I have a Velleman K4040 clone power amp which gets very upset with capacitive loads. I first noticed it years ago when I tried some Litz speaker cables, and it oscillated. Just recently I bought a pair of Quad 2805s and tried it with them. I though I had destroyed the speakers! It made such cracking noises that I leapt up and switched it off as quickly as I could. The speakers are OK, but it reminded me of the earlier Litz episode. A little Ming Da EL84 amp is fine, and of course my Parasound A21 is too.
I have just read something about putting a 1k resistor across the output terminals to tame this; is that a reasonable method, or is there a better way? Or even a different value?
Any advice welcomed, as I built this amp over 25 years ago (one of my first forays into cloning) and it would be nice be able to use it again.
I have just read something about putting a 1k resistor across the output terminals to tame this; is that a reasonable method, or is there a better way? Or even a different value?
Any advice welcomed, as I built this amp over 25 years ago (one of my first forays into cloning) and it would be nice be able to use it again.
Last edited:
You used to find 1k R's across speaker terminals on guitar amps, mainly there though to stop the amp self destructing if someone forgets to connect the speaker. A couple or 3 reverse biased diodes from anode to ground does the same thing, protects your OPT.
A 15r (high wattage) and 0.22u or a 10r and 0.1u across the speaker OP can help, a "zobel network", you may need to try different values though. Something like a 2200p and 1k from anodes to ground reduces HF gain, cap goes to anode, 1k grounded, reduces gain @ 30khz ish, just use the old 1/(2x pi x f x R or C) to find the right combo.
If you can though try and get it to oscillate on a dummy load and or cap in order to find out at what frequency it is oscillating you'll have a better idea of what frequency to roll the gain off.
See here and read the text - 100w-monobloc3-2014
Hope that helps, Andy.
A 15r (high wattage) and 0.22u or a 10r and 0.1u across the speaker OP can help, a "zobel network", you may need to try different values though. Something like a 2200p and 1k from anodes to ground reduces HF gain, cap goes to anode, 1k grounded, reduces gain @ 30khz ish, just use the old 1/(2x pi x f x R or C) to find the right combo.
If you can though try and get it to oscillate on a dummy load and or cap in order to find out at what frequency it is oscillating you'll have a better idea of what frequency to roll the gain off.
See here and read the text - 100w-monobloc3-2014
Hope that helps, Andy.
Last edited:
What OPTs did you use on this amp? You'll need to dial back the feedback or compensate the amp, maybe both.
The easiest solution is just to triode wire the output tubes.
The easiest solution is just to triode wire the output tubes.
The OPTs came from Velleman, the only parts that did. And do you mean reduce the NFB? Rewiring for triode operation is not something I would be comfortable doing.
Adrian-Kingston.com - Velleman K4000 Tube Amp
Adrian-Kingston.com - Velleman K4000 Tube Amp
I think the opt is a toriod which is unusual for a valve opt. These have a wider bandwidth than the normal frame type and is possibly the reason for this amps sensitivity to capacitive loads. With solid state amps a small series inductor (10uh) is added in the o/p lead to compensate for the capacitive load. Ive made these by winding about 20 turns of thick insulated copper wire around a pencil.
Ive never used this technique with a valve amp though !
Ive never used this technique with a valve amp though !
Yes, I'm familiar with that in SS amps, and I had a look at the schematic to check, but no nothing like that here. And they are indeed toroids.
As I may use the 4 ohm or 8 ohm tap, could I try that in the 0 volt line? Or would it be a bad idea with a valve amp?
That's easy to triode strap. R53, 54, 55, 56 get moved to connect pins 3 and 4 of each output tube socket. Problem solved.
Hmm, I've just made a couple of these, 20 turns of 1.6mm wire round a 9mm former, and they calculate out at 1uH. Two different online calculators. A parallel resistor is recommended for damping (in SS amps, of course), but I'm not sure what value. As I get older, my brain struggles more with these things. It's nearing the time to stop, I fear.
I don't know which version you have but check with the latest velleman schematics. The NFB has a dominant pole added C10/R22 so check that and the NFB components C8/R65. I have a K4040 and it was stable. The OPT is not the best at the HF end I did measure it up and I think there's about 5us of lag between primary and secondary.
Last edited:
Thanks. My version started as a K4000 and was modified to K4040 spec. If you mean R10/C22, they are the same; 820R/470nF. But I have C8/R65 marked as C11/R13; they are the same values though, 2n2/820R. But the NFB on mine goes to the 8 ohm tap, whereas the late one goes to the 4 ohm tap. And mine is on three boards (plus input board) where the later one is on one board (plus input board). Everything else is the same, apart from 47k input instead of 33k. The K4000 was 100k.
Could the NFB to 8 not 4 ohms be causing problems? The K4000 went to the 8 ohm tap, but R13 (R65) was 1k5 not 820R.
When you say yours was stable, was that into electrostatics? Mine has been OK into dynamic speakers.
Could the NFB to 8 not 4 ohms be causing problems? The K4000 went to the 8 ohm tap, but R13 (R65) was 1k5 not 820R.
When you say yours was stable, was that into electrostatics? Mine has been OK into dynamic speakers.
Ok a little confusing over references. With reference to your schematic in #8 the latest version has a series R and C placed across R52 - 3k9+100p. Thats important as it adds a dominate pole. I don't like it there as it causes of the driver stage to run out and clip at HF. I would prefer between plate and grid with much smaller components making use of the miller effect.
I swapped out the EL34 for 6550 in the end as they kept popping and the power transformer is very meaty. Just made the bias supply more negative and cut some bigger holes.
I swapped out the EL34 for 6550 in the end as they kept popping and the power transformer is very meaty. Just made the bias supply more negative and cut some bigger holes.
Last edited:
If your on the 8R tap I would expect R13 to be bigger and C11 to be smaller. So R13 being say 1k5 would be more correct. The problem with stability is that without proper mathematical analysis its a big finger in the air. I think velleman must of had issues given the changes made. Electrostatic speakers have very different impedances to that of normal speakers.
Last edited:
Quad Reference ESL-2805 loudspeaker Measurements | Stereophile.com
Looking at this there's a nasty peak at 20KHz you would benefit with zobel network on the output maybe 10R + 100nF in series placed across amp output to make it resistive.
You could use the old square wave test if you have a scope.
Looking at this there's a nasty peak at 20KHz you would benefit with zobel network on the output maybe 10R + 100nF in series placed across amp output to make it resistive.
You could use the old square wave test if you have a scope.
Last edited:
So I think it should be C11 2n2, R13 = 1k5 (for 8R NFB), 100p+39k (not 3k9) placed across R52 (plate to supply) and a Zobel network 10R+100nF across output terminals.
That's really useful info, thanks. It will be easier for me to move the NFB back to the 4 ohm tap, than to lift the pcbs and change C11 and R13. The 3k9 is easier to change. and the Zobel is no problem at all. I'll give it a go when I've finished painting the bathroom wall. My wife is finding things to keep me busy, so kind of her! 🙂
Yep super.
I do have an accurate ltspice model of the output transformer ZD043, I do have some time and could check the NFB components are OK but it will be later today.
I do have an accurate ltspice model of the output transformer ZD043, I do have some time and could check the NFB components are OK but it will be later today.
Right, thanks. Although I can clone stuff, and construct with little difficulty, programmes like LTSpice seem to be beyond me. Computers in general, actually. I always have to get someone else to install stuff as my brain just doesn't hold the information needed. And now my memory (and eyesight!) is getting worse. Ho hum, at least I can still enjoy listening to music. Except my hearing is also worsening! 😀