I was wondering what sort of performance one could get from an amp built with Hitachi 2SK1058 and 2SJ162 output transistors, if one took a "no holds barred" approach to the design.
I figured that with a Vdsmax of 160V, setting rails at +/-70V was reasonable. In order to get ample power into a typical 8 Ohm speaker, one could drive a pair of amps in a bridge configuration. Running the numbers, it's reasonably easy to do 800W, with +/-57V swing...
So I set to designing half that amp - something that's able to push 400W into a 4 Ohm load. I've attached the results of my couple-of-evening effort.
It's clearly Tilbrook inspired, but I've added a few output pairs. That meant a reasonable amount of gate current, so I added emitter followers to the VAS. I'm still playing with other transistors in the input to the VAS as well - this time I thought I'd try Phillips PBSS8110Y and PBSS9110Y duals. They've got an fT of 100, with hFE<80. Alas they're limited a little by their 100V Vceo, but in this spot that's not an issue. Being duals, I imagine they won't suffer from the thermal tracking issues of the SOT-23 MMBTA06/56 transistors I used in my 50W amp.
I'd have used 2SA1209 and 2SC2911 transistors in place of the MJE340/350s, but I haven't been able to find a spice model for these - their specs imply that they'd improve the performance significantly though.
It simulates okay - 0.002% THD pushing 400W into 4 Ohms at 1KHz... At 200W, that drops to 0.0008%. Alas, LTspice gets a little confused when I put two back-to-back to drive 800W into 8Ohms. I presume it would be better, with all those even harmonics being canceled.
I probably won't build it though, as that would be somewhat silly. But it was fun as an exercise.
I figured that with a Vdsmax of 160V, setting rails at +/-70V was reasonable. In order to get ample power into a typical 8 Ohm speaker, one could drive a pair of amps in a bridge configuration. Running the numbers, it's reasonably easy to do 800W, with +/-57V swing...
So I set to designing half that amp - something that's able to push 400W into a 4 Ohm load. I've attached the results of my couple-of-evening effort.
It's clearly Tilbrook inspired, but I've added a few output pairs. That meant a reasonable amount of gate current, so I added emitter followers to the VAS. I'm still playing with other transistors in the input to the VAS as well - this time I thought I'd try Phillips PBSS8110Y and PBSS9110Y duals. They've got an fT of 100, with hFE<80. Alas they're limited a little by their 100V Vceo, but in this spot that's not an issue. Being duals, I imagine they won't suffer from the thermal tracking issues of the SOT-23 MMBTA06/56 transistors I used in my 50W amp.
I'd have used 2SA1209 and 2SC2911 transistors in place of the MJE340/350s, but I haven't been able to find a spice model for these - their specs imply that they'd improve the performance significantly though.
It simulates okay - 0.002% THD pushing 400W into 4 Ohms at 1KHz... At 200W, that drops to 0.0008%. Alas, LTspice gets a little confused when I put two back-to-back to drive 800W into 8Ohms. I presume it would be better, with all those even harmonics being canceled.
I probably won't build it though, as that would be somewhat silly. But it was fun as an exercise.
Attachments
Hi,
I have a set of old Sugden P128 that use these Lateral Fets on +-70Vdc supply rails.
The original pair still run on the 2pair output stage.
They tested well into 8ohms but due to collapse of supply rail voltage they survived 4ohm testing with mediocre results.
I looked at SOAR and concluded that 2pair is stretching these devices a bit far, so I modified a pair to triple the smoothing capacitance and doubled the outputs to 4pair. They are now capable of properly driving 4ohm loads but only for medium term transients (i.e. music). The supplies still collapse into 4ohm loading. They do have a touch of sibilance and fail to rail on the output if a fuse blows so are out of use until I can sort both problems.
My conclusion for your use is, 4pair on +-70Vdc will drive 4ohms loads adequately even with a really stiff power supply. But there is a problem @ Vds between 5 & 6V and output current >10Apk. Your 5pair solves this.
Bridged this would make an enormous 8ohm amplifier.
But the heatsink will also need to be enormous if you go for Ib=100mA per device. I have to keep mine at 75mA and it still runs >50degC when the weather's warm.
Forget bridged into 4ohm unless you are prepared to go to >=6pair (24 FETs in each monoblock).
The Sugden runs the front end and drivers from +-78Vdc.
I have a set of old Sugden P128 that use these Lateral Fets on +-70Vdc supply rails.
The original pair still run on the 2pair output stage.
They tested well into 8ohms but due to collapse of supply rail voltage they survived 4ohm testing with mediocre results.
I looked at SOAR and concluded that 2pair is stretching these devices a bit far, so I modified a pair to triple the smoothing capacitance and doubled the outputs to 4pair. They are now capable of properly driving 4ohm loads but only for medium term transients (i.e. music). The supplies still collapse into 4ohm loading. They do have a touch of sibilance and fail to rail on the output if a fuse blows so are out of use until I can sort both problems.
My conclusion for your use is, 4pair on +-70Vdc will drive 4ohms loads adequately even with a really stiff power supply. But there is a problem @ Vds between 5 & 6V and output current >10Apk. Your 5pair solves this.
Bridged this would make an enormous 8ohm amplifier.
But the heatsink will also need to be enormous if you go for Ib=100mA per device. I have to keep mine at 75mA and it still runs >50degC when the weather's warm.
Forget bridged into 4ohm unless you are prepared to go to >=6pair (24 FETs in each monoblock).
The Sugden runs the front end and drivers from +-78Vdc.
suzy, maybe consider KSA1381/KSC3503 from Fairchild. These look like excellent transistors for VAS/predrive duties - and Fairchild have a good SPICE model 
Actually, I think it's a bit too good for ltspice.. i noticed that its quite easy to make the models exhibit bad rail sticking when pushed into clipping.. which im sure they don't in real use.
If you need more current for drive duty, they also have KSA1220/KSC2690... the capacitance is higher but they still look good.
Actually, I think it's a bit too good for ltspice.. i noticed that its quite easy to make the models exhibit bad rail sticking when pushed into clipping.. which im sure they don't in real use.
If you need more current for drive duty, they also have KSA1220/KSC2690... the capacitance is higher but they still look good.
ESP has the P101 which is a nice simple amp using the 2SK1058 and 2SJ162.
http://sound.westhost.com/project101.htm
Out of all Rod's amps, I like this one the best.
http://sound.westhost.com/project101.htm
Out of all Rod's amps, I like this one the best.
You might find that ditching the drive circuitry and using the LM4702 to be quite advantageous -- in a recent incarnation I was able to get the THD of a LM4702 + Lateral MOSFET amp down to 0.001% THD at 120W -- this by tweaking the compensation and simplifying the bias -- and with no additional drive transistors (the lateral MOSFETs were matched, however, and the load balancing resistors 100 milliOhms.) This was 0.002% better than the result I had obtained previously, and getting down to the resolution of my test setup.
On the bench right now I have an LM4702 with a higher current drive transistors -- 2SB649A + 2SD669A -- going through the paces so there's no data to report. The KSA1381/KSC3503 look very good as well with higher Ft, hfe, Vceo etc. and i will check these out as well.
On the bench right now I have an LM4702 with a higher current drive transistors -- 2SB649A + 2SD669A -- going through the paces so there's no data to report. The KSA1381/KSC3503 look very good as well with higher Ft, hfe, Vceo etc. and i will check these out as well.
suzyj,
I used my old tool again.
It appears that with rails of +-70V under load you will be able to pull 210W in 8Ohm and 400W in 4Ohm out off it. But beware, having rails of +-70V with a 4Ohm load will mean that you will have +-82V when there is no load connected!!
With a load lower than 8Ohm the chances are bigger that you're working out of the SOA of these nice devices.
This is simulated with:
* 5 devices in parallel
* 50°C heatsink temperature
* 45° phase shift due to the reactance of the speaker.
What do you think of 2SA1349 (2x2SA970) and 2SC3381(2x 2SC2240) devices for the diff stage?
Are the MJE15034 and MJE15034 not better than the 340/350 for the emitter followers?
Greetz
Ben
I used my old tool again.
It appears that with rails of +-70V under load you will be able to pull 210W in 8Ohm and 400W in 4Ohm out off it. But beware, having rails of +-70V with a 4Ohm load will mean that you will have +-82V when there is no load connected!!
With a load lower than 8Ohm the chances are bigger that you're working out of the SOA of these nice devices.
This is simulated with:
* 5 devices in parallel
* 50°C heatsink temperature
* 45° phase shift due to the reactance of the speaker.
What do you think of 2SA1349 (2x2SA970) and 2SC3381(2x 2SC2240) devices for the diff stage?
Are the MJE15034 and MJE15034 not better than the 340/350 for the emitter followers?
Greetz
Ben
jaycee said:
Hey, thanks for that. They look really good.
rabbitz said:
Yeah, Greg's amps are beautifully designed and really well thought out. I didn't design this though as something to build - more just an exercise to see what can be done with these transistors. Bear in mind I already have a pair of 100W amps using these transistors (and a similar circuit) that I'm very attached to.
jackinnj said:
That's very cool. I noticed the LM4702 app note about using it with MOSFETs, and was quite impressed at the figures they were getting. It looks like a really amazing chip.
jackinnj said:
Any chance of bunging up a schematic for that? It sounds very neat. I don't suppose there are models kicking around gor the LM4702...
Actually, I just found out the duals aren't. Duals, that is. The PBSS8110Y and PBSS9110Y are listed in Farnells catalogue as duals, but when you get the data, they're actual single transistors (though in six pin packages).
I went on a search for high-voltage duals, and found the IMX8, from Diodes inc. 120V Vceo, hFE>180, 140MHz fT... Even better than the NXP devices. They're only good for 50mA, and 300mW, but that shouldn't be a worry in the earlier stages.
I went on a search for high-voltage duals, and found the IMX8, from Diodes inc. 120V Vceo, hFE>180, 140MHz fT... Even better than the NXP devices. They're only good for 50mA, and 300mW, but that shouldn't be a worry in the earlier stages.
@suzyj
What do you think about 3 pairs on 4 ohms in this link:
http://www.turneraudio.com.au/solidstateamps1mosfets.html
The amp is meant to be used with above 4 ohm loads, but will tolerate normal home use with loads down to 2 ohms when the power is unlikely to average high levels.
What do you think about 3 pairs on 4 ohms in this link:
http://www.turneraudio.com.au/solidstateamps1mosfets.html
The amp is meant to be used with above 4 ohm loads, but will tolerate normal home use with loads down to 2 ohms when the power is unlikely to average high levels.
felixx said:
That would frighten me. I think it falls outside the SOA.
It's an interesting amp though - I'll bet the schematic would work a lot better if he connected the feedback loop up
I really disagree with his rationale for ditching the JFET input stage. I reckon it's one of the nicest bits about Tilbrook's topology. My 100W amp (with its dual JFET input stage) manages 12 nV/sqrtHz input voltage noise. I'll bet his is tens of dB above this.
Also, the high input impedance of the JFETs lets me completely do away with the capacitor in the feedback path, and DC couple the whole box and dice. It makes for some impressive thumps at power on and off, but is well worth the trouble, I reckon.
AndrewT said:
It's at
http://www.diyaudio.com/forums/showthread.php?s=&threadid=72993 .
I've also got a web page talking about it, at http://www.littlefishbicycles.com/poweramp/.
SUZYJ
Was this screen-shot of the 1kHz sim performed without the low pass output filter?
When measured the noise of the amplifier did you have the 400Hz filter of the SG505 switched in?
Was this screen-shot of the 1kHz sim performed without the low pass output filter?
An externally hosted image should be here but it was not working when we last tested it.
When measured the noise of the amplifier did you have the 400Hz filter of the SG505 switched in?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.