[BUILD] GIX-51X tube preamp

[gemini86]

Okay, I've done 2 things which has fixed the LF oscillation.

This is a no-no when it comes to trouble shooting- because now I don't know which thing I did was actually necessary and now I'm too lazy to go back and try one or the other.

1) Added .1uF cap from FB pin3 to ground (across R503,504)

2) Lowered impedance of feedback network. R502=270k R503=680R, R504=1k

Try the cap first. If that doesn't stabilize the output, try changing the feedback network.

With this combo I'm running at about 238V, only noise is about 20mV at around 79Khz, which is knocked down to a couple milivolts after L503.

Science be praised!

Update on shipping: I was planning to have everything shipped out Saturday, but my boss called me into work. Monday is a federal holiday for the post office. So Tuesday will be the final exodus.

Thanks for your patience, everyone.

 

[irfrench]

Rodney,

Thanks for updating so quickly.

Glad to see my unhelpful 'help' didn't derail you!

I'm looking forward to this!


Regards,
Ian

 

[bruce0]

I used R502=332K R503=249R, R504=1.8K which I think gets me to the same place you are.  I built it without the cap to stabilize FB (I know you told me to try the other thing first, but lets just consider this diagnostic).  Finished build on unit 1 except transformers, I will strap in a couple of transformers I have laying around and fire it up tomorrow, and let you know.

I have to say the fit and finish on this board, metalwork and faceplate is outstanding, no issues other than drilling the faceplate.  The tube RA mount is really solid.

 

[gemini86]

that should be about right, bruce.

I'm glad you're pleased with the boards...I spent quite a lot of time getting everything to line up just right. Well worth the effort, I think.

 

[bruce0]

You need the cap I think.

With no cap, I also got a slow hunting behavior of a couple of hundred milliamps in the B+ supply, in addition to noise at the 109K, 300K ish, and 1+mhz range.

I have 4 boards, so I made 4 different HV Boards using different size caps, and tested them on the bench.

I think the cap value must be related to the R502+R504 value which in my case is 2049 Ohms (yours was 1680, and you used .1uf).

Unboxed (so I don't know if there is any issue from shielding) my results were.

I tried .01uf, .047uf, .22uf and .67uf value caps.

I got the slow hunting except for the .22 and the .67, and I could even get the .22 to do it sometimes (it seems to be ok, and then "lock in" to that Low frequency noise and get stuck there.  I could kick it off by putting a signal through and turning the gain up. It would remain even when I turned it down.

So for my case I think I am putting a .67uf cap in there (i used polyester), I might try a 1UF.

Note that i also did some tests generating low frequency signal output at approximately 1.2V RMS (0dbu) and I found that at any frequency below around 60 Hz it is easy to see the sag in the B+ supply, so maybe it would be good to up the caps on the B+ rail (or maybe the drain of the lower value resistors in the voltage divider is a bad idea.  Any thoughts on this?

My current config is
R502=332K R503=249R, R504=1.8K with a .67uf film cap across R503, R504.

The cards generate between 247V and 245V (1% resistors, hand matched).  So I am getting about 0.7milliamps across the feedback divider.  What current can the card put out at 245 Volts?

Gemini86:  What about larger caps for C507 and C25? Any thoughts on which one needs to be expanded to avoid the power supply getting sagging at low frequencies (it doesnt exactly sag, it matches the output side wave in frequency but looks more triangular.) At low output levels it is not such a problem, this may be because I wired the transformer at 3:1 to get more inductance.  At 2:1 it would have to output less.  Even wired this way it can crank out a pretty big undistorted signal, but I am worried about what the PSU is doing.

Any thoughts on the feedback stabilization cap, and how big it can be? A .1uf cap in parallel with a 1680R resistor has a corner frequency around 900hz, so I can't figure out why it helps at all. 

 

[bruce0]

Build PICs.

I got to use my hot air rework station to assemble these.  I put the big inductor down first, because it takes so long to heat, but you really have to shift it to make room for the diode, so the big electrolytic fits (as gemini86 notes, but did i listen?) and I didn't, but I got it all on the card.

4 little High Voltage cards, with different bypass caps tacked on. 

Card set up to test,  I don't plan to leave these makeshift transformers, but I made Lundahl adapter cards for some scavenged transformers (and worked very well actually) 1:11 step up Altec on the input, 3:1 step down on the output, got good response all the way down to 20HZ with about 1 db rolloff, although I got some rolloff on high frequencies when the line switch was in LINE position, worked fine in MIC.  Maybe the test generator? I will look into that.

I could permanently fit the output tranny but I would have to carve up the card... and it is such a pretty card, so not going to do that.  The input however fits, and I might use that on some cards.

Note that I lucked out on the gain resistors.  I had to use a 1/2 watt one in there, and it fit with some trouble, but only because of it's position.  Some of them must be 1/4 watt where the main card comes close to the control card, or it won't fit.  You can see one sticking up off the back of the control card.

Also note my choice of Polyprop for C3 fits, but I had to shift it, and I suggest using something smaller.  Also you can see that I wedged in a 24mm output pot, works fine, and feels nice, I have the spec'd part on order.

Big honking 4.7uf Polyprop on output barely fits, I had to shift the electrolytic cap some (that cap has 250 volts on it so be careful with the leads).

Set the Lorlin for 11 positions, 10 clicks or ticks.

I needed to drill out the faceplate output hole, this may be because it is too small but it may be because I used a 24mm pot.

In insulated the back mounted MOSFET with foam tape, not sure this is such a good idea, hard to remove things once this is on, and it does soak up the flux cleaner.  But it is secure.

Here is how I mounted the socket for the board mating.  Because my sockets were a bit too long i thought, I bent the bins a bit and mounted it elevated, which worked out fine (but lowers the above board clearance a hair, so if you are putting some tall transformer  or cap on then lookout.

You can see the socket peeking out here from below the board.  you can see my C3 is too big (and I had to bend the pins on it to keep it from contacting the tube sockets.  Also note that C10 could interfere with soldering the bottom socket if you install it early (as I did).

Another view of the socket.  Note C10, if it was a WIMA I would have had to remove it.

Bent pins on the socket, to help it stay back under the board, hope it doesn't mess up the reliability of the socket, not sure it was necessary.

Here is how it goes together, and with standoffs and screws it stay firmly supported.

On the second build I used OEP's, and cans.  The cans come disassembled, and I tacked to ground pin to the bottom plate and the plate to the case for solid grounding.  Not sure it is needed.  Note... label the trafo's or you might have to take em apart to see which one is input.  Also not sure I am going to stick with these Trafo's they were cheap (15 bucks a piece) but I am getting really wacky stuff with the low frequency and the frequency response I am measuring is not very flat. Maybe I need to terminate them differently?  Should these work down to 50 Hz in a reasonable sine wave at 0.5Vrms?  Mine don't ... I might have some other problem.

Also NOTE on the BOM the Input and Output are switched.  They are correct on the schematic.  The A2E is output the A3E is input.

It looks nice, even with Radio Shack knobs.

 

[braeden]

Very cool, good job!

 

[okgb]

In retrospect , was your Bom good ?  not that there's room but would the converter benefit from heatsinking ?

 

[gemini86]

I did some testing with my unit. I'm using OEPs input and out, which are small and don't have a huge low end.

I've measured only a 200mV sag in the HV when putting out a little over 20v peak to peak @ 100Hz, terminated into 1k. Of course the sign wave is totally distorted at that point, but this should be a worse than worst case scenario for the output stage.

I wasn't able to get the converter to start oscillating, but I started messing with my other converter to see if we can make things better.

The fet and the inductor are the biggest sources of noise. Try this and see what happens.

-Moved the feedback network to the underside of the PCB
-Moved the FET to the top side of the pcb (cut the mounting tab off)

I want to experiment with separating the control side from the high current side, but that's to come later. I'll need to have a new proto pcb made and I just don't have the cash flow right now. I also want to try the maxim 668 chip for my next designs. tech support tells me it's much more stable and less sensitive to feedback contamination.

Bruce, I think the inductance on the output transformer you're using is maybe a bit much to drive for this output circuit. The G9 historically has a hard time driving heavy loads... It's been recommended elswhere to use a 3:1 or even 4:1 output for better impedance matching when driving low Z inputs, but I can't remember who said that. I personally love the sound of this pre. Mine is not totally flat, and has some low end roll-off, but it suits the source materials that are used for this pre to really shine.

Also, bruce. you've set the gain switch to 10 ticks... should be 11. I used a '0-10' scale on the faceplate, because I prefer that for the output pot and wanted them to match.

okgb, the bom has been updated today to reflect what I've used for my dc converter. Also, the converter itself is doing some pretty light lifting. There is very little heat generated (I've selected the FET to be as low RDS on resistance as possible) and the inductor will be the only thing warming up due to core losses. A heatsink is not needed.

 

[bruce0]

In retrospect , was your Bom good ?  not that there's room but would the converter benefit from heatsinking ?

The bom had some problems, but I have been correcting them and updating as i gain experience.  I will make another correction today.

Here is a sheet with comments from my build, it is based upon the layout that was posted.  Does anyone have an up to date PCB layout PDF?

 

[gemini86]

Here's an updated layout

https://docs.google.com/file/d/0B6s_klsk9wMDV3M0RGg1cUlJdDA/edit

 

[bruce0]

I did some testing with my unit. I'm using OEPs input and out, which are small and don't have a huge low end.

I've measured only a 200mV sag in the HV when putting out a little over 20v peak to peak @ 100Hz, terminated into 1k. Of course the sign wave is totally distorted at that point, but this should be a worse than worst case scenario for the output stage.

I just got some of those OEP's I will try that out. 

Was your sag in "time" with the music?  I was only putting out 1.2V rms

Looking at the schematic it looks like maybe the issue might be the 1K resistor with 10uF low pass filter in the B+ rail.  That has a rolloff at 15hz, and maybe what is happening is that the HV supply cant keep up with the load of the second stage and that is varying B+ in the first stage, and that gets amplified thus the oscillation.

I have ordered the Lundahls and I have the OEP's I will try them out and see what gives, and report back.

The big G9's have a plentiful B+ supply and the 10uf cap C25 is not called upon to stabilize large audio band noise I think.  I am considering a larger C25, maybe a larger R31, I can't fit a larger C507 or C503, but might be able to replace C506 with something.

I will report back

Also, bruce. you've set the gain switch to 10 ticks... should be 11. I used a '0-10' scale on the faceplate, because I prefer that for the output pot and wanted them to match.

okgb, the bom has been updated today to reflect what I've used for my dc converter. Also, the converter itself is doing some pretty light lifting. There is very little heat generated (I've selected the FET to be as low RDS on resistance as possible) and the inductor will be the only thing warming up due to core losses. A heatsink is not needed.

I set it for 10 clicks which is 11 positions. I think that is what you want.

 

[bruce0]

Sorry my mistake, set the switch on 11, for 10 clicks.. You are correct, I will remove the incorrect note.

 

[bruce0]

I did some testing with my unit. I'm using OEPs input and out, which are small and don't have a huge low end.

I've measured only a 200mV sag in the HV when putting out a little over 20v peak to peak @ 100Hz, terminated into 1k. Of course the sign wave is totally distorted at that point, but this should be a worse than worst case scenario for the output stage.

I wasn't able to get the converter to start oscillating, but I started messing with my other converter to see if we can make things better.

The fet and the inductor are the biggest sources of noise. Try this and see what happens.

-Moved the feedback network to the underside of the PCB
-Moved the FET to the top side of the pcb (cut the mounting tab off)

I want to experiment with separating the control side from the high current side, but that's to come later. I'll need to have a new proto pcb made and I just don't have the cash flow right now. I also want to try the maxim 668 chip for my next designs. tech support tells me it's much more stable and less sensitive to feedback contamination.

I think a generic HV card for 51X tube projects is a great idea, maybe with some right angle tube socket boards.  I would spring to make those cards, pm me if you are interested.

To clarify my situation, I don't think I was clear, and I don't think moving the cap helps.

I should explain that the oscillation I was getting I measured at 4hz, 5Hz and 7 Hz ish, my scope isn't that clear, and there was other noise, and the scale of the oscillation in some cases was only 100mv or 200mv, sometimes larger.

It occurred before I put a large enough cap in.  And when I put in the .22uf cap, I could cause it to start by turning up the volume on a low sine wave, but it wouldn't stop when I turned it off.

I think the kinds of noise that might radiate to my cap antennae would be high frequency where I think things are under control.  With the .68uf cap in there all is well.


Separately I had a "sag" problem.
As far as the HV "sag".  I saw the HV voltage at the HV jumper vary substantially in time to a 20-60HZ sine wave while the unit was putting out 1.2vRMS.  THe filter made by R31/C25 has a corner frequency of 18 (assuming that the electrolytic is only 10% under it's stated value... they rarely are better).  That would only attenuate the variation in HV supply by about 3db, and I think the remaining signal could be being amplified and causing distortion.  (And incidentally this is just the kind of long feedback loop that would causie a  Low Frequency oscillation like what we are seeing, but I have no proof of that). 


I am building a unit with OEP's and will test it.  I will build it with C25 as 150uf for a low pass with a 1 Hz corner (it's pretty tall!, but it fits for testing).  I will report back.

 

[gemini86]

Thanks for the explanation.

I read in another forum about somebody developing a HV converter using this chip, and that the HF noise causes this low frequency instability. In their design they simply put more space between the inductor and the controller, which obviously isn't a choice for this application. But, just like us, they were a bunch of amateurs poking things with sticks until the desired result is achieved.

What I did was run a sign through my unit @ 100hz (even tried it at 50hz) and cranked the gain and output up to max. The clipping on the output was a very weird, non square wave. Sort of like a double sine, or something like a camel hump on the sine. I'll post pics maybe tomorrow. I'll also look for a signal appearing on the HV supply, see if the overall DC voltage is dropping or if it's moving in sync with the signal.

But for tonight, I sleep... I have a cold and I'm trying to be as big a baby about it as I can so my wife will take care of me. (Already scored some tasty soup for dinner and got out of doing the dishes...)

 

[bruce0]

I built a unit with OEP transformers.  They are very small.

I also saw the weird distorted wave on output, anywhere below 60 HZ or so it begins to go sour.  I think that is a result of the OEP 262A2E transformer, but I don't know what it could be because I took a separate one of the same transformer and drove it loaded and unloaded and while it doesn't handle much level, it doesn't do that.

Both builds pass signal, the one with the big transformer passes signal down to 20HZ at really high output levels (it distorts at around 18 dBu 6V rms 17V peak to peak approx).  I ordered a couple of LL5402's and LL1538's last night.  Maybe they will work better.

I am going to swap an OEP Trafo into that build and see what happens, I used a 11uF output cap on the EOP build, don't know if that could effect anything.

Using the larger C25 did not help anything (but it did take a long time to discharge), I still see variations in voltage in time with the sine wave.  Looks like the first stage is using 0.5mA because the voltage drop across R31 is 5 volts.

There is always noise but the thing is what is "normal" noise for this HV supply.

I did hook it up to an amp on the bench and listened a bit and it didn't sound too noisy, i couldnt hear the B+ noise

 

[tonycamp]

I just got my 2 yesterday, quite nice indeed Rodney!, I gotta say thanx to bruce0 also for working through some of the pains, if this was left to my newb arse, the project would be finished in the shipping box ! I will be watching the progress with bated breath.

Rodney, I noticed your comment on what the G9 was best for, i'm new to the G9, if you dont mind me asking, on what sources does it shine?

thanx fellas

 

[bruce0]

Ok, so I spent a bit of time trying to get the HV supply working.  Here are things that I found that make it better:  1) Put more load on it (3 150ohm resistors to ground), I thought this might be because the on-time or off-time of the chip was limited, looked at the data sheet, but the way this PSU runs it is on for most of time and off for only 1 usec which is shorter than the minimum off time.  So maybe more load fixes that.

Then I began reading the data sheet. They do mention shielding and radiation. They say if you are going to make an adjustable supply to make it with R1 (or on our schematic R503+R504) from 10K ohms up.  Ours is below that.

Further they suggest a cap across R2 (R502 on our schematic) rather than across R1 which is what we did.  And the values of that cap are much lower 47pf to 220pf the smaller the better.

So I pulled all my power supplies apart, and built them as follows:

1) I used 2 resistors not 3, because the three resistors make a sort of coil for the radiation to excite, and a smaller coil is better.
2) I used 2.2M for R502, and 13.7K for the combined R503+R504. (This produced 4 units between 244V and 245V but I DVM matched the resistors so the units would be comparable)

3) I put 47PF across R502, which fit now that there are only two resistors.  There is 243  volts across there so a high voltage cap is called for.
4) I experimented with various shields.  I found that the best arrangement was a shield of .006" thick AD-MU-80 high permeability Mu Metal.  Shielding the inductor didn't help much.  It is easy to measure the field, it is HUGE and can be measured with a bare oscilloscope probe all the way over at the input transformer.  I will send pictures of the shield.

With the above I get a stable, oscillation free supply.  I built 4 of them, and on 2 I get 7mv noise, and on 2 I get 12mv noise, and I can't for the life of me tell why those two are different.  Any ideas are welcome.

With the stable supply the OEP's behave much better, but still cant produce a sine wave at 0dB 20Hz, although they can produce something that vaguely resembles a sine wave now. I may chop the boards up to fit my salvaged traffo's after all because they run down to 20 Hz no problem.

The power supply on both sides of the R31 resistor varies in time with the sine waves to a pretty large degree, which is disturbing. 

Shield - AD-MU-80 .006" thick, I think it works.

Notch to clear the sense resistor so you don't short it.

Installed it is pretty nifty, tacked to the sense resistor at the ground end, and one of the ground pins.  But watch clearances there are a lot of things at 250 Volts on the board the FET in particular

 

[gemini86]

Hey bruce...

first off, just wanted to say thanks for your help on this...

I've been messing with this for a while as well... My first thoughts were that adding the cap from the feedback pin to ground ruined the output regulation, making the feedback sense too 'slow' to see the output ripple. I see that you've removed it, so that disproves my theory.

One thing I should note, you mention the datasheet calls for a feed forward cap across R502. We have that, it's C509, 10nF. The reason it's so big is because when I tried using smaller caps (in the past) it would become unstable and oscillate at low frequency. So, take out C509 if you've put a parallel cap in place. See what the effects are.

Also note that the fet is a big source of radiation. That big to 220 tab will radiate amost as much as the fet.

I think you're right to increase the feedback network impedance. My original thought was to lower to combat the HF interference but that is probably not solving anything.

 

[bruce0]

Ok, I did not see that.  I feel like a bit of a dope.

So c509 is already doing what i am trying to do.

I think the impedance change or the shield must explain the improvement I got, because my added caps weren't changing C509 by more than 3%.

I go back to the data sheet and look at the FB-Ground resistor (r1) contemplated therein.  They spec it to be 10K to 500K, so if they use the device to double triple or do 10x voltage they must assume the R2 (output to fb) resistor is going to be 20K  to 5M ohms.  So they are contemplating corner frequency of the R2/feed forward cap at between 36000Hz and 677Hz.    The notion being that it won't regulate in response to changes above that frequency because they are damped out.

But with a 10,000pf cap in there (C509) we are putting a corner frequency of 15.9 hz as (from the schematic 1M ohms) and when you lowered that to 270,000 ohms then the corner frequency moved to  59hz (I used 330,000 so I had 48Hz.  So those are far from what the data sheet contemplates, and basically damp out voltage regulation for changes faster than that.  That could explain why the voltage varies in time with the audio sine wave... ( I think that means it It reduces regulation 6db per octave above 50hz! )  For an audio B+ circuit I think we need regulation to respond to changes as fast as 2000 hz minimum because I think below that B+ sag due to load in response to signal will exceed the noise level of 7 mV.  And apologies but I am not doing math here I am doing this by the seat of my pants from what I saw on the scope.

So I think I need to lower C509 to fix that problem.  At 2.2M for R502 that means I need 47PF or less, I am going to rebuild the 4 HV boards I have with 47pf, 250pf, 1000pf and something bigger and I will report back if any of those are stable.  I am wondering if I further increase the impedance will I have a problem? Even if the feedback circuit is 10M ohms I still have 25000 nA flowing through there with 250 volts, and at 24 volts 2500nA so it should be fine right? (the data sheet says the FB pin needs 20 nA max).

Sounds like you have been down this same path and got LFO (low frequency oscillation) so if you can tell me it wont work, go ahead and save me the work.

And while I am thinking about it I really want to thank you for spec'ing this HV board with sockets and headers, I could never have tried these variations if it was soldered! and would have given up long ago!  As it is I have been pretty busy with my Hakko 808 desoldering gun (favorite tool) and the boards are well made and are holding up well.

As for the MOSFET tab is it drain source or gate (data sheet didn't say)?  Even if it is drain it is just a switch right, I mean can it be creating the kind of magnetic flux that the inductors do?

The inductor is radiating magnetically like a son of a gun.  I set my oscilloscope probe to 5 mV sensitivity and I get a signal with the probe waving in the air an inch or two away from the inductor!  It is 20 - 30 mV signal when I get within a half inch.  You figure the resistors and caps have big conductive rings formed by their leads (and a coil in the case of the metal film resistors which are like a foil slinky inside).  I will try to measure the field inside the shield, and see what the tab is radiating on the bottom.

I don't know what our B+ noise target should be, do you?