Please critique my plans for a WE91 clone.

[kevinkr]

Here’s an update on the frequency response of this amp.

Tested at 1VRMS at the 8R load…. 1 watt of power at the speaker.

20Hz -1.3 DB
20KHz -1.95DB

I’ve changed the circuit some what since my last post. I’ll post the current schematic later.
As of today, the ultra-path cap is out, and the 2A3 cathode bypass cap=48uf
(With the 2A3 cathode bypass cap at 18uf, the freq response at 20Hz was -3.6DB)

If I replace the EF37A with a 5842 loaded with a 20mA CCS, the frequency response at 20KHZ is -.7DB.

Hi Sqerus,
Quick question:

Was it 1Vrms or 1W into an 8 ohm load? 1Vrms is actually 125mW, for 1W you would need 2.83Vrms..

-1.95dB is still quite a lot of roll off at 20kHz, did you find the sound preferable with the 5842 or not?

Consider using the 48uF cathode bypass with a ~10uF film cap in the "ultra-path" location for some additional hum canceling.. This can improve low frequency output noise up to 20dB or so, and in addition the added capacitance will flatten out the LF response, and if very high quality will also open up the highs.

Make sure the 2.5V filament supplies are balanced around ground - i.e. use the filament transformer center tap if present for the bias network or create with 22 ohm resistors in series across the filament supply - connect the cathode bias resistor, cathode bypass cap and "ultrapath" cap to the center connection between them.

 

[sgerus]

Thanks for all the replies!

KM: Please see the new schematic, I’m not using the 0c3. Interesting method for the cathode bypass… thanks for that schematic.
My hum is less than 1mV… I thought that’s pretty low for a 6.5 watt amp with AC heaters.

Andrew: It’s a regular 5842, 175V@plate, -2V, 20mA CCS
The opt’s are Electro-Print

KevinR: Right now I can’t say if the 5842 sounds better… I’ll try to do some blind listing test’s and let you know. I’ve been avoiding the 9 pin tubes just because I don’t think they look as good as an octal based tube (great technical reason…)

OK, I thought I understood the power calculation, but maybe I don’t. Here is how I came up with 1Vrms=1 watt

1V rms (at the 8R speaker load) = 2.82 V pk-pk

2.82V x 2.82V / 8R = 1 watt

 

[kevinkr]

Thanks for all the replies!

<snip>

OK, I thought I understood the power calculation, but maybe I don’t. Here is how I came up with 1Vrms=1 watt

1V rms (at the 8R speaker load) = 2.82 V pk-pk

2.82V x 2.82V / 8R = 1 watt

1Wrms requires 2.83Vrms, not Vpk-pk, that is where your confusion lies. You actually ran the test at 125mWrms based on your description.

 

[sgerus]

Thanks Kevin... I'll redo my frequency tests.

So let me tell you how I did the frequency test so I don’t screw that up as well!

1. Set the sine wave from the single generator to 1KHz
2. Set the volume for 2.83Vrms at the 8R speaker load
3. Set the scope where I can see 1 or 2 cycles, and record the pk-pk voltage
4. Measure the new pk-pk at the given frequency

Say I start out at 5.5 (units on the scope) at 1KHz and have 4.7 units at 20KHZ

The frequency response is 20 * log(4.7/5.5) = -1.36DB

Do I have this procedure correct?

 

[kevinkr]

Thanks Kevin... I'll redo my frequency tests.

So let me tell you how I did the frequency test so I don’t screw that up as well!

1. Set the sine wave from the single generator to 1KHz
2. Set the volume for 2.83Vrms at the 8R speaker load
3. Set the scope where I can see 1 or 2 cycles, and record the pk-pk voltage
4. Measure the new pk-pk at the given frequency

Say I start out at 5.5 (units on the scope) at 1KHz and have 4.7 units at 20KHZ

The frequency response is 20 * log(4.7/5.5) = -1.36DB

Do I have this procedure correct?

Sounds fine to me, and is probably better than trying to interpret what the actual voltage is.

Eventually one of the true rms Fluke multimeters with dB relative mode would be a worthwhile acquisition. The better ones are flat out to well beyond 20kHz. Then you can do a direct analysis.

The other way presupposing you already have a good meter flat to beyond 20kHz is to measure the rms value at the two frequencies and take the ratio of those as you have done with your scope.

 

[kmaier]

Scott,

Give the split balance a try... it might help. You can also use a second balance pot with the cathode bias resistor and have independently adjustable DC and AC balance. I've done that as well... gives you the ability to tweak the tube to it's lowest hum level. Do note that all tubes balance differently, i.e., you can have a near perfect match on gm, current, etc. and one will balance down very low and the other won't... it's just the simple fact that the coating on the filament is never completely perfect combined with the internal alignment, grid wire tolerance, plate purity, etc.

For measurements.... I would highly recommend finding an older HP distortion analyzer. I use a 334A... excellent piece of equipment and you can trust them. The have a range of 5Hz - 600KHz and make frequency response, output noise and distortion measurements nice, simple and reliable. I also use a Fluke true RMS 8060A, which is also vintage these days. I've seen clean 334A units on ebay for under $200 these days, so certainly a good deal. In any case, I tend to prefer vintage measurement/test equipment for designing vintage type equipment ;-)

Regards, KM

 

[sgerus]

Here is another update on the frequency response of this amp, this time tested with 2.9Vac (RMS) at the 8R test load at 1000Hz.

EF37A Pentode
20Hz= -.87Db 20KHz= -1.34Db

EF37A Triode, 33K plate R
20Hz= -2.00Db 20KHz= -.41Db

EF37A Triode, 25K plate R
20Hz= -1.87Db 20KHz= -.20Db

EF37A Triode 5.5mA CCS
20Hz= -1.87Db 20KHz= -.20Db

5842, 22mA CCS
20Hz= -2.00Db 20KHz= -.20Db

Right now I can’t test the triode at full power (I’m using a PC for the single gen). As that as soon as I buy a real single generator I’ll test at full power and then decide if converting this to triode mode is the way to go.

With the ERF37A wired as a pentode, the input sensitivity is about .6Vrms, with the same tube wired as a triode, the input sensitivity is about 2.1Vrms, which is fine for me because I run this amp off a preamp.

 

[sgerus]

Regarding the hum level... 1mv is pretty high, especially for a few watts output. I prefer AC filaments when possible and have implemented a split balance technique for all of my 45 and 2A3 designs. I prefer to use a center-tapped filament winding for the DC path to ground, ie, the bias resistor. This provides a very low impedance thru the winding to ground. For AC balance, I use a padded 100 ohm pot (use a pair of 12 ohm resistors) which yields a 20 ohm balance pot without the knife-edge adjustment. From the wiper, just use the bypass cap by itself to ground. This splits the return path, DC is fixed at center, AC is adjustable.

KM,

This worked like a charm…. Hum went from .5mV down to .1mV

I’m using 10 turn 100R pot’s without the extra resistors. The cathode bypass resistor is grounded by the B+ power supply cap, and the cathode bypass capacitor is grounded right on the input/driver ground location.

So now, I wonder, 300B with AC heaters using this method????

 

[kmaier]

Glad it worked out well for you. I would also put the 12 ohm pad resistors on... not only do they expand the adjustment band, they also significantly lower the series resistance of the AC path to ground. With a 100-ohm pot, you effectively have a 25 ohm resistor in series with your bypass cap. Adding a pair of 12 ohm resistors will drop that to 5 ohms and will probably drop your output hum level even more.

As for the 300B... sadly not... these use a center-tapped filament so you have two filaments running in parallel with the center being common and the far ends tied together. This compounded with a 5-volt supply makes it ugly. I recently purchased a new manufacture pair of Western Electric 300Bs... no way to get the hum to anything near acceptable with an AC supply. I have also found the 2A3 single plate triode to also be a center-tapped filament design and the same applies. I wish it were different.... but as a result, I have no choice but to use electrolytic caps for a DC supply.

Regards, KM

PS - I have quite a few 45 pairs that can achieve below 130 microvolts hum level using the split balance approach... maybe my only good idea this century

 

[kevinkr]

<snip>

As for the 300B... sadly not... these use a center-tapped filament so you have two filaments running in parallel with the center being common and the far ends tied together. This compounded with a 5-volt supply makes it ugly. I recently purchased a new manufacture pair of Western Electric 300Bs... no way to get the hum to anything near acceptable with an AC supply. I have also found the 2A3 single plate triode to also be a center-tapped filament design and the same applies. I wish it were different.... but as a result, I have no choice but to use electrolytic caps for a DC supply.

Regards, KM

PS - I have quite a few 45 pairs that can achieve below 130 microvolts hum level using the split balance approach... maybe my only good idea this century

Unfortunately I can confirm the 300B comments from experience, although I have found most 2A3 and all 45 acceptable on ac heating without any special precautions.

I generally use fixed bias so the above clever technique doesn't work for me, but generally I have been able to get hum well below 1mV with 45 and just a little worse with 2A3.. - although the Sovtek sounds and measures a bit noisier than others I have tried.) It hasn't really been a problem for me with speakers of under 103dBSpl efficiency..

CCS based heating with the 300B seems to work quite well and sounds good. I've also had good luck with chokes (both CM and not) with voltage regulation, and not so good with a CCS and chokes.

 

[kmaier]

Hi Kevin,

I think I recall you saying you had a single-plate 2A3... true? In addition to the inability to get acceptable hum balance on the single-plate 2A3, I have also found them to be exceptionally sensitive to mechanical vibration compared to the later dual-section types. Have you noticed this?

I found the same on the EML 45 (massive envelope and also an atypical filament arrangement vs NOS 45) for mechanical sensitivity. In general they can balance pretty quiet but need higher cathode current to sound good.

Regards, KM

 

[kevinkr]

Hi KM,
My single plates were just lowly Sovtek types and they did have just a bit more hum than some of the dual plate types I had. (Kenrad) I have just one single plate RCA and no longer have my 2A3 amplifier to try it in unfortunately..

The Sovtek was not very microphonic, but the RCA one looks like it might well be.

None could be as bad as a 26 for microphony..

 

[kmaier]

Hi Kevin,

Thanks for the quick reply. Despite having a pair of filament tension springs at the top of the assembly there's not enough tension to minimize microphonics, but then
again, that wasn't the design goal... oh well... DC it is for these too.

Regards, KM

 

[rcavictim]

Hi Kevin,

Thanks for the quick reply. Despite having a pair of filament tension springs at the top of the assembly there's not enough tension to minimize microphonics, but then
again, that wasn't the design goal... oh well... DC it is for these too.

Regards, KM

I might have missed something above but a DC filament supply does not reduce microphonics in a tube that has this problem. It can however make a microphonic tube with hum quieter and a better microphone from a s/n point of view.

 

[kmaier]

Hi ...victim,

Sorry if my last post was misleading... you're correct, there is no connection between hum and microphonics... just a comment that in addition to the center-tapped filament arrangement being difficult to balance the hum with an AC filament, that the tensioned filaments are still affected by mechanical vibration. The tension springs are only to prevent sag as they (filament wires) expand from being lit.

Needless to say, a DC supply has no effect on this (as it's mechanical) but the result is different as the bias at any given point on the filament wire is fixed versus a floating 60Hz (or whatever your line frequency is) along the length of the filament wire.

Regards, KM

 

[sgerus]

Here’s an update on my WE91 Clone.

For the past 2 months I had the EF37A’s wired in triode mode. That worked out better for my needs, in terms of input sensitivity and better frequency response.

In the last couple of weeks I got the urge to re-try something using 6sn7’s
As of 8/24/08
It’s no longer a WE 91 clone, the EF37A pentode has been retired (for Now) and replaced with a pair of 6sn7’s…..

The first 6sn7 is set up as 5.5ma SRPP direct coupled into the 2nd 6sn7, a 6ma white cathode follower, the CF is then C-L coupled to the 2A3-40.

I found the 6sn7 SRPP by itself was about equal to the EF37A (triode mode).
The 6sn7 had slightly better frequency response, the EF37A (triode) had slightly lower input sensitivity.

Adding the white cathode follower is that really made the difference. With the CF stage, the sound seems to be more detailed had has more depth…. You can hear the difference.

I run the amp from a pre-amp, and do lot of listing in the evening at low volume.
So for MY needs this seems to be the best circuit I have tried.

I have a question about the White CF stage. How important is the cap that that connects the upper plate to the lower grid? Ie should this be a high quality “Audio” capacitor.
(right now I’m using an orange drop)

 

[kmaier]

Looking at the schematic, the 0.1uF is tied to the plate which is a power supply rail... hence it is effectively an AC ground. This makes the 0.1uF cap a bypass cap. You could probably tie it to ground instead and avoid any supply ripple being fed into the grid. As for quality... use the same type of cap you would use for a screen bypass on a pentode stage... I'd say an orange drop is fine.

On a different note... did you pad the 100 ohm balance pot with a pair of 12 ohm resistors? This should drop your output noise a bit more and improve the lower frequency response a bit.

Regards, KM

 

[kevinkr]

Looking at the schematic, the 0.1uF is tied to the plate which is a power supply rail... hence it is effectively an AC ground. This makes the 0.1uF cap a bypass cap. You could probably tie it to ground instead and avoid any supply ripple being fed into the grid. As for quality... use the same type of cap you would use for a screen bypass on a pentode stage... I'd say an orange drop is fine.

On a different note... did you pad the 100 ohm balance pot with a pair of 12 ohm resistors? This should drop your output noise a bit more and improve the lower frequency response a bit.

Regards, KM

No White CF here.. Add a small resistor in series with the plate of the follower section in order to implement a White CF.. You need just enough signal to assure symmetrical drive from the both sections. You can use a pot to find the optimum drive level - as a rough guide the overall gain should not be significantly higher than it is now, if you reach unity gain you have gone too far. (It will probably be much less than a couple of hundred ohms.)

I would expect you will see further gains in that sink/source capability of the driver will be more symmetrical.

 

[kmaier]

Ah, good catch! Wasn't looking to verify the actual circuit... here's a link that might be useful however:

http://www.tubecad.com/october99/page4.html

Regards, KM

 

[sgerus]

KM, thanks for the link to the tubcad site. I will print and read that article tonight.

KevinR, I will test with a resistor on the CF plate…. According to the tubcad article, around 100R seems to be the magic number.

BTW, If it’s not a White CF (without the plate resistor), what would you call the CF as shown?

Getting back to the cap used from the CF upper plate to the lower grid, should that be a high quality "Audio" cap?