Great idea, I think I'll do that in the next one I build. Do you think I should also match my source resistors. When measuring the bias, they were all slightly different and I'm assuming this was down to the resistors as the mosfet were matched. I think I would have to come up with a different way of measuring the resistance as my multimeter isn't going to be much use at 0.47 ohm directly. Maybe use a regulator and another larger resistor (say a 3w 180r on a 18v reg, obviously I need to check the reg can handle that) to limit current and measure the voltage across the 0.47? Or is there a better approach like using some form of opamp comparator type circuit or some sort of bridge? I thought about using a ccs but at 10mA that leads to a voltage drop of 4.7mV so same problem.
I would think the absolute value is of little significance but they would want to be the same.
Or is it that the variance of the resistors is so small compared to the varience of the mosfet that they need to be matched and that's close enough.
Of course just buying a better multimeter would probably work.
I would think the absolute value is of little significance but they would want to be the same.
Or is it that the variance of the resistors is so small compared to the varience of the mosfet that they need to be matched and that's close enough.
Of course just buying a better multimeter would probably work.
Matching the mosfets and source resistors is what separates a good amplifier from a great amplifier. Nelson matches the mosfets much closer than most diyers. His standards are higher. He buys his parts by the thousands not hundreds and matches them. Having been involved in car racing this is what separates a racing motor from a production one. In a racing motor for example they buy the best valves made by the thousands and then only use the best matches out of the thousand for an engine. The difference is one reason a drag racing motor can make 10,000 hp vs a production 300 to 500 hp so called high power motor. It will take a very good lab ohm meter to accurately measure less than 1 ohm consistently though.
I wonder how good the sub ohm vaping coil meters are. The one on my battery pack displays to 3 decimal place and they are specifically for 0.1 ohm to a few ohms. I seem to get reasonably consistent readings but of course when putting power through it the resistance varies due to the temp, but I wonder if one designed to testing when winding the coils would be much better as it would be low current and not the 12amps or so when in use.
connect your source/emitter resistor is series. A pair of long insulated terminal strips works well.
Pass a typical source/emitter current through the string.
A Lab Supply set to EITHER voltage reg, or current reg does well.
Give the test system a few minutes to warm up and settle. Then measure the Vdrop on each resistor. Repeat to see what changes have ocurred, if big repeat again.
Now match up those that have very similar Vdrop.
If you only match transistors, then the variations in resistances will leave you guessing at why you measure different currents.
Once you know you have a good matched set of source/emitter resistors you then know that current variations in the amp are due to inaccuracies of the transistors.
Pass a typical source/emitter current through the string.
A Lab Supply set to EITHER voltage reg, or current reg does well.
Give the test system a few minutes to warm up and settle. Then measure the Vdrop on each resistor. Repeat to see what changes have ocurred, if big repeat again.
Now match up those that have very similar Vdrop.
If you only match transistors, then the variations in resistances will leave you guessing at why you measure different currents.
Once you know you have a good matched set of source/emitter resistors you then know that current variations in the amp are due to inaccuracies of the transistors.
Yes, that's a much better idea. I have a pretty robust bench supply. That'll do both voltage and current. The transistors were bought from a member here who matches them and I think that he does a good job of it, I'm fairly sure that the different voltages I get when biasing are in this case down to 5% resistors but that's rally useful to know for my next amp. Probably a ba3 (? basically an f4 with built in gain stage) to power my studio and another f4 most likely to biamp my main stereo.
Opps! xxxxxx Sugar!
Hi guys. My F4 build has been moving along and I got to the point of first powering it up with the boards.
Apparently, I hadn't done anything stupid enough recently, so I started this one out with a big one. I managed to connect the power supply to the boards backwards. In other words, V+ went to V- and V- went to V+. Only for a few seconds, but that was plenty to create failed components on the board. Happily, the power supply is OK.
Here is what I know.
- D3 and D4 (1N4148) failed in short circuit mode (at least one per board). One even burned the lead into two pieces. I may have parts to replace these on hand.
- P1 and P2 seem to be fine. They can be measured in place and have their normal range of resistance and smooth, linear change with adjustment.
- As far as I can tell checking them in circuit, all resistors are correct and not damaged.
- D1 and D2 show reasonable forward voltage (0.668 V on a DVOM diode tester). That said, I don't have any tricks to check their zener function in circuit. Replacing them would be smart on principle.
Here is what I think I know.
- The TL431 shows no signs of burning and shows Vka = 1.9 V, Vak = 0.557 V with Vref = 0. Looking at the datasheet, I haven't found a way to test these, but I would probably be smart to just replace them on principle.
In this condition, the DC offset is about 8.8 V (not surprising) and bias is low (30 mV).
Do you have any suggestions for checking the health of the jFets and Mosfets? Especially while on the board? What other components should I check or replace?
Thanks in advance for your help. Hopefully, I be able to avoid bone head moves like that in the future.
Jac
Hi guys. My F4 build has been moving along and I got to the point of first powering it up with the boards.
Apparently, I hadn't done anything stupid enough recently, so I started this one out with a big one. I managed to connect the power supply to the boards backwards. In other words, V+ went to V- and V- went to V+. Only for a few seconds, but that was plenty to create failed components on the board. Happily, the power supply is OK.
Here is what I know.
- D3 and D4 (1N4148) failed in short circuit mode (at least one per board). One even burned the lead into two pieces. I may have parts to replace these on hand.
- P1 and P2 seem to be fine. They can be measured in place and have their normal range of resistance and smooth, linear change with adjustment.
- As far as I can tell checking them in circuit, all resistors are correct and not damaged.
- D1 and D2 show reasonable forward voltage (0.668 V on a DVOM diode tester). That said, I don't have any tricks to check their zener function in circuit. Replacing them would be smart on principle.
Here is what I think I know.
- The TL431 shows no signs of burning and shows Vka = 1.9 V, Vak = 0.557 V with Vref = 0. Looking at the datasheet, I haven't found a way to test these, but I would probably be smart to just replace them on principle.
In this condition, the DC offset is about 8.8 V (not surprising) and bias is low (30 mV).
Do you have any suggestions for checking the health of the jFets and Mosfets? Especially while on the board? What other components should I check or replace?
Thanks in advance for your help. Hopefully, I be able to avoid bone head moves like that in the future.
Jac
Do you have a high current bench supply that you can slowly ramp up the voltage on the boards. I initially tested my boards that way and made sure that my biasing was at the lower range and worked my way up to plus minus 24v.
I think the mosfets should be off when unpowered because there's no gate voltage (enhancement mode or similar?) so might be worth measuring the resistance across the plus 24 to ground and minus 24v to ground.this wants to be disconnected from the psu when you do this. If it's very low then you have probably done your output devices. Im doing this from memory but hopefully will give you some avenues to explore.
I think the mosfets should be off when unpowered because there's no gate voltage (enhancement mode or similar?) so might be worth measuring the resistance across the plus 24 to ground and minus 24v to ground.this wants to be disconnected from the psu when you do this. If it's very low then you have probably done your output devices. Im doing this from memory but hopefully will give you some avenues to explore.
Thanks for the idea. I don't know what would be a low resistance, but everything seems even and reasonable, at least to me. I also measured resistance drain to source on each device in circuit.
V+ to GND 104k/146k boards A/B respectively
V- to GND 101k/145k
Board A
Q3 37.6k
Q4 37.6k
Q5 37.6k
Q6 39.7k
Q7 39.7k
Q10 39.8k
Board B
Q3 37.7k
Q4 37.7k
Q5 37.7k
Q6 40.8k
Q7 40.6k
Q10 40.8k
Jac
I take it that's after replacing the 1n4148 diodes that had shorted. Bit strange that there's a 40k discrepancy between boards. I think there's a strong chance the power mosfets are ok. Had a look on the the data sheet for 2sk170 and I can't seem to find a maximum opposite voltage for vgs. But you can put -40 on that device (assuming you don't melt the drain source!). Might be worth systematically measuring across each component in turn to try and figure what's causing the 104k to 145k difference between boards.
Thanks.
Unfortunately, I didn't find replacement 1n4148 diodes in my junk drawer. The measurements are with the bad parts. Since one has gone open circuit and the other board has a short circuit, that may explain the 40k difference. Tomorrow, I will pull the bad diodes and put in some smaller diodes, just to compare the boards. I'm afraid the little ones aren't high enough current to live under power. I will have to wait for Mouser before powering anything.
R2 is good on both boards. I'm using 100k instead of 47k because I am going to feed it with the dedicated Impasse. The 100k was one of SY's suggestions to keep the output caps on the Impasse reasonable in size while maintaining the upper bandwidth.
I looked up the Toshiba 2SJ74 datasheet (which these jFets actually are) and it has a breakdown voltage of 25 V which is close, but should be higher than the power supply voltage. I haven't seen a definitive statement of whether exceeding the breakdown voltage creates permanent damage or just goes outside of the usable range of the transistor.
Thanks again for the input. It's good to have someone else thinking about the problem.
Jac
Thanks. I will do that. I like a good sleep and I should order all of the parts I need at one time.
Jac
That is precisely what the the Mains Bulb Tester + protection diodes (on the load) protects against.
Just to close the loop. I removed all the diodes and the TL431 in preparation for replacing them. I also measured all components and compared them board to board to BOM. All resistors are fine and the semi's are the same board to board.
Finally, I remeasured V+ to GND and V- to GND on both boards and they are all now about 145k. Clearly, the bad diodes were affecting one board.
Thanks again. Off to place a parts order.
Jac
You are right of course. Low frequency response. SY recommended the 100k combined with a 0.47 uF Impasse output cap. With the original 47k input resistor on the F4, we would need a 1 uF output cap and it's hard to find one at 400 VDC and small enough to fit on the Tech-DIY board.
I got myself confused because I had been recently tuning the high frequency input response of the Impasse and had high frequency on the mind.
Jac
A bit off topic - sorry
This is Impasse related, so off topic, but I will try to keep it short.
Regarding high end extension, it may be a case for 6SN7 tube rolling. Paraphrasing SY, the 6DJ8 is fully degenerated and doesn't change the sound. We know the F4 is extremely neutral too. That means the 6SN7 will contribute the most to the sound character of the amp. Perhaps something like this, if you can find one locally.
Early Russian 6H8C / 6SN7 – Upscale Audio
The F4 has an upper band width limit of 200 kHz. I knew with the input transformers, we wouldn't get the Impasse that high, but I wanted to get as close to 100 kHz as possible. There are 3 things to trade off the upper frequency limit at the input of the Impasse, a clean square wave, upper frequency limit of the input impedance and tube Miller capacitance, and the upper frequency limit through the input transformers.
I am not using a pot for volume control as originally designed. In my case, the pot is replaced by a single resistor and my input impedance is constant.
Looking at the datasheet for the Cinemag input transformer, it was clear that the upper frequency limit of the transformer depended on how it's loaded.
I contacted Cinemag and they confirmed that the upper frequency limit continued to climb with higher load impedance, at least through 18k ohms. Assuming a linear relationship (a big assumption), 18k Ohms should give 92 kHz upper limit and 20k should give 100 kHz.
I am using a 5692 in the 6SN7 position. Using the capacitance for the 5692, the upper frequency limit based on Miller capacitance gives 120 kHz for an 18k Ohms and 108 kHz for 20k.
Square wave testing led me to 18k but I've got more testing to do.
Jac
This is Impasse related, so off topic, but I will try to keep it short.
Regarding high end extension, it may be a case for 6SN7 tube rolling. Paraphrasing SY, the 6DJ8 is fully degenerated and doesn't change the sound. We know the F4 is extremely neutral too. That means the 6SN7 will contribute the most to the sound character of the amp. Perhaps something like this, if you can find one locally.
Early Russian 6H8C / 6SN7 – Upscale Audio
The F4 has an upper band width limit of 200 kHz. I knew with the input transformers, we wouldn't get the Impasse that high, but I wanted to get as close to 100 kHz as possible. There are 3 things to trade off the upper frequency limit at the input of the Impasse, a clean square wave, upper frequency limit of the input impedance and tube Miller capacitance, and the upper frequency limit through the input transformers.
I am not using a pot for volume control as originally designed. In my case, the pot is replaced by a single resistor and my input impedance is constant.
Looking at the datasheet for the Cinemag input transformer, it was clear that the upper frequency limit of the transformer depended on how it's loaded.
I contacted Cinemag and they confirmed that the upper frequency limit continued to climb with higher load impedance, at least through 18k ohms. Assuming a linear relationship (a big assumption), 18k Ohms should give 92 kHz upper limit and 20k should give 100 kHz.
I am using a 5692 in the 6SN7 position. Using the capacitance for the 5692, the upper frequency limit based on Miller capacitance gives 120 kHz for an 18k Ohms and 108 kHz for 20k.
Square wave testing led me to 18k but I've got more testing to do.
Jac
Last edited: