update. The sound is starting to grow on me, it seems the amp has more bounce and seems more lively than before. It's not as grey sounding and I'm less aware of it being an amp with feedback. All subjective I know!

Ouch, that's a broad swipe at feedback Neal when the rest of your work seems so analytical

Seriously, many thanks for your evaluation here. It's often hard to decide what might be worth doing and what will have little bearing on the big picture isn't it.

From the magnitude of the differences you mention I'd say the difference of the valves must be greater as, to me, the difference between a 5751 and ecc83 at this point is chalk and cheese.

Incidentally I was amazed how well matched the sections of the handful of nos 5751 that I've tested are. I gather section matching was part of their spec. Perhaps that's (partly?) why they seem good in this position.

Rich

[NealG]

Sorry Rich if that came over as a swipe at feedback, it wasn't meant to! It's just an observation, when listening to the amps I have with feedback it seems to rob the music of air and life. I much prefer the 300B for instance with it switched out. Of course I can't do that with K5881.

Feedback tightens the sound and many like it that way. With K5881 in it's new configuration, I'm less aware of this effect and yes the difference may well be down to the 5751. I have the Sovtek's in there, out of the 4 I had to hand, three of them are very well balanced with the fourth a little out. The Sovtek 7025 on the other hand seem to vary much more.

I would like to think that getting a better balance is also helping, allowing the feedback to function in the way intended, reading some of Noel's comments on this it seems the amp needs to be as linear as possible before loop feedback is applied for it to function well.

Anyhow, the K5881 is still in my system, the longest ever I think! so something is right with it!

Cheers!

[Richard]

Hi Neal

Only joking, I don't know the truth about feedback. Some amps I like have it and some not. It does seem to "tidy up" the sound of some amps but I do wonder if those amps were previously "out of focus" and really had false "air".

I say this based on listening the other way round to your example. When I removed all feedback from Kit88 and had a listen, yes, it was more "open" sounding but it was also a mess sound-stage wise and bass was very woolly. Now, in some systems in which the owner doesn't put great store by soundstage or imaging and which may be a bit bass light I could see the new amp sound being attractive and maybe more "natural".

The point about making the amp good before applying feedback is undeniably the best route. A bit like saying car suspension will have less to do on smooth roads. Both are fitted to overcome a problem which shouldn't be there in a perfect world and both can only do so much can't they.

MJ has a bit about feedback and I found the first bit most interesting. I read it as feedback being blamed for the poor sound of 70's transistor amps when really it wasn't working as the designers thought it was.

I've not searched too hard but can't really get my head round it. Bro did some feedback calcs for regulators he was building years ago and decided feedback can't work he he . MJ gives approx equations (which didn't work for those 70's designers) but doesn't explain how it actually functions.

Bits on the web suggest there's some sort of delay in feedback correcting errors. I've even seen one chip amp builder proudly extolling his 1/4" long feedback loop ..eer.. like it makes a difference. Electricity doesn't flow - there is no delay - it's like a rod, move one end and the other moves instantly. Caps might upset it. With that in mind feedback seems more friendly somehow like it's a solid part of the circuit acting instantly in real time to correct errors before they get to the speakers. Hey, how romantic can feedback be eh

Rich