DCAudioDIY.com

DC Area Audio DIYer's Community
It is currently March 28th, 2024, 12:29 pm

All times are UTC - 5 hours [ DST ]




Post new topic Reply to topic  [ 54 posts ]  Go to page Previous  1, 2, 3, 4, 5, 6  Next
Author Message
PostPosted: May 12th, 2020, 1:29 pm 
Offline

Joined: June 4th, 2013, 2:39 pm
Posts: 487
David McGown wrote:
Grover,

One thing I noticed about the Kiebert article is that he used the 5687 at the same current at the 6SN7 it replaced. Plate resistors are 50K. That is really a poor operating point for that tube, it really need more current for linearity. Which is why I currently have 20K plate resistors (for the original planned ECC99). What did you use, if you remember?

David


I built it as he designed it because I wanted to hear it. I've never been a huge fan of the 5687. I think it actually sounds better at lower currents but that's just me. At high current it sounds harsh to my ears. What I like about the the 6SN7 is the utter naturalness of the sound.

DavidB, I agree that fiddling with high feedback and phase-lead/lag caps can be a fool's errand, but you are a magician with tubes and do things that would be difficult for the average hobbyist to duplicate. ;-) I've got your 807 schematic which you generously shared a while back, but I don't know how well it would work without the audio-bias board and some of the other details I would not be skilled enough to implement. I've a pair of Acro TO-290s, 12K p-p at 20 watts, and the big Peerless, of course. I'm game to give it a try but I'd have to simplify it a bit.

David McGown, I'd be interested in your reactions to the stock Haffler circuit, and also interested in how well it holds up to some rigorous stability testing. With bandwidth out to 200kHz I wonder, but I'd imagine it's good enough, probably better than the early Heathkit amps, which need quite a bit of tuning to get them stable. Maybe I'll snag a pair of repro A-431's and put them on the breadboard so we can compare notes.


Top
 Profile  
 
PostPosted: May 26th, 2020, 8:37 am 
Offline

Joined: February 28th, 2013, 1:19 pm
Posts: 914
Grover,

Well, I have had an interesting time trying to troubleshoot the amp, going through an iteration of the fixed bias network and reversion from my 5687 differential amp with current sink in the tail to the stock 6SN7 with 375 ohm resistor in the tail.

I am using the Dynaco Modified Williamson schematic with the following changes:
1. Since I am using KT88 outputs, I am using a 47K gridleak resistor for the output tubes.
2. The resistor bypassing the coupling cap between the driver and the output tube has be decreased from 1M to 470K
3. Increased the value of the coupling cap from 0.22uF to 0.47uF (keeping the same frequency as 0.22uF||1M).
4. Put in carbon comp grid stoppers for the input section, 220 ohm at the paraphase inverter grid, and at each section of the differential amp grid

With the driver tube removed, the amp powers up fine to full voltage (470V). I can set a stable bias at 70mA per tube (at around -49V at the grid). This is what I would expect.
With the output tubes removed, and the driver tube replaced in circuit, the amp powers up fine to full voltage (a little higher due to reduced load on the transformer. Everything appears to be OK from a DC perspective.

With all the tubes in, the amp goes haywire. The output tubes start drawing 200mA of current each, my 8ohm 100W load resistor across speaker terminals heats up. B+ drops to around 380V due to the current draw. The bias voltage actually decreased from -49V (with the output tubes isolated from the driver per above) to around -75V or less. This happens both in the 5687 driver configuration and the original 6SN7 driver. Needless to say, I kill the power pretty quickly rather than run an amp that is so heavily loaded to avoid burning out an output transformer and/or output tubes.

Things I have tried.

1. Disconnecting the driver feedback loop (through 100nF cap). - no change.
2. Disconnecting the global feedback loop AND driver feedback loop - no change.
The two cases above SHOULD have told me there was a problem with the feedback being misapplied.
3. I removed the resistors bypassing the driver coupling caps (to output stage) - no change
4. My layout has the filter choke positioned adjacent to the input/driver tubes. I disconnected the leads and hooked up a choke off chassis (in case there was magnetic coupling). - no change.
5. I checked my wiring of the output transformers (based on the lead color markings, and all is per the schematic.
6. I checked my driver feedback connection, and is per the schematic.

What I have not done yet:
1. Swap output tube transformer leads (in case the transformers were miswired). These were new Dynakit Parts A431 transformers, so not a case of misreading color codes due to fading. Pulling the feedback network and still having the problem SHOULD have eliminated positive feedback being the cause.
2. Hook up a scope to see what the hell is going on. Hesitant about leaving the amp on for extended probing due to overload, but could test out the input/driver stage with the output tubes removed.

Any ideas? (addressed to all)

Thanks.

Roscoe, if you feel this should be moved to a new thread, feel free to do so, sorry for taking over Grover's discussion on his Williamson amp project.

David


Top
 Profile  
 
PostPosted: May 26th, 2020, 11:55 am 
Offline

Joined: June 4th, 2013, 2:39 pm
Posts: 487
Sure sounds like it's oscillating.

1) Could be the circuit is out of phase. While disconnecting the feedback elements should have stopped the oscillation, check your circuit again to make sure the phase is correct from the phase splitter outputs to the output tubes. It's unlikely to be the output transformer, since those are well-known drop-ins for the originals.

2) I would restore *all* original values--no grid stoppers, no changes. I honestly think that until you build it exactly as he designed it, you won't know what's going on. Seems to me that the 1M resistors bypassing the driver stage caps combine with the 100K grid resistor to form a voltage divider at the KT88 grid, no? Don't know what changing those would do.

Things like choke placement etc. shouldn't affect stability (noise floor, yes, but not stability). Built as shown, it should be stable into a fixed load.

Good luck, and keep us posted!


Top
 Profile  
 
PostPosted: May 26th, 2020, 12:44 pm 
Offline

Joined: February 28th, 2013, 1:19 pm
Posts: 914
Grover,

Thanks. I will check the phase. It is really easy to get mixed up soldering a layout together directly off a schematic, and would not be surprised if I got the phase splitter outputs reversed. I rechecked the differential stage a number of times so I know it is OK.

The 1M bypass into 100k gridleak is for EL34 outputs. The change to 470K bypass into a 47k gridleak is a documented change from a couple of late 1950's sources that addresses use of 6550s or KT88 outputs. These need a smaller gridleak to keep happy, and the voltage divider is the same. The stated purpose is to provide some positive grid bias from the driver in event of bias circuit failure, so this should result in the same. But it also serves to direct couple at very high frequencies, which is why I changed the cap value to maintain the same effect.

I don't believe the grid stoppers impact the issue I have, it actually should address parasitic oscillation. Maybe a 6SN7 doesn't need them, but the 5687 would, and I just did not take them out. Still, I put in a grid stopper at the input to the phase splitter out of good measure. Note that the original schematic does use a 10K grid stopper at the input.

Thanks for your help, I will let you know when I get this working. I have the other amp built as well (and probably will check to see if it has the same issue).

David


Top
 Profile  
 
PostPosted: May 26th, 2020, 1:16 pm 
Offline

Joined: June 4th, 2013, 2:39 pm
Posts: 487
Fair enough on the documented changes. I believe the 10K at the input is more to dampen HF's than a grid-stopper per se. But I guess they can't hurt. Let me know what you find out.


Top
 Profile  
 
PostPosted: May 26th, 2020, 3:29 pm 
Offline

Joined: March 2nd, 2013, 2:43 pm
Posts: 216
Location: Potomac, MD
It is oscillating, use a scope to find out the frequency.

Be glad you are not using TV sweep tubes. Once I had an oscillation that resulted in a 40- volt drop in a one-inch length of wire as detected as dc by an audio scope that could not respond to the hundreds of MHz that the oscillation was likely to be. Did not have a high-bandwidth scope at the time.


Top
 Profile  
 
PostPosted: May 26th, 2020, 9:09 pm 
Offline

Joined: June 4th, 2013, 2:39 pm
Posts: 487
I'm trying to think what would cause this oscillation. You did keep the 10K in-line resistor at the input, right? You didn't swap in a pot at the input?

Here's another thought--check the OPT primary leads to make sure they are correctly tapped--in other words, make sure blue is plate, green is 43%, etc. Because if they're reversed you'll get a giant oscillator for sure.

I feel like maybe there's some answer in the bias supply. Why would the bias voltage *drop* during oscillation?

This coupling cap bypass thing--I thought it was intended to stabilize the LF response.

Okay, now wait a minute. Look at the Keroes article, Dave:

http://www.technicalaudio.com/pdf/Gordo ... n_6550.pdf

He's using the reverse tranny for bias, and he's reducing the coupling cap bypass resistors to 470K. He's also reduced the grid resistors on the KT88s to 47K, BUT he's also got another 100K between the bias pot legs and ground. Are you only using 47K to the 10K bias pot to ground? Are you shunting too much driver plate voltage to the KT88 grids? If you are, then are the KT88 grids are going positive, and the bias supply is taking a weird hit? In other words, in Keroes's design, the *effective* grid resistance is 47K, but he's actually got 147K between each grid and ground, am I looking at that right?

Also, Keroes does not increase the coupling cap value to the KT88 grids, so where did you get that? Might be creating some LF oscillation.


Top
 Profile  
 
PostPosted: May 27th, 2020, 9:20 am 
Offline

Joined: February 28th, 2013, 1:19 pm
Posts: 914
Grover,

Thanks for the link to the Keroes article, I was trying to find it to sent you with respect to my bias/balance scheme, but could not find it (should have bookmarked). I have been using that article as well as the Hafler Modified Williamson as the basis.

With respect to your comments/questions,

I am using a 10K input resistor to the input stage (grid stopper). No pot.

I checked the transformer primary windings, and they are correctly tapped.

My bias arrangement is similar to the Keroes, only I used as 25K pot with the wiper tied to one end to work as a variable resistor instead of the fixed resistor. Balance pot arrangement same as Keroes (10K pot), however, using two 47K resistors to ground instead of 100K resistors in the Keroes circuit. This is to increase current through the bias network since I included a 150H/8mA choke inline with the bias pot (between the two bias supply caps) to improve filtering. (Note to self, I need to try bypassing this to ensure it is not inductively coupled and injecting noise into the bias supply). The two resistors above are effective only in DC, since the last bias capacitor bypasses these, I calculate a corner frequency of 0.14 Hz. From an AC perspective, it should have no effect on the voltage divider formed by the resistor bypass of the coupling cap and grid leak resistor.

I did some quick calculations of the effect of the coupling cap values and voltage divider values, compared with the Keroes and Hafler Williamson variants.

This circuit is a LF step (cut) circuit (Grover, you are right about that, I mistakenly thought the effect was at high frequencies). It basically limits the LF rolloff of the coupling cap, wiht the cap acting into the grid leak resistor defining the bass cut frequency, and the cap in parallel with the bypass resistor defining the stop.

The Hafler circuit used 0.22uF || 1M -> 100K, which gives a 7.2 Hz corner for bass roll off, stopped at 0.72 Hz
The Keroes circult used a 0.22uF || 470K -> 47K, which gives a 15.4 Hz corner for bass roll off, stopped at 1.54 Hz
My substitution of the 0.47uF cap results in the same bass roll off and stop frequency as the Hafler circuit. Therefore, I believe I am OK with respect to the original Hafler circuit (using the Dynaco A-430/A-431 iron, whereas the Keroes is using Acro TO-330).

So I am pretty sure that my modifications are technically sound, still, the amp is oscillating... :crazy:

David


Top
 Profile  
 
PostPosted: May 27th, 2020, 9:38 am 
Offline

Joined: February 28th, 2013, 1:19 pm
Posts: 914
David McGown wrote:
This circuit is a LF step (cut) circuit (Grover, you are right about that, I mistakenly thought the effect was at high frequencies). It basically limits the LF rolloff of the coupling cap, wiht the cap acting into the grid leak resistor defining the bass cut frequency, and the cap in parallel with the bypass resistor defining the stop.

The Hafler circuit used 0.22uF || 1M -> 100K, which gives a 7.2 Hz corner for bass roll off, stopped at 0.72 Hz
The Keroes circult used a 0.22uF || 470K -> 47K, which gives a 15.4 Hz corner for bass roll off, stopped at 1.54 Hz
My substitution of the 0.47uF cap results in the same bass roll off and stop frequency as the Hafler circuit. Therefore, I believe I am OK with respect to the original Hafler circuit (using the Dynaco A-430/A-431 iron, whereas the Keroes is using Acro TO-330).



To avoid confusion, by "stop" I meant the roll off is stopped at around -20 dB level reduction.


Top
 Profile  
 
PostPosted: May 27th, 2020, 11:05 am 
Offline

Joined: April 22nd, 2013, 12:58 pm
Posts: 285
Do you have a cheap set of speakers you could attach to the outputs? It might help you identify whether this is a low frequency or a high frequency oscillation. The culprits would be quite different, I imagine. If it's motor-boating for instance, that's something you would hear right away.


Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 54 posts ]  Go to page Previous  1, 2, 3, 4, 5, 6  Next

All times are UTC - 5 hours [ DST ]


Who is online

Users browsing this forum: No registered users and 53 guests


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
cron
Powered by phpBB® Forum Software © phpBB Group