PMsix61 Limiter DIY

[alexc]

Investigating noise some more :

And it depends very much on the primary termination.
Below 10K much less. Above 20K a lot more. O/C is the biggest.
And the higher the term, the more mid-freq hash picked up.

Looking again at my rack unit, powered down I have a loopback of -84dBu.
Powering up with no psu secondaries  -58dBu
Powering up with everything but the heaters (no signal amp gain) -58dBu
Adding heaters so signal amp is giving 17dB of gain  -41dBu

So it looks like I get a basic 26dB of noise from proximity and then 17dB for the 17dB of gain across the signal amp+output traffo.

I can confirm by placing the identical but disembodied transformer next to the rack installed one and showing that with 10K primary termination and the same Motu on the outputs, I getmore or less the same -84dBu with rack powered off and -58dBu with power on.

So : what I'll do now is rebias the signal amp for less current, redo the gain and noise measurements then try some other output transformers : first the Sowter and then an Edcor XSM. I can adjust the gain more or less with the bias to try to get as like to like gains as I can.


In a clean system, 17dB of gain can give less then that much added SPL noisefloor, but with a dirty floor that's what you expect. Like added for like and sometimes even more.

By way of comparison, my original Poorman which uses the same PSU and the XSM has a unity gain noise floor of  around -80dBu and around 72dBu with 14dB of added signal amp gain.

 

[alexc]

So I did the comparisons and didn't come up with anything useful really.

The XSM seems to be a lot quieter in the powered-on (some 10dB quieter), no signal amp current case and also showed a higher gain, despite the bigger step down ratio.

The Sowter has a step down ratio too high to be useful here. (I have only 135V B+ in the signal amp)
But it is around 13dB quieter in the no signal amp current case.

Back to the drawing board.  :

 

[alexc]

So my next move is to strip back the signal amp to the beginning and check as I go for noise.

I have another chassis with similar parts so that's where I'll do it first.

Hopefully I can determine exactly what is my major malfunction

In the meantime, I'm looking for an LS - I sold off some more stuff so hopefully I can find an affordable one!

 

[MartyMart]

Is this all "flying leads" style Alex or do you have it "in case" during your testing ?

Just wondering how that might effect the noise readings, have had that kind of issue myself and
when all buttoned up with good ground and case connections, it went away ?
( or am I way off base here )
M.

 

[alexc]

The 'substitute traffo' tests were flying leads and yes, that alone could cause high noise induced because it is a relatively hi-Z wired coil.

But I don't think that would account for the major noise levels I'm seeing. 27dB of basic induced noise is huge. Unprecedented for me.

However, I'm not yet able to conclusively show the cause of my issue - I do feel it is basic inductive thing in a hi-Z winding context, but cannot yet rule out ground loops.

Surely I've never seen this level of induced hum, so before I write off this traffo I need to do more investigation - rebuilding the circuit from the 'ground up' in a more basic form but as close as possible. Eliminating all the dc reference voltages and extraneous grounds.

I do think in the end it will be this traffo is not for this circuit. But I'll know more in a few days.

One things for sure - the way it is, it is a 'led zeppelin'. Fly's like a brick. Straight down.

 

[DaveP]

Alec,
Have you tried Star grounding each section to the same point?
best
DaveP

 

[MatthisD]

What transformer is it exactly, is it one that is rated for DC push-pull?

Do you need something equivalent to the 670/Triad HS-52 output, I know it is a mono channel you're building and that twin secondaries are not necessary but are you flowing roughly the same current as the 670 signal amp?

The HS-52 catalogue data states that "low frequency loss will result from use of unbalanced DC in windings other than where specified" which isn't specified for HS-52 so in the 670 signal amp there must be very little unbalanced current through the windings. If your signal amp is reasonably well balanced then your output transformer should not need to be specifically rated for unbalanced DC.

I think you mentioned that your low frequency response is good, if we knew the inductance of your output transformer that would tick another box for whats needed here with 4x 6K4P.
Matt.

 

[alexc]

More or less, I have (a star point)

At least I have seperated the high current return grounds with their own heavy gauge ground to chassis star point.
In some low return current cases there are local ground star points which then go back to the main star.

I have not eliminated ground loops as the cause of my probs just yet.

I think I have reasonably satisfied myself that it is not the issue, but am going to strip down the signal amp to it's most basic form to do just that. 

ie. remove the -ve reference voltages with their grounds and run the signal amp as a simple pp cathode biased amp.

That should be conclusive.

I also have a second chassis with same psu traffo, same output traffos that I will further investigate.

Either I have one mother of a ground loop or one mother of psu induction.

I hope I'm wrong, but I think it is the latter - these guys seem to suck up 50Hz and harmonics like a sponge.

Couple of days, maybe a week or so  I should know what to do.

In the meantime, I'm saving the bux for a good, shielded traffo.

All my 'off the beaten tracks' builds end up being an oddessey 

Cheers

 

[alexc]

Hi Matthis,

I'm using an Edcor 10W 8K:600 traffo CXPP series. It is intended for PP applications, HV CT fed.

I don't have all the specs, such as tolerated unbalanced DC or even balanced dc rating.
It is generally the kind of thing one would use in a 6V6 PP plates to line application.
Meaning around 40mA or so per phase DC balanced. Which is more or less what I am running.

Not sure about the Triad spec or model. Not even sure about what the orig 660 dc currents are.
This is a mongrel, prototype build loosely based on the topology of the original.

My issues are not related to dc unbalance - I am testing at one current really  - at idle, and I have good balance.
To within less than 1mA.

What I have is big hum - I think it is basic psu radiated induced.
I'm in the process of proving that to myself.
I have not fully  ruled out ground loops, but nearly have.

The inductance could well be an issue - the traffo is reasonably heavy, lots of iron and so on.
But I have not explicitly measured the inductance.

I will say that the cheaper, smaller XSM series appears to be much quieter in this app. And I don;t think it has more inductance.

I don;t think I am seeing excessive bass roll off - if anything, I thought it was a big hump at the lower end.

But I did do a freq response test - I don't see a generalised hump at the low end which would exacerbate the hum.
The response showed a predicatable roll off from 100Hz down.
Nothing huge, just not a big peak at the bottom.

 

[MatthisD]

  I realise the noise issue but that itself is probably linked to the transformer being so large and not having magnetic shielding. The rolloff and low frequency peak suggest inadequate inductance/impedance mismatch in which case the size of this transformer may not mean it has high inductance. The DCR measurements of the primary taps would give an idea of the winding structure which may be another problematic feature.

  Establishing that the transformer doesn't need to handle more any more than 1mA means the ideal transformer can be much smaller. The XSM 10K:600 has roughly 70H primary inductance. If you have that XSM you can compare frequency response to the large 8K:600 to see if there are changes with the rolloff curve and the peak in the low frequency of your circuit. Theres no insulation between the windings in these transformers. I'd imagine DC should not be fed through the windings in normal use in case of primary shorting to secondary.

  The HS-52 has something like 100H total primary inductance and the 670 final B+ filter cap is 80uF which may be minimum size for no low frequency resonance. Sorry to mention the 670 but you haven't posted a schematic that I can see and it would seem more relevant than PM660 which uses 2 6BC8.

 

[alexc]

  Theres no insulation between the windings in these transformers. I'd imagine DC should not be fed through the windings in normal use in case of primary shorting to secondary.

The CXPP is certainly designed as a push pull centre tap fed B+ config with significant balanced dc current at quite high voltages.

The XSM is also able to work in that config, even at quite high B+. I've used that config a number of times with no issues at even 30mA + of balanced dc operating in a vari-gm limiter where undoubtedly imbalances happen dynamically.

It could well be the case that the insulated magnet wire windings (primary and secondary) are not further seperated by insulating materials - I think that is a feature of higher quality transformers. But there is no question of primary-secondary shorting. That would not work at all.

When this circuit is working as a limiter, there will certainly be imbalances in the dc - probably up to 10% of peak current.
In this case probably up to 5mA or so under the dynamic of shifting high current-low current in the vari-gm operation.

That's the reason for needing a beefy transformer here - generally, but not always , big lumps of iron can tolerate transient unbalanced currents better than smaller lumps of iron, in ungapped transformers.

But for now, at this one operating point, the dc imbalance is tiny. There does not appear to be any rolloff beyond what I expect to see, and that observed rolloff is mostly a function of the input transformer in this circuit.

But I will review how to measure inductance and take a stab at it - it is clearly a very important parameter.

What I find strange is that I have used bigger output transformers, with higher primary impedances closer to even larger PSU transformers with very little directly radiated mains induced noise. My EL34pp amps built from edcors are a case in point.

Very similar in many ways to what I'm doing here, seperated by only 15cm or so and operating at sustained larger currents.
I guess the turns ratio being larger means the secondary noise is much lower. Or something


Thanks
Alex

 

[MatthisD]

The CXPP is certainly designed as a push pull centre tap fed B+ config with significant balanced dc current at quite high voltages.

I've no experience with that one but the datasheet is missing DCR and inductance values.

The XSM is also able to work in that config, even at quite high B+. I've used that config a number of times with no issues at even 30mA + of balanced dc operating in a vari-gm limiter where undoubtedly imbalances happen dynamically.

It is wound with 36AWG primary and secondary. The rule of thumb gives 35mA for 36AWG which is well within what you mentioned.

It could well be the case that the insulated magnet wire windings (primary and secondary) are not further seperated by insulating materials - I think that is a feature of higher quality transformers. But there is no question of primary-secondary shorting. That would not work at all.

I'm not suggesting that the windings are shorted together but simply that it is not sold as a transformer to operate with DC in the windings. There is no insulation between the primary and secondaries, not even a single layer of tape.

When this circuit is working as a limiter, there will certainly be imbalances in the dc - probably up to 10% of peak current.
In this case probably up to 5mA or so under the dynamic of shifting high current-low current in the vari-gm operation.

That's the reason for needing a beefy transformer here - generally, but not always , big lumps of iron can tolerate transient unbalanced currents better than smaller lumps of iron, in ungapped transformers.

Isn't that lessened and essentially the point of using 4 valves per side, we know from CJ's interview that everything is carefully balanced to minimize all sources of distortion. I assumed your signal amp is intended to emulate the 670 signal amp; output using low-nickel core/21EI laminations/little or no unbalanced DC. From what you describe you're running higher current. Simply scaling up the core to the next size up with 35 or 36AWG wound primary should do the job, depending on inductance requirement. Or a bigger core again would be more than adequate...an example here by member vari-mu:
http://www.groupdiy.com/index.php?topic=35664.0

Obviously something like that wouldn't be cheap or easy to get but an idea of what is optimum for the task. If you deicde to replace what you're using now then determining the inductance of it will help to choose a replacement.

Lundahl 1630/5mA model, made for push pull or single ended and is 7.2:1 ratio. It has the advantage of DCR balanced windings.

 

[alexc]

Isn't that lessened and essentially the point of using 4 valves per side, we know from CJ's interview that everything is carefully balanced to minimize all sources of distortion. I assumed your signal amp is intended to emulate the 670 signal amp; output using low-nickel core/21EI laminations/little or no unbalanced DC. From what you describe you're running higher current. Simply scaling up the core to the next size up with 35 or 36AWG wound primary should do the job, depending on inductance requirement. Or a bigger core again would be more than adequate...an example here by member vari-mu  http://www.groupdiy.com/index.php?topic=35664.0

Lundahl 1630/5mA model, made for push pull or single ended and is 7.2:1 ratio. It has the advantage of DCR balanced windings.

Yes the averaging out of differences  in each GR tube's current characteristics is a major advantage of using parallel stages, as is achieving sufficient current delivering capability.

Even so, and with careful tube matching, I do believe there will be significant mismatches under deep GR, particularly when rapidly shifting from hi to lo current. I don't really know how much - it's pretty difficult to measure dynamically, at least for me.

But I have no reason to expect that currents or traffo windings will all be finely balanced under those conditions.

Having said that, I am happy with the edcor I have except for this game breaking hum I have.

It may be that it is a result of the poor match for this app or it it could be my psu grounds or it could be just plain EM induced noise. At least I have narrowed it down to those 3 possibilities

Stripping the signal amp down now - simple cathode biased pp class A amp  into the 8K:600, terminated in 600R.
It's not very complicated! Should be able to get the damn thing to do as I expect at the one operating current!

Thanks for making some good points that I can think about further - I'm just reviewing my CJ notes on measuring inductances and hopefully I can get a measure of some more of the main traffo parameters soon.

----

btw - do you now what a real 660 or 670 signal amp idle current is? (I don't).

I think the the 6386 have some 240V of regulated B+, grid at -6.2V or so  and a cathode at some +12 to +14V giving a grid-cath difference of -18 to -20V ? That would indictate a current of 7mA per stage, 28mA per phase and 56mA of B+  ? Into the 60K:600 output (Sowter) or 20K:600 (HS52) terminated in 600ohms or so ?

It may well be I have my target idle current too high. I can rebias it easily enough.
I can also switch to +245V reg B+ if I want.

In any case, I have tried lower currents and yes, it works even at my lower B+, just gives me very low output, even with a low stepdown. Same noise, tho

So - I'm still quite a noob in this application!

Cheers

 

[alexc]

OK - looks like I'm making some progress.

Did the strip down for 22mA per phase at +135V cathode bias 5.9V and grid at 0V for 12.5dB of gain.
Mains harmonics  at around -70dBu for an SPL  of around -59dBu which is not crazy bad. Just very bad

Mostly the combination of 600R termination ensuring the reflected impedance is the spec'd relatively low 8K as well as biasing to lower the gain of the sig amp.

Seems that a secondary only terminated by the Motu (10K ?) reflects a really big Z which causes the noise pickup to shoot up. .  Running higher gain  exacerbates it.

Anyway, it's a start.

Also, I am recalibrating my RTA - I think I may have lost my settings in a computer 'reboot'.

So I did the recal and see pretty much the same as always :

-> some 3.8mVpp persistant 50/100/150Hz hash across the output

I can definately see it is not noise amplified from the source - it is completely independant of input signal attenuation.
Noise level is depending solely on the gain of the signal amp and the termination value of the output transformer.

I can't seperate it from ground noise or traffo induced noise.

It is significant that the ground noise I can measure is also 3.8Vpp and has been all along  :-[

The gain of the signal amp is roughly the same as the step down of traffo and so it could actually be the ground noise showing up at the same level at the secondary. The signal amp itself goes easily < 50Hz in response before rolling off.

So the odds are once again  50-50 for ground noise vs traffo induced noise  ;D


Next : I'm taking the output traffo out of the circuit and running the signal amp as the same simple cathode biased pp class A, but with plate resistors ac coupled to a resistive load similar to the output traffo.

ie. configured similar to a phase splitter in a hifi amp.

That will at least eliminate the possibilty of psu noise induced in the output traffo windings and leave with me with just the possibility of ground noise.

Also, I measured the freq response at -3dB @ 35Hz and 45KHz very flat with the exception of a 1-2 dB gradual peaking 20KHz - 40KHz

The input traffo has a lower freq spec of 40Hz. The output traffo goes down to 20Hz.

 

[DaveP]

Hi Alec,
I did the maths on the signal amp from the figures given and the correct voltages are on the schematic attached.
Incidently, the signal amp and the control amp are connected via a U pad (150/680/150) to give the correct termination/amplication as you have just discovered yourself.
best
DaveP

 

[DaveP]

I did this test (attached) to select the 6BA6's for balance.
I made a simple circuit with a 10k load and measured the current draw as the grid voltage went down.
As you can see, the 10 samples were all over the place.  This is basically a gm chart mA/V, but it shows how important it is to get the balance right throughout the range, it can be done if you have enough tubes to select from
best
DaveP

 

[alexc]

Thanks for both of those Dave. That confirms and corrects what I thought about the original.

I do hope one day to be able to worry about the match     I have 25 tubes to pick from so far.
I'll do something similar and pick my best ones.

Looks like I'm going to go for a Sowter signal amp output and get closer to what you describe.

Higher voltage, lower current, higher load. And the shielding won't hurt either. 

Just out of interest Dave, do you recall what idle current signal amp gain you were getting?

Thanx again

 

[MatthisD]

Yes the averaging out of differences  in each GR tube's current characteristics is a major advantage of using parallel stages, as is achieving sufficient current delivering capability.

Even so, and with careful tube matching, I do believe there will be significant mismatches under deep GR, particularly when rapidly shifting from hi to lo current. I don't really know how much - it's pretty difficult to measure dynamically, at least for me.

But I have no reason to expect that currents or traffo windings will all be finely balanced under those conditions.

Cheers

Hi,
In my posts I'm trying to point out that if you're wanting to use the 670 signal amp then the transformer's requirements are already known. Maybe at the design stage he selected a transformer ratio to match the number of tubes or instead maybe the number of tubes to match a chosen ratio but in the end the transformer is not a gapped transformer has no special allowance for unbalanced DC and is not particularly big. It has well balanced windings and equal coupling between primaries and secondaries. One thing that is sure is that minimum distortion was a priority.

Dave's voltages indicate winding DCR of 335ohms per half, HS-52 would have a bit higher DCR with 39AWG wire but much the same voltage drop across the windings.

I refer to the 6386 datasheet rather than 6k4p or 6ba6. The paralleled plate resistances with each plate at 5mA or less show that it is loaded very lightly in the 670. If i'm right with the loads then the twin secondaries have each 980R, in parallel with the loads connected to the output connection then calling that a minimum 200ohms x ratio 11:1 for each secondary, creating a primary load of 24.2K || 24.2k or 12.1K total and then 3K to each primary half. Total plate resistance must then be 10 times less for light loading.

If you compare to your circuit using 8K:600 ratio then it looks like; 2K to each primary half if loaded with 600ohms but with your configuration of higher current, lower plate voltage this would look to be a much different impedance match.

1K secondary load with that ratio 3.65:1 would equal 3.65k to each primary half which could be altered for optimum value with the signal amp in the 670 configuration.

Good luck with the build, thanks for sharing it. 

 

[alexc]

Thanks for the detailed info Matthis.

You are quite right - I am running a configuration which is not only different from the original, but also not very optimal in it's own right.

My purpose here was to loosely emulate the orig with cheap, easily obtainable parts whilst I learn the issues. So that's what's happening here. I am well aware that I'm not running a transformer anything like the original. Or voltages or currents. Or input traffo, or control traffos or anything really!

So now I'll move to the next level and get the higher spec'd traffo with the higher impedance and well balanced windings so as to minimise the unbalanced dc, along with careful matching.

And the higher voltages and lower currents.

In the meantime, I'm just doing some learning here. The 8K primary that I have, when loaded with 600R, I expect is giving each phase 4K, and each stage 1K (in Class A) which is too low for a triode connected 64KP, especially using the rule of thumb of loading.

But I'm guessing I have that wrong - you indicate each phase is 2K. Which would make it worse.
I was going to try to load the secondary higher, but that seems to be causing more probs in other areas right now.

I just like to see what can be done with what I have. So I am learning my lessons. I expected to find problems of distortion and what not a bit later in the piece, but I'm being sidelined by other probs.

So -back to the drawing board while I await some new part from Sowter. I don't know of any other available that is more suitable for the orig type application. I'm no traffo winder! - hence my starting point with this cheap one which I thought I could get something useful from, albeit NOT like the original setup.

Anyway - mea cupla! I'll try something else now.

In the meantime, I'll rejig my regulated psu for a higher voltage and set up my tube matching rig.

Thanks all for the assists - I'll post again when I have a different config or something more interesting.

Regards
Alex

 

[alexc]

I did the remove of the output transformer, setting up the signal amp as a 'phase inverter' with plate loads.

Surely there is ground noise - I see it at the grid, and it is amplified and appears at the plates. Basically amplified by the signal amp gain less a bit of rolloff.

All those who said ground resistance and/or loops - CORRECT!
So I have a way forward - find the source of the ground voltage and fix it!

Then I can revisit the circuit with transformer.

Cheers