As I intimated previously, I have completed the filament power supply design for the 300Bs. The 300B filament supply has to do several things. First it has to supply two separate supplies of 5vdc@1.2A, one for each output channel. We decided on DC filament supplies for the 300Bs because we are going for a very high fidelity amp. Balanced AC secondaries simply can’t provide the ripple rejection necessary for high fidelity. Secondly, the supply has to maintain at least 70dB of isolation between the two supplies (The standard stereo channel separation specification used in most high fidelity equipment). This is to ensure that channel crosstalk is kept to a minimum. Without this isolation, the channels would bleed together to some extent, and the width of the amp’s perceived soundstage would suffer greatly. The third is that the supplies must not pump “too much” (definition to follow) noise into the signal chain.
The first requirement is easy to meet. We simply design two identical supplies, one for each channel. The second requirement is a little more difficult to meet. If the supplies are passive, then getting 35dB of isolation in each supply is going to be difficult. This implies separate transformers. However a single transformer with dual secondaries may be used if we use an active regulator to provide additional ripple rejection. The third requirement is easy to specify. The supplies must meet the noise requirements of the B+ supply. However, this may be easier said than done.
Working backwards, we’ll deal with the ripple specification first. The ripple from the filament supply is superimposed on the cathode bias voltage; in our case ~71v. Now the ripple specification from the power stage is -90dBv (or ~0.003%). Applying this to the cathode bias voltage yields a ripple spec for the filament supplies of Vr = 71*10^(-90/20) = 2.2mV. Obtaining this ripply level using only rectification and passive filtering is going to be almost impossible. However, using a modern linear (i.e. non-switching) voltage regulator it should be possible to meet this. In addition, most linear regulators provide excellent line regulation such that getting the channel separation should be possible as well. Finally, getting 5v@1.2A out of a modern linear regulator is easy.
So, we have arrived at a set of dc supplies using modern linear regulators, running off of separate secondaries on the same transformer. It’s time to select some components. The first should be what regulator to use. Now one of the things to keep in mind about tube filaments is that when cold, the presented resistance is about 10% of what it is when the filaments are up to temperature. This means that whatever regulator we choose must be able to effectively start up with this very low resistance and must current limit without “crowbar-ing” back to some low voltage state. We also need a regulator designed to operate with a very low input-output voltage differential (i.e. a “low dropout” regulator). This is for two reasons; first, we want to limit the voltage differential to limit the power dissipated by the regulator (to limit thermal design problems) and second, because we will likely be using 6.3vac filament transformers which when rectified and filtered will supply less then 7.9v (i.e. 2.9v max regulator input-output differential). An excellent regulator that meets all of these requirement is the Linear Technologies LT1085 low dropout 3A adjustable linear regulator. (
http://www.linear.com/pc/productDetail.jsp?navId=H0,C1,C1003,C1040,C1055,P1283) Line regulation is better then 70dBv so channel isolation shouldn’t be a problem.
For the rectifier, the ability to attach a heat-sink will be important. The current of 1.2A means that the regulator may well be dissipating 1.2W. Also seeing as how the primary filter cap will be relatively large (~6800µf) the startup surge current will be high and the conduction angle will be small. I settled on the Micro Commercial Components GBU4A. (
http://61.222.192.61/mccsemi/up_pdf/GBU4A-GBU4M(GBU).pdf) This is a 4 amp 50volt bridge rectifier with a 150A surge rating, 1v forward drop, and a case with a heat-sink mounting hole.
Finally, for the transformer, I chose the Hammond 266L12B (or the 266L12). This is a dual secondary 6.3v 2A (2.5 for the 266L12) filament transformer. The 300B heater current load of 1.2A with derating requires at least a 2A secondary. The final circuit is shown in the following figure.
Attachment:
Schematic Fliament Final.png
The circuit includes a trimmer for setting the final filament voltage and the primary can be wired for either 240v or 120v mains (50Hz or 60Hz). The supplies are not grounded except at AC via the 300B cathode bypass capacitors. Also, the positive side of the supply is connected to the same side of the filament as the cathode resistor and bypass capacitor such that the DC supply does not artificially affect the bias on the 300Bs.
There is one more thing to discuss with this supply; the thermal design parameters. Using the stated Vf of the GBU4A from the data sheet we get a power dissipation of 1.2W. Using the bridge Vf we can get a worst case regulator dissipation. The worst case input voltage is Vin=6.3v*sqrt(2)-1=7.9v which gives a power dissipation of (7.9v-5v)*1.2A=3.48W. These “back of the envelope” numbers indicate that a more complete thermal check is in order.
Starting with the GBU4 I have made the following assumptions. Total dissipation is 1.2W at load. I am going to allow a maximum ambient temperature (where the rectifier is located) of 75˚C (167˚F). This is actually a pretty standard design number for vented enclosures without forced air cooling. I am also using a thermal resistance for the case to whatever heat sink used of 1.5˚C/W. This is reasonable for a plastic case to anodized aluminum heat sink using a good quality thermally conductive paste. Finally. I am assuming a junction to ambient thermal resistance of 50˚C/W for the rectifier without a heat-sink. This is also a pretty typical number for plastic cases of this type. The results of the thermal calculations are shown in the following spreadsheet.
Attachment:
Thermal Design GBU4A.png
You’ll notice that with these assumptions, the maximum thermal resistance of a heat sink to ambient is ~38˚C/W. For my prototype board I used a Comair Rotron 822202B00000 heat-sink with a thermal resistance = 13˚C/W. (
http://www.alliedelec.com/search/productdetail.aspx?SKU=5990351) This gave almost 25˚C/W margin in my design. And given that it was in open air, this may even be rated as overkill. However, please note that the maximum allowed thermal resistance junction to ambient is 41.67˚C/W which is less then the 50˚C/W without a heat-sink so, even at this load, a heat sink of some type is required.
For the regulator, the design is a little more critical. I used much the same assumptions as made above. However, I calculated the regulator dissipation directly from the circuit parameters assuming an input DC voltage of 7.9v. The thermal design spreadsheet is shown below.
Attachment:
Thermal Design LT1085.png
Here the maximum allowed thermal resistance junction to ambient is only 14.82˚C/W. This means that with the other parameters we need a heat-sink with a thermal resistance to ambient of no more than 10.3˚C/W. In this case for my prototype I chose the rather substantial Aavid Thermalloy 529802B02500G heat-sink with a thermal resistance = 3.7˚C/W. (
http://www.alliedelec.com/search/productdetail.aspx?SKU=6190109) This gave me only 6.62˚C/W thermal margin so clearly this heat-sink is a good choice for this application. It should also be noted that if the trimmer is used to lower the 300B filament voltage to less then 5Vdc, the regulator dissipation will go up and the thermal design will have to be revisited.
This is it for the 300B heater design. Although I may also post an alternate design post for those who would like to try the build using a balanced AC heater approach. It’s not quite as clean as the DC heater option, but it is a much simpler circuit to be sure.
Well, this is the last of the circuit design work for the 300B stereo amp.
I hope that people have gleaned some good information from everything posted on this thread. In my next post (probably tomorrow) I’ll post a summary with all the design schematics and the overall specifications for the amp.
Comments anyone?
Login
Register
FAQ
This means that we need a Hammond 266L12B or 266L12 instead of the 266K12. Sorry Mark. I hope this doesn't cause a problem.
