Optimum Voltages

[mcandmar]

I am trying to work out what effect different voltage would have on the operation of the tubes, and how to interpret the graphs in the datasheets. For example if you were to power the quickie from 36v to 72v or 100v what effect would each have on the power output, linearity, or distortion. I guess what i'm really asking is, are there any idiots guides to reading datasheets and working out optimum plate voltages and bias levels?

What i am really interested to work out is there any advantage to higher voltage levels when driving output transformers to use the quickie as a headphone amp.  I want to build an AC supply but need to figure out what voltages i should use.

Thanks,

Mark

 

[Paul Joppa]

As an approximation, the available power increases rapidly with plate voltage. To give you an idea, here are three examples:

1) The stock quickie can generate about 1mW into a 4000 ohm load.

2) Using a choke (or series feed transformer) plate load into an 8K impedance, and re-optimizing the operating point, you can get about 4mW.

3) Running a 72v supply with a choke plate load and parafeed into a 5600 ohm output transformer, you might get as much as 25mW output.

These are three design point that I have worked out and can provide some details. A general rule for the engineering design of an amplifier is beyond the scope of a post or an article!

Notice that all the above degigns require an output transformer to match the optimum tube load impedance (4000 to 8000 ohms as above) to the headphone impedance (usually between 30 and 600 ohms). I am not aware of any ideal transformer to use for this application. Here are three less-than-ideal choices which have appeared on this forum in the past:

A) the "original" SPECO T7010. This transformer has not been available for a few years. It was used in earlier versions of the SEX amp and the Paramour, and those customers who upgraded the transformers had these left over; some of them used it with their Quickie. It can be wired for phones of 16-32 ohms and 300-600 ohms, but there is nothing in between. The bass capability is questionable.

B) the "new" SPECO T7010. We learned that the SPECO had been redesigned when an order came in with little tiny transformers instead of the larger ones we had been using. The new one was very different, which led to our use of our own design transformers for the amplifiers. This new one can be configured for use with the Quickie and drive any headphone, using four impedance taps. The bass is still questionable at least until someone tries it.

C) Our own OT-2 as used in SEX, Stereomour, and the Mainline headphone amp; only sold with the matching PC-3 plate chokes. Because of Quickie's low power, it is really only suitable for low impedance phones - less than 50 ohms. Very high sensitivity phones of higher impedance might work acceptably, but you'd need over 100dB/mW sensitivity to get even average levels. (Mainline has much more power and can drive higher impedance phones from its relatively low impedance output.)

 

[2wo]

Try this one

http://www.tubesandmore.com/products/B-916

 

[mcandmar]

Paul,

I have a pair of the new style Speco T7010 transformers (cute little things) that i have been experimenting with. They work better than i was expecting but there is a compromise depending on the taps used. It varies from decent sounding but too quite, to decent volume level with too much low end roll off.  Hence i'm wondering if/how i can get more output power from the quickie to drive them to the level i am looking for.  Its real close, but just needs that little bit more oomph.

Building an AC supply isnt a problem if i have a target voltage to shoot for, i just dont know how to work out what that target should be.

@2wo, is that book written in laymans terms or do you have to be an electrical engineer to decipher it?

Cheers,

Mark

 

[charger]

could one of this be suitable to feed the Q ? 

http://www.trcelectronics.com/View/Mean-Well/EPS-15-36.shtml

 

[Paul Joppa]

I think I posted this before, and it may be what mcandmar is using, but for the interest of others who may not have seen it.

 

[mcandmar]

Yep you posted that in the headphone thread here http://www.bottlehead.com/smf/index.php/topic,2.0.html

From memory using the red tap was ok volume wise, but it was the brown tap that sounded best. That was with 32ohm headphones.

From your above post, if the stock quickie outputs 1mv the 2nd option for 4mv would be more than enough power for my needs, i would say even doubling to 2mv would be enough.  I assume its not as simple as just replacing the 4k resistor with an 8k?

 

[Paul Joppa]

Replace the 4K with a plate choke - the hammond 150H/8mA that is widely used is a great candidate. Then double the cathode resistor to 2K, which reduces the cathode current and optimizes for the higher load impedance. Best output capacitance plate to OPT is then 5uF but 2.5 to 10 is a good range.

 

[mcandmar]

Replace the 4K with a plate choke - the hammond 150H/8mA that is widely used is a great candidate. Then double the cathode resistor to 2K, which reduces the cathode current and optimizes for the higher load impedance. Best output capacitance plate to OPT is then 5uF but 2.5 to 10 is a good range.

Just to confirm, are you talking about the 156C at the bottom of the list?  http://www.hammondmfg.com/153.htm

I assume the PJCCS is redundant and gets discarded when using a plate choke with an AC supply power source?

Also do i aim for stock ~38ish volts with the above or can i go higher?  I have a small toroid here with dual 18v taps but by the time i wire them in series, rectify, and smooth i'm up to ~58v DC. I also have a single tap 24v laminated core which should get me just under 40v. I can of course buy a transformer if needed, would just be nice to use what i've got.

Thanks again,

Mark

 

[Paul Joppa]

You asked for the 4mW design, which happens to be what you can do with the stock 36v battery supply. So that's what I posted. It incorporates the 156C.

As long as you keep the cathode resistor bias, you can vary the voltage with pretty good success as long as it doesn't get too far from the design point. No, I can't be more precise without doing a design study.

 

[mcandmar]

Thanks Paul, i will experiment with a few configurations and see how it performs.

 

[mcandmar]

A) the "original" SPECO T7010. This transformer has not been available for a few years. It was used in earlier versions of the SEX amp and the Paramour, and those customers who upgraded the transformers had these left over; some of them used it with their Quickie. It can be wired for phones of 16-32 ohms and 300-600 ohms, but there is nothing in between. The bass capability is questionable.

B) the "new" SPECO T7010. We learned that the SPECO had been redesigned when an order came in with little tiny transformers instead of the larger ones we had been using. The new one was very different, which led to our use of our own design transformers for the amplifiers. This new one can be configured for use with the Quickie and drive any headphone, using four impedance taps. The bass is still questionable at least until someone tries it.

I managed to pick up a set of the original Specos on ebay so i have one of each hooked up to compare them. Using the 10w outputs with 32ohm headphones the difference is very subtle, it sounds like the newer ones have slightly less low end roll off, however i think the older ones sound a little clearer in the mid/high end.  I ended up resorting to test tones and switching between cups on one ear to really pin point any difference.

I am thinking of using the new versions for that little bit more low end but was wondering if you have any insight as to which is technically more suitable?

Thanks,

Mark

 

[Paul Joppa]

I'd go with the new one for sure. Here's the reasoning:

The 10-W tap on the old one is optimum for 500 ohm phones, in terms of maximum power. That's important when the maximum Quickie power is so low. However, using the 10-W tap does increase the bass capability. This is probably why it sounds as good as it does using 32 ohm phones. But giving up 90% of the available power to get a bit more bass is not a great compromise, in my opinion. That ability to optimize for phones between 32 and 300 ohms makes the deal, technically.

 

[mcandmar]

As an approximation, the available power increases rapidly with plate voltage. To give you an idea, here are three examples:

1) The stock quickie can generate about 1mW into a 4000 ohm load.

2) Using a choke (or series feed transformer) plate load into an 8K impedance, and re-optimizing the operating point, you can get about 4mW.

3) Running a 72v supply with a choke plate load and parafeed into a 5600 ohm output transformer, you might get as much as 25mW output.

These are three design point that I have worked out and can provide some details. A general rule for the engineering design of an amplifier is beyond the scope of a post or an article!

Notice that all the above degigns require an output transformer to match the optimum tube load impedance (4000 to 8000 ohms as above) to the headphone impedance (usually between 30 and 600 ohms). I am not aware of any ideal transformer to use for this application. Here are three less-than-ideal choices which have appeared on this forum in the past:

A) the "original" SPECO T7010. This transformer has not been available for a few years. It was used in earlier versions of the SEX amp and the Paramour, and those customers who upgraded the transformers had these left over; some of them used it with their Quickie. It can be wired for phones of 16-32 ohms and 300-600 ohms, but there is nothing in between. The bass capability is questionable.

B) the "new" SPECO T7010. We learned that the SPECO had been redesigned when an order came in with little tiny transformers instead of the larger ones we had been using. The new one was very different, which led to our use of our own design transformers for the amplifiers. This new one can be configured for use with the Quickie and drive any headphone, using four impedance taps. The bass is still questionable at least until someone tries it.

C) Our own OT-2 as used in SEX, Stereomour, and the Mainline headphone amp; only sold with the matching PC-3 plate chokes. Because of Quickie's low power, it is really only suitable for low impedance phones - less than 50 ohms. Very high sensitivity phones of higher impedance might work acceptably, but you'd need over 100dB/mW sensitivity to get even average levels. (Mainline has much more power and can drive higher impedance phones from its relatively low impedance output.)

Hi Paul, i have been revisiting this in an effort to squeeze a bit more power out of the amp and would love to hear more about your option 3?    Just to recap the current configuration is a plate load of 150H DCR 3.7k, 5.1uf cap (~7.5hz cutoff?), into the new style Specos .25w tap (20k i think?), bias resistor of 2k with ~4volts across it with a 45v plate, so i calculate a little over 2ma plate current.  Output is connected to the 10w Brown tap.

I found if i reconfigure the B+ supply into a voltage doubler it gives me ~110v which just happens to be the ignition voltage of a 7pin 0C2 gas regulator, which in turn gives me 75v out.  With the B+ supply at 75v i tried a few different bias points from 2ma to 5ma and i am not seeing any significant changes of voltage or power output at the headphone output. I picked up maybe .5v into 300ohm, and 5mw into 32 ohms. Am i right in saying to get any appreciable voltage gain i would need a lower plate load to increase the voltage swing of the plate? or abandon the chokes and switch to a CCS instead?

Reading your last post again it has also occurred to me that i am running the output transformer as an autoformer in the current setup so i will reevaluate that setup too.

Also your comment above about the 72v operating point running into a 5600ohm transformer has me thinking about the 1w tap on the Speco as it should be around 5k, but i am well and truly lost when it comes to output transformers. Either way i would love to hear your thoughts as even if there is little to be gained its all good fun and educational at the end of the day

Cheers,

Mark

 

[Paul Joppa]

Gain will not change with bias current, only the maximum output before clipping distortion will change. To get more voltage out, put more voltage in. Changing to a current source will not have any effect on gain compared to a choke load, but will reduce the maximum voltage output. Changing the plate load impedance will have little effect on gain, it just unbalanced the clipping limit on the positive or negative side. That balanced clipping is essentially the definition of an "optimum load impedance".

With a 72v plate supply, use a 1400 ohm cathode resistor. You should see about 4.25mA current, 6v across the resistor. That is optimum with a 5600 ohm primary load, which is 70 ohms on the brown 10W tap, 140 ohms at the red 5W tap, 280 ohms at the orange 2.5W tap. The higher impedance taps will have more voltage output. I think it's reasonable to use headphones that are between half and twice the calculated load for that tap.

While the blue primary tap is labelled 0.25 watts, which calculates to 20,000 ohms with a 70.7v line, I suggest using this tap as the 8000 ohm (36v power) or 5600 ohm (72v power) point so that you use all the available inductance and get the best bass possible. The load impedances are scaled by the same factor (5600/20000=0.28 times the "normal" impedance). This will make the leakage inductance more important, possibly reducing the treble a bit.

This is an off-spec use of the transformer, intended to give the best balance between bass and treble results. If you discover you have missing treble and tons of bass, you can experiment with the 0.5W and 1.0W taps as the primary, but remember to switch the output tap as well, to maintain optimum primary impedance. Normally one should avoid off-spec uses, so I am violating my own rule there - I think I know what compromises had to be made to make that transformer so small and cheap, and I think I am compensating for them. But as said, I have not spent a week inthe lab with this little puppy, so I've just posted my best guesses.

Yes, the transformer is configured as an autoformer. This helps reduce the importance of leakage inductance.

 

[mcandmar]

That is fantastic, thank you!

Currently i have 1.5k resistors installed and measured 6.8v across them with a 77v B+ so it’s around 5ma. I was close, even if it was a complete fluke

I couldn't help but notice when i draw the load lines and operating points for a stock quickie, my current 45v quickie, and this 72v operating point onto the datasheet chart i can draw a line from 0 in the bottom left all the way through these three operating points.  And if i continue the line on to my 77v load line it’s pretty much spot on 5ma too.  I assume this is not a coincidence so what does this line represent, and what is the secret to determining it?  I feel like there is another penny about to drop…

I have one more question if you would be so kind, what is the relationship or rule of thumb for plate load vs output transformer impedance?  i.e. when you say 5600ohm is ideal for 72v, where does that figure come from?

I'm going to have to digest all this new information and do some experimentation now, thanks again Paul.

 

[Paul Joppa]

It all derives from the 3/2 power law for tubes - the current goes as the voltage to the 3/2 power. It's a theoretical rule, based on good physics but assuming a lot of real-world effects are not important. From that, if you double the voltage then the current goes up by 2.828 times, hence 2mA becomes 4.25mA. Since resistance is voltage divided by current (Ohm's Law), the 8K load becomes 5.6K.

I post my rule for SE outputs every so often in various forums - the formula is:

RL = Rb - 2.38*rp

RL is load impedance, Rb is beam resistance (plate to cathode voltage divided by plate current), rp is plate resistance. The triode-mode 3S4 at 22v/2mA has a plate resistance of about 4K ohms.

This rule looks odd, but the physics behind it is fairly simple. I assume you have a graph of the plate curves at hand, and you have a decent idea of what a load line looks like. Draw a load line, and note the current at zero grid bias. Then find the place on the load line where the current is half that - that is the quiesvcent operating point. At the low-current end of the load line, you can find the point where the grid bias is twice that of the quiescent point; it will be a bit more than zero current - this keeps operation out of the region near zero current where the curves are all scrunched up together. That's it; I just did the algebra using the 3/2 power law.