A huge thanks, El Arte!!! I will check these today.
In my listening tests, those devices that had a wide and linear Vgs v Id region around the operating point sounded better. So, lets look at DMP45H4D9HJ3. I presume the current running through the device is less than 100 mA. If you look at the curve below, 100 mA is in the knee of the curve, right near the bottom. That part is not linear. You want to be in the region around 0.2-2A for this device where the current changes linearly as a function of Vgs. Now, while this bottom device is operated in follower mode, where the 100% feedback provides a high amount of linearization, I still heard this effect during listening (with the SCG). And when I switched to devices where the operating point is in the linear range of the curve, it sounded better. The problem is that there aren't any devices in the TO-220 package or any kind of package that can take some heat AND are also linear, by this definition, near the operating point we want. In P-channel devices, there may be none that are linear between 10-100 mA, say. You can check. I switched over to JFETs because they are linear around 10 mA and that is enough for a preamp application.
The caveat with all of the above is that the follower mode somewhat covers for this problem and in reality you may or may not hear this as a problem at all. It all depends on the overall architecture and how things play together. This is also just my theory that I have confirmed in listening, but I have not fully established that there are no other effects.
How much power dissipation is carried by that P-ch device in the UNSET architecture? You may be able to switch to paralleled TO-92 devices if the Pd is not that high (say, less than 1W). Also, I don't want to sound like a downer. For the most part, the follower mode covers this issue, and the gain device is a tube that is linear in the operating region. So, it is perhaps not a big problem. Just find a P-channel device that works and you are good to go. I am simply sharing what I've learned playing with this topology.
The UNSET and SCG bottom device operates as a source follower, which means it has no voltage gain and 100% feedback (these are the properties of a source follower). The middle gain device, the 'Schaded' device, is amplifying the signal coming from the source follower. Looking at simulation models, the distortion looks pretty low coming out of the source follower, but whatever is coming out gets magnified by the gain device. If you look closely, while the distortion products are low, but they are not always monotonic, i.e., H2>H3>H4 and so on. Sometimes, you get products H7, H9 and so on that are highest. Overall THD looks low from the follower alone, but then it gets magnified and coupled with the gain device's distortion products.
In my listening tests, those devices that had a wide and linear Vgs v Id region around the operating point sounded better. So, lets look at DMP45H4D9HJ3. I presume the current running through the device is less than 100 mA. If you look at the curve below, 100 mA is in the knee of the curve, right near the bottom. That part is not linear. You want to be in the region around 0.2-2A for this device where the current changes linearly as a function of Vgs. Now, while this bottom device is operated in follower mode, where the 100% feedback provides a high amount of linearization, I still heard this effect during listening (with the SCG). And when I switched to devices where the operating point is in the linear range of the curve, it sounded better. The problem is that there aren't any devices in the TO-220 package or any kind of package that can take some heat AND are also linear, by this definition, near the operating point we want. In P-channel devices, there may be none that are linear between 10-100 mA, say. You can check. I switched over to JFETs because they are linear around 10 mA and that is enough for a preamp application.
The caveat with all of the above is that the follower mode somewhat covers for this problem and in reality you may or may not hear this as a problem at all. It all depends on the overall architecture and how things play together. This is also just my theory that I have confirmed in listening, but I have not fully established that there are no other effects.
How much power dissipation is carried by that P-ch device in the UNSET architecture? You may be able to switch to paralleled TO-92 devices if the Pd is not that high (say, less than 1W). Also, I don't want to sound like a downer. For the most part, the follower mode covers this issue, and the gain device is a tube that is linear in the operating region. So, it is perhaps not a big problem. Just find a P-channel device that works and you are good to go. I am simply sharing what I've learned playing with this topology.
Last edited:
Checked all Mouser projects and made a couple of modifications in some places.
The most important change was: I put in terminal blocks for JP1, as I figured these can be easily jumped and also give you the option to connect leads for a switch that include or exclude the preamp buffer.
All orders are in.
The most important change was: I put in terminal blocks for JP1, as I figured these can be easily jumped and also give you the option to connect leads for a switch that include or exclude the preamp buffer.
All orders are in.
Thank you for the analysis, looking at the part I mentioned.
Looking at Transfer Characteristics for other candidates, including the FQPF9P25 which was specified, they seem to like a logarithmic chart, starting from 1A, so not deemed to be a useful range.
I guess the proof of the pudding will be in the eating.
Looking at Transfer Characteristics for other candidates, including the FQPF9P25 which was specified, they seem to like a logarithmic chart, starting from 1A, so not deemed to be a useful range.
I guess the proof of the pudding will be in the eating.
Yeah, it is hard to tell with the log graph. What these devices have going for themselves is immense transconductance. This means there is a ton of feedback available to reduce nonlinearities. So, it kind of works.
Built up my first board and connected it to my bench supply at 64V.
With C102 centered at 32V, roughly looks like it's swinging about 10Vrms before it starts to clip on the top end of the waveform. By reducing that down to around 28V I was able to get ~13Vrms before clipping.
I plan to run it on a ~70V supply and look forward to tweaking it once I have it set up properly.
I have had a bunch of these parts around since Dec of 2023, feels good to get the latest iteration of the board set up and working.
Thanks Rahul and Rick!
With C102 centered at 32V, roughly looks like it's swinging about 10Vrms before it starts to clip on the top end of the waveform. By reducing that down to around 28V I was able to get ~13Vrms before clipping.
I plan to run it on a ~70V supply and look forward to tweaking it once I have it set up properly.
I have had a bunch of these parts around since Dec of 2023, feels good to get the latest iteration of the board set up and working.
Thanks Rahul and Rick!
Last edited:
Beautiful! Thanks for sharing! Very eager to hear your sound impressions.
I have measured up to 15+ Vrms before clipping on mine, though it is powered by about 60V. I would lower the buffer bias current with a 70 V supply. The sweet spot is between 50-60V.
Edit: which amp will you pair with the SCG?
I have measured up to 15+ Vrms before clipping on mine, though it is powered by about 60V. I would lower the buffer bias current with a 70 V supply. The sweet spot is between 50-60V.
Edit: which amp will you pair with the SCG?
Last edited:
Metaphile - listen to what you have a for a week or so and then consider something other than that polystyrene cap.
I have long trusted Allen Wright's advice to avoid those.
I have long trusted Allen Wright's advice to avoid those.
I’ve had good success Wima FKP-4 in C101 and Wima DC link in C102. Other verified performers are Dayton Audio polypropylene and Janzten Audio. BTw, cap differences are easily heard. I am also hearing clear differences in transformers. I tried a no name toroid and it had great bass but the mids and highs were better with a Hammond EI.
Maybe you can also reverse Q101 when you swap caps and hear the effect 🙂
Maybe you can also reverse Q101 when you swap caps and hear the effect 🙂
Rahul - I am planning to use these SCG as the voltage gain stage in a version of the TDV. I have been doing some bench testing with the 2SK180's to find the sweet spot for Vds and Id. The plan is to run the 2SK180's at a low enough dissipation for passive cooling w/ the 4U heatsinks, so I'm looking at currents <2A and voltages from 22-34V or so.
For the amp: last night, I tried the common source sweet spot measuring circuit that lhquam posted and found that the lowest distortion was at 22V Vds (32V supply with 10V being dropped across the 8 ohm load resistor) and 1.28A current with -2.76V of bias voltage. So in a TDV implementation that would be 2.15 ohm of resistance, or a ~1.5ohm resistor in series with the .7 ohm DCR of the inductor. Interestingly, when I reconfigured the test setup to common drain mode and removed the 8 ohm load resistor from the drain of the SIT, I found that the distortion was higher and barely changed with current. That said, I don't have the inductors on hand so I am not really simulating the TDV circuit conditions and frankly I think there may have been something wonky going on with my test setup.
Once I get the inductors and start testing with those I'll post over in the TDV thread instead of posting here. 🙂
Rick - the capacitors I am using are Kemet C4G series and they are polypropylene, not polystyrene. Here is the datasheet. I have not used these particular ones before. I would have liked to use the CDE 940 or 942 series, which I used in a Tubelab TSE-II and which sound quite nice, but they are not available in high enough values.
I am interested to see how these ones sound in comparison and I'm also interested to hear the SCG + TDV amp and how it compares with the TSE-II w/ 300B's.
As for the power xformer for the SCG, I have a Hammond EI on hand that I was planning to try in this circuit. I can probably reconfigure the zener string to drop the voltage down to 60V if that is indeed the sweet spot. Will do some experimenting and report back.
For the amp: last night, I tried the common source sweet spot measuring circuit that lhquam posted and found that the lowest distortion was at 22V Vds (32V supply with 10V being dropped across the 8 ohm load resistor) and 1.28A current with -2.76V of bias voltage. So in a TDV implementation that would be 2.15 ohm of resistance, or a ~1.5ohm resistor in series with the .7 ohm DCR of the inductor. Interestingly, when I reconfigured the test setup to common drain mode and removed the 8 ohm load resistor from the drain of the SIT, I found that the distortion was higher and barely changed with current. That said, I don't have the inductors on hand so I am not really simulating the TDV circuit conditions and frankly I think there may have been something wonky going on with my test setup.
Once I get the inductors and start testing with those I'll post over in the TDV thread instead of posting here. 🙂
Rick - the capacitors I am using are Kemet C4G series and they are polypropylene, not polystyrene. Here is the datasheet. I have not used these particular ones before. I would have liked to use the CDE 940 or 942 series, which I used in a Tubelab TSE-II and which sound quite nice, but they are not available in high enough values.
I am interested to see how these ones sound in comparison and I'm also interested to hear the SCG + TDV amp and how it compares with the TSE-II w/ 300B's.
As for the power xformer for the SCG, I have a Hammond EI on hand that I was planning to try in this circuit. I can probably reconfigure the zener string to drop the voltage down to 60V if that is indeed the sweet spot. Will do some experimenting and report back.
Sorry - in the photo they looked like those nasty old polystyrenes. Glad to hear they are not. Nothing wrong at all with polystyrene as a dielectric but those things are throwaways.
Just looked at the datasheet and I sure did not see them clearly in the photo!
Rahul recommended I use a WIMA for this position and I have been pleased. Have not tried anything else. I did notice - when i got my boards - that the WIMAs pitch is smaller than what has been given on the PCB.
Another of my goofy things is to have the zeners drop as little voltage as possible to attain regulation. Rahul's fifth order filter before the zeners is not going to do anything about the noise of the zener string if there is a large drop. In fact, in my opinion, it is throwing their goodness away. Maybe, if you have enough current delivery in your transformer - see if you can lower the voltage with larger resistors. In fact, I would think using different values for the resistors could be advantageous instead of them all being the same. Vary the time constants. Use DUNCAN AMP Power Supply tool.
Those CDEs have become much harder to get, at least, at MOUSER. I use two of the huge 5 uF high voltage ones between my SCG and TDV.
I have found in my my weird system that no resistor is best for TDV. Just the choke. Using the LUNDAHL LL2733 (parallel) works well though I would like less resistance and a bit more Henry. But there are not many choices.
I have four of the HAMMOND 100mH 5A chokes that have been replaced by TOMBO's R25 power supply. You should look into this. I find it complimentary to the simple circuit. Has soft start as a built in feature and is very quiet. I had thought a buffer was immune to power supply noise but R25 proves this is not the case. Plus the marvel allows for a neater layout - no need for a big choke - much more compact.
I am tempted to try the HAMMOND in place of the LUNDAHL - for the bass channel. I figure the superior winding of the LUNDAHL is almost required for the tweeter amp.
You are going to be pleased.
Just looked at the datasheet and I sure did not see them clearly in the photo!
Rahul recommended I use a WIMA for this position and I have been pleased. Have not tried anything else. I did notice - when i got my boards - that the WIMAs pitch is smaller than what has been given on the PCB.
Another of my goofy things is to have the zeners drop as little voltage as possible to attain regulation. Rahul's fifth order filter before the zeners is not going to do anything about the noise of the zener string if there is a large drop. In fact, in my opinion, it is throwing their goodness away. Maybe, if you have enough current delivery in your transformer - see if you can lower the voltage with larger resistors. In fact, I would think using different values for the resistors could be advantageous instead of them all being the same. Vary the time constants. Use DUNCAN AMP Power Supply tool.
Those CDEs have become much harder to get, at least, at MOUSER. I use two of the huge 5 uF high voltage ones between my SCG and TDV.
I have found in my my weird system that no resistor is best for TDV. Just the choke. Using the LUNDAHL LL2733 (parallel) works well though I would like less resistance and a bit more Henry. But there are not many choices.
I have four of the HAMMOND 100mH 5A chokes that have been replaced by TOMBO's R25 power supply. You should look into this. I find it complimentary to the simple circuit. Has soft start as a built in feature and is very quiet. I had thought a buffer was immune to power supply noise but R25 proves this is not the case. Plus the marvel allows for a neater layout - no need for a big choke - much more compact.
I am tempted to try the HAMMOND in place of the LUNDAHL - for the bass channel. I figure the superior winding of the LUNDAHL is almost required for the tweeter amp.
You are going to be pleased.
Rick - I am actually using one polystyrene right now -- the 100pF for the optional feedback capacitor is a polystyrene. Had some left over from a headphone amp project. I assumed you were asking about the coupling capacitors, which are the polypropylene ones I linked above.
Thanks for sharing what you've learned from experimenting with the power supply, I will look into all of this!
I happen to have a 100VAC center tapped (so 2x 50VAC) transformer that, when I tested it unloaded at my line voltage, produced 79VDC on one side of the secondary. That's why I was thinking to run the SCG stage at around 70V so I'd only be dropping 9V in the regulator, maybe less under load.
I will experiment a bit (and btw I am not against the idea of using a power resistor to drop some voltage, I have done this in a tube headphone amp where I am using a 0D3 regulator with good results).
(Oh, and I already bought all the parts in the BOM for the PSU for the non-JFET version of the SCG so I am going to start by building a power supply with those first and go from there)
Thanks for sharing what you've learned from experimenting with the power supply, I will look into all of this!
I happen to have a 100VAC center tapped (so 2x 50VAC) transformer that, when I tested it unloaded at my line voltage, produced 79VDC on one side of the secondary. That's why I was thinking to run the SCG stage at around 70V so I'd only be dropping 9V in the regulator, maybe less under load.
I will experiment a bit (and btw I am not against the idea of using a power resistor to drop some voltage, I have done this in a tube headphone amp where I am using a 0D3 regulator with good results).
(Oh, and I already bought all the parts in the BOM for the PSU for the non-JFET version of the SCG so I am going to start by building a power supply with those first and go from there)
Last edited:
Just curious - are you saying that you're self-biasing the TDV but only using the DCR in the choke, no resistor? Or do you also have a separate bias supply to put some negative voltage on the gate?
And now I realize I might be confused because we're talking about the TDV and the SCG in the same thread - were you suggesting the Tombo R25 power supply for the SCG? Or for your TDV build?
For TDV and I would bet any MOFO style amplifier. Tombo's R25 has impressed me. Enough to convert four amplifiers - even though the fourth will have to wait until Wednesday.
For SCG I think what Rahul has provided is good and going any further with regulation would likely harm the sound. As he said early on the current source is doing a good job of isolating the gain part. Of course, if someone finds out otherwise I would be glad to know about it. Plus R25 would not work at the voltages with SCG - it would be close but the LT4320 is the limiter. Would be a waste since the thing can deliver lots of current and that is not what is needed with SCG.
No resistor in parallel with the choke. I am running about 39 volts. Big heatsink with fans that unfortunately generate noise into the amplifier. I have no idea how to stop that. Something to work on ...
The 6 to 10K input "load" resistor is the bias circuit.
I found out the hard way (even though I knew) by forgetting to replace the wire from this resistor to the choke/B- in my case, and you would not believe the amount of amps that were asked of R25 to provide a THF51. Luckily, it seems, only the output mosfet was destroyed. Waiting for a replacement to arrive tomorrow.
Have any others noticed how MOUSER is making shipments from their Canada warehouse? Of course, one could understand with the bad weather in Dallas doing this but it seems to have happened with all of recent orders and I make more than I should. You get them the next day even though you did not ask for next day. I doubt they are doing this just for me.
Playing around with larger resistors in PSUD and one would have to use really large ones. The supply would be supremely quiet but you wonder if it would be strangled for current? I wish I knew how all of this works!
For SCG I think what Rahul has provided is good and going any further with regulation would likely harm the sound. As he said early on the current source is doing a good job of isolating the gain part. Of course, if someone finds out otherwise I would be glad to know about it. Plus R25 would not work at the voltages with SCG - it would be close but the LT4320 is the limiter. Would be a waste since the thing can deliver lots of current and that is not what is needed with SCG.
No resistor in parallel with the choke. I am running about 39 volts. Big heatsink with fans that unfortunately generate noise into the amplifier. I have no idea how to stop that. Something to work on ...
The 6 to 10K input "load" resistor is the bias circuit.
I found out the hard way (even though I knew) by forgetting to replace the wire from this resistor to the choke/B- in my case, and you would not believe the amount of amps that were asked of R25 to provide a THF51. Luckily, it seems, only the output mosfet was destroyed. Waiting for a replacement to arrive tomorrow.
Have any others noticed how MOUSER is making shipments from their Canada warehouse? Of course, one could understand with the bad weather in Dallas doing this but it seems to have happened with all of recent orders and I make more than I should. You get them the next day even though you did not ask for next day. I doubt they are doing this just for me.
Playing around with larger resistors in PSUD and one would have to use really large ones. The supply would be supremely quiet but you wonder if it would be strangled for current? I wish I knew how all of this works!
You can use one half of a transformer to find out where you need to be - but my guess is that is far from ideal. I wonder if it would be worse with a center tapped transformer than a dual secondary one. My instinct is that it would be.
ZEN MOD would recommend always using the secondaries in series unless you want a separate rectifier for each secondary. This is just a bit of good advice from someone who does know what they are talking about. Somthing to consider in choosing a transformer - or a single secondary, which is what I am using.
It would be good to hear from Rahul or Vunce what they think of this.
ZEN MOD would recommend always using the secondaries in series unless you want a separate rectifier for each secondary. This is just a bit of good advice from someone who does know what they are talking about. Somthing to consider in choosing a transformer - or a single secondary, which is what I am using.
It would be good to hear from Rahul or Vunce what they think of this.
For now I'm just playing around on the bench using what I have; I was going to connect one leg of the secondary and center tap to a bridge rectifier so as to decrease the DC voltage to an acceptable level for this project (as I mentioned above this gives about 79VDC out when unloaded).
When the SCG goes into the real amp I will give it a proper transformer with an appropriate single secondary voltage (I like the small antek donuts with the steel covers).
When the SCG goes into the real amp I will give it a proper transformer with an appropriate single secondary voltage (I like the small antek donuts with the steel covers).
He is talking about using one half of a center tapped transformer - that is what concerned me.