1st May 2020, 1:25 pm | #61 |
Dekatron
Join Date: Sep 2005
Location: Seaford, East Sussex, UK.
Posts: 5,997
|
Re: Mullard 5-20 high DC Voltages
I would connect the shield pins 2 and 7 and the metal socket base of the EF86 to your busbar. Not using a metal chassis does introduce risk as it removes the stray capacitance to chassis and EMI protection.
|
1st May 2020, 6:39 pm | #62 |
Triode
Join Date: Apr 2020
Location: Newport, South Wales, UK.
Posts: 33
|
Re: Mullard 5-20 high DC Voltages
I've now connected pins 2 & 7 to ground for EF86. I've been looking at the Claus Byrith modifications made for his 4-30 to try and get an idea of what changes need to be made to the mullard design in order to connect the EF86 up in triode.
This is his 4-30 schematic. [img]https://***********************/65535/49842209896_f3ae6a4860_h.jpg[/img] Now I know that the feedback will need adjustment and can see he has used 3.3K with 560pF. He also ditches the step network. I'm trying to get my head around the changes to all the resistor values around EF86. This is as far as I got with altering the mullard schematic before getting a tad confused. [img]https://***********************/65535/49841671928_b57cf81caf_h.jpg[/img] |
1st May 2020, 8:50 pm | #63 | |
Dekatron
Join Date: Sep 2005
Location: Seaford, East Sussex, UK.
Posts: 5,997
|
Re: Mullard 5-20 high DC Voltages
Quote:
Using a triode wired EF86 will dramatically reduce the open loop gain. To maintain the low distortion and high damping factor of the Mullard deign you will need to maintain the loop gain of 30dB (difference between open and closed loop gains) by significantly reducing the sensitivity of the amp. I am not an amp designer but I would be sceptical of designs found on the web unless accompanied by evidence of comprehensive testing. |
|
1st May 2020, 10:45 pm | #64 |
Triode
Join Date: Apr 2020
Location: Newport, South Wales, UK.
Posts: 33
|
Re: Mullard 5-20 high DC Voltages
It's a well documented design based off the original 5-20. The Claus Byrith design was on the lundahl transformer site for many years and there are the elektor (part 2) and maplin millennium derivatives too. Notably, the maplin design has some differences compared with the other two.
Obviously I'm not going to go to town on mine and make all the modifications as I don't have the necessary parts nor am I wanting to replace things I already bought. So fixed bias and AC balance adjustments are out. I've had a good dig through my component stash and found I do have enough of the right values to make the changes if they are necessary. |
2nd May 2020, 6:30 am | #65 |
Dekatron
Join Date: Jul 2009
Location: Sleaford, Lincs. UK.
Posts: 7,670
|
Re: Mullard 5-20 high DC Voltages
Ah, sorry, I was looking at the schematic here - http://www.r-type.org/articles/art-003d.htm and missed the g2 stopper's, what I said still applies, it means pulling out resistors though, tag boards do not lend themselves to making changes, hope you haven't wrapped the leads round the turrets.
On that schematic there's a 4k7 R and 47p cap across the EF86 Ra, that's cuts HF gain at 72khz, ****** high, I'd knock that back to 40khz at least, maybe even 35khz by using a 3k9 and 1n or bigger. These are found by the usual formula 1/(6.28 x r x c). I had a similar issue yesterday, cascode E88CC front end, amp completely unstable with even a whiff of NFB, so used a cap decade box to sub C, but you can just get a load of caps then tack them in and test. In the end the cascode came out, too much gain, too sensitive. To strap the 86 in triode mode use a 100r to tie anode to g2, then the fun starts. Had a quick sqizz at the KB doc, he uses a 47k Ra, a strapped directly to g2, then changes the Rk's, can't see the values he uses well , looks like 390euro and 100euro, so lets say 390r and 100r. He leaves the cathode unbypassed and leaves the 4k7 stopper in place, which I think will need looking at. Strapping the 86 in triode mode, will change gain, which will change g1 miller C, which will change the roll off. Now that the gain is lower, maybe the 4k7 isn't needed, or can be reduced. I don't like stoppers on the IP stage, knackers your frequency response, backs you into a corner. If it were me I'd use an active load instead of a resistor = better gain, better linearity, thus better THD + use a bypass cap. I'd help with this by knocking up the IP stage but sold all my EF86's I think, though can help with the active load if you want. I noticed there was a thread recently on this KB amp on diyaudio, might be worth a butchers. there's quite a few very good amp builders on there though getting them to agree on anything is tricky. Andy.
__________________
Curiosity hasn't killed this cat...so far. |
2nd May 2020, 5:24 pm | #66 | |
Dekatron
Join Date: Sep 2005
Location: Seaford, East Sussex, UK.
Posts: 5,997
|
Re: Mullard 5-20 high DC Voltages
Quote:
The Elektor is the Claus Byrith design and the specification is not very detailed but does show it is less sensitive and the loop gain is lower. In the Maplin specification https://livinginthepast-audioweb.co....ln-page-37.jpg you will see a figure for 'Overshoot and ringing'. You should make sure you are not trying to eliminate something that is inherent in a valve hi-fi amplifier designed to keep distortion low and damping factor high. |
|
3rd May 2020, 12:26 am | #67 |
Triode
Join Date: Apr 2020
Location: Newport, South Wales, UK.
Posts: 33
|
Re: Mullard 5-20 high DC Voltages
This appendix has the performance details for the Claus Byrith modifications to the mullard design. (page 12 onwards)
I did some tinkering between Andy's suggestions and the changes made by Claus Byrith. Since my intention is to drop the gain as I really don't need the sensitivity of the pentode. I tied pin 1 of EF86 to pin 6 via 100R, pulled out the step network and changed the feedback network to 3K3/540pF. Test results looked very promising, but I am no expert. 100Hz as standard [img]https://***********************/65535/49847612242_53e0f44265_h.jpg[/img] 100Hz with the triode mod described above [img]https://***********************/65535/49847608672_ea45762688_h.jpg[/img] 20KHz as standard [img]https://***********************/65535/49847304661_6677ff0bd8_h.jpg[/img] 20KHz with the triode mod described above [img]https://***********************/65535/49847604382_b432376934_h.jpg[/img] 1KHz square with the mods [img]https://***********************/65535/49846770828_a7728f770f_h.jpg[/img] 5KHz square with the mods [img]https://***********************/65535/49847606207_de5ddb7c01_h.jpg[/img] Mains hum at the speaker output. Oddly, there is no hum with just a phone as the input source. (probably because it's a battery powered device that's not grounded) The only earthing differences between the Mk1 & Mk2 is that the transformer secondaries all terminate at the IEC earth pin which then links to the busbar in one place. (like the standard mullard did) and C12 earth passes across C15 rather than going directly, which is also following the mullard wiring. The Mk1 had everything earthed on the bar. [img]https://***********************/65535/49847609547_dfed78cb2f_h.jpg[/img] Whole gallery of pictures for those interested. https://www.***********/photos/399591...10461998/page2 I am strongly considering going full CB 4-30 mod on the front end just to see what the results are like. |
3rd May 2020, 8:00 am | #68 |
Dekatron
Join Date: Jul 2009
Location: Sleaford, Lincs. UK.
Posts: 7,670
|
Re: Mullard 5-20 high DC Voltages
Interesting oscillograms, very envious of your scope. I was really surprised at the standard 20khz sine, that is absolutely terrible, anyroad, the triode modded sine looks a lot better but not perfect.
You really have ringing issues there though I've seen worse. One thing I do is to sub the FB RC and bung in a 10k pot, start at "0" then wind it up, adding more FB, this whilst checking frequency response, amplitude and stability. I apply NFB till the amp goes unstable, then back it off, obviously making sure sensitivity is ok etc. I then sub the pot for the nearest fixed value, then bung in a square wave with a capacitance decade box across the NFB R and play with that switching between 800hz sq and 10khz sq. I find what happens is you crank the capacitance up to get a good 1khz response, sw to 10hhz and find you have too much, so have to dial it back a bit, the knack being to get a good compromise. Having done a bit of research recently into the 5-20 and "the Mullard topology" it's clear a lot of folk found the standard 5-20 lacking and some even going as far as to condemn Mullards approach full stop. I'm not one of those but in my brief time of messing about with valve amplifier design there is one thing I'm pretty certain of, the drivers for the OP stage have to be as substantial as possible, IE plenty of grunt, high current capability. If your driver can drive the OP stage cleanly, your onto a winner. With this in mind Mullard's choice of the ECC83 as the LTP is what I would change, it just doesn't have grunt, that and the ECC83 makes a poor LTP. I'd change it for an ECC82/6CG7 at least, far better 12BH7 or ECC99 or even better two EL84's. If you do this it's then a simple job of knocking up a front end to provide enough V gain. You end up with an amp that's stable has very good performance with THD well under 1%, 0.02% at low OP is achievable. This is not what you want to hear though I suspect, it would mean a complete re-design, a Mk3. Andy.
__________________
Curiosity hasn't killed this cat...so far. |
3rd May 2020, 9:55 am | #69 |
Dekatron
Join Date: Sep 2005
Location: Seaford, East Sussex, UK.
Posts: 5,997
|
Re: Mullard 5-20 high DC Voltages
What are the test conditions? Load, input amplitude, output amplitude.
It might be worth trying this modified amp with no feedback. A 20KHz sine showing the phase shift and gain and a 20KHz square wave to see if it still has ringing. |
3rd May 2020, 11:14 pm | #70 | ||
Triode
Join Date: Apr 2020
Location: Newport, South Wales, UK.
Posts: 33
|
Re: Mullard 5-20 high DC Voltages
Quote:
Quote:
In any case, I've had a good look in my stash of components and found I have all of the necessary parts to convert the front end to the Claus Byrith redesign, so tomorrow I plan on making the change. Honestly, I don't expect a great change vs what I've already seen, but it's worth trying. The benefit of turret board/point to point is at least showing through here as it's very easy to modify. This is the schematic for it [img]https://***********************/65535/49851468863_4efa4a800d_h.jpg[/img] This shows the change to the turret board [img]https://***********************/65535/49852006761_ee2220cdd5_h.jpg[/img] I'm thinking that a big part of the problems I'm seeing is likely down to me choosing the 4/8/16 ohm tapped secondary output transformer rather than the multi-secondary or just the single 8 ohm secondary. It's only now having done a lot more reading up on the matter that I have found a few notes saying the tapped secondary limits HF bandwidth due to increased leakage inductance. (a costly mistake) I'm also looking at getting some 100R dropping resistors for the HT and also some larger resistors for the EL34 cathode bias to try and curb the red plating issues. I'll be getting both 560R and 680R to try. I'm guessing that the higher value here will cause a drop in output power. Will the bypass capacitors need adjusting too? |
||
4th May 2020, 7:13 am | #71 |
Dekatron
Join Date: Jul 2009
Location: Sleaford, Lincs. UK.
Posts: 7,670
|
Re: Mullard 5-20 high DC Voltages
I can knock up a LTP with CCS in my sleep, sorry, not bragging, just done it a hell of a lot, also built two 120w monoblocks with 6 EL34's, so happy to help there, I think it'll improve things a lot, if it doesn't I'll eat my socks : )
First off can you measure AC RMS at full 20w going into g1 EL34's, that'll save some calculations, though would think something like 30-40v ish would be about right, so just off the top of me head an ECC82/6CG7/6SN7 with two 47ks as anode loads, HT @ 300v, 275v might be better, therefore we'd need to change R17 to a 12k or something near, Id up C5 to 50/100u, same for C4 but not essential. Grid leak on g1 (1), ie IP - 470k, g1 (2) grounded, but this means you'll need another tfmr really, 0 - 70v @ 50mA, but you can do it without but it's nowhere near as good. You want a stiff, low ripple supply, so bridge rectified, two 200v 470u caps out of a SMPSU, with something like two 680r R's to form two RC filters will give you -80v with only a few 10's of mV ripple. That should mean we need to bias the cathodes at 6/7v for mid bias, each cathode will be at 3mA, so total = 6mA, so for the CCS a MJE340, a 6v2 zener, 15k, two 100r's and a 100r or 250r preset with a 820r fixed R as the Re. I can send you one already built if you want. If your not keen to plonk another tfmr in your amp we can re-jig the CCS and LTP, will just have to do some more calculations. Give me a few hours and I'll draw you a schematic. Apologies if all this is a bit much, I like designing amps and had the dynamic characteristics for the 6CG7 and 6SN7 on the table in front of me. Re the OPT, I think your right, it does seem to point towards it being difficult, I highly recommend limiting the HF gain with a big 1k R and 2200p cap from each anode to ground, this will improve stability and help the ringing a bit. Though the values might need tweaking, a snubber like this is found on lots of amp designs, cuts the HF gain, cuts the overshoot, cuts the ringing. Another thing you can try is taking the NFB off the 4 or 16r sec taps, same with grounding, IE ground the 8r tap, use the 16r as +, it might help, it might not. Other things not suggested so far is using inductors on the LTP etc, something like 1mH, or even a bit of current FB, just thinking out loud. Using bigger cathode R's does drop quiescent current, they bias the OP valve at higher V, so more into cutoff, you shouldn't need to change the de-coupling C's, though one final tweak would be to match the C's, both cathode and coupling. Andy.
__________________
Curiosity hasn't killed this cat...so far. |
4th May 2020, 9:49 am | #72 | |
Dekatron
Join Date: Sep 2005
Location: Seaford, East Sussex, UK.
Posts: 5,997
|
Re: Mullard 5-20 high DC Voltages
Quote:
I also think it would be useful to take some open loop scope measurements of phase shift of a 10KHz sinewave and overshoot/ringing on a 10KHz square wave. If I get the time and can find an 8 ohm load, I will do some measurements on the Avantic DL7/3. |
|
4th May 2020, 11:25 am | #73 |
Dekatron
Join Date: Jul 2009
Location: Sleaford, Lincs. UK.
Posts: 7,670
|
Re: Mullard 5-20 high DC Voltages
Schematics for CCS as promised, though as PJL says would be a good idea to see what the OPT is up to at HF. I've never come across as terrible a 20khz sine and that's using all sorts of odd OPT's & OP stages. Adjust pot for 900r (pot - 820r fixed) then power up and adjust to 6/7v or best OP. Missed out the 100r's on the CCS for standard supply, though may not be needed.
Anyroad, find attached schematics, both are starting points, not actual finished design's. Andy.
__________________
Curiosity hasn't killed this cat...so far. |
4th May 2020, 4:12 pm | #74 | |||
Triode
Join Date: Apr 2020
Location: Newport, South Wales, UK.
Posts: 33
|
Re: Mullard 5-20 high DC Voltages
Quote:
Quote:
10KHz square (CH2 I believe was connected to the output of EF86 rather than the input. [img]https://***********************/65535/49854196396_395bc68804_4k.jpg[/img] Phase inversion point (70KHz) [img]https://***********************/65535/49804511603_3d884ca476_4k.jpg[/img] I'd looked at the TL25 as someone mentioned it earlier in this thread. I also noticed that it uses ECC81/12AT7's instead of ECC83/12AX7's and the cathode bias resistors are 440 ohms. Quote:
|
|||
4th May 2020, 5:49 pm | #75 |
Moderator
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 22,902
|
Re: Mullard 5-20 high DC Voltages
There looks to be something wrong with that amplifier, even on the first picture of the scope, one waveform shows a burst of oscillation on the positive-going edge of the negative half cycle on one of the channels
Something is unstable, and that will relate to your ringing problem. It could be a problem with overall feedback, or it could be in the circuitry around a single output valve. You may need some serious stoppers in an ultra-linear stage. Being on one half-cycle and not the other points to output stage, but that isn't 100% trustworthy. David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done |
4th May 2020, 8:36 pm | #76 |
Triode
Join Date: Apr 2020
Location: Newport, South Wales, UK.
Posts: 33
|
Re: Mullard 5-20 high DC Voltages
This behaviour has been consistent across two different versions using completely different parts. (transformers, turret boards and all valves) I'm left with the feeling that the wooden chassis (which is only temporary) may be the cause as I can't see why anything else would cause this. I do have some 19" aluminium panels that I'm loathed to hack up as an experiment but it's looking like it's going to become a necessity to prove where the problems lie.
Anywho, I've completed and tested the Claus Byrith front end modifications bar the 0.47uF capacitor that he places in series with the signal input resistor. My feedback also has 540pF rather than his 560pF as I don't have a 560pF on hand to try. The performance appears to be quite a bit better, however there are some problems that remain. Ringing is almost gone entirely and sensitivity is massively improved, requiring quite a lot more at the input for full output. (which is a big part of why I wanted to try it) The oscillation with large LF signals still exists but it's a lot harder to push it there. The cap on the input might well be there to eliminate this issue. I will add it and see. The main problem that still stands is the excessive phase shift at 20KHz. Due to this, I cannot get more than about 10W out at this point before it distorts to hell. Here are the scope shots for the various tests I ran this afternoon. 8 Ohm feedback tap with 8 ohm dummy load 16 ohm feedback tap with 8 ohm dummy load (load on 8 ohm tap) Open Loop EDIT: just had a good look and noticed that I forgot to link the cap from the anode side of EF86 to ground. (C2 in this altered schematic I drew up) Not sure it would make a big difference, but better to point out my mistake. [img]https://***********************/65535/49851468863_4efa4a800d_h.jpg[/img] Last edited by Kei-1986; 4th May 2020 at 9:01 pm. |
4th May 2020, 11:34 pm | #77 |
Dekatron
Join Date: Sep 2005
Location: Seaford, East Sussex, UK.
Posts: 5,997
|
Re: Mullard 5-20 high DC Voltages
The open loop square wave is showing a very slow rise time but no ringing. I would guess the reduction in ringing amplitude compared to the 5-20 is most likely to be due to reduced loop gain.
Earlier you mentioned one of the EL34 appeared to be 'red plating'. Although the DC conditions seem OK, have you tried other EL34's? It does not show ringing with the 10KHz open loop sq. wave even though you are running at 4x the power of the 8 ohm feedback test, so the fault appear to be propagation delay and that doesn't seem like a layout problem but I guess anything is possible. Stay with the open loop 10KHz square wave and scope the EF86 anode and each of the ECC83 anodes just to make sure the drive looks OK. |
5th May 2020, 6:32 am | #78 |
Dekatron
Join Date: Jul 2009
Location: Sleaford, Lincs. UK.
Posts: 7,670
|
Re: Mullard 5-20 high DC Voltages
I agree the CCS looks a bit scary if you've never used one, if anything though it simplifies the LTP meaning you don't have to sacrifice 1/3 HT for the tail, grids aren't hanging in the air etc.
Been having a think about this and glad to see things have improved a bit, is one EL34 still glowing red? I'd want to find out why that's red plating, it might be at the bottom of a lot of issues. What could it be? Duff valve - swap valves over. Duff base - check all pins are making contact, those Chinese ceramic bases are not like vintage ones. I have to pull like bu**ery to get 34's out of the vintage bases I used on my amps, the octal bases Tek used in their scopes need 10 people all gripping each round the waist as in a comedy sketch, anyhoo, worth checking. Could be a bad joint like an intermitant gnd or cathode. Re sensitivity aim for 0.5v closed loop. Next be worth trying bigger g2 R's as already suggested. Lastly putting in bigger Rk's might help, but then that will increase HT, swings and roundabouts... Lastly, lastly would forget about applying NFB till things are sorted, scope EF86 and OP of LTP, when ok then try a pot as suggested, with fixed R's your just guessing really. Andy.
__________________
Curiosity hasn't killed this cat...so far. |
5th May 2020, 7:18 am | #79 |
Moderator
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 22,902
|
Re: Mullard 5-20 high DC Voltages
If you swap the EL34s around, does the glowing anode move with the valve, or stay with the socket?
The rising output at higher frequencies fits with that burst of oscillation. One valve running much hotter than the other is a big asymmetry in the output stage. Now the oscillation is asymmetric as well, tied to one half cycle and not the other. Are these connected? Does the oscillation stay or swap to the positive going half-cycle when the EL34s are swapped? Those aren't high voltage scope probes, so I'd build a little attenuator 10k from each EL34 anode driving 1k to ground will give an 11 to 1 attenuator so you can examine what's going on on the anode. I wonder if that oscillation is a side effect of a much higher frequency RF oscillation which doesn't make it through your output transformer? This isn't a probable cause, but it is something to check when peculiar things seem to be happening. If it is happening, it'll drive you round in circles looking for other things, and some serious oscillation will generate that heat anomaly. David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done |
5th May 2020, 11:30 am | #80 | |||
Triode
Join Date: Apr 2020
Location: Newport, South Wales, UK.
Posts: 33
|
Re: Mullard 5-20 high DC Voltages
Quote:
Quote:
Quote:
These Tek P6137 probes are rated to 500VDC + AC offset up to 1.3MHz, so may be ok. I've been using the bandwidth limiter set to 20MHz to avoid spurious noise on the waveforms. That does make me wonder whether it's an RF issue as the "noise" on the waveforms is above 50MHz. |
|||