It is true that for lowest distortion in an SRPP there is an optimum load. It is also true that the higher the idle current the greater the drive capability. Also, neither the transformer nor the headphone is a constant load. Because the top triode represents a relatively low impedance load to the bottom triode, the SRPP topology always produces quite a high level of distortion. For these reasons it makes more sense in this application to design for drive capability. The NFB reduces distortion to tolerable levels.
Cheers
Ian
Ian,
Thanks, I've come full circle it seems. Attempts to optimize the SRPP stage are not needed because the feedback is taking care of the distortion and this allows us to use a sub-optimal output stage purely for it's drive capability.
I pulled the trigger on parts, included the Sowter transformers. Fingers crossed I can get this project to the finish line and I like what I find at the end!
Thanks again,
Brian
Thanks, I've come full circle it seems. Attempts to optimize the SRPP stage are not needed because the feedback is taking care of the distortion and this allows us to use a sub-optimal output stage purely for it's drive capability.
I pulled the trigger on parts, included the Sowter transformers. Fingers crossed I can get this project to the finish line and I like what I find at the end!
Thanks again,
Brian
Considering the load (being both inductive, capacitive and resistive) what could be done to optimize it to the drive capabilities of the SRPP? In this setup for a five times (pre) amplifier:
An externally hosted image should be here but it was not working when we last tested it.
I tried to shift towards resistive, optimizing for a 10K resistive load. The coupling capacitor C1 sets a low frequency pole, so it should be out of the AF "working range". Would limiting the bandwith to AF be beneficial against distortion?
For what amplification is your setup, two times?
It could be raised by reducing the FB resister I guess...
Last edited:
A couple of points to bear in mind. First you can obviously wire the amp for 6.3V heaters. If you are building a stereo amp then you can wire the heaters of the two ECC99s in series and run it off 12.6V.
Secondly, you might need to use heater elevation. With a 300V HT supply, the top cathode of the ECC99 is at about 150V. The specified Vhk of the ECC99 is 200V which leave just 50V spare for the positive signal peaks. This means the maximum safe output signal from the SRPP is about 35V rms which might be marginal. The normal way to improve this situation is to elevate the heaters. So, instead of connecting one side of the heaters to HT 0V, you connect it to a decoupled pot divider across the HT. It is common to raise the heaters to about 75V. My normal method of doing this is to use a 4:1 pot divider consisting of a 22K in the bottom arm and 66K in the top arm. The 22K is a 2W resistor and the 66K consists of two 33K 2W resistors in series. I decouple the 22K with a 47uF 100V electrolytic.
Cheers
Ian
Sorry for intruding
The big difference is my load is only 4mW (6V in 10K) while yours is a manyfold, having a big impact on distortion.
The datasheet I have indicates 12.6v operation @ 400mA. Is there a reason I can't wire them in parallel at 12.6V? Just want to make sure I didn't misread or you aren't tipping me off to a new trick!
Doh! I should have known that.... makes running a DC regulated filament supply a little more tricky. I'll go and check tonight, hopefully the PS ground on the PCB kit I have isn't grounded to the screw holes. If not, I can float the whole circuit... if they are then I need to isolate the PCB from whatever chassis I use. Fingers crossed. I'd skip the DC heaters (I'd imagine I might be able to dodge hum with good layout) but the transformer I set aside for this project runs the heaters high specifically so they can be regulated down.
Hi Disco,
I think the trouble here is that your circuit is optimized to a resistive load, which is nice, but precludes it from use in the circuit we're discussing. A 100k pot won't do a good job driving a small impedance like headphones with any available power.
I frankly don't know what kind of voltage gain I'll need yet, though if this circuit works like a standard op amp circuit the gain seems to be set to 100, or 40db. I'm thinking 300mW of peak power into 600 ohms requires 13Vp-p; which would require 53Vp-p at the primary side with a 4:1 turns ratio; and that would require .5V at the input if the closed loop gain is 100. Seems quite reasonable provided I'm not mistaken anywhere in that logic.
Brian
Hi Brian,
It confused me reading that outrageous distortion is inherent to SRPP while others claim SRPP has negligiable distortion. Because of different jobs to consider both statements can be true I guess. In the meantime I found Ian's recipe, being more in line to what I'm cooking.
Rgds, Jaap
It confused me reading that outrageous distortion is inherent to SRPP while others claim SRPP has negligiable distortion. Because of different jobs to consider both statements can be true I guess. In the meantime I found Ian's recipe, being more in line to what I'm cooking.
Rgds, Jaap
If you run a triode in a circuit that allows it to perfom the best it possbly can, with the circuit making no contribution to the distortion, then the distortion you measure is the intrinsic distortion of the tube below which you can never go. Morgan Jones did this for a variety of tubes. In simple terms, a triode gives its lowest distortion when it has the lightest of anode loads. Jones achieved this using a CCS plate load.
The next best distortion perfomance comes with a mu follower, which is a crude plate load multiplier on top of the triode. However, it works surprisingly well and gives distortion figures within a dB or so of the instrinsic values and it can also do this while drving a reasonable load.
Next comes the SRPP which has a relatively low impedance active anode load. This low value of anode load means it creates a lot more distortion than the other two circuits, no matter what load it is fed into.Many peole have spent a lot of time trying to find the optimum load for lowest distortion or to find how to calculate the circuit values for a given load in order to get the lowest distortion. The bottom line is that no matter what you do with it, the SRPP always generates a lot more distortion than the other two configurations above. However, and it is a big however, it is a true push pull circuit and it beats the pants off the other two for the amount of current it can drive into a given load. If drive capabilty is what you need then the SRPP is the answer. If the distortion is a problem then negative feedback is the way to reduce it.
Cheers
Ian
Last edited:
Ian,
You know what I'm not really getting yet? When people say the SRPP is poor in terms of distortion, do they mean relative to other, better (perhaps more exotic) tube circuits or in comparison to a regular resistive load common-cathode gain stage... or relative to what current technology is capable of? It looks like Pete's design was about 1% THD at 1V into 100 ohms from input to output (ie: a 12au7 gains stage and the transformers involved too) with no feedback. In the grand scheme of things is that really all that bad compared to what another tube topology would be able to do without feedback? I've only ever made high volume overdrive machines so I'm very new to the pursuit of linearity in tube circuits, no frame of reference I'm afraid.
Either way if raising the loop gain and closing the feedback as you have done gets it down to the .2 - .3% range at full output, that seems pretty reasonable. What would do better with 3 triodes and transformer per channel?
Brian
You know what I'm not really getting yet? When people say the SRPP is poor in terms of distortion, do they mean relative to other, better (perhaps more exotic) tube circuits or in comparison to a regular resistive load common-cathode gain stage... or relative to what current technology is capable of? It looks like Pete's design was about 1% THD at 1V into 100 ohms from input to output (ie: a 12au7 gains stage and the transformers involved too) with no feedback. In the grand scheme of things is that really all that bad compared to what another tube topology would be able to do without feedback? I've only ever made high volume overdrive machines so I'm very new to the pursuit of linearity in tube circuits, no frame of reference I'm afraid.
Either way if raising the loop gain and closing the feedback as you have done gets it down to the .2 - .3% range at full output, that seems pretty reasonable. What would do better with 3 triodes and transformer per channel?
Brian
I think they mean relative at what can be achieved with a common cathode stage. A simple equivalent circuit for a triode is a voltage generator connected to the anode via the anode resistance ra. The simplest way to model distortion in a triode is as non-linearity in ra. If you do this then it is clear that the higher the anode load the less effect ra has on the output. If the anode load is infinite then the output is just the generator voltage and there is no distortion. A CCS load looks almost like an infinite load.
In a mu follower, the anode load can be many times bigger than ra so its effect is quite small. As a rule, if you have an anode load of at least 5 times and preferably 10 times the value of ra then the effect of the non-linearity in ra will be very small. A mu follower can easily achieve this and also drive reasonable external loads.
The SRPP has quite a small anode load so the effect of ra non-linearities is quite large, hence its reputation for high distortion. The one thing the SRPP can do is drive relatively heavy loads.
No, it's not bad at all. I built lots of different topologies for my headphone amp before I settled on the SRPP. There are plenty of topologies that offer lower output impedance or potentially lower distortion. However, none of them perform very well when presented with a heavy load like headphones, except the SRPP. I am not sure why people obsess about minimising the SRPP distortion by designing it for the load. In practice, its performance is not much worse at loads a long way from the optimum.
Precisely.
Cheers
Ian
I would go for a mu follower as this offers a simple way of achieving low distortion with good drive capability. The mu followers I designed using the 6CG7 could output 10V rms into a 10K load at about 0.2% distortion and a gain of 26dB. You could improve the distortion by using NFB.
Cheers
Ian
Thanks for the insights, Ian
Since I have time before the parts arrive I've been immersing myself in all things linearity. I'll be building exactly to the schematic for the first go around but I wonder if trying some of the CCS plate loading stuff that dsavitsk mentioned would be a potential upgrade down the road. Have you ever considered trying a CCS plate load on the 12AX7 to increase open-loop gain and linearity? The design "as is" seems so pure and simple I'd hate screw around, but making some BJT CCS wouldn't be too much strain and they look to be plug and play in exchange for the 220k resistors. Or for the price of a second 12AX7 one could active-load with a triode.
If i'm doing the math right- and by math I mean plugging numbers into this: Adam's Amplifiers: Triode Calculator - it looks like open loop voltage gain of the AX7 would go from 68 to 77 with an active load (Assumption of infinity ohms || 470k), and up 82 if the 470k grid resistor of the ECC99 could be raised to 1M (Not clear that it could be).
I wonder if 9 extra db of loop gain would net much in the distortion realm?
Brian
Since I have time before the parts arrive I've been immersing myself in all things linearity. I'll be building exactly to the schematic for the first go around but I wonder if trying some of the CCS plate loading stuff that dsavitsk mentioned would be a potential upgrade down the road. Have you ever considered trying a CCS plate load on the 12AX7 to increase open-loop gain and linearity? The design "as is" seems so pure and simple I'd hate screw around, but making some BJT CCS wouldn't be too much strain and they look to be plug and play in exchange for the 220k resistors. Or for the price of a second 12AX7 one could active-load with a triode.
If i'm doing the math right- and by math I mean plugging numbers into this: Adam's Amplifiers: Triode Calculator - it looks like open loop voltage gain of the AX7 would go from 68 to 77 with an active load (Assumption of infinity ohms || 470k), and up 82 if the 470k grid resistor of the ECC99 could be raised to 1M (Not clear that it could be).
I wonder if 9 extra db of loop gain would net much in the distortion realm?
Brian
Adding a CCS would increase the open loop gain. Every extra dB of open loop gain reduces the distrotion by a dB.
However, raising the gain from the existing 68 to 82 is not 9dB of gain. It is 1.6dB. (20 log 82/68).
Look at it this way. Raising it from 50 to 100 is double, which is 6dB. 100 is the best it could possibly be. We currently have close to 70 which is only 3dB below 100. That's the best improvement a CCS would bring.
Cheers
Ian
Progress Report
Well, my Sowter transformers have a 3-4 week lead time (+ transit) so I'm going to mock this whole thing up first and work out the bugs. I have a bad habit of making the enclosure first - before everything is tested and optimal - and this usually leads to tragedy… so I'm taking my girlfriend's advise and prototyping first.
I got through most of the power supply today…. death trap:
I haven't sorted out how to elevate the 12.6VDC regulator board, but I did determine that there is no common connection on the screw holes. Floating it above chassis ground won't be tough but I'm open to suggestions. Do I just tie the common of the board to a bypassed resistor divided? Does that put my low voltage filter caps in series with the higher voltage cap bypassing this +50ish volt rail (and therefore reduce there capacitance)? And should I float the AC lines AND the DC ground? Here's the data sheet on the kit I have: http://www.audiowind.com/pdf/A-230LM.pdf
Close up of the 300vDC series regulator:
I wired it up and crossed my fingers - worked the first time! 302vDC at the output with a 250k bleeder resistor. The raw filtered B+ is 400vDC, so worst case dissipation would be well under 10W. I'm planning for maybe 42mA at idle so really less than 5W continuous assuming no sag under load. Maybe I can get away without a heat sink (This FQA19N60 is rated for 300W before derating for temperature).
And a draft layout:
I went an bought fancy new tube sockets for this project but then realized I had a 3 x Dual triode prototyping board that I'd made for for experimenting with tube pedals. It looks like a good fit for this design. You can see the board in the first picture. The dashed line is a ground bus; I accounted for, though may not use, a separate supply line for each channel; and space for a resistor/cap decoupler on the first gain stage; Unmarked resistors are grid stoppers; ? resistors are where anode resistors would go on the SRPP, leaving room just in case; huge 4.7uf Solen caps will be out at the extreme ends. The plate resistors for the 12AX7 are next to each other in the middle with a little breathing room in case I ever do want to try CCS loading.
Since the filtering on the regulator board is a bit light for the heater current (only 1000uf before and after the regulator) for a full amp of heater draw, I'm going to cross my fingers that the input ripple doesn't trigger any drop-out and add filters at each tube socket across pins 4 and 5. One will be 12.6vDC and the other common (referenced to a higher DC potential) and I found 3 small 2200uF @16vDC caps that I can hang right off of the pins.
Tomorrow I'll press in the turrets and start wiring the board.
Brian
Well, my Sowter transformers have a 3-4 week lead time (+ transit) so I'm going to mock this whole thing up first and work out the bugs. I have a bad habit of making the enclosure first - before everything is tested and optimal - and this usually leads to tragedy… so I'm taking my girlfriend's advise and prototyping first.
I got through most of the power supply today…. death trap:
An externally hosted image should be here but it was not working when we last tested it.
[/URL]
I haven't sorted out how to elevate the 12.6VDC regulator board, but I did determine that there is no common connection on the screw holes. Floating it above chassis ground won't be tough but I'm open to suggestions. Do I just tie the common of the board to a bypassed resistor divided? Does that put my low voltage filter caps in series with the higher voltage cap bypassing this +50ish volt rail (and therefore reduce there capacitance)? And should I float the AC lines AND the DC ground? Here's the data sheet on the kit I have: http://www.audiowind.com/pdf/A-230LM.pdf
Close up of the 300vDC series regulator:
An externally hosted image should be here but it was not working when we last tested it.
[/URL]
I wired it up and crossed my fingers - worked the first time! 302vDC at the output with a 250k bleeder resistor. The raw filtered B+ is 400vDC, so worst case dissipation would be well under 10W. I'm planning for maybe 42mA at idle so really less than 5W continuous assuming no sag under load. Maybe I can get away without a heat sink (This FQA19N60 is rated for 300W before derating for temperature).
And a draft layout:
An externally hosted image should be here but it was not working when we last tested it.
[/URL]
I went an bought fancy new tube sockets for this project but then realized I had a 3 x Dual triode prototyping board that I'd made for for experimenting with tube pedals. It looks like a good fit for this design. You can see the board in the first picture. The dashed line is a ground bus; I accounted for, though may not use, a separate supply line for each channel; and space for a resistor/cap decoupler on the first gain stage; Unmarked resistors are grid stoppers; ? resistors are where anode resistors would go on the SRPP, leaving room just in case; huge 4.7uf Solen caps will be out at the extreme ends. The plate resistors for the 12AX7 are next to each other in the middle with a little breathing room in case I ever do want to try CCS loading.
Since the filtering on the regulator board is a bit light for the heater current (only 1000uf before and after the regulator) for a full amp of heater draw, I'm going to cross my fingers that the input ripple doesn't trigger any drop-out and add filters at each tube socket across pins 4 and 5. One will be 12.6vDC and the other common (referenced to a higher DC potential) and I found 3 small 2200uF @16vDC caps that I can hang right off of the pins.
Tomorrow I'll press in the turrets and start wiring the board.
Brian
Last edited:
Hi Brian,
Your layout looks OK to me. I checked out the heater regulator board. it is pretty standard. Just connect the transformer secondary to the AC input terminals. Connect the heaters to 0V(GND) and DC output pins.
Across you HT supply connect a 33K 2W, another 33K 2W and a 22K 2W in series with the bottom of the 22K connected to HT 0V and the top of the first 33K connected to HT +. The heater elevation voltage of about 75V appears across the 22K. Connect a 47uF 100V electrolytic across the 22K. Take a wire from the junction of the 22K and the second 33K to to 0V (GND) pin of the heater regulator. This will elevate the heaters to about 75V above ground. here is a link to another project where I do a similar thing but with different resistor values:
http://www.ianbell.ukfsn.org/EzTubeMixer/docs/PMEQP1A/PMTGMUcctjpeg
The power supply schematic is at the top.
Cheers
Ian
Your layout looks OK to me. I checked out the heater regulator board. it is pretty standard. Just connect the transformer secondary to the AC input terminals. Connect the heaters to 0V(GND) and DC output pins.
Across you HT supply connect a 33K 2W, another 33K 2W and a 22K 2W in series with the bottom of the 22K connected to HT 0V and the top of the first 33K connected to HT +. The heater elevation voltage of about 75V appears across the 22K. Connect a 47uF 100V electrolytic across the 22K. Take a wire from the junction of the 22K and the second 33K to to 0V (GND) pin of the heater regulator. This will elevate the heaters to about 75V above ground. here is a link to another project where I do a similar thing but with different resistor values:
http://www.ianbell.ukfsn.org/EzTubeMixer/docs/PMEQP1A/PMTGMUcctjpeg
The power supply schematic is at the top.
Cheers
Ian
Ian,
Thanks for checking it over. I put it together today and only found a couple of small errors, nothing irreversible. And thanks for talking me through the heater elevation… I'd done it with simple AC heaters (Voltage divider as you describe, to an artificial center tap made of 100 ohm resistors) but never with a floating regulator. It looks like the AC will float up with the rest of the circuit because the DC bias will have access so the winding via the bridge rectifier - I think this is exactly as simple as you describe.
Here's what it looks like so far:
The only parts missing are the coupling caps between the first stage and the SRPP (2 X 0.1uF), they're in the mail. The wiring went so well I forgot to put grid stoppers in anywhere so I'll see how it goes and modify as necessary. Per my layout I strung the feedback resistors under some other components to keep the path short… the 2 empty spots on each end are for bypass caps on the SRPP cathode resistor - I want to double check and make sure really need a big electrolytic in those spots or if the impedance of the upper triode will permit a smaller poly cap to be used and still maintain a reasonable -3db point. Things were getting tight in those spots (just south of the ECC99 sockets) so I opted to run a jumper on the "top" side of the board. Once the coupling caps are in I'll finish wiring up the power supply and power it up.
Transformers are quite a ways out so after a quick power up/debug I'll start thinking about a chassis for a while. I'm thinking about 8" wide at the front, 12" front to back and 3" high. I might do wood sides, Im not feeling the Hammoond 1441 CWW or HWCHAS offerings.
Brian
Thanks for checking it over. I put it together today and only found a couple of small errors, nothing irreversible. And thanks for talking me through the heater elevation… I'd done it with simple AC heaters (Voltage divider as you describe, to an artificial center tap made of 100 ohm resistors) but never with a floating regulator. It looks like the AC will float up with the rest of the circuit because the DC bias will have access so the winding via the bridge rectifier - I think this is exactly as simple as you describe.
Here's what it looks like so far:
An externally hosted image should be here but it was not working when we last tested it.
[/URL]
An externally hosted image should be here but it was not working when we last tested it.
[/URL]
The only parts missing are the coupling caps between the first stage and the SRPP (2 X 0.1uF), they're in the mail. The wiring went so well I forgot to put grid stoppers in anywhere so I'll see how it goes and modify as necessary. Per my layout I strung the feedback resistors under some other components to keep the path short… the 2 empty spots on each end are for bypass caps on the SRPP cathode resistor - I want to double check and make sure really need a big electrolytic in those spots or if the impedance of the upper triode will permit a smaller poly cap to be used and still maintain a reasonable -3db point. Things were getting tight in those spots (just south of the ECC99 sockets) so I opted to run a jumper on the "top" side of the board. Once the coupling caps are in I'll finish wiring up the power supply and power it up.
Transformers are quite a ways out so after a quick power up/debug I'll start thinking about a chassis for a while. I'm thinking about 8" wide at the front, 12" front to back and 3" high. I might do wood sides, Im not feeling the Hammoond 1441 CWW or HWCHAS offerings.
Brian
Last edited:
AC heaters should be fine. I usually choose a 12AX7LPS for the 12AX7 as its spiral wound heater helps reduce hum.
Sowter tend to be conservative with the lead time estimates. They always quote quite long times but often I have had them arrive in two to three weeks. Of course, I am in the UK so the shipping time is rather less.
I really like your prototype tag board. I think many people would find it useful. Did you ever think about making it available commercially?
Cheers
Ian
Sowter tend to be conservative with the lead time estimates. They always quote quite long times but often I have had them arrive in two to three weeks. Of course, I am in the UK so the shipping time is rather less.
I really like your prototype tag board. I think many people would find it useful. Did you ever think about making it available commercially?
Cheers
Ian
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.