D3a Triode Square Wave shows HF boost

[RhythMick]

After spending way too much time and effort with work I finally got back to making progress on the phono. I'm starting with a real simple D3a Triode stage and want to get the output from that as clean as I can get it before I start experimenting with what comes after. I've set up the D3a as follows, all clip-leaded for now except the filaments which are wired.

Cathode biased with 110R, expecting 15mA OP so roughly 1.5V bias. Unbypassed.
G3 wired to Cathode.
G2 100R to Anode. Anode Load (various, see below) to B+. 100K Load resistor Anode to Ground.
G1 47K to Ground. 200R grid stopper.

PSU is transformer, SS Bridge Rectifier, LCLC delivering approx 250V clean unregulated DC.

1KHz square Wave input 100mV - grid (47k/200R junction) to Ground. Measuring the same on B probe (red).
A probe measuring across the 100K load resistor (Blue).

The square wave output shows a real HF boost. I need to get this right before I move on, but the setup is really straightforward and I'm struggling to understand what might be giving that HF boost ? I've tried changing the anode loads - picture below is with a simple 10K resistor load, but changing to choke load or choke + resistor makes no difference (choke load gain is much higher but the shape is the same). This is the first pentode I've worked with - is it possible I need a bigger grid stop, or bigger g2 resistor ?

Measurements (10K R Anode Load)
Ik = 14mA
Ia = 11mA
Ig2 = 3mA
Vak = 140V
Vg2k = 139V

All inputs welcomed

D3a Triode 10K R Load.PNG (26.58 KiB) Viewed 6043 times

[Nick]

I would measure with a sine wave sweep to see if there is a HF boost that you need to worry about. Also, 100nv is 30dB more than it will normally see, do you see the same effect with 1mv input?

[RhythMick]

I would measure with a sine wave sweep to see if there is a HF boost that you need to worry about. Also, 100nv is 30dB more than it will normally see, do you see the same effect with 1mv input?

Thanks Nick I'll report back tomorrow.

[izzy wizzy]

I thought a droop on the top of a square wave was LF rolloff not HF boost? I'm not seeing HF overshoot. Or am I missing something here?

Many use the d3a as input and driver in a power amp so that could see a 1V input. 10K anode load and 1.5V bias should be fine for HT around 300V

I'm not understanding why there is a 100k load for this test. What is the power supply capacitance feeding the stage?

My understanding is an unbypassed cathode resistor will degrade the performance of the stage by increasing Rp and reducing gm.

Edit: Done some homework on scope traces. I see the curve droop is indeed HF boost rather than straight droop being LF cut. My bad.

[Nick]

How is the scope input isolated from the anode voltage?

[RhythMick]

I thought a droop on the top of a square wave was LF rolloff not HF boost? I'm not seeing HF overshoot. Or am I missing something here?

Many use the d3a as input and driver in a power amp so that could see a 1V input. 10K anode load and 1.5V bias should be fine for HT around 300V

I'm not understanding why there is a 100k load for this test. What is the power supply capacitance feeding the stage?

My understanding is an unbypassed cathode resistor will degrade the performance of the stage by increasing Rp and reducing gm.

Edit: Done some homework on scope traces. I see the curve droop is indeed HF boost rather than straight droop being LF cut. My bad.

100K load arbitrarily chosen, representing the input load into the next stage (eg 100K grid resistor between next stage grid and ground). My understanding is there's little point measuring the unloaded output, so I loaded it as if the D3a was connected to the grid of the next valve.

My understanding is that an unbypassed cathode reduces the gain of the stage but reduces distortion as it's providing local negative feedback into the cathode.

PSU is is LCLC : 17H/200uF -> 17H/100uF

[RhythMick]

I would measure with a sine wave sweep to see if there is a HF boost that you need to worry about. Also, 100nv is 30dB more than it will normally see, do you see the same effect with 1mv input?

Thanks Nick I'll report back tomorrow.

I've reconstructed the test build so I can get directly to the pins. Producing the same result with 100mV square input.

10mV input same shape but much noisier, likely limitations of the test equipment (picoscope - I have a nice Siglent shamefully sitting on the desk because I haven't familiarised myself with it - doh).

002.PNG (31.89 KiB) Viewed 5985 times

I set up a sine wave sweep from 10Hz to 20Khz at 10Hz steps ...

003.PNG (17.2 KiB) Viewed 5985 times

...and set the scope for FFT...

004.PNG (63.3 KiB) Viewed 5985 times

The peak moves with the sweep frequency. If I stop it sweeping, set it to 1KHz the FFT shows the peak at 1KHz.

[izzy wizzy]

100K load arbitrarily chosen, representing the input load into the next stage (eg 100K grid resistor between next stage grid and ground). My understanding is there's little point measuring the unloaded output, so I loaded it as if the D3a was connected to the grid of the next valve.

My understanding is that an unbypassed cathode reduces the gain of the stage but reduces distortion as it's providing local negative feedback into the cathode.

PSU is is LCLC : 17H/200uF -> 17H/100uF

Good call that there is a load. How is it coupled to the anode?

The EQ which will be dangled off the end of it should be a lot lower in Z than 100k. More like 20k; probably less at the higher frequencies. I'd probably try to pull more current at nearer 1V bias and 4k7 on the anode for a triode but all good for now.

The idea IMO is to provide a low Z drive to the EQ so no unbypassed resistors here. I think the idea should be to maximise the gain and the performance (gm) of the valve so some form of low Z in the cathode is required. Feedback at this point isn't needed, again IMHO.

I realise you're only testing which is more than I've ever done. I'm just urging to test a higher performance circuit. PSU looks fine.

[RhythMick]

Ok my previous post prompted me to hook up the Siglent.

PXL_20210724_105843192_compress15.jpg (249.77 KiB) Viewed 5984 times

Looks better. The square wave is still coming from the Picoscope.

[Nick]

Much better. The remaining slope is probably the ac input protection. Try at 10khz

For the frequency sweep I meant an amplitude against frequency plot. Not an fft.

[RhythMick]

Much better. The remaining slope is probably the ac input protection. Try at 10khz

For the frequency sweep I meant an amplitude against frequency plot. Not an fft.

Ok thanks Nick I'll try and figure out how to do that.

[Mike H]

I was going to say, looks to me like a classic case of a series capacitor with not large enough value.

[RhythMick]

I was going to say, looks to me like a classic case of a series capacitor with not large enough value.

I think the Picoscope doesn't have probe calibration while the Siglent does.

[Nick]

I think Mike (and me) meant the DC blocking capacitor stopping DC from showing a horizontal line instead of calibrating for input capacitance.

If you used a 10kHz square, the flat top would not be so affected by the blocking cap.

[RhythMick]

I think Mike (and me) meant the DC blocking capacitor stopping DC from showing a horizontal line instead of calibrating for input capacitance.

If you used a 10kHz square, the flat top would not be so affected by the blocking cap.

Ok I see ... But there wasn't a blocking cap in the test circuit, either before or after the D3a.