some tradeoffs to consider.

Design and use of Chokes for PS, anode and filament use.
Post Reply
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

some tradeoffs to consider.

Post by dave slagle »

new topic, new person...

This is from email and a few interesting topics were starting to come up so I brought it here.
I'm looking for some real decent plate chokes to load type 112A, as an input tube. (5k Rp, 8mA Ip)
All you guys and your Low current High Rp tubes are making me think about a few things differently.

Given that it is an input tube swinging very low voltages it seemed logical to me to shoot for 80% nickel as a core and I pointed him to these plots for a starting point.
Image
There seems to be a range of current for which the inductance is at least fairly constant, with some of the air gaps. 0.016" with the 80% nickel for example, 8 to 13mA or so
The .016 gap will indeed net you a nice linear inductance. Assuming your 5k source simple math suggests a -1dB point of around 15hz with that particular gap (125hy) you can also interpolate the graphs a bit. For example I would expect a gap of .012 to fall between and give you around 160hy and show the onset of saturation a bit earlier. The one thing often overlooked is the actual AC flux in use plays a large role in what part of the curve you can operate in. For this choke your max of 2V @ 20hz represents an AC flux of only 65 gauss! this around 1% of the Bsat of the material and remember that as voltage goes down and frequency goes up flux decreases rapidly we really are talking about tiny signals here (which is why I suggested 80% nickel)

Look at it this way, with a .012 gap increasing the standing DC current 1 ma form 7ma to 8ma will change the DC flux 400 gauss which is almost 7X the largest AC signal the coil could ever see. Some people will claim this to be headroom, others will suggest that the gap could be made much smaller and due to the small AC involved the device will still be linear. Going to the .008 gap would net you 200hy's and even though the line appears to be on the edge of saturation the small ac signal will not even register unless you zoomed in 100 times.

We can theorize about this all day, but when it comes down to it, it's your amp and not mine. So rather than picking an arbitrary value and slapping the device in a can, I encourage you to listen ot all three gaps and push it even more to see what actually works best for you. With this feedback combined with the details of the rest of your system I suspect we will start to see some patterns emerge.

Inductance is not so important
What a fresh diversion from the "grunt grunt, I need a billion henries" world audio has become :-)
I would prefer to go with DCR as a higher priority. I can detail the reasons for this when the time is right, at the forum; LF poles and overall response needed from the amp etc, but I am in no hurry at all, so take your time.
Do tell... And post a schematic if you can.
What are the consequences of an uneccessarily high DCR
I can give you a few theories and a couple of hunches but I honestly do not know.
apart from voltage drop (non issue).
The relationship of DCR to the source impedance plays a role on how your LF rolloff occurs. Spice can tell you what it will look like, but who knows how that relates to sound. Interestingly enough DCR is also a barometer of other design choices so
By linearity, are you meaning the higher (inductance)
impedance load having a more horizontal load line?
This is where it gets screwy. A CCS will have a horizontal load line, does that make it lower distortion than a choke? By linearity I mean a device that behaves as expected over the widest range of voltages and frequencies. While I agree that the simple models suggest a higher inductance will make a circuit more linear, they miss the fact that what you do to increase that single parameter makes other parameters suffer. I tend to step back and consider the big picture.

When talking about choke loads, the truth is you start out with an inductive load and slowly transform to capacitive load as frequency goes up. At the point where the inductive and capacitive loads are equal (resonance) the load actually appears as resistive! What a mess. The problem is this "mess" always seems to sound better than anything else I have tried.

Well I see this getting long winded and better end it now... Attached below is the schematic.

dave
Attachments
this is the current overall schematic and the chokes we are talking about are for the first stage
this is the current overall schematic and the chokes we are talking about are for the first stage
amplifier.gif (52.19 KiB) Viewed 7777 times
shane
Posts: 2
Joined: Wed Jun 08, 2005 5:08 am

Post by shane »

[i]What a fresh diversion from the "grunt grunt, I need a billion henries" world audio has become [/i]


Inductance of the IT and Rp of the driver (40H,3k5) will set the overall LF performance of the amp, -1dB at 54Hz, -3dB at 15Hz.

Any idea how close the -3dB points can be before they add?.

If a few Hz difference is enough for it to be ok, using fixed 0.1uF for the RC filter, consider 120k = 13.3Hz, and then using 5k as Rp for the input tube;

80H, -1dB at 38Hz, -3dB at 11Hz
100H, -1dB at 26Hz, -3dB at 9Hz
125H, -1dB at 25Hz, -3dB at 7Hz
150H, -1dB at 19Hz, -3dB at 5Hz

Leaves a bit of room for suitable inductance.

I'd be interested to know how you see the DCR affecting the rolloff here, if Rp/4 is close to ideal, which way does Rp/2.85 (5k/1k75) move things ?

shane
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

Inductance of the IT and Rp of the driver (40H,3k5) will set the overall LF performance of the amp, -1dB at 54Hz, -3dB at 15Hz.
Wow, -1dB at 54hz?? Are you using an active sub or some other way to get that back?

I can see an entire system being -1dB at 40hz and still being musical, but for a single sage it seems a bit much (assuming full range whatever that is)
Any idea how close the -3dB points can be before they add?.
Don't they just add?

80H, -1dB at 38Hz, -3dB at 11Hz
100H, -1dB at 26Hz, -3dB at 9Hz
125H, -1dB at 25Hz, -3dB at 7Hz
150H, -1dB at 19Hz, -3dB at 5Hz
Leaves a bit of room for suitable inductance.

Given the plots above, I'd start out towards the higher inductances and then play with the gap to see where that gets you.
I'd be interested to know how you see the DCR affecting the rolloff here, if Rp/4 is close to ideal, which way does Rp/2.85 (5k/1k75) move things ?
Its just a spice thing, at one point I though if your DCR = Rp you would never go below -3dB (as it turns out the dcr is resistive so the max was -6dB) but it never seems to sound as good as a lower DCR. Not sure whay that is, but I arbitrarily picked the Rp/4 as a number ot start with when I begin a design and then see how long that lasts before it becomes compromised. Hopefully in the future, more data will come in and give a clearer picture of how the DCR relates to the source Z.

Below is a graph from steve bench showing how the DCR relates to Rp for an imaginary tube with a 1K Rp and a 30hy choke. If nothing else, it suggests that a low DCR will actually improve your LF response. Being a big fan of the ideal 6dB per octave roloff from a simple filter, the DCR= 1/4 Rp seems to give pretty close to that. I'm not saying any of this matters mind you:-) Just making some observations.

dave
Attachments
sB did this for me a number of years ago.
sB did this for me a number of years ago.
choke-rollof-with-DCR.jpg (64.43 KiB) Viewed 7765 times
shane
Posts: 2
Joined: Wed Jun 08, 2005 5:08 am

Post by shane »

Wow, -1dB at 54hz?? Are you using an active sub or some other way to get that back?

I can see an entire system being -1dB at 40hz and still being musical, but for a single sage it seems a bit much (assuming full range whatever that is)
Yes, active sub.
Don't they just add?
I was thinking they needed to be close in order to do so. Measured BW of the amplifier suggests they are not adding as things are now.
Its just a spice thing, at one point I though if your DCR = Rp you would never go below -3dB (as it turns out the dcr is resistive so the max was -6dB) but it never seems to sound as good as a lower DCR. Not sure whay that is, but I arbitrarily picked the Rp/4 as a number ot start with when I begin a design and then see how long that lasts before it becomes compromised. Hopefully in the future, more data will come in and give a clearer picture of how the DCR relates to the source Z.

Below is a graph from steve bench showing how the DCR relates to Rp for an imaginary tube with a 1K Rp and a 30hy choke. If nothing else, it suggests that a low DCR will actually improve your LF response. Being a big fan of the ideal 6dB per octave roloff from a simple filter, the DCR= 1/4 Rp seems to give pretty close to that. I'm not saying any of this matters mind you:-) Just making some observations.
It seems that its not going to matter a great deal in this situation, that end of things will be well below the passband at the IT.


From the charts I get the impression that if a winder represents his chosen gap for a particular off the shelf model, in mA.. then that would more than likely be a centralised value (so far as plate current vs inductance), and the core should then be able to accommodate more than the 'gapped for xmA' rating, up to a certain point, at the expense of inductance (generally), and so then saturation need not be any real issue, if inductance is the limiting factor?

Looking to the chart for 0.016", 7 to 24mA. If you were to express this model in mA, like they all seem to, say "125H, gapped for 12mA", it could take almost double the plate current before the core satuated, and if you can live with less inductance then all good and well.

If that is correct, then using 8mA with a '120H/5mA' plate choke will reduce the amount of inductance, but it wont saturate the core.

How can you measure to see if you are saturating the core ?

thanks,
shane
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

Yes, active sub.
This opens up a number of possibilities to you that are not available to the average builder. You can vary the gap of the inductor to help tailor your LF response.

dave wrote:Don't they just add?
I was thinking they needed to be close in order to do so. Measured BW of the amplifier suggests they are not adding as things are now.
If the 3dB points are an octave away, you will get a 4dB loss at the original -3db point... Going further apart makes the lower point have less impact on the upper. I'm not sure how phases come into play here.

But if your -3dB point is say 50hz prior to inserting the plate choke, and then you gap the choke to be -3dB at 50hz also, you should get a nice 6dB/octave rolloff which will allow the amp to run full range. Then you can just "blend" your subs in.

It seems that its not going to matter a great deal in this situation, that end of things will be well below the passband at the IT.
It's easy enough to increase the gap to see what things sound like.
From the charts I get the impression that if a winder represents his chosen gap for a particular off the shelf model, in mA.. then that would more than likely be a centralised value (so far as plate current vs inductance)
The problem is nobody knows how things are rated. My hope is the family of L vs. I plots in conjunction with ears on experience will start to fill in some of the blanks.
and the core should then be able to accommodate more than the 'gapped for xmA' rating, up to a certain point, at the expense of inductance (generally)
But at what expense does that inductance come? This is a question I seriously do not know the answer to.
so then saturation need not be any real issue, if inductance is the limiting factor?
This brings us back to the linearity of the inductance. If we determined we needed 125hy @ 13ma we could choose either an .008 or the .016 gap to meet our spec. Does this mean they will both sound the same?
Looking to the chart for 0.016", 7 to 24mA. If you were to express this model in mA, like they all seem to, say "125H, gapped for 12mA", it could take almost double the plate current before the core satuated, and if you can live with less inductance then all good and well.
I really don't see the point in using such a convoluted method to spec, when the plots and a user adjustable gap opens up so many more doors. Plus you actually get to listen to (and measure) the results which is far more valuable than any single spec.

Saturation is not a hard wall, much like tubes soft clip, so does iron. Below I'll attach the plots for smaller gaps for reference. Remember these are actual measurements. All of the specs typically given for iron are based on math. Nothing like numbers to turn a slow occurring event into black or white. I'll also attach the plots of what the math says for this device. Now the only question is where do you put saturation :-)
If that is correct, then using 8mA with a '120H/5mA' plate choke will reduce the amount of inductance, but it won't saturate the core.
Maybe, maybe not... It depends on what convoluted method was used to spec the choke ;-)
How can you measure to see if you are saturating the core ?
A scope and a signal generator should show you where it happens, but it's sometimes tricky.

dave
Attachments
just for completeness sake... Calculated AC flux
just for completeness sake... Calculated AC flux
image001.gif (16.39 KiB) Viewed 7716 times
Calculated values of Bdc for various currents.  Saturation for 80% nickel is typically quoted between 6-8KG.
Calculated values of Bdc for various currents. Saturation for 80% nickel is typically quoted between 6-8KG.
image001.gif (12.46 KiB) Viewed 7716 times
Calculated inductance for various gaps.
Calculated inductance for various gaps.
image001.gif (9.69 KiB) Viewed 7716 times
showing the effect of smaller gaps. (real world measurements)  The AC flux for all of these was ~300 gauss
showing the effect of smaller gaps. (real world measurements) The AC flux for all of these was ~300 gauss
image001.gif (11.16 KiB) Viewed 7716 times
Post Reply