Re: Hmmmmm....add a harmonic generator
You mentioned that you redesigned the hardware with discrete circuitry. Is that the fix you're referring to? Or was there a fix that gave the 846-based hardware improved subjective performance? If so, I'd love to hear what that fix was.
BTW, I've used the 846 in a radar signal processing application. Its performance was outstanding. It was processing fast rise time detected RF pulses. Its performance far surpassed a full custom solution that the government had already spent untold amounts of money on. I read an article about it by its designer, Wyn Palmer, and was really impressed with what he had done. The combination of low noise, low offset and high speed was so much better than anything we had used at the time that some specs we thought were impossible to meet became a walk in the park.
Jocko Homo said:
You mentioned that you redesigned the hardware with discrete circuitry. Is that the fix you're referring to? Or was there a fix that gave the 846-based hardware improved subjective performance? If so, I'd love to hear what that fix was.
BTW, I've used the 846 in a radar signal processing application. Its performance was outstanding. It was processing fast rise time detected RF pulses. Its performance far surpassed a full custom solution that the government had already spent untold amounts of money on. I read an article about it by its designer, Wyn Palmer, and was really impressed with what he had done. The combination of low noise, low offset and high speed was so much better than anything we had used at the time that some specs we thought were impossible to meet became a walk in the park.
Yes, it is a great chip, and Wyn is one sharp guy. I hear that the main reason they are DC'ing it is because it is intended for military applications, and us industrial types should use the '844.
All I did was to basically make one with discrete parts (obviously not as exotic, accurate, etc.) and find out how to make it work without feedback. That is about all I did.
Sounded great.
Jocko
All I did was to basically make one with discrete parts (obviously not as exotic, accurate, etc.) and find out how to make it work without feedback. That is about all I did.
Sounded great.
Jocko
Hi!!!
Just run some test on the amps.
First i looks classical right know.
I have tested it without Q3 and Q4. this is done by raising the current in the current source to vbe/100R + Iidle And connect the posetive currentsource to emitter of Q2.
It works but it drifts a lot. To much heat in the transistor and they have to thermally coupled.
560mW in the currentsource when running from the rail. 45V and 12.6mA.
So this solution was rejected.
So i went back using Q3 and Q4 running at ~ 555uA with a RE of 1k Q1 and Q2 with a RE of 100R => Iidle = 5.56mA
....
Ccomp is ~ 100pF (47pF*2 plus stray.)
I ended up with a bandwidth of ~ 300KHz with Rf = 5k and ~ 40V/us
with Rf of 2.5K, 600KHz and 60V/us.
It is stable.
The output is a CFP wich it was necessary to compensate with a 3.3nF from output to base of the output drivers.... This i am not happy about.
Without the output BJT's and only double Emitter follower i can push it without problems to 100V/us and 750KHz and it is stable.
The feedback path is ac coupled to ground.
The currentmirrors is standard with the output cascoded.
All for now .. back to more testing.
Sonny
Sonny, as I mentioned earlier, I messed around with SwitcherCAD and had current gain in the current mirror and the benefit of this was more speed with the same Rf. Why don't you try 1-2 mA in the - input and then have approx 10 mA in the high impedance node?
I know one person (you know the one?) who doesn't think 40 V/us isn't enough but if you want the amp faster, change to ring emitter transistors maybe?
BTW: Interesting to hear about your progress.
I know one person (you know the one?) who doesn't think 40 V/us isn't enough but if you want the amp faster, change to ring emitter transistors maybe?
BTW: Interesting to hear about your progress.
Yes i do think i know who!
:O)
Actually i am running at 60V/us with a bandwidth of 600KHz.
I will try your suggestion..
The reason i builded them was.. That we needed an amp fast who was stable... So i thought after reading you thread .. .Why not!?
go for the CFP design which i already have tried once.
When running at 1 - 2 mA @ in- you will switch it into class AB, also @ 20kHz with 100pF there will be a peak current draw of 500uA at ~ 40Vpeak.!?!?
What do you think??. Right know i am running at 5.56mA in both nodes.
:O)
Actually i am running at 60V/us with a bandwidth of 600KHz.
I will try your suggestion..
The reason i builded them was.. That we needed an amp fast who was stable... So i thought after reading you thread .. .Why not!?
go for the CFP design which i already have tried once.When running at 1 - 2 mA @ in- you will switch it into class AB, also @ 20kHz with 100pF there will be a peak current draw of 500uA at ~ 40Vpeak.!?!?
What do you think??. Right know i am running at 5.56mA in both nodes.
I would suggest 10-15 mA (or as much the transistors can take) for the high impedance mode and 1-2 mA for the input.
You could also tweak different degeneration resistors values of the current mirror. Start with 300 mV and go up to 1 V.
Have you tested my QRV-03 design with cascodes for the emitter followers?
Do you use Wilson current mirror?
You could also tweak different degeneration resistors values of the current mirror. Start with 300 mV and go up to 1 V.
Have you tested my QRV-03 design with cascodes for the emitter followers?
Do you use Wilson current mirror?
peranders said:
For some reason i haven't read your comment about currentgain in the currentmirror....
But it sounds like a real good idear.
then we can raise the size of Rf 5 - 10 times.... nice.
I will test it within this week.. if not then there will go another week (holiday) before i can test it again.
This i will do too if the time is on my side this week.
no but i will take a look at it
not in this PCB, right now i use a standard currentmirror with a cascode on the output of the currentmirror.. the next PCB will use a wilson mirror.
I have done some simulating on a SMD current feedback amp with Diamond buffer and the performance seems to be rather good. Bandwidth up to 25 MHz but how about real world influence? I expect over 1 MHz is very possible, maybe 10 MHz.
I notice that working point aren't very sensitive. 0.5 mA or 5 mA doesn´t mean much.
I'm very keen on making a small cute all SMD headphone amp, mayby with Jung Super Regulators.
I have attached the simulation files if somebody wants to fool around.
I notice that working point aren't very sensitive. 0.5 mA or 5 mA doesn´t mean much.
I'm very keen on making a small cute all SMD headphone amp, mayby with Jung Super Regulators.
I have attached the simulation files if somebody wants to fool around.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.