Asher Guitars WD Music Products Amplified Parts Mod Kits DIY Nordstarnd Pickups darrenriley.com
Asher Guitars WD Music Products Amplified Parts Mod Kits DIY Nordstarnd Pickups Warmoth.com
Asher Guitars WD Music Products Amplified Parts Mod Kits DIY Nordstarnd Pickups Warmoth.com

Cathodyne Phase inverter - Design issue ?

Discussion in 'Amp Tech Center' started by HF1600ie, Aug 2, 2018.

  1. HF1600ie

    HF1600ie Tele-Meister

    Age:
    37
    147
    Sep 26, 2016
    Portugal
    Hi,
    I don´t know this PI format too well. I wonder if it is designed well and if there´s any component that should/could be added or changed ?

    By the way, V3A seems to be a gain stage right after the FX loop / reverb and right before the cathodyne PI (V3B)

    I meanwhile changed R73 to 100K, removed C51 and C56. The sound got a lot less compressed, more detailed and I lost a bit of overall volume (and hiss!). Not too bad. It´s a little more open sounding now, but still feels very compressed and mushy at high master volume settings.

    Just wondering about the PI configuration...

    I found this link below with some suggestions, but some input would be greatly appreciated for improved design.

    http://www.valvewizard.co.uk/cathodyne.html

    This is my 20W amp´s Cathodyne Phase inverter :


    upload_2018-8-2_15-34-9.png
     

  2. Bendyha

    Bendyha Tele-Afflicted Ad Free Member

    Mar 26, 2014
    Northern Germany
    You have a direct coupled cathodyne there.; no cap between the gain stage V3A and the grid of the PI V3B. The bias for the PI is set by the voltage at the grid. By changing R73 from1M to 100K , you have raised the voltage on the grid by a very large amount. The cathodyne is very good adjusting its self to make it sort of self biasing, but you are stretching the limits way past what could be expected as acceptable tolerances.....TAKE THE 100K OUT ......and......PUT THE 1M BACK IN. You have a very bad design now!.
     

  3. Bendyha

    Bendyha Tele-Afflicted Ad Free Member

    Mar 26, 2014
    Northern Germany
    Okay, C56 might well be cutting the highs a bit to much. If it was sounding muddy, worth a try.
    C51..47µ...surprisingly large at first glance, but I guess for the 2K2 it was bypassing, probably right for full range boost.....-3dB = 41Hz. Removing the cap will drop the voltage gain of the stage from about 72 down to 36...or dB gain from 37 to 31.
    Put the 1M back in and see what it sounds like then..............if you are not happy with the sound, maybe you are tring to fix it in the wrong place.
    What is the amp? Could we see the rest of the schematic.
     
    Last edited: Aug 2, 2018

  4. Inglese

    Inglese Tele-Meister

    Age:
    39
    107
    Mar 7, 2018
    Italy
    I quote the previous posts, biasing a DC coupled cathodine PI isn't that easy.
    And you'd better keep under control the cathode-heater voltage!
    However I assume R73 can be decreased but you need to check the DC bias and to know what you want.
     

  5. Bendyha

    Bendyha Tele-Afflicted Ad Free Member

    Mar 26, 2014
    Northern Germany
    You are right that the PI cathode to filament voltage is important, but the 12AX7 can handle 180V difference, and you are probably within that limit.
    But to get the cathode voltage up high enough to make direct DC coupling of the grid possible, you need to use big cathode resitors, and there are 100k plate and cathode resitors there now, which for a cathodyne are about as big as you would want to go if you want a decent output from it. If you made the plate resistor any smaller than 1M on the V3A plate, the voltage would rise higher, and increase the need for the cathode resistor on V3B tro be even bigger.
    The only way to solve this would be to use a voltage divider between V3A plate out-put and the V3B grid, to drop the voltage.
    If it was my project, I would probably ditch the DC coupling, and go for a standard cathode biased PI, rather than a fixed bias with a bit more drive...which for a pair of EL84, you really don't need.
    But then again...it seems like whoever design the amp, did just that, and designed it with knowledge of the parameters, in which case I would leave everything alone if it was all working okay, and seek to solve any weaknesses, or rather, things not to my needs of taste, somehow else.
     
    MitchMiami and JD0x0 like this.

  6. JD0x0

    JD0x0 Poster Extraordinaire

    Age:
    28
    Feb 22, 2009
    New York
    IMO, part of the mush is the 47nF coupling caps from the PI to the EL84's. You've got about -2dB reduction @ 10hz. Way over coupled, if you're having issues with low end mush. Also, schematic doesn't show power supply. You could have insufficient filtering giving you a relatively 'loose' power supply, which, IME, tends to make the lows mush out more, and the amp will sound more compressed at high volumes, especially with a tube rectifier.

    I'd suggest also reducing the grid leak resistors value in front of the EL84's. This will help reduce blocking distortion and will hit the EL84's with a bit less signal. EL84's generally don't like to get with a ton of gain and bass, it doesn't take much to drive them compared to some of the bigger bottle power tubes, so that's something to keep in mind.

    Try 22nF couplers with 220k grid leaks on the power tubes. Should tame the mush a compression a bit. Get us a more full schematic, complete with power supply, too, if you can.
     
    robrob likes this.

  7. bwacke

    bwacke Tele-Meister Gold Supporter

    355
    Oct 28, 2007
    Toledo, OH
    All good suggestions above. Here are another couple:

    R70, the grid leak for V3A, is only 33k which really loads the preamplifier stage preceding it. Something from 330K to 1 Meg would be more in keeping with a grid circuit that doesn't excessively load the output of the preceding stage and minimizes signal loss and distortion.

    The 2.2k (R69 & R78) grid stopper resistors should be moved as close to the EL84 grid pins as possible and the grid leaks (R66 & R77) should connect between the coupling caps (C53 & C54) and these grid stoppers (R69 & R78) to reduce the chance of the power tubes self oscillating.
     

  8. HF1600ie

    HF1600ie Tele-Meister

    Age:
    37
    147
    Sep 26, 2016
    Portugal
    Hi everyone,
    Thank you so much for the input and sorry for my late reply.
    The circuit is a Hayden HGTA-20W from a 1x12" combo.

    The first time i plugged it in I was very very disapppointed with the tone. So much gain (look at those cathode bypass caps - C28, for instance!), mush and lack of tone. No bass, no "air", just plenty of bad-quality mids and mud. I understand the purpose of the manufacturer, but it was just too much.

    The first thing I though I should try was to change the speaker which was by itself a great improvement. I scored a used Celestion G12H 30w Greenback and that by itself was an audible improvement right there !

    This circuit has 3 footswitchable channels, being V1A dedicated to the single-triode clean channel.
    Channel 2 and 3 are actually the same ( you can find them in the layout through the V1B path with GAIN 1 and GAIN 2 - it is a bit misleading, but you will see that the main difference is the GAIN POT value !! ). Their interaction is quite useful if blended correctly !!

    So, I wanted to build something with a clean channel (V1A), a good overdrive channel and a full distortion channel via V1B path. So I thought about a typical Marshall sound and used JCM800 specs as my base, as it is versatile enough.
    But, one of the issues is that the JCM800 combo (4010) has only 3 gain stages plus a cathode follower and a different PI and Poweramp. The Hayden has what it seems to me 5 gain stages and the tone stack is plate followed !! There will have to be some interstage attenuation ! I can still clean it up very nicely rolling back the volume of my guitar.

    I like the sound of the amp, but currently it is lacking volume (probably because R73 is only 100K - I had to max P1, a small trim pot to recover for a part of the lost volume) and at loud volumes - when I push the master pot it should be where a tube amp shines, but in fact I am loosing the preamp gain feel (the hair-spray JCM800 kind of tone, hehehe) and it gets mushy as if I am loosing current. The picking dynamics also get very loose. It is like the opposite of a cranked VOX AC15. It doesn´t bite anymore and it looses treble.
    I have also removed C51 and C56 which have helped opening up the sound and with R73 = 100K I have lost some "bad" compression and hiss (which was a good thing). Probably it was the global effect of the 3 mods altogheter, but I see in the comments that R73 should be put back in to factory speced 1M.

    So far, I have made the modifications on the attached file. I have some mods to to this weekend too as I have placed the 470+470pF network in the wrong places. The first one should go into the R9 place, right before the gain pot. And the second one, after the gain pot in R44 (where it is already). The 470+470pF network I have in R39 must be removed. R39 will stay as a 470K resistor and C22 will be removed.

    Then, it is not visible in the mod file, but I have also jumpered C31 (47nF cap before the tone stack). That helped a lot in making the tone come alive. At least from low to moderate volume levels. A "blanket" has been removed...

    I have used voltage dividers reasonably between the gain stages to help smoothing things out. Typical 470k + 470k for 50% attenuation.

    The clean channel is not too bad. I have increased the bass cap to 100nF, because I lost the warmth with the 680nF cathode bypass on V1A. This is a shared triode with channel 2 and 3 (gain 1 and gain 2), so it has to be a compromise.
    A capacitor to ground (C7) from the treble pot was also removed and I made a jumper trying to copy another tone stack which has brought back some high-end clarity. Not sure how it works, but it has helped. With a 2.7K / 680nF combo in V1A cathode, I can hardly have a better clean " full bodied" channel with "sparkle". I´m quite happy with the clean sound, i guess...
    The new 500K volume pot VR3 also helped with some extra gain and clarity. Still, my AC30 normal channel distorts more with just a single preamp triode.

    The amp should sound better and better as the volume increases. It gets better and after 50% on the master pot it just becomes worse.
    It is a lot better than factory specs at the moment. A LOT has been improved, but right now, I am stuck on the improvements. I thought the PI or the poweramp stage could be the culprit.

    Also thinking in changing the master volume pots (VR10 + VR11) from 100K to 500K.

    Bendyha made a good observation : Improving the weaknesses of the schematic is the key, specially using the already available component slots and change the values to the correct ones. I think the amp has got a good base for that !

    What do you all think ?
     

    Attached Files:

    Last edited: Aug 3, 2018

  9. HF1600ie

    HF1600ie Tele-Meister

    Age:
    37
    147
    Sep 26, 2016
    Portugal
    Hi, meanwhile i have placed R73 back to 1M. Volume has increased a lot again and also hiss. R36 was jumpered and R9 is my first 470k/470pf treble peaking circuit. The one at R39 was normalized by removing 470pf C22.

    Any suggestions about this design? PI and poweramp? Preamp should be fine, but the hiss is still a bit above my expectations. Could be better...
     

  10. Bendyha

    Bendyha Tele-Afflicted Ad Free Member

    Mar 26, 2014
    Northern Germany
    If you have still left out the 1M bypass C56, this may well be the cause of some rise in hiss. If you found it was bleeding away too much treble, try making it a bit smaller, but it is there to limit the hiss, (amongst other things) I would guess the 1M will be adding thermal noise.
    For a less bottom, but still good gain...hence more top-end punch...you could try halving C51 from 47µ to 22µ, or even 15µ.
     

  11. JD0x0

    JD0x0 Poster Extraordinaire

    Age:
    28
    Feb 22, 2009
    New York
    4 gain stages will hiss.. What resistor wattage and composition are you using for plate resistors?



    Did you 'fix' the bass in the PI yet? (ie. Couplers and grid leaks)

    You could also change the 470k grid leaks in the OD stage(s) to trimmer pots. This will let you tailor your gain and compression a bit more in the clipping stages, and you can individually tame stages at your choosing.
     

  12. HF1600ie

    HF1600ie Tele-Meister

    Age:
    37
    147
    Sep 26, 2016
    Portugal
    Hi, in fact, the amp is feeling treblier now that i have reallocated the first treble peaking circuit 470k/470pf into R9 slot. Is 47nF as PI coupling caps too much? Should i decrease the EL84 grid leaks from 470k to 220k too? Never changed those.

    I think i have 5 gain stages, not 4. They are V1a, v1b, v2a v2b and v3a and I have other amps with 4, they dont hiss. The Hayden hisses just by being ON, even with the gain and master at zero! No cable, nothing. Just ON. When i had R73 at 100k, it was a lot less, but something wasnt working right that way in the cathodyne PI, as said previously. So i put back the 1M plate resistor.

    I use normal light brown resistors. They should be carbon, but nota vintage. All 1/2 watt rated. R73 was and still is 1W.

    Thats a good idea to place trim pots in the OD stages grid leaks!! I have to try that!
     

  13. JD0x0

    JD0x0 Poster Extraordinaire

    Age:
    28
    Feb 22, 2009
    New York
    Carbon will tend to hiss a bit, add in 5 stages amplifying any of that hiss, and it can add up fast. That may be an area to look at for hiss, especially on higher value resistors. Also, you could probably experiment with NFB to tame the PI and some noise there, since this amp doesn't seem to have any..

    https://www.ampbooks.com/mobile/amplifier-calculators/coupling-capacitor/

    Yes, for the grid leaks and PI output couplers try 22nF and 220k resistors. You can plug in the values to the calculator I posted above, and see for yourself, it actually shouldn't really trim up the bass too much, but it will attenuate some lower frequencies not reproduced by guitar (or even bass) which will make the bass sound a bit tighter, without actually losing much that you can hear.

    The smaller caps and grid leaks will also help improve bias excursion, resulting in a tighter faster response from the power amp. It should reduce blocking distortion, which is a very compressed, farty, fizzy unpleasant sort of distortion that happens when stages are over coupled, and pushed hard into clipping. If you check on the calculator, you will see with the 22nF/220k you only lose about .5dB of gain @ 82hz (low E) vs 1kHz, compared to the previous values, but it's trimming out a bunch of 'garbage' from 1hz to ~40hz that no guitar speaker (and most bass speakers) would not even be able to reproduce.
     
    HF1600ie likes this.

  14. Old Tele man

    Old Tele man Friend of Leo's

    May 10, 2017
    Tucson, AZ
    If it worked BEFORE you changed something...put that SOMETHING back.
     
    LudwigvonBirk likes this.

  15. HF1600ie

    HF1600ie Tele-Meister

    Age:
    37
    147
    Sep 26, 2016
    Portugal
    It worked, but it didn´t seem to work so well...
     

  16. HF1600ie

    HF1600ie Tele-Meister

    Age:
    37
    147
    Sep 26, 2016
    Portugal
    Hi,
    Ok, I will give it a try to the grid leaks (R66+R77) and PI output couplers (C53 + C54) at 220k/22nF. I believe some Marshall use 22nF PI output couplers such as the JMP50 lead and the JCM800.

    What kind of resistor composition should I use for less hiss ? They´re all 100K.

    I also have the preamp plate coupling capacitors set at 22nF except the last one which is 1nF. I believe I will have to raise this one to t least 10nF as the amp is feeling trebly due to the re-allocated 470k/470pF network into the R9 slot. It feels right in "texture", but I need get some warmth back. Another option would be to remove the second 470k/470pF network and just leave the 470K resistor in its place alone. I´ll swap the 1nF cap out first because it´s a lot easier.
     

  17. Bendyha

    Bendyha Tele-Afflicted Ad Free Member

    Mar 26, 2014
    Northern Germany
    You should not look at the coupling caps so individually. The limiting of frequency width starts at the start....and goes through to the end.
    There is no point amplifying things that will muddy or overload the sound up-stream, if the frequency is going to be cut out down stream.
    The original schematic had had the first couple cap C4, as a moderate bass limiter at 4n7, this was followed by 22n 22n 47n 47n....so ever increasing, but they were not, and can not, let any bass through that was cut-out by the first cap. This is good practice for an open, clean and balanced sound, first remove what is not wanted, then don't squeeze or limit the sound developed in the amp.

    Some marshalls have a 2n2 as the first couple to the gain channel, followed by 18n or 22n.

    What you now seem to have created is counterintuitive to conducive, congenial circuitry.
     
    Last edited: Aug 6, 2018

  18. HF1600ie

    HF1600ie Tele-Meister

    Age:
    37
    147
    Sep 26, 2016
    Portugal
    I wonder if the gtid leaks of the poweramp have to be 1watt rated. I believe i only have 1/2 watt around. Unless i use a 420k 1/2 watt resistor in paralel with the ones in the PCB which are 470k 1watt ratted. This gives me 220k, but mixed wattage.
     

  19. wanderin kind

    wanderin kind Tele-Meister

    Age:
    63
    253
    Apr 24, 2018
    Olympia WA
    anybody tried the choke tail inverter?

    don't know the choke value as the pic is not so good, maybe 30 H ? bhl-15-values.gif
     

  20. Bendyha

    Bendyha Tele-Afflicted Ad Free Member

    Mar 26, 2014
    Northern Germany
    30H, 595 Ohm..that would be the Hammond 157G
    Never tried it, but find the idea interesting. I was looking at the idea of a cathodyne PI with two chokes not so long ago, but a pig-tailed-pair sets a certain minimalistic appeal.
    With the 6SN7 combined with the (sort of) constant current cathode, there will be plenty of current to overdrive the grids of the 6V6's
     

IMPORTANT: Treat everyone here with respect, no matter how difficult!
No sex, drug, political, religion or hate discussion permitted here.