OK, I have been giving this some thought today. Here is what I have come up with.
I will follow Tubelab's LTP driver board as linked above pretty much to the letter. Drivers will be Octal 6SN7.
Power requirements are:
B+ 400 Volts - planning Maida style regulator from 800VCT 250mA transformer
-10 volts (CCS supply for tail of first LTP, about 10mA). Power transformer has two unused 5.0VCT secondaries. Paired up, these should provide enough voltage to feed a LM7910 -10 volt regulator.
-150 volts (feeds -ive supply to MOSFETS and bias to output tubes). Full wave rectify from 800VCT to a 0D3 regulator tube 150 volt 5-40mA (have always wanted to use one of these). Max grid current draw will be about 30mA, so should be OK.
Anything grossly wrong with this approach?
Thanks for your thoughts...
Chris
I will follow Tubelab's LTP driver board as linked above pretty much to the letter. Drivers will be Octal 6SN7.
Power requirements are:
B+ 400 Volts - planning Maida style regulator from 800VCT 250mA transformer
-10 volts (CCS supply for tail of first LTP, about 10mA). Power transformer has two unused 5.0VCT secondaries. Paired up, these should provide enough voltage to feed a LM7910 -10 volt regulator.
-150 volts (feeds -ive supply to MOSFETS and bias to output tubes). Full wave rectify from 800VCT to a 0D3 regulator tube 150 volt 5-40mA (have always wanted to use one of these). Max grid current draw will be about 30mA, so should be OK.
Anything grossly wrong with this approach?
Thanks for your thoughts...
Chris
The power supply sounds good to me. Just knowing how Hammonds tend to have high output, I would suggest a heatsink for the pass device since you will be dropping so many volts in the regulator, unless you have a monster thick chassis.
Just be aware that Tubelab's circuit will be complete overkill for driving 7591s. Don't let that discourage you, though. Driver distortion will probably be obscenely low, and you can rewire and sub in virtually any tube type(I would try triode connected 6L6 and KT88) and you will never be limited by your driver.
I just did a board layout for two Maida regulators on a single board and am willing to share gerber files if you are interested.
Just be aware that Tubelab's circuit will be complete overkill for driving 7591s. Don't let that discourage you, though. Driver distortion will probably be obscenely low, and you can rewire and sub in virtually any tube type(I would try triode connected 6L6 and KT88) and you will never be limited by your driver.
I just did a board layout for two Maida regulators on a single board and am willing to share gerber files if you are interested.
Thanks for the encouraging words SpreadSpectrum. It was the Maida regulator diagram kindly posted by you earlier in this thread that I was going to use! I would be very interested in using your board layouts! I have been doing a bit of a search on line seeing if anyone had some boards to offer to save me some space and to try to keep my layout orderly. You may have solved two problems for me 
As for heat sinking, I noted the advice you gave earlier in the thread. The chassis I am using is heavy aluminium. Top plate is 230 x 360 x 5 mm, so should be able to dissipate some heat. I have purchased some 100 x 60 x 18 mm heat sinks (3mm base with 16 x 1.5mm fins) to place on the top plate for the pass device. Hopefully this will be adequate. I have some 2SK1119 1000v 4A MOSFETs for the pass device.
A question or two regarding your reg design if I may... I see you have designed an output voltage of 460v, I would like 400v. I think that R37 and R35 are a voltage divider to set the output voltage. Have you a suggested value (R37?) for output of 400v? I note you have an inrush current limiter feeding 200uF at he input. What are you using after your rectifier prior to the regulator? I think there was some advice that a choke prior to the reg would be good, but your post recommending a reg stated that it would save using a choke if I recall correctly.
Yes, I know that the Tubelab driver solution is overkill, but I agree that it will open up some options for future experimentation, and that is what most of us are here for I guess
Once again, thanks for the support and help.
Chris
As for heat sinking, I noted the advice you gave earlier in the thread. The chassis I am using is heavy aluminium. Top plate is 230 x 360 x 5 mm, so should be able to dissipate some heat. I have purchased some 100 x 60 x 18 mm heat sinks (3mm base with 16 x 1.5mm fins) to place on the top plate for the pass device. Hopefully this will be adequate. I have some 2SK1119 1000v 4A MOSFETs for the pass device.
A question or two regarding your reg design if I may... I see you have designed an output voltage of 460v, I would like 400v. I think that R37 and R35 are a voltage divider to set the output voltage. Have you a suggested value (R37?) for output of 400v? I note you have an inrush current limiter feeding 200uF at he input. What are you using after your rectifier prior to the regulator? I think there was some advice that a choke prior to the reg would be good, but your post recommending a reg stated that it would save using a choke if I recall correctly.
Yes, I know that the Tubelab driver solution is overkill, but I agree that it will open up some options for future experimentation, and that is what most of us are here for I guess
Once again, thanks for the support and help.
Chris
It's not really my design. I just lifted it from Jones's Valve Amplifiers and the Michael Maida paper. I used the input bypass cap (C33) that Jones uses but that Maida states must not be used if the regulator is to survive a short circuit. There is a trade-off here between robustness and performance. I guess don't bypass the input of the LM317 if you want it to have a chance of short circuit survival.
I chose the divider values to pass enough current to guarantee regulation with no load. The 50k is a 1% caddock thick film power resistor bolted to the chassis. They don't offer many values, so I change R35 to change output value, adjust on test. LM317 is supposed to have 1.24 or 1.25V between output and adjust pin so just use Ohm's law to figure out a new value. Let me know if you have trouble.
After my rectifier prior to the regulator I have...a capacitor, that's it. I never used the inrush limiter and I used 100uF for the filter cap. I stacked two 200uF caps with equalizing resistors because these 800VCT Hammonds put out a lot of voltage.
Would you like me to whip out a board with a single Maida on it or are you interested in the double board? It would be trivial as I just have to delete half the board.
Just to let you know, I typically remote mount heat-sinked components on the chassis (not far away, just not on the board). I curl up the leads into a loop and solder wires and cover in shrink tube. Gate stoppers get soldered directly onto pins and covered in shrink tube as well, so they are not on the board. I guess I could solder mosfets on the back and use short stand-offs to mount them directly to the chassis, but that seems like more of a pain than extending the leads to me.
I chose the divider values to pass enough current to guarantee regulation with no load. The 50k is a 1% caddock thick film power resistor bolted to the chassis. They don't offer many values, so I change R35 to change output value, adjust on test. LM317 is supposed to have 1.24 or 1.25V between output and adjust pin so just use Ohm's law to figure out a new value. Let me know if you have trouble.
After my rectifier prior to the regulator I have...a capacitor, that's it. I never used the inrush limiter and I used 100uF for the filter cap. I stacked two 200uF caps with equalizing resistors because these 800VCT Hammonds put out a lot of voltage.
Would you like me to whip out a board with a single Maida on it or are you interested in the double board? It would be trivial as I just have to delete half the board.
Just to let you know, I typically remote mount heat-sinked components on the chassis (not far away, just not on the board). I curl up the leads into a loop and solder wires and cover in shrink tube. Gate stoppers get soldered directly onto pins and covered in shrink tube as well, so they are not on the board. I guess I could solder mosfets on the back and use short stand-offs to mount them directly to the chassis, but that seems like more of a pain than extending the leads to me.
Thanks again for the reply. I think I can work out the required resistance for the divider...
I was going to do pretty much what you suggest regarding the remote mounting of the heat-sinked components, so that sounds good.
Single board would be great!
So to confirm, you are running rectifier - 100uF (off board) - CL140 - then the 200uF (2 x 200uF with current balancing resistors) on the board. This is good, as it saves valuable chassis real estate, weight, expense etc.
Thanks again, I really appreciate your advice, assistance and patience!
Regards,
Chris
I was going to do pretty much what you suggest regarding the remote mounting of the heat-sinked components, so that sounds good.
Single board would be great!
So to confirm, you are running rectifier - 100uF (off board) - CL140 - then the 200uF (2 x 200uF with current balancing resistors) on the board. This is good, as it saves valuable chassis real estate, weight, expense etc.
Thanks again, I really appreciate your advice, assistance and patience!
Regards,
Chris
Actually, I am just running it as rectifier -> CL140 -> 100uF -> regulator. The 100uF is actually two 200uF stacked but the capacitance is cut in half since they are in series. I am using solid state diodes, though. If I remember correctly you are going to use a tube rectifier, right? That will require less capacitance. Check the data sheet. I did the capacitors off board because I used those JJ can capacitors mounted on top of the chassis.
I'll try to whip out a board soon. Gotta do some yard work today. My wife is crackin the whip already.
You can download a free Gerber viewer here:
http://www.pentalogix.com/Download/download.html#item8
I looked at your gerbers and they look like the picture you posted in an earlier post.
http://www.pentalogix.com/Download/download.html#item8
I looked at your gerbers and they look like the picture you posted in an earlier post.
Thank you all very much!
Yes I am going to be using a tube rectifier for the slow turn on of B+. I was thinking of about 47uF after the rectifier, resistor, then the caps in your design. I was intuitively thinking that 47uF only might not be enough capacitance. I need to get in to the books
I will post a schematic of final design soon. There was another thread discussing schematic drawing programs for Mac. I have looked at a few and need to train myself. Tried to use the 'GIMP', but steep learning curve. I guess I will have to persist...
Thanks again for all the help guys, I very much appreciate your patience!
Chris
Yes I am going to be using a tube rectifier for the slow turn on of B+. I was thinking of about 47uF after the rectifier, resistor, then the caps in your design. I was intuitively thinking that 47uF only might not be enough capacitance. I need to get in to the books
I will post a schematic of final design soon. There was another thread discussing schematic drawing programs for Mac. I have looked at a few and need to train myself. Tried to use the 'GIMP', but steep learning curve. I guess I will have to persist...
Thanks again for all the help guys, I very much appreciate your patience!
Chris
I believe that the coupling capacitors between the first and second stages can be eliminated (shorted out) for driving 6L6 type tubes. This will bias the second stage up at a higher voltage requiring a different cathode resistor value. It also allows the cathode resistor to be connected to ground instead of a negative voltage since the cathodes will be at about 100 to 125 volts. Of course this reduces the available voltage swing by 100 to 125 volts, but we are driving 6L6 control grids, not sweep tube screen grids. I have not tried this yet since I have not had time to even turn my workbench on since I returned home.
This should work well. I used Maida type regulation in my original 300Beast amp, but the regulator blew up twice, creating a bright red glow in the 300B's, so I ditched the regulator and switched to a choke input filter which gave me about the same voltage. I don't think that is possible in your case. My mosfets were rather wimpy, so I think that's why they fried. The Maida I have now has a monster 1000 volt fet that costs $7. It hasn't fried yet.
Add an ordinary diode (1N4007) from the regulator output to ground in the non-conducting direction. This prevents the B+ from being pulled negative before the rectifier tube warms up.
Keep in mind that you will be dissipating some serious power in the mosfet. Your transformer will put the B+ in the 480 to 500 volt range, maybe more (it is a Hammond!). The idle current for each channel is around 150 to 170 mA (100 mA for output tubes, 10 mA for the input tube (total), 20 mA for the second stage, and 10 to 20mA for each mosfet. Current at full power will increase to 220 to 250 mA per channel. Assuming a 100 volt drop across the mosfet, and 200 mA of average current you will burn 20 watts per channel in the mosfets. I would use two Maidas, one for each channel. This will spread out the heat and improve the channel separation. You are still going to need something more that the chassis for this.
Probably big enough. Heat sinks work best when oriented such that natural convection will allow for air movement. Lying flat on their backs in the center of a flat horizontal plate is worse case. Mounted vertically on the rear panel is the best. This kind of stuff is why we have PHD thermodynamics guys at work. I just do some serious testing before actually building anything.
You can do this. If you are going to use an LM334 this is the way to go since it can only handle 40 volts. If you are going to use a 10M45, don't bother with a -10 volt supply, just tie the CCS to what ever negative supply that you have up to -450 volts. A small heat sink (like on the Simple SE) will be needed for -50 to -150 volts, a bit larger for -450 volts.
There is a potential problem with with the use of negative voltages above -75 volts on either stage. The cathode voltage will settle down to a small positive value (1 to 4 volts) after the tubes are warmed up. However when the tubes are not yet warmed up, but the supply voltage is present the tubes are not drawing current. The cathode voltage will equal the negative supply voltage. This can exceed the heater to cathode breakdown rating. There is a simple cure. Use some sort of clamp device from the cathodes to ground. I used two 75 volt zeners diodes in series (cathodes tied together) in my 300Beast amp, but those who don't like silicon have used a NE2 neon bulb. Either will work, and neither should conduct in normal operation. The NE2 will glow on power up, and extinguish as the tubes warm up.
I prefer to use a low voltage source for both the positive and negative supplies to feed the PowerDrive fets.
It is possible to use the main +400 volt supply to feed the mosfets but there are two reasons not to do this. If for some reason a mosfet were to short out, it will put the full B+ on the control grids of your output tubes. Trust me they won't like it. Since the fet will be happy with +50 volts or so, the rest is just burned up as heat. You could use a regulator to drop 350 volts or so, but this just moves the heat. Of course +400 volts is actually needed for screen drive applications so I have my fets tied to B+ and they do put out some heat, but they work just fine.
The negative supply feeds three circuits.
The tail of the first stage. -10 volts or so is all that is needed here. -450 volts is the maximum rating for the 10M45. A voltage clamp is needed on the cathodes if a voltage over -75 volts is used.
The tail of the second stage is happier with about -50 volts or more if coupling caps are used. Any negative voltage can be used if a voltage clamp is used on the cathodes.
The negative supply to the mosfets must be more negative than the voltage required to fully cut off the output tubes being used. I have found some evidence that a higher voltage makes the mosfet happier. The negative voltage can be quite high if the source resistor is increased in resistance and power rating.
I use a transformer with two 120 volt secondaries for amps of this type. Small transformers (often toroids) with 4 seperate 120 volt windings are very common here. Most are made in China or India, so they should be available world wide. If they are not available in Australia, look for a small filament transformer with two 120 volt primaries. 6.3 volts at one or two amps would be ideal. Connect the 6.3 volt secondary to one of your extra 5 volt windings, and you will have two 100 volt or so windings (the original primaries). Connect them in series to make a 200 volt CT winding, ground the CT and use a SOLID STATE bridge rectifier to make plus and minus 140 volts or so. CRC filtering is usually adequate. Why SS rectification? You want to be absolutely sure that the bias voltage is present on the output tubes before the B+ voltage shows up. If you use a seperate transformer the entire mosfet circuit is up and running instantly, insuring that the bias voltage is at the proper level long before the output tubes need it.
I'm thinking it might be too big! Why? Rectifier tubes have a maximum capacitance spec. 47uF is the limit for a 5AR4. We have all seen the threads with people losing 5AR4's in their simple SE's with a 47uF cap. You are looking at a bigger transformer than a Simple SE uses, so your rectifier tube is in for a shock on power up. I would use a 22 or 27 uF, then the CL140 then a reasonable cap (100uf). The Maida should clean up any hum.
My 278CX(800VCT 465mA) puts out obscenely high voltage. I got 565V out and that was after going through a 193Q choke (55Ohm DCR). Even thought it was a pain and expensive, I built a dummy load for the power supply and probed around to see what voltages I would get at the filter capacitors and measure ripple, etc., before I connected the PS to the tubes.
You don't need big input capacitors. You'll have plenty of volts to spare. The Maida works quite well on the 100-120 Hz ripple. Solid state diode switching noise is another story...
The CL140's resistance will be quite low(6-7 Ohms?) once it is warm. Would a little more resistance between the capacitor banks be advisable for long rectifier tube life? I have minimal experience with rectifier tubes.
I'll add this some time in the next few days.
There are three things that kill rectifier tubes (or SS diodes).
The big one is the initial surge that is drawn by the filter caps which are empty as the rectifier is just beginning to conduct. The capacitors are at zero volts and electrically nearly a short circuit. The tube heater started cold and is in the process of warming up. The cathode will start with zero emission capability which rises slowly. So when the cathode is at maybe 20% of its capability it will be looking into a really hungry input cap asking for more current than the cathode can supply. This is what causes the sparking that you see in rectifier tubes at turn on. It seems that new production tubes are much worse in this regard than older tubes and this is likely related to the care taken when coating the cathode (or filament). All it takes is some hot spots that heat up quicker than others and all of this start up current happens at one little spot on the cathode. This is why the sparking rectifiers seem to be batch related. An inrush current limiter anywhere in the secondary circuit of the transformer (I usually put them in the center tap lead) gives the tube a few extra seconds to get hot before the hungry caps want to be fed, and ramps the current slowly.
The second issue is high peak current after everything has warmed up. The input voltage (transformer secondary) is a sine wave. The desired output is pure DC. We have been taught that the DC output voltage is equal to the peak AC voltage under ideal conditions. This doesn't happen, and things would be bad for the retifier if it did. Visualize this hypothetical example:
Lets assume for a minute that the transformer secondary puts out 100 volts. We are using a rectifier that is perfect (no loss) and a CLC filter with perfect components (zero resistance choke). Our power transformer is wound with magic wire that has no resistance. It is also magnetically perfect.
Under no load conditions the input cap would charge up to 141.4 volts DC. If the DC output voltage was equal to the peak AC voltage, when does the rectifier conduct? It DOESN'T! The rectifier conducted extremely heavilly during the time it took to charge up the capacitors (see above), and then there is no place for current to flow.
Now we will draw current (assume 100 mA) from this perfect power supply and see that the DC output voltage drops to 140 volts. Why, all of our parts are perfect, right? Yes, they are, but if the voltage didn't drop, when would the rectifier conduct? During the time that the AC sine wave is below 141.4 volts, the rectifier is reverse biased, it can't conduct. If the output voltage is equal to the peak of the sine wave there will be no current flow.
Now with the output voltage at 140 volts, when does the rectifier conduct? It conducts ONLY for the brief time when the instantaneous value of the sine wave is above 140 volts. But we are asking for 100 mA ALL OF THE TIME! Where does the current come from? There is energy stored in the charged capacitors and in the choke's magnetic field. It is this stored energy that actually runs our amp for most of the time. But this energy gets depleted. It can only get replenished during the time the rectifier conducts. If the peak of the sine wave is 141.4 volts it will be above 140 volts for a very small fraction of the total sine wave period. This "time of rectifier conduction" is referred to as the conduction angle, and is expressed in degrees.
If the rectifier is replinishing our storage system for a fraction of a cycle, but we are drawing 100mA for all of the cycle, it is easy to see how the peak rectifier current can easilly be AMPS! Now think about how the transformer feels, you draw no current for most of the time and then ask for an amp for 3 or 4 milliseconds and then wonder why it buzzes! The input filter cap has a tough job too. The amps of peak current is dumped into the cap and then sucked right out, 100 or 120 times a second. These large currents being drawn for brief periods of time can easilly couple noise into nearby circuits.
This is admitedly the worse case. How do we reduce this peak current? We must increase the conduction angle, or limit the peak current with series resistance (which increases the conduction angle). The biggest influence on the conduction angle is the value of the input cap. The smaller the cap, the larger the angle, and the larger the ripple. A choke input filter has no input cap and the largest conduction angle. Since the choke is now a primary energy storage device it must be designed for input duty and sized for the current draw.
Any resistance in the circuit reduces the current, including the DC resistance of the transformer. Tubes have inherently far more internal resistance than solid state diodes.
Figuring out the exact conduction angle and the peak currents during conduction requires knowledge about the components being used, and some rather advanced math. There are modelling programs like PSUD2 that handle the math.
The third cause of rectifier death is excessive reverse voltage. In our "perfect" example above with no load the output was 141.4 volts. At the positive peak of the sine wave the rectifier is fed +141.4 volts and just begins to conduct. At the negative peak of the sine wave the rectifier is fed -141.4 volts, but its output is still at +141.4 volts. The rectifier sees 282.8 volts in the reverse direction.
So with a 400-0-400 volt transformer the rectifiers will see 1131.2 volts if everything is operating normally. Given a Hammond transformer's propensity for over voltage, 1500 volts is a given.
What happens when you switch off the power? If you happen to hit the switch at the peak of a cycle there is a large current peak flowing through the transformer, which you attempt to interrupt. The inductance in the transformer will attempt to keep this current flowing resulting in a voltage spike. I have measured 2500 volt spikes in a Simple SE with a Hammond transformer. This has to go somewhere. The spike usually contains little energy and is dissipated in the filaments of the tubes (they are resistors connected across the other transformer windings) but it can eat solid state diodes that are not designed to handle such spikes (IXYS FREDS).
As we have seen above, resistance is your friend in a power supply. Resistance does kill regulation, so it must be used judiciously in an unregulated supply. If you are regulating the supply, yes, add some resistance between the first and second caps.
This is not required in most amplifier designs. It is a welcome addition in an amplifier design (like the Tubelab SE) where there are positive and negative supplies with resistive paths connecting them together. I ALWAYS design the negative supply to come up first (good for the tubes) but this can cause the positive supply to recieve a negative voltage before its rectifier tube is warm. This will be the case in the amplifier design discussed here.
Sometimes I am not the sharpest tool in the box, but this is twice now that Tubelab has advised a separate transformer for lower-voltage supply to MOSFETs and bias. I would be foolish not to take that advice
Transformers are not the easiest or cheapest things to come across here is Australia. Fortunately I will be travelling to Hong Kong with work on the 11th. I know where DIYCLUB is in Mong Kok. They have a transformer R26-42 primary 115 X2 :0-50-55(0.45A)X2. If I wire up the two primaries for 115 volt input, but plug them in to my 240 mains, I should get about 114 VAC out of it, for about 160 VDC rectified. That should be fine to do as suggested. My only problem is I have ordered a handful of 0D3 VR tubes for the -150 supply that are no longer required . I would love to try some VR tubes, so maybe I can use a 0C3 for 105 +ive supply and a 0A3 for a 75v -ive supply (from what I have seen, 0A3 is an orange glow, but some 0C3 have an interesting violet glow)?
As for the heat sinks on the Maida regs, I have done a little research on heat sink requirements and I suspect that the heat sinks I have may be marginal. I will get some better ones that will guarantee adequate cooling. Yes, I will have sep regs for each channel, as I am building monoblocks.
I will be using 10M45, so will tie LTP CCS to the above -75v supply. 75v chosen due to issues mentioned with higher value -ive supply.
At this stage planning on direct coupling 1st and 2nd stages unless there is a good reason not to.
Once again, thanks for the assistance!
Chris
Transformers are not the easiest or cheapest things to come across here is Australia. Fortunately I will be travelling to Hong Kong with work on the 11th. I know where DIYCLUB is in Mong Kok. They have a transformer R26-42 primary 115 X2 :0-50-55(0.45A)X2. If I wire up the two primaries for 115 volt input, but plug them in to my 240 mains, I should get about 114 VAC out of it, for about 160 VDC rectified. That should be fine to do as suggested. My only problem is I have ordered a handful of 0D3 VR tubes for the -150 supply that are no longer required
. I would love to try some VR tubes, so maybe I can use a 0C3 for 105 +ive supply and a 0A3 for a 75v -ive supply (from what I have seen, 0A3 is an orange glow, but some 0C3 have an interesting violet glow)?As for the heat sinks on the Maida regs, I have done a little research on heat sink requirements and I suspect that the heat sinks I have may be marginal. I will get some better ones that will guarantee adequate cooling. Yes, I will have sep regs for each channel, as I am building monoblocks.
I will be using 10M45, so will tie LTP CCS to the above -75v supply. 75v chosen due to issues mentioned with higher value -ive supply.
At this stage planning on direct coupling 1st and 2nd stages unless there is a good reason not to.
Once again, thanks for the assistance!
Chris
You can't do that. The 115 volt primary will not have enough inductance to run on 230 volts. It will draw too much current and usually leads to a rather smokey result. Someone told me this a long time ago when I wanted to use a 115 to 75 volt trasnformer in reverse to make 150 volts. I didn't believe them and I stunk up the whole house. The 2 X 55 volt windings can be used as is to make about + and - 75 volts which is enough for the 7591's but may be marginal for 6L6GC's or 6550's.
OK, don't get me wrong. I use the seperate transformer approach because it is the easiest way for ME to get the job done. In fact my version of this amp may indeed use one transformer. It just happens to have a pair of 70 volt taps which will make an easy + and - 100 volts.
You have OD3's, and want the cool glow. I would go ahead and use one as the reference for a mosfet follower type regulator. Feed it with a solid state rectifier for instant on negative voltage. I used this approach in my 845SE amp. The Allied power transformer feeds a 5AR4 into a cap input filter for driver B+ (about 450 volts). The same winding also feeds a pair of solid state diodes into a choke input filter for the negative supply. I get about - 400 volts which is used directly by the PowerDrive board to feed the 845's. You can use a negative voltage regulator here, which could be as simple as a resistor, an OD3 and a P channel mosfet.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.