Hi,
I have just completed the first of a series of full power tests on the 3pair Leach clone using the final transformer bought for the job.
I tested one channel of a stereo amp into 8r0 and 4r0.
37.05Vac into 8r just below visible clipping (=170W = 22.34dbW)
35.31Vac into 4r0 just below visible clipping(=311W = 24.94dbW)
changing from 8r to 4r reduced the maximum output voltage by 0.4dbV.
Absolutely no latch up as it comes out of clipping and no sign of instability at 3W & 300W into resistive load.
The rails run at +-58.55Vdc from a 1kVA 40Vac transformer when mains is at 240V
The test was carried out at 241Vac with the rails @ +-58.9Vdc
The 311W result pulls the supply voltage at the PCB fuses down to +-57.65Vdc. The F3.1A rail fuses survived the test.
Much better than I had been led to believe.
It agrees closely with the spreadsheet model, which shows the 4ohm (60degree) half way between 1second and 100mS SOAR.
At the worst point 2.2Apk above the protection locus.
That 4ohm 311W coincides with the DC SOAR and with the DC protection locus at 46degrees phase angle.
I was cautious enough to never publish my model figures, they appeared unbelievable, until now.
Delay through the amp averages about 1.5uS and HF response is 1db down @ 62kHz.
I have just completed the first of a series of full power tests on the 3pair Leach clone using the final transformer bought for the job.
I tested one channel of a stereo amp into 8r0 and 4r0.
37.05Vac into 8r just below visible clipping (=170W = 22.34dbW)
35.31Vac into 4r0 just below visible clipping(=311W = 24.94dbW)
changing from 8r to 4r reduced the maximum output voltage by 0.4dbV.
Absolutely no latch up as it comes out of clipping and no sign of instability at 3W & 300W into resistive load.
The rails run at +-58.55Vdc from a 1kVA 40Vac transformer when mains is at 240V
The test was carried out at 241Vac with the rails @ +-58.9Vdc
The 311W result pulls the supply voltage at the PCB fuses down to +-57.65Vdc. The F3.1A rail fuses survived the test.
Much better than I had been led to believe.
It agrees closely with the spreadsheet model, which shows the 4ohm (60degree) half way between 1second and 100mS SOAR.
At the worst point 2.2Apk above the protection locus.
That 4ohm 311W coincides with the DC SOAR and with the DC protection locus at 46degrees phase angle.
I was cautious enough to never publish my model figures, they appeared unbelievable, until now.
Delay through the amp averages about 1.5uS and HF response is 1db down @ 62kHz.
Thanks for that info Andrew.
You've just provided me with kick in the pants I needed to go and pull the amp ou of the box and continue onwards.
Was that with the values we descussed some time back post 1344?
Quote: I am trying a different set of values (not tested yet) 200r, 1k3, 11k, 22k, 0r337 and give 8.8A into a short and use 160W of capacity.
Regards
Ian.
You've just provided me with kick in the pants I needed to go and pull the amp ou of the box and continue onwards.
Was that with the values we descussed some time back post 1344?
Quote: I am trying a different set of values (not tested yet) 200r, 1k3, 11k, 22k, 0r337 and give 8.8A into a short and use 160W of capacity.
Regards
Ian.
I bought a box (1000) of 20r 5W radial wirewounds.TomWaits said:
Three in series=60r then fifteen sets in parallel=4r0 225W.
Four sets allow 8r 450W and 2r 450W or 4r 900W when used in combination.
Assembled on 42mm X 151mm PCB with 300mm of 2.5mm^2 speaker wire to 4mm banana plugs. 90 1mm diam holes to radial pitch took a bit of time. Strip board would be much quicker.
Have not measured inductance and have no idea if I need a cap or snubber to correct it to true resistance.
Anyone in the UK want/need 20r 5W for soft starts- going cheap.
Three in series allow transformers between 300VA and 1kVA to start on a small fuse.
Yes, those same values, I have not tested the protection yet. I would not expect the protection to cut in at 1kHz on full power resistive. It is miles away from both the protection locus and from the SOAR and there is the effect of the filter that delays triggering. It should be completely inauduble if it operates correctly.Loboone said:
I found a method in ESP site and Janneman has described Walker's (Quad) method.
I am still thinking how to do this without blowing up my amps if the protection fails to cut in.
Hi,
45Vac to 50Vac will put a LOT of extra stress on the output devices.
It may become just an 8ohm amp with that level of PSU voltage.
Jens' five pair PCB would be more suitable for these higher voltages.
That was why Jens originally designed the smaller board, to better suit lower voltages and higher impedances.
The sink (lying face down on the bench) got to about the limit for hand touching, I'd guess about 45degC.
As soon as I lifted it back up it started cooling and bias returned to usual warm current level within 30minutes.
My usual testing uses a 20db switch to turn up the power for a second or so to check and adjust for clipping.
Once this level has been found I then switch and hold at the higher level for about 2.5seconds to get an accurate output voltage. The 4.75digit DMM takes a bit longer than 3.5digit to settle.
In total this method puts full power through the load for about 6seconds over a period of about a minute or so. Then cooling while I swap loads etc.
I also use a 600r attenuator in front of the amp set to exactly the same gain as the amp. Input voltage and output voltage match until +20db is turned on. Makes scope display easy since voltages match or back off three clicks.
45Vac to 50Vac will put a LOT of extra stress on the output devices.
It may become just an 8ohm amp with that level of PSU voltage.
Jens' five pair PCB would be more suitable for these higher voltages.
That was why Jens originally designed the smaller board, to better suit lower voltages and higher impedances.
The sink (lying face down on the bench) got to about the limit for hand touching, I'd guess about 45degC.
As soon as I lifted it back up it started cooling and bias returned to usual warm current level within 30minutes.
My usual testing uses a 20db switch to turn up the power for a second or so to check and adjust for clipping.
Once this level has been found I then switch and hold at the higher level for about 2.5seconds to get an accurate output voltage. The 4.75digit DMM takes a bit longer than 3.5digit to settle.
In total this method puts full power through the load for about 6seconds over a period of about a minute or so. Then cooling while I swap loads etc.
I also use a 600r attenuator in front of the amp set to exactly the same gain as the amp. Input voltage and output voltage match until +20db is turned on. Makes scope display easy since voltages match or back off three clicks.
AndrewT said:
Understood. I'd use three or more pairs of TO3s, and probably a fan.
Jens was planning on running at 70V.
See also:
http://www.diyaudio.com/forums/showthread.php?threadid=1033
Hi,
I ran 3pair of 230W devices through Bensen's spreadsheet.
comparing 40Vac to 45Vac, the 8ohm capability on reactive load fell from an easy (well inside DC SOAR) 60degree phase angle to just 40degrees.
into 4ohm it fell from 46degrees to 23degrees.
I would consider it marginal on 4 to 8ohm speakers if driven hard.
I did not look to checking 50Vac.
If you do want serious power you will also need to check the drivers against the output gain (hFE) at worst current and phase angle.
Jens,
the bias falls from Vre=25mV to Vre=15mV when I let the sink get hot during testing. Still within the range suggested in other threads for optimum ClassAB bias.
The reduced bias allows the sink temp to recover more quickly after working hard and I suppose that is an advantage.
I tried a variety of caps on the output (700mVac) ranging from 100nF to 2.2uF//8r0. Still sending the squarewave test signal through the input filter.
There was very slight squaring of the corner into 100nF//8r0 but the bigger caps caused considerable oversheet (upto 30% at worst) lasting about 2 damped cycles before disappearing. At 1kHz the overshoot had disappeared into the dead flat calm within about 10% of the flat top. At 10kHz it lasted about 50% of the square wave top. I have not measured the ringing frequency but guessing from the ripple wavelength it could be around 100kHz.
Should I inject the test signal after the input filter?
Where should I scope the output signal? before or after the Thiele network?
3pair is not going to cut it on 45Vac unless you seriously undersize the PSU to protect the output stage on continous output.
I ran 3pair of 230W devices through Bensen's spreadsheet.
comparing 40Vac to 45Vac, the 8ohm capability on reactive load fell from an easy (well inside DC SOAR) 60degree phase angle to just 40degrees.
into 4ohm it fell from 46degrees to 23degrees.
I would consider it marginal on 4 to 8ohm speakers if driven hard.
I did not look to checking 50Vac.
If you do want serious power you will also need to check the drivers against the output gain (hFE) at worst current and phase angle.
Jens,
the bias falls from Vre=25mV to Vre=15mV when I let the sink get hot during testing. Still within the range suggested in other threads for optimum ClassAB bias.
The reduced bias allows the sink temp to recover more quickly after working hard and I suppose that is an advantage.
I tried a variety of caps on the output (700mVac) ranging from 100nF to 2.2uF//8r0. Still sending the squarewave test signal through the input filter.
There was very slight squaring of the corner into 100nF//8r0 but the bigger caps caused considerable oversheet (upto 30% at worst) lasting about 2 damped cycles before disappearing. At 1kHz the overshoot had disappeared into the dead flat calm within about 10% of the flat top. At 10kHz it lasted about 50% of the square wave top. I have not measured the ringing frequency but guessing from the ripple wavelength it could be around 100kHz.
Should I inject the test signal after the input filter?
Where should I scope the output signal? before or after the Thiele network?
bremen nacht said:
Yes I did, and have not see any problems with it.
I have an other amp (non Leach) it runs at +-68V DC with 2 x 300VA transformers pr. channel and lots of capacitors on the rails. It has run in a 6 Ohm system since 1996 and it has seen some abuse at times.
The output devises are Thoshiba 2SC3281/2SA1302A, they have never failed. The heatsink is not huge by any measure compared to what is often displayed on these pages...
\Jens
Jens' 3pair PCBs - plugable INPUT options.
Hi,
anyone out there that still has to assemble the PCBs?
The three input holes at the edge of the PCB can be filled with 0.1inch PCB pins. Signal ground, AC input, DC input.
Attach the input cable on the back of the pins, screen to signal ground, core to AC input pin.
Attach a 0.1inch shorting plug from screen pin to AC input pin for testing (Rs=0r0).
Move the shorting plug to engage one end for normal AC input.
Move the shorting plug to DC input + AC input to bypass the DC blocking cap.
This allows all combinations of input using the plugable 0.1inch pins. Very convenient.
I wish I had spotted this before I did my first pair of PCBs, now done to PCB 3 and gone back and added pins to PCB 1 & 2 (with difficulty).
From post1512
Hi,
anyone out there that still has to assemble the PCBs?
The three input holes at the edge of the PCB can be filled with 0.1inch PCB pins. Signal ground, AC input, DC input.
Attach the input cable on the back of the pins, screen to signal ground, core to AC input pin.
Attach a 0.1inch shorting plug from screen pin to AC input pin for testing (Rs=0r0).
Move the shorting plug to engage one end for normal AC input.
Move the shorting plug to DC input + AC input to bypass the DC blocking cap.
This allows all combinations of input using the plugable 0.1inch pins. Very convenient.
I wish I had spotted this before I did my first pair of PCBs, now done to PCB 3 and gone back and added pins to PCB 1 & 2 (with difficulty).
From post1512
Any clues forthcoming?
Andrew,
IMO You should exclude the output inductor and RC filter from this type of measurement, so you will have to get rid of them if you want to see the true step response of the amp...
The only problem is that the filter will be present when the amp is in use so I'm not really sure what the best way is to measure the step response.
What info do you want to get from the measurement? slewrate? stability?
\Jens
IMO You should exclude the output inductor and RC filter from this type of measurement, so you will have to get rid of them if you want to see the true step response of the amp...
The only problem is that the filter will be present when the amp is in use so I'm not really sure what the best way is to measure the step response.
What info do you want to get from the measurement? slewrate? stability?
\Jens
Hi Jens,
stability, with and without a source connected.
BTW, have you experienced a surge in the front end at switch on?
I am using 820uF caps and F200mA fuses and I lost a second one and don't know why.
Oh, how I wish I could find the European plug-ins for the fuse location. When I ordered mine the fuses arrived but the plug-in mounting had been de-stocked.
But, on both occasions the output offset stayed below 150mV, when one front end supply rail was down.
stability, with and without a source connected.
BTW, have you experienced a surge in the front end at switch on?
I am using 820uF caps and F200mA fuses and I lost a second one and don't know why.
Oh, how I wish I could find the European plug-ins for the fuse location. When I ordered mine the fuses arrived but the plug-in mounting had been de-stocked.
But, on both occasions the output offset stayed below 150mV, when one front end supply rail was down.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.