[Maxx134]

No use looking on the first page for these, they aren't mentioned!

I didn't know this...
Also I believe in both this thread and the other mk6 thread there were owners with slightly different values...
I don't believe exact value is critical.

Just read or measure them by disconnecting one side from board to test...
The reason I don't say remove totally before test, is because there is a strong chance these resistors will break once removed.

Anyways they should still be visible to read their values,
AND the PSU schematic on first page should have the values,
Which were taken from an MK8.

 

[baronbeehive]

I didn't know this...
Also I believe in both this thread and the other mk6 thread there were owners with slightly different values...
I don't believe exact value is critical.

.....

I just reread the first page to check if I was mistaken, but it's not there. I remember it coming up way back and I can't remember why it was mentioned except to say that it was decided to replace the existing resistor with 2 in parallel to double the wattage. I believe there was some uncertainty about the values on our boards, due to someone having a different one or something.

 

[SonicTrance]

Hi guys!

I had some time today so I took some measurements of the MK6. Thought I'd post them here for you guys to see!

Amp in question:

I started by measuring all important voltages of course (I haven't used this amp in quite some time). All voltages were good!

As some of you know I've built @MrCurwen LTP design which is a forced balanced design by the CCS (constant current source) tail. If I inject a signal in one phase of one channel in the LTP and ground the other phase, both phases output the exact same amplitude!

So, I thought it'd be fun to do just that with the mk6. I had to leave one phase input floating or disconnected for this to work. Could barely get a reading with one phase grounded.

Input signal is 700 mVpp @ 1kHz

A 100 ohm resistor as load across the right channel outputs. Scope and DMM probes hooked up across it.

This is what came out at full volume. 2.28 Vpp from a 700 mVpp input signal gives me a total gain of 3.26. Remember, the ECC35 has µ68 and 6AS7G has µ2, so full gain would be 136. That's the NFB. Oh yes, I have the gain set to low, so maximum amount of NFB!

Here's the measurement from the phase without input signal (same channel). I just flipped the scope leads on the load R.

As you can see there's barely any signal there at all.

Also thought it'd be fun to do a power output measurement. Remember, gain still set to LOW!

This is right before the sine wave starts to round off at the bottom.

7.1 Vrms ^2 / 100R (load) = 0.5 Wrms or.....
10 Vp ^2 / 100R (load) = 1 Wpeak

I don't know how much the low gain setting affected power output but for fun I also simmed the MK6 in spice:

Also with low gain setting. If I drove it any harder than this the signal would clip. This is about 15 Vpp with 100 ohm load. I don't know how accurate spice is when doing these types of measurements but use it as entertainment at least.

Both phases have input signal here as you can see. Nice and balanced outputs. The voltages at V4 and V5 are adjusted to get 0 VDC at output. Don't know how accurate those are either.
Edit: I got curious and measured the voltage at the top triodes grids. It was in fact very accurate at -23 V!

I did not post this to bash on the MK6 in any way. I still think it's a great sounding amp!

 

[MrCurwen]

Excellent work Sonic! Really interesting to see.

This kind of amp would be quite perfect for driving a pair of high impedance horn speakers. Plenty of output power for any kind of speakers really, but the output impedance is too high for normal speakers.

Regarding balance;

Exactly as theory would predict. The biggest impedance in the series (between B+ and ground or B-) determines the balance in any LTP (which the LD input stage has, there's two triodes and a common tail in series with both triodes), so which is it? It's not the 1.5k tail, it's the tubes.

So the tubes determine the balance. This is why tube rolling works, and also why it's not a good solution.

This is what came out at full volume. 2.28 Vpp from a 700 mVpp input signal gives me a total gain of 3.26. Remember, the ECC35 has µ68 and 6AS7G has µ2, so full gain would be 136

I think the WCF has a negative gain with any real load, something like 0.8 if I remember correctly from way back when I simmed this circuit a lot. Doesn't change the point though, gain is suppressed to a significant degree (within tube amps, SS amps have a lot more NFB).


By the way, I recommend running sims for at least 100mS preferably 200mS if you wish to have any degree of accuracy in FFT or THD results. The longer the sim runs the better the resolution becomes. But, as you said, these are for entertainment purposes only. You can make a lot of 'perfect' sims that come out less than perfect or even unuseable (Spice is suspiciously stable from a builders perspective). Only way to become a decent designer is to build build build, sim afterwards for entertainment and food for thought.

If you sim that very LTP balance test with my circuit, you will not get nearly as good a result as you will in real life measurements. Sometimes sims give worse than real results, sometimes better than real results. Oftentimes the fault lies in tube models; they are very much simplified. In reality a tube doesn't have a set of curves that are set in stone and static. Spice tube curves are static.

 

[SonicTrance]

By the way, I recommend running sims for at least 100mS preferably 200mS if you wish to have any degree of accuracy in FFT or THD results. The longer the sim runs the better the resolution becomes.

Thanks!

Edit: Miss information. Correct info in later post.

 

[SonicTrance]

If any of you guys need me to take any measurements of the MK6 now is the time! Just let me know and I'll do it!

 

[MrCurwen]

Did you use the .four statement that is visible in the schem you posted? If so, it's invalid. It's "SE output" THD, not the actual balanced signal. I'd suggest changing node 10 to I(RL).

 

[SonicTrance]

Did you use the .four statement that is visible in the schem you posted? If so, it's invalid. It's "SE output" THD, not the actual balanced signal. I'd suggest changing node 10 to I(RL).

You're right! I didn't think about that.

With 1.6 V signal (near clipping) THD = 0.5%, 15 Vpp
With 0.5 V signal, THD = 0.045%, 4.85 Vpp

With gain set to high (low NFB) 0.5V signal, THD = 0.40%, 10 Vpp

So, THD increases with lower NFB and gain increases. To be expected.

 

[Maxx134]

If I inject a signal in one phase of one channel in the LTP and ground the other phase, both phases output the exact same amplitude!

So, I thought it'd be fun to do just that with the mk6. I had to leave one phase input floating or disconnected for this to work. Could barely get a reading with one phase grounded.

I believe the single ended output jack is also giving this signal.

I did not post this to bash on the MK6 in any way. I still think it's a great sounding amp!

Thanks for info and it would ne interesting to see how iy looks at high gain setting which uses leads NFB which I use.
I believe this setting you can offset gain with a lower MU tube choice,

But chose not to.
I balanced it another way by different choice of output tube.

there's two triodes and a common tail in series with both triodes), so which is it? It's not the 1.5k tail, it's the tubes.

So the tubes determine the balance. This is why tube rolling works, and also why it's not a good solution.

I agree and also want to still keep optimising this amp, so I have planned to place CCS at the driver tube cathode which is probably the last things I could do,
Except for putting a lot in the place of the gain switch, to see how low NFB I can go without messing up the sound .

If any of you guys need me to take any measurements of the MK6 now is the time! Just let me know and I'll do it!

That would be great not only to fill in the schematic with measurements, but to analyze and learn both the flaws and compromises in these designs.
I am interested in how MrCurwen also thinks about designs.
It is great to get you thinking.

So, THD increases with lower NFB and gain increases. To be expected

If possible take in high gain as it would be interesting to see what type of distortion increases with less NFB.
Probably dependant on type of tube and makes this amp "tube rolling" sensitive...

 

[SonicTrance]

I believe the single ended output jack is also giving this signal.

I don't have any SE jacks in my MK6.

I agree and also want to still keep optimising this amp, so I have planned to place CCS at the driver tube cathode which is probably the last things I could do

Yes, might try this myself just for fun. I'm wondering how much it would lower the B+ though. I definitely think that we should go for 6SL7/ECC35 or 6C8G which doesn't require much current to be linear. About 2mA. Otherwise we wouldn't have much B+ left! We got plenty of B- for FET voltage headroom, so should not be a problem there. Not sure how the CCS would interact with retro style resistor plate load though.

If possible take in high gain as it would be interesting to see what type of distortion increases with less NFB.

What measurements would you like to see with high gain? Same as above or something in particular?

 

[coinmaster]

Not sure how the CCS would interact with retro style resistor plate load though.

As long as the cathodes are current source loaded it won't matter much, the cathode feedback is what matters if you want amplitude matching. The stock plate resistors can only handle less than 1ma before the b+ gets too low, so that needs to be taken into account.

 

[SonicTrance]

The stock plate resistors can only handle less than 1ma before the b+ gets too low, so that needs to be taken into account.

If you pull 1 mA through 220k you drop 220 V. That's more than the available voltage.

We need to use something like 33k as I'm using now. From previous experiments I've found that if I pull 3.5 mA / tube, so 7 mA total from the driver stage psu I get a B+ of 160 V. If we draw more current than that we loose to much voltage swing.

0.00175A * 33k = 58 V drop across Ra

Operating point would be:
101.4 Va
-0.6 Vg
1.75 mA

 

[coinmaster]

If you pull 1 mA through 220k you drop 220 V. That's more than the available voltage.

Indeed, that's why I said "less than", I think the tubes run at like 700ua stock or something. It wouldn't be overly difficult to just use a gyrator. Bartola sells gyrator pcbs that can be slotted right into the chassis. A p2p gyrator should be doable without too much difficulty as well. Just gotta be careful how much current you draw, the RC filter might need to be adjusted and I don't remember the current rating on the winding.

 

[SonicTrance]

Here's a schem using a CCS tail and ECC35's with the above operating point. It sims very well. Ground one input and both phases output same amplitude.

 

[coinmaster]

A simple low voltage jfet current source would be simpler and more practical than a scheme like that for a current source implementation in the mk6. I'm not sure how good of an idea it is to attach it to B- either since it won't be clean and will hurt the impedance of the current source to some extent and introduce some measure of crosstalk.

It would probably be better to simply attach it to ground, as long as your tubes aren't wildly unmatched the voltage at the cathodes should have enough swing to not need a b- connection if you run them balanced. If the tubes are perfectly matched there will only be pure DC at the cathodes during balanced operation, so 1.5v+ grid bias should be plenty of theoretical headroom for voltage swing with mismatched tubes, 0.6 is probably okay too but I wouldn't want to place a bet on that especially if you account for the headroom required for the jfet if you're grounding it. I'm not sure how low the LV jfets can go.
1.5v would need a gyrator to work as well. Something for you to think about.